quality and utilization of raw milk retentate produced at the dairy...

Quality and utilization of raw milk retentate

produced at the dairy farm

PhD thesis by

Ida Sørensen

May 2017

Aarhus University – Foulum

Department of Food Science

Denmark

ii

Preface

This PhD thesis concludes on the results obtained by Ida Sørensen from October 2014 to May 2017.

Experimental work was conducted at the Department of Food Science, Aarhus University, Foulum,

at Danish Cattle Research Center, Aarhus University, Foulum and at Danmark Protein, Arla Foods.

The project was a collaboration between Aarhus University, Arla Foods and GEA; and funded by

GUPD.

Main Supervisor

Lars Wiking

Associate professor, Department of Food Science, Aarhus University

Co-supervisor

Lotte B. Larsen

Professor, Department of Food Science, Aarhus University

Assessment Committee

Niels Oksbjerg (chairman)

Senior researcher, Department of Food Science, Aarhus University

Monika Johansson

Associate professor, Department of Molecular Sciences, Swedish University of Agricultural

sciences

Alexander Tolkach

R&D Director, Bayerische Milchindustrie

iii

Acknowledgements

I would like to give very sincere thanks my main supervisor Lars Wiking for guidance and patience;

and most of all, for giving me the opportunity to be at this point in my career. I truly respect the

work it has required on your behalf. I also owe much gratitude to my co-supervisor Professor Lotte

B. Larsen, for all the knowledge and support you have given me.

A special thanks is given to Gitte Hald Kristensen and Rita Albrechtsen for help and assistance on

some of the more complicated analysis. Without your help, I would not have been able to reach this

level. Moreover, I would also like to thank the rest of the technicians: Caroline, Hanne and Jens for

advice and inspiration; and, naturally, bringing a good atmosphere to the labs.

I would also like to thank all the people from DKC, AFI and GEA whom I had the chance to work

with, and who taught all I wanted to know about cows and membrane filtration; and saved my day

on several occasions. In addition, deep thanks to my family for being supportive and understanding

when the times have been tough. Furthermore, thanks to Helle for helping me with much needed

graphical assistance.

Finally, a truly heartfelt thanks to all my wonderful friends and colleagues in Foulum, whom I have

had extraordinary fun with, and who have always been there if I needed help or company.

iv

Abstract

Because of the increased demand on the dairy industry to be competitive on a global market and to

have focus on environmental issues, it has long been a wish to explore the option to concentrate

milk at the farm before transport to the dairy. When considering that the size of the average dairy

cowherd is increasing and the number of dairies is decreasing due to merging, it is an obvious

solution to reduce the transportation costs. By implementing membrane filtration plants for

concentrating the milk directly at the farm, the volume of milk for bulk tank storage and transport

could be reduced to half, and the remaining permeate could be utilized as drinking water for the

cows or for cleaning. Part of the purpose of this project is to process the retentate into e.g. cheese or

milk powder, where pre-concentrating the milk is already part of the standard procedure during

production.

Before implementation of this type of process, it is important to ensure that the milk quality is not

affected. Mechanical treatments are known to cause damage to the milk, especially concerning the

milk fat globules (MFG); moreover, enzymes and bacteria native to the milk are concentrated as

well. Consequently, it is easy to imagine that the filtration process could potentially be harmful to

the milk quality. The purpose of this study is thus to identify the issues related to concentrating milk

at the farm, how it will affect the milk as a raw material and in relation to further processing, and

evaluate whether implementation of the technology is advisable from a quality perspective.

As part of this study, it was examined whether membrane type (ultra-filtration (UF) and reverse

osmosis (RO)) and several membrane filtration process parameters had an influence on the MFG

size distribution, free fatty acid (FFA) concentration and extend of proteolysis. The results showed

that the temperature during the filtration process and feed pressure had a significant influence on the

FFA concentration in the retentate. However, this effect was only observed on milk from automatic

milking system (AMS); not in milk from conventional parlour milking system. It was there for

concluded that the concentrating process only had an impact on the milk, when the mechanical

aggregation during AMS had predisposed the milk for further damage. In spite of the much higher

pressure required to perform RO compared to UF, no difference in e.g. FFA concentration was

observed. In order to avoid waste of lactose, it was decided to focus on RO during the following

studies.

v

The rennet coagulation properties of RO retentate, as is essential during cheese production, was

compared to that of raw milk during the entire coagulation process. The study showed that the

coagulation onset time is later in retentate, but the gel firming rate was higher. Moreover, the

kinetics of chymosin was included as the hydrolysis rate of κ-casein (κ-CN) to caseinomakropeptide

(CMP), was evaluated based on varying concentrations of rennet and retentate dry matter. The study

showed that the ratio between chymosin and substrate was the determining factor for the reaction

rate of the hydrolysis, and thus, the time for coagulation onset. The distribution of calcium between

colloidal and serum phase was compared between retentate and raw milk was studied, and the

results showed that the retentate contained a higher portion of colloidal calcium and less serum

calcium. The change in ion equilibrium may have contributed to the changed coagulation

properties.

As part of the thesis, it was examined whether whole milk powder pre-concentrated by RO at the

farm had different properties and storage stability, compared to whole milk powder made from

regular raw milk. The study was affected by unintended changes to the milk composition (fat and

protein content) during the further processing prior to spray drying. This made it impossible to

conclude on differences directly caused by the RO process. Thus, several commercial whole milk

powders were included in the study as references in order to draw conclusions based on production

scale and composition on powder quality and storage stability. The results showed that the protein

content in particular had in influence on development of Maillard products during storage, reflected

on furosine concentration but not colour. The commercial powders had an initial higher

concentration of products from proteolysis than the experimentally produced powders from both

retentate and raw milk.

The contribution from this thesis has been a demonstration of how RO affects the milk and what

consequences that may have during further processing. From a quality point of view, there is no

hindering to implement membrane filtration at the farms. The overall status is however that there

are no economic and environmental benefits from this technology in Denmark, as the herd sizes and

distance to dairies is not enough to compensate for the extra cost of running many small filtration

plants. The implementation of the technology will however be an advantage in countries with

different conditions regarding transport distances.

vi

Sammendrag

Grundet det øgede præs på mejeriindustrien til at kunne være konkurrencedygtig på verdensplan, og

det samtidige fokus på miljø, har det længe været et ønske for mejerierne at undersøge muligheden

for at opkoncentrere mælken på gården, inden det bliver transporteret til mejerierne. Når man tager i

betragtning at den gennemsnitlige malkebesætning bliver større, og antallet af mejerier nedbringes

via sammenlægning, er det oplagt at forsøge at skære ned på omkostninger der er forbundet med

transport. Ved at implementere opkoncentrering direkte hos mælkeproducenterne vil volumen af

mælk til opbevaring og afhentning kunne halveres, og det resterende vand kunne bruges til

drikkevand til køerne eller til rengøring. En del af formålet med denne proces er at retentatet kan

videreforarbejdes til produkter såsom ost eller mælkepulver, hvor opkoncentrering af mælken

allerede er en del af standardprocessen under produktionen.

Før et system som dette kan implementeres er det nødvendigt at sikre sig at mælkens kvalitet ikke

belastes. Da mekanisk forarbejdning er kendt for at forvolde skade på mælken, især i form af

mælkefedtkugle (MFG) -beskadigelse, og mælken desuden bliver opkoncentreret med et intakt

indhold af enzymer og mikroorganismer, er det let at forestille sig at behandlingen måske kan have

en forringende virkning på kvaliteten. Således er formålet med dette studie at afgøre hvilken

betydning og konsekvenser opkoncentreringen vil have på mælken som råvare og under den

efterfølgende forarbejdning af retentatet, og ud fra et kvalitetsmæssigt synspunkt vurdere om denne

proces er tilrådelig.

Under studiets forløb har det været undersøgt hvorvidt membrantype (ultra-filtrering (UF) og

omvendt osmose (RO)) og proces-parametre tilknyttet membranfiltreringen, har haft betydning for

størrelsesfordelingen af MFG, indholdet af frie fedtsyre (FFA) og niveauet af proteolyse.

Resultaterne viste at procestemperaturen, og fødetryk havde indvirkning på dannelse af FFA, men

at denne effekt kun gjorde sig gældende i mælk fra automatisk malkningssystem (AMS), og ikke i

mælk fra en konventionel malkestald. Det blev derfor konkluderet at opkoncentreringen kun havde

en effekt når mælken i forvejen havde været udsat for hårdhændet behandling under malkningen og

derfor var prædisponeret for kvalitetsfejl. På trods af det væsentligt højere tryk som anvendes under

RO membran filtreringen, var denne proces ikke mere skadelig for mælken end UF. Så med henblik

vii

på at udgå spild af laktose gennem UF permeatet, blev det besluttet udelukkende at anvende RO i de

fremadrettede studier.

Retentatets løbe-koaguleringsegenskaber, der er essentiel ved fremstilling af ost, blev sammenlignet

med rå mælk i forhold til selve koaguleringsforløbet. Forsøgene viste at koaguleringen af retentate

starter senere, men med en hurtigere udvikling af koagelets fasthed. Yderligere, blev løbens

reaktionshastighed, baseret på hydrolyse af κ-casein til casein makropeptid (CMP), undersøgt under

varierende forhold mellem løbens koncentration og mælkens opkoncentreringsgrad. Det viste sig, at

forholdet mellem enzym og substrat var direkte bestemmende for reaktionenshastigheden af

hydrolysen, og dermed hvornår koaguleringen påbegynder. Desuden blev det undersøgt hvorledes

opkoncentreringen påvirkede fordelingen af calcium mellem micelle, serum og ionisk fase.

Resultatet viste at retentatet havde en højere andel af calcium bundet til caseinmicellerne og en

mindre andel som frie ioner sammenlignet med fordelingen i den rå mælk inden opkoncentrering.

Den ændrede mineral- og ionbalance kan meget ved være en medvirkende årsag til, at retentatet

opnåede hurtigere fasthed af koagelet.

Som en del af studiet blev det også undersøgt, om pulver lavet ud fra mælk der havde være

opkoncentreret på gården havde samme egenskaber og holdbarhed som pulver lavet ud fra

almindeligt rå mælk. Studiet bar præg af utilsigtede ændringer i råvarernes sammensætning (i form

af forholdet mellem fedt og protein), opstået under den videre forarbejdning, og det var derfor ikke

muligt at se en direkte konsekvens af opkoncentreringen på kvaliteten af det færdige pulver. Derfor

blev der i studiet inddraget adskillige kommercielle sødmælkspulvere til sammenligning, for på den

måde at konkludere sammenhængen mellem pulverets produktionsskala og sammensætning på de

endelige kvalitetsegenskaber. Resultatet viste, at især proteinindholdet havde betydning for

udvikling af Maillard produkter under lagring, hvilket afspejlede sig i indholdet af furosine men

ikke som farveændringer. De kommercielle pulvere havde som udgangspunkt en højre

koncentration af proteolyseprodukter end forsøgspulverene. Ligeledes blev der også udviklet

mindre eller tilsvarende mængder oxidationsprodukter i forsøgspulverene – både fra retentat og rå

mælk, sammenlignet med en kommerciel reference.

Denne afhandling har bidraget med en anskueliggørelse af hvorledes RO processen påvirker

mælken, samt hvilke konsekvenser det har for den videre forarbejdning. Der er ud fra et

viii

kvalitetsmæssigt synspunkt ikke noget til hindre for at implementere opkoncentrering af mælken på

gårdene. I det samlede billede af projektet, har det dog vist sig at den miljømæssige og økonimiske

gevinst kræver at gården har en vis minimum størrelse i forhold til afstanden til mejeriet som bliver

svær at imødekomme i Danmark. Implementeringen vil derfor være fordelagtig i lande med større

geografiske afstande mellem mælkeproducenter og mejerier.

ix

List of publications included in the thesis

Paper I

Chemical Quality of Raw Milk Retentate processed by Ultra-filtration or Reverse Osmosis

at the Dairy Farm

Ida Sørensen, Søren Jensen, Niels Ottosen, Tommas Neve & Lars Wiking

International Journal of Dairy technology, February 2016, Volume 69, Issue 1, Page 31-37

Paper II

Caseinomarcropeptide release in relation to rheological properties during rennet coagulation

of raw milk reverse osmosis retentate

Ida Sørensen, Thao T. Le, Gitte Hald Kristensen, Lotte Bach Larsen & Lars Wiking

Manuscript in preparation, intended for publication in International Dairy Journal

Paper III

Storage stability of whole milk powder produced from raw milk reverse osmosis retentate

Ida Sørensen, Tommas Neve, Niels Ottosen, Lotte Bach Larsen, Trine Kastrup Dalsgaard

& Lars Wiking

Dairy Science & Technology, 2017, Volume 96, Page 873–886

x

List of abbreviations

AMF Automatic milking system

CMP Caseinomacropeptides

FFA Free fatty acids

MF Micro filtration

MFG Milk fat globule

MFGM Milk fat globule membrane

NF Nano filtration

RO Reverse osmosis

TAG Triacylglycerol

UF Ultra filtration

VCF Volume concentration factor

ii

Contents

Preface ii

Main Supervisor ii

Co-supervisor ii

Assesment Committee ii

Acknowledgements iii

Abstract iv

Sammendrag vi

List of publications ix

List of abbreviations x

1. Project overview 1

1.1 Introduction 1

1.2 Aim and objectives 3

1.3 Experimental outline 3

1.3.1 Concentrating milk at the farm 3

1.3.2 Analytical methods 6

2 Raw milk quality 7

2.1 Milk composition 7

2.1.1 Fat 7

2.1.2 Protein 8

2.1.3 Lactose 10

2.1.4 Enzymes 11

2.1.4.1 Lipolysis 11

2.1.4.2 Proteolysis 12

2.1.5 Minerals 12

2.1.6 Somatic cells in milk 14

2.2 Primary production 15

2.2.1 Milking systems and pumping 15

2.2.2 Bulk tank storage 16

3 Filtration process 17

4 Cheese making process 20

4.1 Cheese production overview 20

4.2 Milk coagulation process 22

4.2.1 First stage 22

4.2.2 Second stage

4.2.3 The role of calcium

5 Milk powder 24

5.1 Milk powder manufacture 24

5.2 Milk powder quality 25

5.2.1 Powder composition and physical properties 25

5.2.2 Storage stability 26

6 Summary of included papers 28

6.1 Paper I: Chemical Quality of Raw Milk Retentate processed by Ultra-filtration or Reverse Osmosis at the Dairy Farm 28

6.1.1 Study objectives 28

6.1.2 Experimental setup 28

6.1.3 Summary of results 28

iii

6.2 Paper II: Caseinomacropeptide release and rheological properties during rennet coagulation of raw milk reverse osmosis retentate 29




6.3 Paper III: Storage stability of whole milk powder produced from raw milk reverse osmosis retentate 32




7 General discussion 34

8 Conclusion 41

9 Perspectives 43

10 References 45

Paper I

Paper II

Paper III

Appendix 1

Appendix 2

Appendix 3

Appendix 4

1

1. Project overview

1.1 Introduction

During the production of several dairy products such as milk powder, cheese and yoghurts, membrane

filtration is used to remove water, decrease volume in the cheese vats and standardize the protein

content of the milk. By transferring this process to the dairy farm, it would be possible to save costs for

transport and cooling, and reduce CO2 emission. As the heard sizes are currently increasing, and

certain countries have long distances between farms and dairies, this technology becomes even more

relevant. In cases with implementation of reverse osmosis membrane filtration it would even be

possible exploit the permeate as drinking water for the cows or for cleaning the equipment. Similar

practices have been implemented in the United States, Australia and New Zealand with little

documentation of the effect on milk quality. Generally, even though this technology is very well

known, the impact on the raw milk quality has not been fully studied.

Most literature on the subject of concentrating milk at the farm dates back to the 1980’s. Garcia III and

Medina (1988) described that reverse osmosis might have more advantages compared to ultrafiltration,

since ultrafiltration changes the composition of the milk by losing e.g. lactose in the permeate. This

makes the retentate less versatile for application in the dairy industry, and either value is lost in the

permeate or it needs subsequent treatment. Even so, a number of studies describe how ultra-filtration

could be implemented (Slack et al. 1982; Zall 1987a; Zall 1987b), with focus on the hygienic aspect of

microbial growth and handling and cleaning of the filtration equipment. An early study by de De Boer

and Nooy (1980) on RO showed that processing temperatures below 7.5°C lead to non-significant

increase of FFA concentration compared to higher processing temperatures. They concluded that the

main challenge would be to keep a high retentate quality, while applying enough pressure to prevent

fouling of the membranes. Kelly (1987) shared this concern, stating that the fouling effect would be

greater in raw milk compared to skim milk due to the presence of fat globules, which in turn would

cause a decreased flow and thus need of higher pressure. Pumping and general mechanical treatment

has been linked to milk fat globule damage, and thereby increased FFA formation (Wiking et al. 2003;

Fonseca et al. 2013).

2

This would result in quality defects such as off flavor (Nielsen 2005), and perhaps impair proper milk

fat separation (Evers 2004). Yet another major concern is that microorganisms and enzymes from the

raw milk would be concentrated as well, and perhaps affect the storage stability of the retentate.

This PhD study is structured in two parts according to the dairy processing route – from raw material to

final product. The first part is centered on the quality or raw milk retentate as directly affected by the

filtration process. Different membrane types (UF and RO) and processing parameters such as

temperature, pressure and concentration factor were studied in relation to the chemical quality of the

retentate (paper I). The second part used the knowledge gained from the first part to focus on the

functional properties of RO retentate, especially concerning cheese making properties (paper II) and

powder production (paper III). These products were chosen as obvious processing pathway of RO

retentate. Figure 1 shows a schematic overview of the study structure.

Quality of raw milk retentate in relation to processing parameters (Paper I)

Cheese making properties of raw milk RO retentate (Paper II)

Quality of whole milk powder made from raw

milk RO retentate (paper III)

Part 1

Part 2

Figure 1: Main structure of study elements included in this PhD thesis.

3

1.1 Aim and objectives

The general objective of this study was to examine and characterize changes to the milk after

membrane filtration of raw milk at the farm, both as a direct effect of the mechanical process, and due

to derived compositional and enzymatical changes. Thus, the aim of this study has been to correlate the

observed retentate attributes and quality defects to specific variables during the filtration process, such

as temperature, pressure and membrane type, and investigate how pre-concentrating raw milk will

reflect properties of further retentate processing into cheese curd and milk powder.

1.2 Experimental outline

1.2.1 Concentrating milk at the farm

The membrane filtration pilot plant was provided by GEA, assembled from existing parts. A model

drawing of the unit can be seen in figure 2. Danish Cattle Research Centre (DKC) provided the in situ

farm facilities and milk for the experiments. The filtration plant was built to be as versatile as possible,

with pumping capacity for both UF and RO, a plate heat exchanger for temperature control, and the

option of either continuous process or batch process. The specific setup differed with each experiment

(paper I -III). The plant was built with milk volume capacity fitting a medium size farm of

approximately 200 cows to accommodate the animal number at DKC. The milk used for the

experiments was mainly from Danish Holstein cows, and in some situations from a mixture of Danish

Holstein and Jersey. Both milk from automatic milking system and from herringbone parlour milking

was tested depending on the specific experiment. The cows were from herds of mixed lactation state

and feeding – especially the heard milked by automatic milking system had individual feeding

programmes depending on other parallel studies conducted at the farm. The raw milk was generally of

very high quality standard, with very SSC and CFU.

4

Figure 2: Drawing of the complete membrane filtration plant used for the experiments in this thesis,

including a small balance tank, plate heat exchanger and pumps.

Ideally, the RO permeate would have been utilized for either cleaning or drinking water. A drinking

water analysis was conducted (the approval failed on a phosphorus level of 1.8 mg/l and pH of 5.7),

and due to lack of resources, utilization of the RO permeate was not implemented.

Figure 3 shows how the filtration unit can be implemented in the process line between farm and dairy,

as it was done in this study.

5

Figure 3: Overview of the process line – from cow to final dairy product, where the membrane

filtration plant has been included as a continues process at the dairy farm.

6

1.2.2 Analytical methods

Table 1: overview of experimental techniques used and a reference to the paper where a detailed description is found.

Aim Method Paper

General milk composition FT-IR, MilkoScan I, II, III,

Appendix 3

FFA concentration BDI I

MFG size distribution Dynamic laser light scattering I

Proteolysis Flourescamine assay I, III

Total calcium Titration II

Ionic calcium Horiba electrode II

Rennet coagulation properties Convnetionel rheometer and ReoRox II

CMP concentration LC II

Fat content Rose-Gottlieb III, Appendix 4

Protein content Kjeldahl III, Appendix 4

Moisture content of powder Drying chamber III

Insoluble particles Dissolving, washing and centrifuged III

Surface free fat Petrolium ether washing III

Powder particle size distribution Sieving III

Oxidation HPLC III

Colour Minolta colourmeter III

Furosine concentration HPLC III

SSC Flow cytometry, Fossomatic Appendix 1

CFU Flow cytometry, BactoScan Appendix 2

Viscosity Rheometer Appendix 3

7

2 Raw milk quality

2.1 Milk composition

2.1.1 Fat

Fat is the milk components that differ most in content and composition through cow breed, feeding and

state of lactation, with an average content of 4 % (Walstra et al. 2006a). Milk is generally regarded an

oil in water emulsion, where the lipids primarily exist as triglycerides (TAG). Furthermore milk is

considered the fat source of most diverse TAG composition (Fox and McSweeney 1998c; MacGibbon

and Taylor 2006). Only very small parts of the milk fat exist as free fatty acids or as mono- and di-

glycerol (Jensen, 2002). The main fatty acids are palmitic, oleic, stearic and myristic acid. During

summer season a decrease of palmitic, myristic and caproic acid concentration can be observed

together with a higher concentration of oleic and stearic acid, thus giving a less saturated lipid

(MacGibbon and Taylor 2006; Huppertz et al. 2009). Organic milk has been found to have a higher

content of poly unsaturated and n-3 fatty acids, which is linked to changes in fodder composition in the

same way as differences between summer and winter season (Ellis et al. 2006). Milk fat synthesis is

divided into different pathways – each dominant during certain circumstances and will affect the fatty

acid composition (Palmquist 2006): Blood plasma lipids and de novo synthesized fatty acids. Blood

plasma lipids are the lipids directly absorbed from the blood stream into the mammary glands. This

account for nearly all the long chain fatty acids in the milk - the C18 and longer fatty acids, and some

of the C16. The blood plasma lipids are mostly in a TGA form, and only a small portion arises from

non-esterified fatty acids (Grummer 1991). The proportion of plasma lipids, and the form in which they

are taken up, is highly dependent on the cow’s energy balance and stage of lactation. During the

negative energy balance that is associated to the first state of lactation, a higher net uptake of non-

esterified fatty acids occurs, together with a higher level of body fat mobilization (Palmquist 2006).

Microorganisms in the rumen will, as part of the feed digestion, ferment the main part of the dietary

fats. Especially due to the microbial bio hydrogenation, lipase and esterase activity. Rumen

microorganisms will furthermore novo synthesize fatty acids, yielding mainly C18 and C16 or C18:1

and C16:1 if the pathway is anaerobic (Jenkins 1993).

8

When the fat is secreted from the mammary glands into the milk, parts of the apical plasma membrane

and parts from the cytoplasm of the secretory cell conceals the fat droplets and herby form a membrane

layer (Evers 2004). The milk fat globule membrane (MFGM) has a loose structure, and consists mainly

of bipolar material such as proteins, enzymes, phospholipids and cholesterols (Jensen 2002).

Depending on breed and state of lactation, the MFG size can range between 0.1 to 20µm (Fox and

McSweeney 1998c).The MFGM suppresses fat droplet merging due to decrease of surface tension

between fat droplets and water phase (Fox and McSweeney 1998c). Additionally, the MFGM acts a

protective barrier against milk lipase enzymes (Deeth 2006), thus preventing off flavor caused by free

fatty acids and lipid oxidation products.

2.1.2 Protein

Milk consists of 3.5% protein on average. However, like the case of fat, milk proteins is a highly

diverse group, where the amount and composition is influenced by state of lactation and genetics (Fox

and McSweeney 1998b; Ginger and Grigor 1999). The proteins are normally considered as two groups:

caseins and whey proteins, with an average ratio of 80:20 %. This fractionation was discovered as early

as 1830 based on acid precipitation, where caseins would precipitate at pH 4.6 (iso-electrical point) and

whey proteins would remain dissolved (Fox 2008). This division remains, but many sub-groups of both

whey proteins and caseins have later been discovered, together with several genetic variants and other

serum proteins (MFGM proteins, blood serum proteins, enzymes etc.).

Whey proteins consists mainly of β-lactoglobuline (10% of total milk protein) and α-lactalbumin (4%

of total milk protein). Early on, whey as regarded a waste product from cheese production, but

development of advanced dairy technology made it possible to utilize the whey proteins as functional

ingredients and infant formulas (Smithers 2008).

Caseins consist of αS1-CN, αS2-CN, β-CN and κ-CN connected by calcium-phosphate linkages to a

colloidal spherical micellar structure (Rasmussen et al. 1999; Farrell Jr et al. 2006). The actual structure

has been the topic of much debate, and considerable research has been conducted in order to shed light

on that matter, especially on the matter of sub-micelles compared to a random structure (Phadungath

9

2005; Horne 2006; Fox and Brodkorb 2008). Schematic models and electron microscopy pictures of

casein micelles is seen in figure 4. αS1-CN is the most abundant milk protein, and together with αS2-CN

construct the “framework” of the casein micelles due to the high content of sulphuric bridges formed.

The second most abundant protein is β-CN, which has the highest hydrophobicity. It has dynamic

abilities and can drift from the casein micelle into the serum phase at low temperatures, this tendency

can be reversed over time if the temperature is increased (Walstra et al. 2006b).

Figure 4: Casein micelle structure models and pictures. A: model based on the sub-micelle theory

(Walstra 1999). B: Tangled open structure (Holt 1992). C: Field Emission Scanning Electron

Microscopy image of casein micelle (Dalgleish et al. 2004). D: Transmission electron microscopy

picture of casein micelle (Karlsson et al. 2007b).

Pressure treatment of milk at >100MPa has shown to irreversibly change the micelle structure by

increasing the solubility of α-CN and β-CN (Huppertz et al. 2004). κ-CN has hydrophilic properties

10

and is primarily situated on the outer surface of the micelles. The hydrophilicity, together with a

negative electrostatic repulsion, keeps the micelles suspended in the milk (Fox and McSweeney 1998b;

McSweeney and Fox 2013). These properties are the basis in cheese production, where cleavage of the

hydrophilic outer part of κ-CN and reduction of negative charge (by lowering pH and adding Ca2+)

allows the casein micelles to merge.

2.1.3 Lactose

Lactose is the main carbohydrate in milk, with an average concentration of about 4.8%. Lactose is a

disaccharide of galactose and glucose, and regarded the main osmole in milk giving a total osmotic

pressure of 700kPa (Fox and McSweeney 1998d). During secession of lactose, water is drawn to the

Golgi vesicles according to osmolarity. Thus, lactose remains a constant concentration in the milk, but

the net-amount decreases during lactation stages due to overall decrease in milk yield (Fox 2008). An

exception occurs when cows suffer from mastitis; a decrease in lactose concentration is observed as a

regulatory system to compensate for the increased NaCl concentration giving a high osmotic pressure

(Auldist et al. 1995). Feeding regulations influencing blood glucose levels and farm managements can

however lead to an indirect increase in lactose output through overall milk yield (Broderick et al. 2002;

Auldist et al. 2007; Grainger et al. 2009). A high milk yield is associated with a lower protein and fat

content and thus lactose constitute more of the milk solids (Jenness and Holt 1987).

Previously, lactose was mainly considered as a waste product from cheese production as part of the

whey. Advances in technology has however made it possible to utilize lactose (Atra et al. 2005).

Human milk contains 6.9% lactose. The increasing demand for infant formula has thus made lactose a

highly valuable ingredient rather than a waste product. Lactose is the main substrate for starter culture

microorganisms during production of cheese, yoghurt and other fermented dairy products, and hereby

indirectly responsible for pH of the final product (Fox and McSweeney 1998a). The starter culture will

however not be able to utilize all the available lactose before the fermentation process stops at around

pH 4.4. Depending on the buffering capacity of the cheese curd, it might be necessary to restrict the

amount of available lactose in order to remain at the desired pH. This can be done by washing the

cheese grains for excess lactose (Nielsen 2004).

11

The properties and quality of milk powder are highly determined by the crystal state of lactose. The

crystals are formed when the temperature and relative humidity conditions fall between glass transition

state and melting point. Depending on the exact conditions different crystal polymorphs occur. This

will influence e.g. solubility and flowability; but also storage stability, as water can be released from

the crystalline structure. Moreover, lactose plays a major role in formation of maillard products – both

during the drying process but also during subsequently storage (Thomas et al. 2004).

2.1.4 Enzymes

The enzymes found in raw milk have two possible sources: Microbial contamination or indigenous

milk enzymes. The indigenous enzymes enter the milk either through somatic cells (Larsen et al. 2004),

from blood stream through leaky junctions, or secreted by the mammary epithelial cells. Especially the

complex MFGM contains a large number of different enzymes (Fox 2008). Today, a minimum of 70

different enzymes has been identified in raw milk, but only few has a known role during biosynthesis

of milk components or post-secretion. Many of the enzymes do not have access to proper substrates,

needs activation or the conditions in milk are not optimal. However, especially lipases such as LPL and

proteases e.g. plasmin are known to have an impact on the milk quality during processing and storage

(Fox and Kelly 2006).

2.1.4.1 Lipolysis

The main lipase enzyme in milk is LPL (EC 3.1.1.34). LPL is otherwise associated with very low

density lipoproteins in the blood stream where it hydrolyses triglycerides into free fatty acids. It will

hydrolyse fatty acids from sn-1 and sn-3 position in triglycerides, but the specificity is generally very

low. In bovine milk LPL is mostly connected to casein micelles. LPL is generally dependant on being

activated by Apolipoprotein C II in order to hydrolyse triglycerides, and hereby limiting the activity.

Still, LPL is the milk lipase responsible for spontaneous lipolysis of milk during storage. The activation

of LPL seems to be facilitated by rapid temperature changes and aggregation (Stepaniak 2004).

Moreover, the MFGM acts as an efficient barrier, shielding the milk triglycerides from LPL (Deeth

2006). Sundheim (1988) observed that LPL activity varied among individual cows, state of lactation

and even according to time of day, the milking was conducted. This emphasizes the complexity of the

system of activators and inhibitors (Bengtsson and Olivecrona 1980). The major concern regarding

12

lipolysis in milk is the formation of off-flavours from the concentration of FFA (Pillay et al. 1980;

Wiking et al. 2002; Santos et al. 2003). Wiking et al. (2002) showed that the concentration of FFA

during cold storage of milk increased within the first 24 h, with a constant concentration during

subsequent storage. They found this result reflected in the sensory analysis. Later on, Gargouri et al.

(2013) reached a similar conclusion as Wiking et al. (2002); that LPL is indeed the lipase responsible

for FFA emission compared to lipases from somatic cells and microorganisms, and that the major part

of spontaneous lipolysis occurs within the first day of storage. In his review, Deeth (2006) expressed

concern about microbial lipases being more heat stable, and might thus cause quality defect during

further processing rather than during bulk tank storage.

2.1.4.2 Proteolysis

Plasmin is the main protease system in good quality milk. Other protease systems such as cathepsin and

elastase are associated with an increased level of SSC (Kelly et al. 2006). Plasmin is not secreted

through the mammary glands, but only from the blood stream, leading to an increased content during

time periods with leaking tight junctions e.g. during udder infections and late state of lactation

(Nicholas et al. 2002). In milk from healthy cows, the inactive form – plasminogen is most prevalent.

Thus, the proteolytic activity is heavily regulated by the presence of activators, inhibitors and

plasminogen activator inhibitors (Ismail and Nielsen 2010). Plasmin is rather heat stable, and is

therefore known to affect the milk quality after heat treatment. β-CN is the primary substrate of

plasmin, however, αS1-CN, αS2-CN may also be hydrolysed at lower rates. Due to the hydrolysation of

casein, high plasmin activity has been linked to lower cheese yield and longer coagulation time (Lucey

and Fox 1992).

2.1.5 Minerals

Milk contains on average 0.7-0.8 % minerals, measured as total ash content (Fox and McSweeney

1998e), mainly phosphate, citrate, potassium, calcium, chloride, sodium and magnesium, together with

a large number of trace elements such as copper, iron and zinc. The minerals exist in several different

phases in the milk: colloidal – as organic or inorganic, and in serum phase as ions or bound in as

undissolved salts, where an equilibrium with pH between the phases is affected by temperature

(Shennan and Peaker 2000; Gaucheron 2005). Especially calcium and phosphate are part of a very

13

complex system related to the casein micelles as colloidal calcium phosphate nanoclusters, and as

calcium bound to phosphorylation of serine in the caseins (Mekmene and Gaucheron 2011; Bijl et al.

2013). This makes them highly responsible for the casein stability and structure. Table 2 shows the

average content and phase distribution of some of the major milk minerals.

Table 2: Mineral content(mM) and distribution in bovine milk. The table is adapted from (Mekmene et al. 2009).

Pinorganic Citrate Potassium Calcium Chloride Sodium Magnesium

Total 21 9 38 30 29 22 5

Colloidal 11 0.9 0.35 20.9 0 0.22 1.29

Serum 10 8.1 37.7 9.10 29 21.8 3.71

Ionic 7.3 0.2 36.26 2.12 27.8 20.9 1.14

Minerals enter the milk through five main routes: The membrane route, through the apical or

basolateral membranes (e.g. Na+, K+, Pi and Cl-); the Golgi secretory route, where proteins and lactose

are secreted (e.g. citrate and calcium); the milk fat route, where some minerals can be associated to the

MFGM and be part of certain hormones and enzymes; the transcytosis route, through expression of

immunoglobulins, transferrin and prolactin – especially during production of colostrum. The fifth

route, the paracellular route, is highly associated to mastitis where direct transport of components from

the interstitial fluid is seen (Shennan and Peaker 2000).

The total mineral content rarely changes, but the specific content of each mineral can exhibit a large

variation (Fox and McSweeney 1998e). Minerals are essential constitutes to calf nutrition, and thus the

mineral content is highly regulated by the lactation state. Each mineral is differently affected by the

state of lactation, and especially calcium secession increases during early and late lactation. Other

minerals, e.g. citrate and potassium concentration will however decrease during the course of lactation

(Gaucheron 2005). Sodium and chloride concentration are known to increase at early and late lactation,

and during an udder infection, due to weakening of tight junctions and thereby entry of interstitial fluid

into the milk. The cow nutrition has very little effect on the minerals in the milk (Schwendel et al.

14

2015), but other factors, such as breed can indirectly affect the mineral content, as Ca and p

concentration are showed to correlate to high protein and casein (Gaucheron 2005).

2.1.6 Somatic cells in milk

The milk secreting cells in the cow udder are connected via tight junctions, regulating the barrier

between milk and blood. This system can, however, be disturbed. This can be related to either lactation

state – such as during secretion of colostrum or due to infection of the glands. A leakage in the tight

junctions between the mammary epithelial cells leads to an increased amount of somatic cells into the

milk, due to concentration normally being higher in blood than milk, together with other substances

such as plasma enzymes and antibodies. Somatic cells are therefore an indicator of leaky tight

junctions, and since somatic cells are easily detected by flow cytometry, they are often used as a

measure of milk quality on the farm level and udder health. Somatic cells themselves may furthermore

be a source of enzymes, so it is required to keep the level of somatic cells to a minimum, as these

enzymes may contribute to milk quality deterioration due to both proteolysis and lipolysis. The amount

of somatic cells is measured in cells/ml milk and is referred to as the somatic cell count, SCC (Blowey

and P. 1995; Nielsen et al. 2002).

Often environmental infections from e.g. E. coli or Streptococcus species are responsible for the

infection, and in some instances are treated with antibiotics (Blowey & Edmondson 1995). Thus, good

hygiene and milking routine and high welfare for the cows prevents most cases of elevated SCC

(Blowey and P. 1995; Ruegg et al. 2000). During AMS milking, a vacuum can be build up if the teats

are empty. This enables bacteria from dirty tits to enter into the teat canal, and hereby infect the gland

(Rasmussen 2000). The amount of somatic cells is divided into categories according to the severity of

the number. In EU, the allowed limit is currently 400.000 cells/ml, very good quality of the milk may

be down to 200.000 cells/ml, while an acute mastitis infection may reach up to 1.000.000 cells/ml

(Kelly and Larsen 2010).

A consequence of an increased SCC number is that enzymes follows into the milk – either from within

the cells themselves or they come from the blood stream of the cow. Together with an increased SCC

an increase of proteolytic, lipolytic and oxidative enzymes can be found. Through their action on milk

15

substrates these enzymes will contribute to reduce the milk quality, even after mixing with high quality

milk, by e.g. causing casein to degrade. Unfortunately, these enzymes are quite heat stable, and will

probably tolerate low pasteurization (Larsen and Nielsen 2006). In addition, the consequences may be

even more pronounced in certain products, like cheese, compared with fresh milk products, due to

storage time. The milk composition at elevated SCC in general also changes, both due to influx of

blood components, leading to increase in whey protein, as well as degradation of the main milk

components, like triglyceride fat, caseins and lactose. Furthermore, milk mineral composition changes

due to exchange between milk and blood stream (Kelly and Larsen 2010).

2.2 Primary production

2.2.1 Milking systems and pumping

Since the introduction of automatic milking system, it has been known to cause certain quality defects

to the milk (Rasmussen et al. 2002). Klungel et al. (2000) and Abeni et al. (2005) observed a higher

concentration of FFA in milk from automatic milking systems compared to parlour milking, as well as

an increase in bacteria count (Klungel et al. 2000). Both studies correlated FFA increase to milking

frequency, when comparing data to farm with parlour milking. It was suggested that the increased

milking frequency stimulated an enzymatic response to the cow, which could result in higher lipase

activity. Moreover, protein content decrease when cows are milked three times a day compared to

twice a day (Klei et al. 1997). Mechanical stress and agitation has been linked to MFGM disruption

during pumping (Wiking et al. 2005). Especially milk with a high fat content was found to be less

stable towards a high shear rate, and this effect was enhances at temperatures above 20⁰C, which

resulted in MFG coalescence and subsequently an increase in FFA concentration (Wiking et al. 2003).

However, the stabilizing effect of crystalline lipids in the MGF in cold milk will largely prevent

MFGM disruption, and thereby limit the access of lipase. Presence of air in the pipe systems caused by

vacuum during milking, is another issue linked to AMS compared to parlour milking is the (Rasmussen

et al. 2006).

16

2.2.2 Bulk tank storage

At Danish dairy farms it is common to store the milk for 24 to 48 hours at 4⁰C before being collected

by the dairy. Furthermore, the milk can be further stored in dairy silos after arrival for an additional

period of time prior to pasteurization and further processing. The milk does, however, undergo changes

during bulk storage. Within the first 48 hours, β-casein is known to be released from the casein

micelles and diffuse into the serum phase. Furthermore, cooling increases calcium phosphate solubility,

leading to more calcium in the serum phase, rather than integrated into the casein micelles (Gaucheron

2005). This may lead to prolonged coagulation time during cheese production (Maciel et al. 2015). The

effect is, however, reversed, when the milk is heated during further processing (Ali et al. 1980; Connell

et al. 2017).

Microbial growth is one of the major concerns regarding bulk storage. The microorganisms mainly

originates from dirt on the udder and teats, from growth and biofilm in the milking equipment (Latorre

et al. 2010) and from unmanaged mastitis outbreak. This gives a very large range of psychotropic and

coliform bacteria, which can contaminate the milk and give undesirable off-flavour and enzymatic

activities. It is, however, possible to significantly minimize microbial growth with a strict hygiene

management at the farm (Elmoslemany et al. 2010; Velthuis and Asseldonk 2011; Connell et al. 2016).

An issue specific to storage raw milk retentate is the high buffering capacity of concentrated milk

(Salaün et al. 2005). pH is a limiting factor of bacterial growth, and the buffering effect combined with

high concentration of substrates for bacterial growths, has the potential to cause severe storage

problems if the hygienic procedures and maximum storage time is not properly controlled.

17

3 Filtration process

Today, a number of different concentration techniques exist. Cross-flow filtration through a semi

permeable polymer membrane is most common in the dairy industry, and depending on the

circumstances, the membrane can be designed as either a plate and frame or spiral wound system. The

plate and frame design is capable of handling highly viscous substrates, but is quite space consuming

and difficult to handle. The spiral wound membranes have gone through a lot of development in the

past years, and is considered very efficient relatively to size (Lewis 1996; Walstra et al. 2006b).

Membranes can be divided in four groups depending on pore sizes: microfiltration (> 500 nm), UF

(500 nm – 5 nm), NF (5 nm – 0.5 nm) and RO (< 1 nm). Microfiltration will hold back and thus

concentrate microorganisms, fat globules and some larger proteins in the retentate, and let smaller

proteins, lactose, minerals and water through the membrane as permeate. UF will retain caseins and

whey proteins together with fat and microorganisms in the retentate. NF will let through only water and

some minerals to the permeate and in RO, permeate will be almost pure water and perhaps small

molecules such as urea (Walstra et al. 2006b). Figure 5 shows a schematic overview of the membrane

types and the permeability.

Figure 5: Overview of membrane types according to pore size, pressure requirements and

permeability.

18

In order to overcome the hydraulic resistance of the membrane and possible fouling cake layer, a

certain feed pressure is needed. The retentate flux is a result of a combination of membrane spacer size,

product viscosity (due to e.g. VCF and temperature) and feed pressure. The retentate flux will often

decrease over time as fouling layer builds up (Kessler 2002; Walstra et al. 2006b).

Figure 6 General setup of a continuous membrane filtration process.

Generally, the filtration can be conducted as either a continuous or batch process. The continuous

process (figure 6) gives retentate with the desired concentration factor directly from the outlet valve

once the system has stabilized. This is done, more or less automatically, by controlling the ratio

between feed flow and retentate outlet flow through a retentate valve. In practice, the membrane unit

exists as a loop, mixing the feed product with part of the retentate that is held back by the valve

(Mulder 1996). A balance tank is required to ensure that air is not introduced into the system.

During batch processing, the retentate circulates back into the bulk tank until the desired concentration

factor has been reached (figure 7). The batch process allows a lower permeate flux and the ratio

between feed product and retentate is not crucial, as the dry matter of the feed material will slowly

increase. The total volume of the bulk tank will drop as more and more permeate it expelled (Mulder

19

1996). The choice of either continuous or batch process is a matter of the exact conditions. The

continuous process will generally be faster but consume more energy compared to the batch process,

where the concentration factor at the retentate outlet is less important and thus the pump efficiency can

be maximized.

Figure 7: General setup of a batch membrane filtration process.

20

4 Cheese making process

As previously mentioned, cheese dairies are a very likely receiver of farm-produced retentate.

Especially considering that part of the pre-treatment of the cheese-milk is standardization of cheese-

milk according to protein content in order to save space in the cheese vats. The process of cheese

making is however quite complex, and thus a thorough understanding of the chemistry related to the

processing steps is needed in order to accommodate changes that might occur due to the membrane

filtration. Milk can coagulate and form aggregates through different pathways: enzymatically, by

hydrolysing of κ-CN; acidification to the iso-electrical point of the CNs and by temperature (heat)

denaturation of the proteins (Walstra et al. 2006b). During modern cheese production, the

enzymatically pathway is most prevalent, but some cheese are also made purely by acidification. This

thesis will focus on the processes involved in the enzymatically induced coagulation.

4.1 Cheese production overview

When looking at factors that influences quality of cheese – that be texture, flavour and composition, the

pre-processing of the cheese milk will contribute to reflect on the final product. Depending on the exact

type of cheese, many of the processing steps differ slightly. However, the overall principal in cheese

making – and especially during large-scale production, will follow the procedure as described in this

chapter.

Heat treatment: The main purpose of heat treatment is to eliminate pathogenic bacteria. It will

however also eliminate non-pathogenic microorganisms and denature indigenous enzymes, which will

help control the characteristics of the final product. Another effect seen from heat treatment is the

restoration of the changes that happened during cold storage, even by high temperature, short time

pasteurization at 72ºC for 15 seconds (Walstra et al. 2006b). Spores are generally more resilient to

regular pasteurization. However, if the heat treatment is too intense, the ability to form a proper

coagulum will be reduced due to denaturation and absorption of β-CN and whey proteins to the surface

of κ-CN and thus shield from the enzymatic cleavage by rennet (McSweeney 2007). Traditionally,

21

chemicals such as NaNO3 and H2O2 have been added to prevent problems from spore forming bacteria.

Due to higher awareness, mechanical processes such as bacteofugation and microfiltration have

replaced this procedure (Champagne 1994; Saboya 2000).

Standardization: Cheese milk is often standardized according to fat content by separating the cream

from skim milk and recombine to the desired fat content. In modern cheese production, the protein

content will also be standardized by UF in order to utilize the capacity of the cheese vats. To ensure

sufficient amounts of calcium in the milk, a solution of CaCl2 is usually added (Sandra et al. 2012).

Acidification: The acidification of the cheese milk serves several purposes. It will inhibit bacteria that

might have survived the pasteurization, by changing the environment, but will also result in appropriate

conditions for the rennet enzyme. A low pH also advances the syneresis and helps to control the texture

of the cheese by influencing the Ca2+ balance. Acidification can be either direct by adding a food grade

acid to the cheese milk or indirect through starter culture. Direct acidification is more time efficient and

easy to control (Kosikowski and Mistry 1997), but will not contribution further to flavour formation.

Indirect acidification is more difficult to control compared to direct, but the indirect is often preferred

due to the role in flavour development during cheese ripening. When adding starter culture, the final

pH of the cheese curd is highly dependent on factors such as the specific species and strains in the

culture, cheese milk composition - especially concerning lactose content together with temperature and

time, all of which contribute to the acidification rate.

Addition of coagulant: When the appropriate pH is reached, the coagulant is added. The coagulation

mechanism of the first and second stage of the coagulation is further described in section 4.2. When the

milk has coagulated into a continuous gel structure, the syneresis of whey and further structural

rearrangement will start, and this step can be regarded the third coagulation stage(Lucey and Fox 1993;

Kosikowski and Mistry 1997). The rennet also plays a role during ripening, especially when the curd is

made from direct acidification where no starter culture is present to induce proteolysis and form other

flavour components (Harboe et al. 2010).

22

Cutting, stirring and pressing: Further processing of the cheese curd involves mechanical agitation to

accelerate the syneresis, and hereby regulate the water content of the cheese. Factors such as cheese

grain size, temperature and stirring time can all affect the syneresis: smaller cheese grains providing

larger surface area, higher temperature and stirring intensity will enhances syneresis. Pressing is the

final step of mechanically increased syneresis and promotes a structural rearrangement and

strengthening of the gel into a firm and durable mass (Walstra et al. 2006b).

Salting: Salt plays a large role in the final perception of the cheese, as it affects both flavour, texture

and starter culture profile during ripening and to some extend the water content. The desired salt

content depends on the characteristics of each cheese type, but recent studies have researched the

possibility to produce known cheese types with a reduced salt content due to health benefits

(Rulikowska et al. 2013; Murtaza et al. 2014).

4.2 Milk coagulation process

4.2.1 First stage

The first stage of the coagulation, also called the enzymatic stage, where κ-CN is hydrolysed into para

κ-CN, with the casein macro peptide (CMP) residue dispersed in the serum phase. This process

destabilizes the casein suspension by reduction of the electrostatic repulsion, enabling the casein

micelles to collide, aggregate and form a network structure. The hydrolysis can be described as a first

order reaction (Osintsev and Qvist 2004; Harboe et al. 2010). The enzyme commonly added to induce

milk coagulation is chymosin (EC 3.4.23.4). Chymosin, (referred to by the common name rennet) is an

aspartic protease endopeptidase, that has a very well defined and specific κ-CN cleavage site, making it

a very efficient coagulant, with limited protein loss (Whitaker 2002). Previously, the length of the first

coagulation stage was measured as the time from addition of rennet and until small aggregates were

visually observed (Lomholt et al. 1998). With modern analysis equipment such as LC, the first stage is

now evaluated by the concentration of CMP in the serum.

23

4.2.2 Second stage

Second stage is the actual aggregation of the remaining part of the casein micelles, when CMP has been

cleaved off. The second phase can be enabled when the electrostatic repulsion is sufficiently low, when

approximately 70% of the κ-casein has been hydrolysed (Sandra et al. 2012). A model based on

overcoming an energy barrier required for aggregation has been described by (Darling and van

Hooydonk 1981) and later confirmed by (Lomholt et al. 1998). The model suggests that hydrolysis of

CMP reduces the energy barrier, that otherwise would have been too high to form stable aggregates.

The driving force of the casein aggregation is mainly due to hydrophobic interactions and van der

Waals attraction, enabled by the loss of electrostatic repulsion. During the early phase of aggregation,

small casein clusters are formed. This will result in an initial viscosity decrease (Karlsson et al. 2007c).

These clusters will, subsequently, link together, until a continuous gel network is obtained, which is

then reflected as an storage modulus increase (Klandar et al. 2007).

4.2.3 The role of calcium

The casein micelle structure is held together by colloidal calcium phosphate (CCP) which consists of

3Ca3(PO4)2CaH citrate; Ca9(PO4)6; or CaHPO4٠2H2O. Thus, calcium plays an essential role during

cheese production, as it would not be possible to form a gel without (Lucey and Fox 1993). The

primary coagulation stage is considered independent of Ca, but due to ionic equilibria, addition of Ca2+

will decrease pH, and hereby indirectly cause an increase of the enzymatic reaction rate. This effect is

however mostly theoretical, as pH of the cheese milk is adjusted by other means during cheese

production (Sandra et al. 2012). During the second stage, calcium has a dual purpose; both by further

decreasing the electrostatic repulsion by neutralizing negative charges of the casein micelles and by

formation of calcium bridges, increasing the curd firmness (Mellema et al. 1999; Sandra et al. 2012).

The bovine milk calcium exist as an equilibrium between casein, bound as complexes in the serum

phase and as free Ca2+ ions. The distribution is temperature and pH dependent. Calcium will dissociate

from the casein micelles at low temperatures, and reverse when re-heating after cold storage. The

solubility of calcium increases at lower pH, leading to a higher concentration of calcium in the serum

phase (Mekmene et al. 2010; Malacarne et al. 2013; Koutina et al. 2014).

24

5 Milk powder

Worldwide the production of milk powder is very important, in order to provide the possibility of milk

to people without access to daily fresh milk. It also makes it possible to store milk for long time in a

convenient way, and hereby export it easily. Finally, milk powder is in many cases preferable in several

food productions, since it is easy to recombine according to special recipes. Milk powder is an obvious

usage of farm produced reverse osmosis retentate; but naturally, the quality must not be compromised.

Milk for powder production has to be of good quality, in order avoid off flavor and loss of functional

properties. Harsh heat treatment of the milk will cause whey proteins to denature, resulting in lower

solubility (Bylund 1995).

5.1 Milk powder manufacture

The most common way to produce milk powder is through spray drying and roller drying. The

principle in drying is to evaporating away the water so only the dry matters are left, by giving the milk

a very large surface area and apply to hot, dry air or surface. Roller drying is traditionally only used in

the chocolate industry, as it is associated to development of Maillard reaction, which is only desirable

in milk chocolate (Bylund 1995).

Spray drying can be performed in many steps and stages, in order to improve functional qualities and

save energy (Kessler, 2002). Many parameters will affect the final product quality: the construction of

the drying tower and atomizer nozzle, inlet air temperature, outlet temperature, air circulation, droplet

size, and air humidity. The state of the milk for drying is also very important. The quality depends on

temperature of feeding product, dry matter content, age thickening and flow into the dryer (Pisecky

1997).

Prior to drying, the milk is often pre-concentrated to about 50% dry matter. This saves energy (Kessler

2002) and may improve the powder quality (Kim et al. 2009; Lin and Chen 2009). The most common

method is by evaporation in a vacuum chamber (Kessler, 2002); however, this process can be partially

substituted by reverse osmosis. The milk concentrate is preheated and homogenized before spray dried.

25

The preheating provide optimal conditions for both homogenization and drying. Homogenizing the

milk ensures a better distribution of fat in the milk and herby better functional properties of the powder

(Kim et al. 2009). The milk concentrate is pumped through an atomizer nozzle, dried with hot air of e.

g. 200°C until the milk powder contains approximately 6% moist. Then the powder is cooled and

packed in appropriate containers (Bylund 1995).

In order to dry powder in an optimal way, it is necessary to choose the right inlet temperature, outlet

temperature, nozzle pressure and product flow. The aim is to get as high product yield as possible and

to dry the powder to an exact moisture content, without burning/browning the product and without the

product sticking to the surface of the drying chamber. The speed/pressure of the atomizer wheel as a

function of droplet size can also be estimated and fitted with the height of the drying chamber, in order

to calculate whether the droplets of a given size have time enough to dry on the way down the chamber

(Kessler 2002). Pisecky (1997)described that the smallest droplets may lose 90% of their water content

within the first 10 cm from leaving the atomizer, and the big droplets will need 1 m to lose the same

amount of water. If the droplets are too big, or the flow of the feed product to high, there may not be

time for all the water inside the particles to evaporate and they will stick to the surface and create

lumps. An increase in temperature can prevent this, but might then burn the particles. The adjustment

of these variables, is also influenced by air moisture and dry matter in the milk (Kessler 2002).

5.2 Milk powder quality

5.2.1 Powder composition and physical properties

Lactose has a major impact on the milk powder properties. It is desirable to keep the lactose in a glassy

state after drying. If the storage conditions give rise to entering a rubbery state of lactose and create

crystals, the quality of the milk powder will change dramatically depending on the specific situation,

and result in many contrary effects (Thomas et al. 2004). During crystallization, lactose alters the

structures of the powder particles, and creates capillary forms, that allow water to defuse into the

particle. At the same time, fat is expelled from inside the particles, and instead, create a water repelling

surface layer (Thomas et al. 2004; Kim et al. 2009).

26

Particle shape, size and agglomeration has a great influence on the final quality. Nijdam and Langrish

(2005) showed via electron micrographs scanning that particles dried at high temperatures are bigger

and have a smoother and round shape, compared to smaller, more wrinkled and shriveled particles

dried at lower temperatures. Due to the shape of the particles, they were able to pack closer together

and hereby give a higher bulk density.

Wettability is a measurement testing the powders ability to suck up water. Wettability is affected by

particle size and density, and can be improved with a lecithin treatment. It is the small capillary holes in

the powder particles that gives the ability to suck up water, or rather letting water cover all the surface

of the particles, so it can quickly be dissolved (Ji 2016). Whole milk powder normally has a lower

wettability than skim milk because of the higher fat content respells water (Pisecky 1997).

The solubility index gives a value of the amount of un-dissolvable particles in the milk, measured by

the volume of sediment after centrifuging. The most crucial factor controlling the solubility index

during drying is the particle temperature through the first drying state, until the moist content of the

particles is below 10%. In reality, the factors controlling the particle temperature is viscosity of the

milk, droplet size, outlet temperature and dry matter content of the milk (Pisecky 1997).

Bulk density is defined as the mass of a given volume of powder, after it has settled. It is directly

affected by the amount of air is trapped inside the powder particles, and the particles ability to pack

close together (Nijdam and Langrish 2005). Bulk density values are very important when it comes to

transport and storage of the powder. Bulk density is measured in g/ml, and a high bulk density means

that the powder does not take up much space according to the mass, which is often desirable. Nearly all

the production factors affect the bulk density (Pisecky 1997). A large difference in density between

before and after stamping is undesirable, since it will cause a larger amount of air on top of the product

after packaging. This takes up space and uses extra packing material for no use.

5.2.2 Storage stability

Enzymatic activity: As one of the advantages of milk powder is the excessive shelf life, the processes

that might occur during storage has to be controlled. Heat treatment of the milk as part of the pre-

27

processing during powder manufacture might not be enough to denature all indigenous enzymes. Since

microorganisms and enzymes require a certain water activity to function, the moisture content of the

milk powder can accelerate reaction and growths; Chen et al. (2003) described observations of

sensorial detectable FFA concentrations in whole milk powder with a water content of 3%. Likewise,

Celestino et al. (1997) showed that a certain level enzymatic activity occurred during storage of the

powder, that may impair the quality. They did however conclude that many of the products from

enzymatic activity originated from heat resilient microbial sources, and even from storage of the raw

milk. This emphasizes the role of raw milk quality in relation to the processes that leads to the final

product (Fonseca et al. 2013).

Oxidation: Another quality defect associated to milk powder storage is oxidation. Both proteins and

lipids may be subjected to oxidation, and in order to prevent this, much of the commercial powder is

sold in sealed packages with alternated atmosphere, as it will delay formation oxidation products

(Lloyd et al. 2009). The extend of oxidation in milk powder is largely influenced by the storage

conditions, powder moisture content and raw milk quality rather than e.g. fat and protein content of the

powder (Celestino et al. 1997; Stapelfeldt et al. 1997; Scheidegger et al. 2013). Oxidation has an

impact on consumer perception, as it will cause formation of off flavours (Romeu-Nadal et al. 2007)

rather than impair nutritional value (Zunin et al. 2015).

Maillard reaction: Formation of Maillard products is often a direct consequence of excessive heating.

During milk powder storage, the maillard reaction can however proceed at room temperature, and is

regarded a major quality flaw as it affects the powder visually, sensorial and nutritional, and the

reaction is further catalysed by storage temperature, pH and moisture content (Van Renterghem and De

Block 1996). The consequence of extensive Maillard is both sensorial through flavour and colour

changes, and nutritional as the lactosylated amino acids are less digestible (Thomas et al. 2004;

Dalsgaard et al. 2007). The maillard reaction is often referred to as non-enzymatic browning, but in

order to cause browning, the Maillard reaction has to be on a highly progressed level in the reaction

chain. Earlier stages of the Maillard reaction involves condensation of amine-carbonyl (reducing

sugars) groups and formation of Amadori compounds. During the intermediate stages, the sugars are

28

dehydrated and fragmented, and amino acids degraded (Strecker degredation). The browning occurs

during the final stages, where Aldol and aldehyde-amine condensates (Nursten 2005).

6. Summary of included papers

6.1 Paper I: Chemical Quality of Raw Milk Retentate processed by Ultra-filtration or

Reverse Osmosis at the Dairy Farm

6.1.1. Study objectives

The objective of this first study was to describe the extent of mechanical damage that RO and UF

would cause to raw milk, depending on variable processing factors: feed pressure, temperature,

concentration factor, pressure drop across the membrane and membrane spacer thickness. The

parameters for evaluating quality was MFG size distribution, concentration of FFA and extend of

proteolysis, as this is common factors for classifying raw milk quality. We anticipated that especially

the milk fat would be vulnerable to the mechanical process.

6.1.2 Experimental setup

The membrane filtration was organized as an in-line process, with milk poured straight from the AMS

milking line into the balance tank of the filtration plant. The milk was provided by Danish Cattle

Research Centre (Aarhus University – Foulum, Tjele, Denmark), where the filtration plant was

installed. Retentate samples were acquired after 30 minutes of processing, so the system had time to

stabilize and adapt when processing settings were changed. As the FFA values of the raw AMS milk

seemed quite high, it was decided to repeat the experiments using bulk tank milk from a herringbone

parlour milking system. The processing parameters that had no significant influence on retentate from

AMS were kept constant in order to reduce the number of experiments required to evaluate the impact

membrane filtration on parlour milk.

6.1.3 Summary of results

29

Raw AMS milk had an average FFA concentration of 1.38 mmol/100g fat. The FFA concentration of

UF retentate was not significantly different from raw milk. Changing the temperature from 5°C to 10°C

did not have any significant impact on the UF retentate, but a single experiment at 15°C did yield a

significant higher FFA concentration. A feed pressure increase from 0.15 Pa to 0.35 Pa resulted in a

FFA increase of 1.25 mmol/100g fat to 1.48 mmol/100g fat. No difference in MFG size distribution

was found between retentate and raw milk, and neither of the process settings had an impact.

FFA significantly decreased by 0.22 mmol/100 g fat, when the RO feed pressure was increased from

2.0 Pa to 3.0 Pa. By increasing the temperature from 4°C to 10°C the FFA concentration vas

significantly increased by 0.24 mmol/100 g fat. No increase in size distribution of the MFG was

observed during RO processing. Feed pressure had a significant influence on the concentration of

proteolytic products. When the feed pressure was increased from 20bar to 30bar, proteolysis increased

by 13.4%. Proteolysis increased by 23.4% when increasing the membrane spacer thickness from 30mil

to 48mil. Concentration factor did not have significant influence, however, a tendency was observed

toward higher level of proteolysis at 2 times concentration factor compared to 1.5 times concentration

factor. The raw milk from herringbone parlour milking system had an average FFA concentration of

0.580 mmol/100 g fat and a concentration of proteolytic products of 1.069 Leu-Equivalents [mM].

Neither FFA, MFG size distribution nor proteolysis increased significantly by RO filtration.

6.2 Paper II: Caseinomacropeptide release and rheological properties during rennet

coagulation of raw milk reverse osmosis retentate

6.2.1 Study objectives

The aim of this study was characterize the coagulation properties of RO retentate compared to raw

milk. The observed differences explained by including calcium distribution and the effect of varying

rennet concentration. Likewise, CMP release was included in order to determine whether the observed

changes were reflected in the first phase of the rennet coagulation process.

30


For this study, in-line RO membrane filtration on fresh bulk tank milk from herringbone parlour

milking system was used. According to previous results presented in paper I, the processing

temperature was kept at 4ºC and the feed pressure was 3.0 Pa. Retentate of both 1.5 VCF and 2 VCF

was collected. All combinations of sample types and rennet concentrations (0.03, 0.04 and 0.05 IMCU)

were included included in the study. The rheological properties were recorded on both conventional

rheometer and ReoRox. Calcium distribution was characterized by measuring total calcium content of

both full samples and serum phase together with ionic calcium concentration.

The rennet kinetics was analysed by collecting samples at certain time points during the coagulation

process. Subsequently, LC was used to quantify the CMP concentration.


Retentate had a delayed RCT compared to raw milk, but the extent of this delay was dependent on

rennet concentration. Figure 8 shows rheological data during coagulation of raw milk, 1.5 VCF

retentate and 2 VCF retentate with a rennet concentration of 0.05 IMCU. At 0.05 IMCU, the retentate

samples exceeds the gel strength of raw milk, in spite of the longer RCT. The rennet kinetics were not

different when comparing retentate and raw milk, and the CMP concentration was purely dependent on

the ratio between rennet concentration and VCF, thus the enzyme: substrate ratio.

The calcium distribution changed by concentrating the milk (figure 9), so that a larger relative fraction

was located in the colloidal phase.

31

Figure 9: Distribution of calcium as % between colloidal, bound in the serum phase and ionic in raw milk and 1.5 VCF and 2 VCF RO retentate.

0.001

0.01

0.1

1

10

100

1000

0 5 10 15 20 25 30 35 40 45 50

log G' , Pa (elasticity)

Time, minutes

Raw

1.5 VCF

2 VCF

Figure 8: Elasticity during the formation of rennet induced coagulum of raw milk, 1.5 VCF and 2 VCF retentate with a rennet concentration of 0.05 IMCU.

32

6.3 Paper III: Storage stability of whole milk powder produced from raw milk reverse

osmosis retentate

6.3.1 Study objectives

The objective of this study was to evaluate whole milk powder made from 2 VCF RO retentate

compared to whole milk powder made from raw milk of the same heard. The powders would be

evaluated based on both functional and chemical properties during storage. Finally, the powder

properties were compared to a number of commercial reference powders.


During this study, the raw milk was concentrated by a factor of two through a RO batch process, on

bulk tank milk from AMS, and sent to a spray drying pilot plant. Out of the same bulk tank, raw milk

was collected and sent for spray drying together with the retentate. The RO filtration process was

conducted at 4ºC with a feed pressure of 3.0 Pa. The entire process of making powder from retentate

and raw milk was duplicated on two subsequent days.

The milk powder samples were analysed based on composition, surface free-fat, insoluble particles and

particle size distribution, together with the storage stability attributes: proteolysis, oxidation, colour and

furosine, and compared to values found in commercial powders. The storage experiment was carried

out over the course to 12 months, with samples retrieved after 3, 6 and 12 months. The powder was

stored at room temperature in light sealed bags.


The powder composition had a lot of variation, that affected the results of the study (table 3), and thus

making it difficult to correlate the RO pre-concentration directly to the results. An increase in

concentration of oxidation products hexanal, heptanal and nonanal was found during storage (figure

10), but to a lower extend than the commercial reference sample. Proteolysis did not change during

storage, but different levels were generally found among the different powder samples, and the same

situation was reflected on the colour measurements. Furosine concentration was higher in powder

prepared from raw milk than in powder from pre-concentrated milk.

33

Table 3: Composition and physical characterization of small-scale powders produced from non-concentrated (Non-conc) raw milk and raw milk pre-concentrated (Conc) at the farm from reverse osmosis, on two subsequent days (1 and 2) of production; and several commercial reference samples (Ref.) – both instant and regular stored in bulk bags for up to 12 months and two reference powders stored in sealed bags with altered air composition

Figure 10: Oxidation products found in powders produced from non-concentrated (Non-conc) raw milk and pre-concentrated (Conc) raw milk manufactured on a pilot-scale spray drier, compared to a commercial reference (Ref) sample. A: hexanal concentration during storage. B: heptanal concentration during storage. C: nonanal concentration during storage. The error bars show the standard deviation found between the samples of same type.

00

50

100

150

200

0 3 6 12

Hex

anal

ng/

100m

g sa

mpl

e

Storage time / months

A

00

05

10

15

20

25

30

0 3 6 12

Hep

tana

l ng

/100

mg

sam

ple


B

Non‐conc Conc Ref

00

02

04

06

08

10

12

0 3 6 12

Non

anal

ng/

100m

g sa

mpl

e


C

Composition Particle size distribution Fat

% Protein % Water % Surface fat g/100 g fat

Insoluble particles ≥ 630 µm

630-400 µm ≤ 400 µm

Raw 1 17.81 30.78 3.55 0.17 0.2 0.20 1.20 98.60

Conc 1 29.89 26.47 2.17 1.57 0.2 0.90 2.20 96.90

Raw 2 21.89 29.29 2.81 0.85 0.1 0.58 1.00 98.42

Conc 2 31.00 26.26 1.87 2.59 0.2 0.26 1.82 97.92

Ref. regular fresh 28.24 23.63 3.17 0.99 0.1 0.30 53.19 46.52

Ref. regular 12mth 27.97 23.45 3.38 0.83 0.1 0.16 37.36 62.48

Ref. instant 6mth 26.47 26.89 3.31 1.14 0.1 4.49 64.22 31.29

Ref. instant fresh 27.91 23.97 2.71 1.49 0.1 2.30 83.21 14.49

Ref. regular 3mth 25.76 24.17 3.04 0.88 0.1 0.20 44.10 55.70

Ref instant sealed 12mth 28.36 23.54 2.85 1.70 0.1 0.98 68.98 30.04

Ref regular sealed 12mth 26.34 23.99 3.22 1.64 0.1 0.52 6.13 93.35

34

7. General discussion

This thesis has sought to uncover the extend of witch concentrating milk at the farm might affect the

milk quality (Paper I). Moreover, as the retentate is intended for further processing into e.g. cheese or

milk powder, the consequences of pre concentrating needs to be understood in relation to product

quality as well (Paper II and III). The following chapter seeks to discuss to what extend concentrating

milk at the farm will, in itself, affect the quality of milk and milk product when considering the whole

production chain, and thereby put the topic of this thesis into perspective.

The overall results found in Paper I reinforces the importance of raw milk quality, especially

concerning the processing steps of the primary production. The focus was on the mechanical treatment

from membrane filtration; however, the results showed very clearly that factors such as choice of

milking equipment had a bigger influence on the final level of FFA than the membrane filtration.

Lipolysis and the resulting FFA emission is highly unwanted due to off flavour formation, with a

sensorial detection threshold of 0.250 meq of FFA/kg milk (Santos et al. 2003). Several studies have

reported an increased in FFA concentration when converting to AMS from a conventional system.

Klungel et al. (2000) reported an average increase in FFA concentration from 0.38 to 0.53 mmol FFA /

100 g fat, together with an increased level of CFU / ml. Abeni et al. (2005) found a difference of FFA

between parlour milking system and AMS from 0.51 to 0.72 mmol FFA / 100 g fat. These results are in

line with the observations we made on the raw milk in Paper I, before subjection to membrane

filtration. The difference between AMS and conventional milking systems can be assigned to milking

frequency, pumping (mechanical agitation) and air intake due to vacuum in the suction cups

(Rasmussen et al. 2006). Bulk tank storage is another factor that can indeed influence the raw milk

quality. Wiking and Bjerring (2010) described how the cooling efficiency has an impact on the FFA

concentration. In order to prevent increased lipase activity, the milk has to be immediately cooled to

4ºC, and mixing in of warm milk should be avoided. If the milk is cooled too severely, with local areas

of freezing, the MFGM will be damaged, resulting in promoted lipolysis. Stirring in the bulk tank

should, if possible, be prevented until the milk is sufficiently cooled.

35

Bhavadasan et al. (1982) observed that aggregation of milk at 15 ºC increased lipolysis compared to

aggregation at 10 and 20 ºC. An explanation as to why the MFG are more subjectable to lipolysis is due

to a combination of factors. At lower temperatures, the MFG contains a larger fraction of solid fat,

which has a stabilizing effect on the globule, and thus prevents shearing of the MFGM. At higher

temperatures, e.g. 20 ºC, the TAG inside the MFGM are mostly at a liquid stage, allowing the MFG

and MFGM to be flexible and adapt the shape rather than tear apart during mechanical treatment.

However, at 15 ºC, the ratio between liquid and crystalline TAG compromises both stability and

flexibility, making the MFGM highly subjectable to aggregation (Lopez 2011). Another aspect of the

LPL activity during cooling is the changes in MFGM associated proteins. LPL is synthesis into the

milk in association with the casein micelles. Dickow et al. (2011) showed that cooling of milk led to a

migration of proteins from the skim milk phase to the cream phase, and to instability of the MFGM

proteins. This is thought to cause enhanced attachment of LPL to the MFGM and subsequently lead to

increased LPL substrate availability, and hence, more hydrolysis of TGA to FFA.

As the AMS milk used during the experiments of Paper I was collected directly from the milking line,

and cooled in the plate heat-exchanger on the membrane filtration plant, the raw milk had been

subjected to further pumping while still warm, compared to the milk from the parlour milking system,

that had been collected cold from the bulk tank. This might bias the results towards higher FFA

concentration in the retentate from AMS.

One of the initial concerns related to concentrating milk at the farm was the lack of pasteurization of

the milk prior to the membrane filtration. It was speculated, whether this would give rise to microbial

growth, through e.g. biofilm formation. As reported by Klungel et al. (2000), AMS caused an increase

in CFU levels compared to conventional milking systems, and it is natural to think that more complex

processing equipment, such as a filtration plant, requires more awareness in order to secure a hygienic

process (Cleto et al. 2012). As part of unpublished experiments in this thesis (Appendix 2), samples

were collected during the bulk filtration process, and compared to the levels found in the raw milk for

microbial analysis. Results showed that 2 VCF retentate contained an average of 4500 CFU/ml after 12

hours of batch processing time compared to a raw milk starting value of 3000 CFU/ml. Thus, the

microbial load was lower than the concentration factor. The experiment was later repeated with similar

36

conclusion. Whether the result was due to the filtration process conditions caused the microorganisms

to lyse or cluster together, and hereby obscuring the CFU count is not clear. Almeida et al. (2014)

showed that the bacterial clustering was affected by shear stress during growth, and cultures exposed to

shear during long periods yielded a higher number of clusters containing few individual bacteria. This

contradicts the results from this unpublished study, but may support the theory of rupturing and lysing.

It was decided not to research more on this topic, as the result was sufficient to discard the concern of

microbial growth.

The impact of mechanical treatment caused by the membrane filtration might also be highly dependent

on the milk composition and other factors related to the cow. Wiking et al. (2003) and (2004) reported

a connection between cow diet and increased milk fat content; and how that lead to an increase in MFG

size that made them more unstable during pumping. Diet is however not the only cause of changes to

MFG size and fat yield. Jersey cows have a higher fat yield and subsequently larger MFGs (Schwendel

et al. 2015). They are also reported to have a higher concentration of de novo synthesized fatty acids

(Carroll et al. 2006); witch in total may result in MFGs that are less resistant to pumping.

During studies as part of the present PhD project, SCC was recorded (Appendix 1) with the intention

of making a possible correlation to other quality parameters to raw milk retentate, like lipolysis and

proteolysis. The levels found in raw milk were very low with an average of 148000 cells/ml, compared

to the study by Klungel et al. (2000), who reported levels of 220000 cells/ml. The retentate had a

surprisingly low number, compared to raw milk, with 125000 cells/ml in 1.5 VCF retentate and 47000

cells/ml in 2 VCF retentate. These results indicate that the membrane filtration process might disrupt as

least a sub-population of the cells. This brings a concern as to enzymes being released due to SCC

disruption by the filtration process and thereby potentially causing a higher level of enzymatic

degradation of lipids and proteins after concentrating the milk. Experiments on storage of retentate did

not result in presence of higher concentrations of peptides compared to raw milk. It was concluded that

the initial level of SCC was too low to give a significant effect, and the results were not included in

either of the published papers. Thus, when evaluating the retentate quality based on common factors for

evaluating raw milk quality it seems that the membrane filtration is not of any concern compared to

37

how other factors and processing steps during the primary milk production. Provided, that the raw milk

is of high quality, no negative effect of pre-concentrating is found.

Based on the results from Paper I there was no concern of the retentate quality. Since the retentate was

intended for further processing, it was however still necessary to ensure that e.g. cheese making

properties had not been impaired (Paper II). Attributes linked to farm management, cow breed

(Frederiksen et al. 2011) and state of lactation are known to have an impact on the coagulation

properties. The major reason is milk composition changes, which in different ways affect the

coagulation process through genetic variants of proteins (Jensen et al. 2012; Jensen et al. 2015; Poulsen

et al. 2015), casein:protein ratio and SCC (Sorensen et al. 2008). The major difference between raw

milk and RO retentate is the composition. Even though the results from Paper I showed that it is

possible to make RO retentate without increased proteolysis and damage to fat globules, the cheese

making properties of milk is highly dependent on other factors such as calcium distribution.

Experiments showed that the rennet coagulation was indeed affected by the RO pre-processing. Further

studies reviled that the observed delay in rennet coagulation time was mostly linked to changes to the

ration between rennet concentration and protein content. RO retentate coagulum reached a higher

storage modulus than raw milk coagulum, suggesting that other factors than enzyme to substrate ratio

influences the coagulation process. This difference was not reflected on the CMP release, so the results

could not be ascribed to the first phase of the coagulation process. These observations are completely in

line with observations made by Sharma et al. (1993) on UF retentate, and Karlsson et al. (2007c)

assigned the more rapid curd firming rate to the decreased distance between casein micelles and

thereby more sensitive to the electrostatic charges. So is seems that in spite of significant changes to the

calcium distribution, the RO coagulation properties are very similar to those of UF.

As the RO process concentrates the minerals of the milk, the entire mineral and ionic equilibrium will

be changed. Paper II focused on changes to the Ca distribution, but other salts are likewise

concentrated. Studies have shown that addition NaCl to milk prior to rennet-induced coagulation, lead

to increased RCT and lower gel firmness (Karlsson et al. 2007a; Sameh 2007). It has not been possible

to make a direct correlation to the effect of NaCl concentration, and general ionic strength, during our

38

studies of RO retentate coagulation properties. This is a topic that could be relevant during further

studies.

As an on-farm membrane filtration technology, UF is generally not a recommendable option due to

issues of permeate handling. The effects of mechanical treatment were not different from RO (Paper

I), and will result in other issues such as loss of lactose and subsequently handling of the permeate. The

best solution to the question of handling UF permeate, retain lactose and account for the total energy

and cost balance compared to RO is still not found. UF can in some situations be preferred over RO

with regards to standardization during e.g. cheese production, as it leads to a reduced lactose content in

the retentate. The lactose concentration will affect the acidification rate during indirect acidification, as

lactose is the primary substrate for the starter culture producing lactic acid (Moynihan et al. 2016). The

starter culture will be inhibited at approximately pH 4.4. For semi hard cheeses it is however not

desired to reach a pH below 5.25, as it will have a negative impact on flavour and texture (Nielsen

2004). Part of the lactose will be expelled into the whey, but more lactose will also be retained in the

cheese curd when the initial lactose concentration is high (Moynihan et al. 2016). This will result in

net-loss of lactose into the cheese that could have been utilized, and lead to concerns about the quality

of the final cheese. In a situation where the majority of the milk consists of RO retentate, it could be

imagined a need to dia-filtrate the retentate to reduce the lactose content. This would however be a

disadvantage regarding energy consumption, and thus not a desired situation.

Viscosity of 2 VCF RO retentate was measured during cold storage for 3 days (Appendix 3) to ensure

that no age thickening occurred. The viscosity did not change over time and the results were very much

in line as what would be expected concerning the dry matter content (Carr 1999). It was therefor

decided to end further experiments on this subject. Experiments on skimming RO retentate was

conducted (Appendix 4), both in laboratory scale and on a pilot scale centrifuge. It was hypothesized

that concentrated milk would be difficult to skim properly without a remaining fat fraction in the skin

milk. The results showed that the fat was properly removed from the skim milk, but the cream had

increased protein content. This could potentially cause problems due to loss of protein. In order to

avoid this, the cream can be washed and re-centrifuged; this is however contradictory to the purpose of

saving resources through concentrating milk at the farm. Thus, it should be considered to utilize the

39

raw milk retentate for purposes that does not require skimming, or alternatively mix it with un-

concentrated raw milk in the dairy bulk tank to limit the issue.

Finally, the utilization of RO retentate was evaluated as a raw material for production of whole milk

powder (Paper III). Milk powder is a highly complex subject, as the production involves several

quality defining process steps. It is of utter most importance that the raw milk is of prime quality, since

quality defects at this stage will transfer to the final product. Based on the results seen in Paper I, it

was expected that the powder from RO retentate would be able of equal quality to powder made from

raw milk. The study did however not progress as expected, as other processing factors dominated the

intention of the study. This only emphasized that the concept of pre-concentrating raw milk at the farm

did not affect the raw material enough to be the main factor regarding quality defects.

There are two aspect of milk powder quality: the functional properties that influences how easy the

powder is to handle – both at the dairy and at the consumer, and the chemical quality witch might affect

the nutritional value and flavour. Handling of the raw milk prior to spray drying will naturally

contribute to the final quality. Storage of the raw milk should not exceed 3 days at 4ºC, as problems

with microbial growth may arise causing increased lipolysis and proteolysis (Celestino et al. 1997;

Fonseca et al. 2013). Unpublished data collected during this thesis showed that cold storage of retentate

for 3 days did not cause elevated levels of proteolysis, but microbial growth during storage remains a

severe concern point. Before the milk enters the spray drier, it is common to concentrate the milk to a

dry matter content of approximately 50 % as it reduces overall energy consumption and leads to

powder particles with better technical properties (Westergaard 2004). This process is normally done by

evaporation (Singh 2007), and the idea was to substitute part of this procedure by RO processing at the

farm. A study by Liu et al. (2012) describes that evaporation of milk caused changes to the ionic

equilibrium of e.g. calcium distribution and pH decrease in accordance to our findings on RO

membrane filtration (paper II). And even though evaporation is usually conducted at elevated

temperatures, the whey proteins were not subjected to denaturation (Singh 2007). It is a common issue

that the concentrated milk undergoes age thickening (Zisu et al. 2013), which causes fouling problems

during spray drying (Westergaard 2004). Age thickening is caused by casein aggregation, which are

dependent on the soluble minerals (Bienvenue et al. 2003). We had concerns that pre-concentrating

40

milk at the farm might lead to increased age thickening, due to storage of the retentate, and thereby a

longer period of changed ionic balances as observed in Paper II. This was however not observed

during the experiments.

Storage stability of milk powder is a great concern, as this is affected by temperature and air humidity

(Fitzpatrick et al. 2004) – conditions that seen problematic for locations where consumption of milk

powder rather than raw milk is prevalent. Increased temperature and air moisture will accelerate

oxidation and formation of maillard products, but the powder composition does also contribute to the

storage stability as was described in Paper III. Generally, powder with a high fat content is more

subjectable to oxidation. A study from Páez et al. (2006) described variation due to season, and

Romeu-Nadal et al. (2007) observed that a high content of unsaturated fatty acids lead to a decrease in

oxidative stability. Through proper handling of the raw milk retentate, storage stability should not be

impaired. We found no evidence of changes to the milk components that might result in decreased

storage stability as a result of on-farm RO.

41

8. Conclusion

The aim of this study was to identify whether concentrating milk at the farm would impair the retentate

quality, both concerning the mechanical treatment and possible damage during storage, caused by

enzymes being equally concentrated. Moreover, the process-ability of the retentate into cheese and

powder had to be studied to get a full understanding of the consequences of concentrating milk at the

farm.

The study of paper 1 showed that high feed pressure and low temperature during the filtration process

had a slight, but significant effect on limiting the development of FFA in RO retentate. Proteolysis of

the RO retentate was affected by feed pressure and membrane spacer thickness. However, when

changing the origin of raw milk from AMS to parlour milking system, none of the processing

parameters had any significant influence on the emission of FFA. The FFA concentration in the raw

parlour milk was significantly lower that the raw milk from AMS, and there was no increase in FFA

concentration between raw parlour milk and the RO retentate after 24 hours storage. Thus, it appears

that the milking system had more impact on the milk quality than the subsequent membrane filtration,

and milk from AMS was predisposed for further damage by the filtration process. Studies comparing

retentate quality between RO and UF showed no advantage of using UF. Therefor UF was not included

in further studies.

Studies on rennet coagulation properties showed that RO retentate had a longer RCT compared to raw

milk, but the curd-firming rate was higher. The cause of this observation was studied by including the

chymosin reaction rate and calcium distribution. Based on measurements of CMP concentration during

the coagulation process it was observed that the delay of gelation was due to the enzyme:substrate ratio

in the samples during the first phase of the coagulation. This transferred to the results of rennet

coagulation time. The increase in curd firming rate was speculated to be due to shorter distance

between casein micelles and thus higher electrostatic attraction. The calcium distribution changed by

RO filtration, so that a larger fraction of the total calcium content was incorporated into the casein

micelles and the retentate contained relatively less ionic calcium. This resulted in a pH decrease.

42

Changes to calcium and ionic balances might also bear part of the explanation behind the increased

curd firming rate.

Whole milk powder produced form RO retentate compared to powder from raw milk that had not been

pre-concentrated prior to evaporation, showed no direct difference in storage stability. Unfortunately,

during afterwards handling of the milk and retentate, the composition of the two samples were very

different – possibly due to insufficient stirring during storage. To compensate for this, all the powder

analysis were compared to commercial whole milk powder samples. The conclusion was that pre-

concentrating milk at the farm did not lead to and quality or storage defects, as the overall whole milk

powder composition had a much larger impact.

All the studies, that have been part of this thesis, shows no clear disadvantage of performing RO

membrane filtration as a method of concentrating milk at the farm. During all situations, other factors

such as milking system and exact sample composition seem to have a larger influence on the final

product, than the filtration process itself. Changes to the ionic balances was the biggest effect derived

from RO processing, and this aspect has to be considered during further processing, and certain

production steps during e.g. cheese manufacture might have to be revised.

43

9. Perspectives

The objective of this thesis has been the aspects of retentate quality, but the studies are however only

part of a much larger project about implementing membrane filtration in the dairy farms. Project groups

within Arla have focused on the business case and legislation.

In Denmark, the Veterinary and Food Administration require the farm to obtain status as a dairy in

order to process the milk in any other way than milking and cooling. In order for the farm to gain

permission to concentrate the milk through membrane filtration, the legislation has to be reworded and

new rules added. This process is very time and resource consuming, as the food safety has to be

thoroughly documented and the exact circumstances regarding the filtration technology has to be

formulated.

From the farmers’ perspective, an agreement has to be made regarding costs of filtration equipment,

maintenance and value of the milk. Several suggestions have been made:

The dairy owns and maintains the equipment, the farmer is payed the regular raw milk value

and a compensation from the dairy.

The farmer buys the filtration equipment and becomes certified to conduct maintenance, and

in return is given higher value of the concentrated milk.

The farmer leases the filtration equipment, is not responsible for service and maintenance,

and given a higher value for the concentrated milk.

The implementation model has to be agreed upon between the dairies and farmer, based on

recommendations from the authorities.

The business case on the feasibility of farm membrane filtration was based on whole milk powder

production, comparing 2 VCF RO retentate to values known from standard production, using regular

raw milk. Expenses from all the process steps, including cleaning, were included in the calculations.

The end result showed that in order for the membrane filtration at the farm to give a positive

economical outcome, the dairy farm must be of a certain size depending on the distance to the dairy.

However, the combination of farm size and dairy distance exceeds the conditions found in Denmark. It

44

was thus concluded that, in Denmark, the business case would be negative, and there for the technology

was not recommended for implementation. The main factor leading to this conclusion was the principle

of smaller plant not utilizing the energy and capacity well enough compared to large scale production

plants. The data for energy consumption were based on values recorded during experiments conducted

during studies of this thesis. Based on the overall conclusion of the business case it was decided not to

spend further resources on this matter. Nevertheless, the technology is still applicable in other

countries, which are known to have far greater distances between farm and dairy. Thus, the positive

results from the studies of this thesis are still useful knowledge in regards to membrane filtration – both

on farm and dairy.

45

10. References

Abeni F, Degano L, Calza F, Giangiacomo R, Pirlo G (2005) Milk quality and automatic milking: fat globule size, natural creaming, and lipolysis J Dairy Sci 88:3519‐3529

Ali AE, Andrews AT, Cheeseman GC (1980) Influence of storage of milk on casein distribution between the micellar and soluble phases and its relationship to cheese‐making parameters J Dairy Res 47:371‐382

Almeida PL, Portela R, Sobral RG, Grilo IR, Cidade T, Leal CR (2014) Living bacteria rheology : Population growth , aggregation patterns , and collective behavior under different shear flows Physical Reviews 90:1‐9

Atra R, Vatai G, Bekassy‐Molnar E, Balint A (2005) Investigation of ultra‐ And nanofiltration for utilization of whey protein and lactose J Food Eng 67:325‐332

Auldist MJ, Brien GO, Cole D, Macmillan KL, Grainger C (2007) Effects of Varying Lactation Length on Milk Production Capacity of Cows in Pasture‐Based Dairying Systems J Dairy Sci 90:3234‐3241

Auldist MJ, Coats S, Rogers GL, McDowell GH (1995) Changes in the composition of milk from healthy and mastitic dairy cows during the lactation cycle Austrailan Journal of Experimental Agriculture 35:427‐436

Bengtsson G, Olivecrona T (1980) Lipoprotein Lipase ‐ Mechanism of Product Inhibition European Journal of Biochemistry 106:557‐562

Bhavadasan MK, Abraham MJ, Ganguli NC (1982) Influence of Agitation on Milk Lipolysis and Release of Membrane‐Bound Xanthine Oxidase J Dairy Sci 65:1692‐1695

Bienvenue A, Jiménez‐Flores R, Singh H (2003) Rheological properties of concentrated skim milk: influence of heat treatment and genetic variants on the changes in viscosity during storage Journal of agricultural and food chemistry 51:6488‐6494

Bijl E, van Valenberg HJF, Huppertz T, van Hooijdonk ACM (2013) Protein, casein, and micellar salts in milk : Current content and historical perspectives J Dairy Sci 96:5455‐5464

Blowey R, P. E (1995) Mastitis control in dairy herds – an illustrated and practical guide Farming Press Broderick GA, Mertens DR, Simons R (2002) Efficacy of Carbohydrate Sources for Milk Production by Cows Fed

Diets Based on Alfalfa Silage J Dairy Sci 85:1767‐1776 Bylund G (1995) Dairy processing handbook. Tetra Pack, Carr AJ (1999) The functional properties of milk protein concentrates. In: Food Technology. Massey,

Palmerstom, North New Zealand, Carroll SM, Depeters EJ, Taylor SJ, Rosenberg M, Perez‐monti H, Capps VA (2006) Milk composition of Holstein ,

Jersey , and Brown Swiss cows in response to increasing levels of dietary fat Animal Feed Science and Technology 131:451‐473

Celestino EL, Iyer M, Roginski H (1997) The effects of refrigerated storage of raw milk on the quality of whole milk powder stored for different periods. Int Dairy J 7:119‐127

Champagne CP (1994) Psychrotrophs in dairy products: their effects and their control Critical reviews in food science and nutrition 34:1

Chen L, Daniel RM, Coolbear T (2003) Detection and impact of protease and lipase activities in milk and milk powders Int Dairy J 13:255‐275

Cleto S, Matos S, Kluskens L, Joa M (2012) Characterization of Contaminants from a Sanitized Milk Processing Plant PLoS one 7:1‐8

Connell AO, Kelly AL, Tobin J, Ruegg PL, Gleeson D (2017) The effect of storage conditions on the composition and functional properties of blended bulk tank milk J Dairy Sci 100:991‐1003

Connell AO, Ruegg PL, Jordan K, Brien BO, Gleeson D (2016) The effect of storage temperature and duration on the microbial quality of bulk tank milk J Dairy Sci 99:3367‐3374

46

Dalgleish DG, Spagnuolo PA, Goff HD (2004) A possible structure of the casein micelle based on high‐resolution field‐emission scanning electron microscopy Int Dairy J 14:1025‐1031

Dalsgaard TK, Nielsen JH, Larsen LB (2007) Proteolysis of Milk Proteins Lactosylated in Model Systems. Mol Nutr Food Res 51:404‐414

Darling DF, van Hooydonk ACM (1981) Derivation of a mathematical model for the mechanism of casein micelle coagulation by rennet J Dairy Res 48:189‐200

De Boer R, Nooy PFC (1980) Concentration of Raw whole milk by RO and its influence on fat globules Desalination 35:201‐211

Deeth HC (2006) Lipoprotein lipase and lipolysis in milk International Dairy Journal 16:555‐562 doi:10.1016/j.idairyj.2005.08.011

Dickow JA, Larsen LB, Hammershøj M, Wiking L (2011) Cooling causes changes in the distribution of lipoprotein lipase and milk fat globule membrane proteins between the skim milk and cream phase J Dairy Sci 94:646‐656

Ellis Ka, Innocent G, Grove‐White D, Cripps P, McLean WG, Howard CV, Mihm M (2006) Comparing the fatty acid composition of organic and conventional milk Journal of dairy science 89:1938‐1950

Elmoslemany AM, Keefe GP, Dohoo IR, Wichtel JJ, Stryhn H, Dingwell RT (2010) The association between bulk tank milk analysis for raw milk quality and on‐farm management practices Prev Vet Medic 95:32‐40

Evers JM (2004) The milkfat globule membrane—methodologies for measuring milkfat globule (membrane) damage Int Dairy J 14:747‐760

Farrell Jr HM, Malin EL, Brown EM, Qi PX (2006) Casein micelle structure : What can be learned from milk synthesis and structural biology? Current Opinion in Colloid & Interface Science 11:135‐147

Fitzpatrick JJ, Iqbal T, Delaney C, Twomey T, Keogh MK (2004) Effect of powder properties and storage conditions on the flowability of milk powders with different fat contents. J Food Eng 64:435‐444

Fonseca CR, Bordin K, Fernandes AM, Rodrigues CEC, Corassin CH, Cruz AG, Oliveira CAF (2013) Storage of refrigerated raw goat milk affecting the quality of whole milk powder J Dairy Sci 96:4716‐4724

Fox PF (2008) Milk: an overview. In: Singh H (ed) Milk Proteins: From expression to Food. Elsevier, pp 1‐54 Fox PF, Brodkorb A (2008) The casein micelle : Historical aspects , current concepts and significance Int Dairy J

18:677‐684 Fox PF, Kelly AL (2006) Indigenous enzymes in milk: Overview and historical aspects ‐ Part 1 Int Dairy J 16:500‐

516 Fox PF, McSweeney PLH (1998a) Chemistry and biochemistry of cheese and fermented milks. In: Dairy

chemistry and biochemistry. Blackie Academic & profetional, London, UK, pp 379‐436 Fox PF, McSweeney PLH (1998b) Lactose. In: Dairy chemistry and biochemistry. Blackie Academic &

profetional, London, UK, pp 21‐66 Fox PF, McSweeney PLH (1998c) Milk lipids. In: Dairy chemistry and biochemistry. BLACKIE ACADEMIC &

PROFESSIONAL, Fox PF, McSweeney PLH (1998d) Physical Properties of Milk. In: Dairy chemistry and biochemistry. Blackie

Academic & profetional, London, UK, pp 437‐461 Fox PF, McSweeney PLH (1998e) Salts of milk. In: Dairy chemistry and biochemistry. Blackie Academic &

profetional, London, UK, pp 239‐264 Frederiksen PD, Hammershøj M, Bakman M, Andersen PN, Andersen JB, Qvist KB, Larsen LB (2011) Variations in

coagulation properties of cheese milk from three Danish dairy breeds as determined by a new free oscillation rheometry‐based method Dairy Science and Technology 91:309‐321

Garcia III A, Medina B (1988) On‐farm Membrane Concentration of Milk American Society of Agricultural Engineers 31:274‐278

47

Gargouri A, Hamed H, Elfeki A (2013) Analysis of raw milk quality at reception and during cold storage: Combined effects of somatic cell counts and psychrotrophic bacteria on lipolysis J Food Sci 78:1405‐1411

Gaucheron F (2005) The minerals of milk Reprod Nutr Dev 45:473‐483 Ginger MR, Grigor MR (1999) Comparative aspects of milk caseins Comparative Biochemistry and Physiology

Part B 124:133‐145 Grainger C, Auldist MJ, Brien GO, Macmillan KL, Culley C (2009) Effect of type of diet and energy intake on milk

production of Holstein‐Friesian cows with extended lactations J Dairy Sci 92:1479‐1492 Grummer RR (1991) Effect of feed on the composition of milk fat Journal of dairy science 74:3244‐3257 Harboe M, Broe ML, Qvist KB (2010) The Production, Action and Application of Rennet and Coagulants. In: Law

BA, Tamime AY (eds) Technology of Cheesemaking. 2. ed edn. Wiley‐Blackwell, West Sussex, UK, Holt C (1992) STRUCTURE AND STABILITY OF BOVINE CASEIN MICELLES ADVANCES IN PROTEIN CHEMISTRY

43:63‐151 Horne DS (2006) Casein micelle structure : Models and muddles Current Opinion in Colloid & Interface Science

11:148‐153 Huppertz T, Fox PF, Kelly AL (2004) Dissociation of caseins in high pressure‐treated bovine milk Int Dairy J

14:675‐680 Huppertz T, Kelly AL, Fox PF (2009) Milk Lipids – Composition , Origin and Properties. In: Tamime AY (ed) Dairy

Fats and Related Products. Blackwell Publishing Ltd., pp 1‐27 Ismail B, Nielsen SS (2010) Invited review: Plasmin protease in milk: current knowledge and relevance to dairy

industry J Dairy Sci 93:4999‐5009 Jenkins TC (1993) Lipid Metabolism in the Rumen Journal of dairy science 76:3851‐3863 Jenness R, Holt C (1987) Casein and lactose concentrations in milk of 31 species are negatively correlated

Experintia 43:1015‐1018 Jensen HB, Pedersen KS, Johansen LB, Poulsen NA, Bakman M, Chatterton DEW, Larsen LB (2015) Genetic

variation and posttranslational modification of bovine kappa ‐casein: effects on caseino‐macropeptide release during renneting J Dairy Sci 98:747‐758

Jensen HB, Poulsen NA, Andersen KK, Hammershøj M, Poulsen HD, Larsen LB (2012) Distinct composition of bovine milk from Jersey and Holstein‐Friesian cows with good , poor , or noncoagulation properties as reflected in protein genetic variants and isoforms J Dairy Sci 95:6905‐6917

Jensen RG (2002) The composition of bovine milk lipids: January 1995 to December 2000 Journal of dairy science 85:295‐350

Ji J (2016) Rehydration behaviours of high protein dairy powders: The influence of agglomeration on wettability, dispersibility and solubility Food hydrocolloids 58:194

Karlsson AO, Ipsen R, Ardö Y (2007a) Influence of pH and NaCl on rheological properties of rennet‐induced casein gels made from UF concentrated skim milk Int Dairy J 17:1053‐1062

Karlsson AO, Ipsen R, Ardö Y (2007b) Observations of casein micelles in skim milk concentrate by transmission electron microscopy LWT Food Sci & Tech 40:1102‐1107

Karlsson AO, Ipsen R, Ardö Y (2007c) Rheological properties and microstructure during rennet induced coagulation of UF concentrated skim milk Int Dairy J 17:674‐682

Kelly A, Larsen LB (2010) Key requirements for milk quality and safety: Milk biochemistry British Welding Research Association

Kelly AL, O'Flaherty F, Fox PF (2006) Indigenous proteolytic enzymes in milk: A brief overview of the present state of knowledge Int Dairy J 16:563‐572

Kelly PM (1987) Application of membranes to whole milk seperations: a review Irish Journal of Food Science and Technology 11:153‐161

48

Kessler HG (2002) Food and Bio Process Engineering ‐ Dairy Technology. 5. edn. A. Kessler, München Kim EH‐J, Chen XD, Pearce D (2009) Surface composition of industrial spray‐dried milk powders. 2. Effects of

spray drying conditions on the surface composition. J Food Eng 94:169‐181 Klandar AH, Lagaude A, Chevalier‐Lucia D (2007) Assessment of the rennet coagulation of skim milk: A

comparison of methods Int Dairy J 17:1151‐1160 Klei LR, Lynch JM, Barbano DM, Oltenacu PA, Lednor AJ, Bandler DK (1997) Influence of Milking Three Times a

Day on Milk Quality J Dairy Sci 80:427‐436 Klungel GH, Slaghuis Ba, Hogeveen H (2000) The effect of the introduction of automatic milking systems on milk

quality J Dairy Sci 83:1998‐2003 Kosikowski FV, Mistry VV (1997) Cheese and Fermented Milk Foods. 3rd edn. L.L.C., Westport, CT, Koutina G, Knudsen JC, Andersen U, Skibsted LH (2014) Temperature effect on calcium and phosphorus

equilibria in relation to gel formation during acidification of skim milk Int Dairy J 36:65‐73 Larsen LB, Nielsen JH (2006) Mælkekvalitet – betydning af somatiske celler i mælk Mejeribrugets

forskningsfond Rapport nr 2006‐78 Mejeriforeningen Larsen LB, Rasmussen MD, Bjerring M, Nielsen JH (2004) Proteases and protein degradation in milk from cows

infected with Streptococcus uberis Int Dairy J 14:899‐907 Latorre AA, Kessel JSV, Karns JS, Zurakowski MJ, Pradhan AK, Boor KJ (2010) Biofilm in milking equipment on a

dairy farm as a potential source of bulk tank milk contamination with Listeria monocytogenes J Dairy Sci 93:2792‐2802

Lewis MJ (1996) Pressure activated membrane processes In: Seperation processes in the food and biotechnology industries. pp 65‐96

Lin SXQ, Chen XD (2009) Engineering Data of Diameter Change during Air Drying of Milk Droplets with 40 % Initial Solids Content Drying Technology 27:1028‐1032

Liu DZ, Dunstan DE, Martin GJO (2012) Evaporative concentration of skimmed milk : Effect on casein micelle hydration, composition, and size Food Chem 134:1446‐1452

Lloyd MA, Hess SJ, Drake MA (2009) Effect of nitrogen flushing and storage temperature on flavor and shelf‐life of whole milk powder J Dairy Sci 92:2409‐2422

Lomholt SB, Worning P, Øgendal L, Qvist KB, Hyslop DB, Bauer R (1998) Kinetics of the renneting reaction followed by measurement of turbidity as a function of wavelength J Dairy Res 65:545‐554

Lopez C (2011) Current Opinion in Colloid & Interface Science Milk fat globules enveloped by their biological membrane : Unique colloidal assemblies with a speci fi c composition and structure Current Opinion in Colloid & Interface Science 16:391‐404

Lucey JA, Fox PF (1992) Rennet coagulation properties of late‐lactation milk: Effect of pH adjustment, addition of CaCl2, variation in rennet level and blending with mid‐lactation milk Irish J of Agric Food Res 31:173‐184

Lucey JA, Fox PF (1993) Importance of calcium and phosphate in cheese manufacture ‐ a reviewI J Dairy Sci 76:1714‐1724

MacGibbon AKH, Taylor MW (2006) Composition and Structure of Bovine Milk Lipids. In: Fox PFaM, P. L. H. (ed) Advanced Dairy Chemistry Volume 2: Lipids. Springer, pp 1‐42

Maciel GM, Hammershøj M, Frederiksen PD, Sørensen J, Bakman M, Poulsen NA, Larsen LB (2015) Dairy processing and cold storage affect the milk coagulation properties in relation to cheese production Dairy Science and Technology 95:101‐114

Malacarne M et al. (2013) Effects of storage conditions on physico‐chemical characteristics, salt equilibria, processing properties and microbial development of raw milk Int Dairy J 29:36‐41

McSweeney PLH (2007) Cheese problems solved Woodhead Publishing, Cambrige, UK

49

McSweeney PLH, Fox PF (2013) Casein Micelle structure, Functions and Inteactions. In: Advanced Dairy Chemistry. 4 edn. Springer Science and Buisness Media, New York, U.S,

Mekmene O, Gaucheron F (2011) Determination of calcium‐binding constants of caseins, phosphoserine, citrate and pyrophosphate: A modelling approach using free calcium measurement Food Chem 127:676‐682

Mekmene O, Le Graët Y, Gaucheron F (2009) A model for predicting salt equilibria in milk and mineral‐enriched milks Food Chem 116:233‐239

Mekmene O, Le Graët Y, Gaucheron F (2010) Theoretical model for calculating ionic equilibria in milk as a function of pH: Comparison to experiment Journal of Agricultural and Food Chemistry 58:4440‐4447

Mellema M, Leermakers FAM, De Kruif CG (1999) Molecular Mechanism of the Renneting Process of Casein Micelles in Skim Milk, Examined by Viscosity and Light‐Scattering Experiments and Simulated by Model SCF Calculations Langmuir : the ACS journal of surfaces and colloids 15:6304‐6313

Moynihan AC, Molitor M, Jaeggi JJ, Johnson ME, Mcsweeney PLH (2016) Effect of standardizing the lactose content of cheesemilk on the properties of low‐moisture , part‐skim Mozzarella cheese J Dairy Sci 99:7791‐7802

Mulder M (1996) Module and process design. In: Basic principles of membrane technology. 2 edn., Murtaza MA, Huma N, Sameen A, Murtaza MS, Mahmood S, Meraj A (2014) Texture , flavor , and sensory

quality of buffalo milk Cheddar cheese as influenced by reducing sodium salt content J Dairy Sci 97:6700‐6707

Nicholas GD, Auldist MJ, Molan PC, Stelwagen K, Prosser CG (2002) Effects of stage of lactation and time of year on plasmin‐derived proteolytic activity in bovine milk in New Zealand J Dairy Res 69:533‐540

Nielsen JH (2005) Afslutningsrapport for projektet Udvikling af metodikker til rutinemassig evaluering af smagsfejl i malk Mejeriforeningen 70

Nielsen JH, Røntved C, Ingvartsen KL, Larsen LB (2002) Mælkekvalitet – betydning af somatiske celler i mælk Mælkeritidende 18:425‐428

Nielsen WE (2004) Principles of cheese production. In: Handbook of Food and Beverage Fermentation Technology. pp 219‐239

Nijdam JJ, Langrish TAG (2005) An Investigation of Milk Powders Produced by a Laboratory‐Scale Spray Dryer Drying Technology 23:1043‐1056

Nursten H (2005) The Maillard Reaction ‐ Chemistry, Biochemistry and Implications. The Royal Society of Chemistry, Cambridge, UK

Osintsev AM, Qvist KB (2004) Study of the mechanism of the proteolytic stage of enzymatic coagulation of milk casein Colloid Journal 66:223‐227

Páez R, Pensel N, Sabbag N, Taverna M, Cuatrín A, Zalazar C (2006) Changes in free fatty acid composition during storage of whole milk powder Int J Dairy Tech 59:236‐241

Palmquist DL (2006) Milk Fat: Origin of Fatty Acids and Influence of Nutritional Factors Thereon. In: Fox PF, McSweeney PLH (eds) Advanced Dairy chemistry, volume 2. Springer,

Phadungath C (2005) Casein micelle structure : a concise review Songklanakarin J Sci Tech 27:201‐212 Pillay VT, Myhr AN, Gray JI (1980) Lipolysis in milk. 1. Determination of free fatty acid and threshold value for

lipolyzed flavor detection J Dairy Sci 63:1213‐1218 Pisecky J (1997) Handbook of Milk Powder Manufacture. Niro A/S, Copenhagen. Poulsen NA, Buitenhuis AJ, Larsen LB (2015) Phenotypic and genetic associations of milk traits with milk

coagulation properties J Dairy Sci 98:2079‐2087 Rasmussen LK et al. (1999) Disulphide‐linked caseins and casein micelles Int Dairy J 9:215‐218 Rasmussen MD (2000) A Review of Milking Preparation: The Science Mastitis Council Ann Meeting nr 39:104‐

110

50

Rasmussen MD, Bjerring M, Justesen P, Jepsen L (2002) Milk quality on Danish farms with automatic milking systems J Dairy Sci 85:2869‐2878

Rasmussen MD, Wiking L, Bjerring M, Larsen HC (2006) Influence of air intake on the concentration of free fatty acids and vacuum fluctuations during automatic milking J Dairy Sci 89:4596‐4605

Romeu‐Nadal M, Chávez‐Servín JL, Castellote AI, Rivero M, López‐Sabater MC (2007) Oxidation stability of the lipid fraction in milk powder formulas. Food Chem 100:756‐763

Ruegg P, Rasmussen MD, Reinemann D (2000) The seven habits of highly successful milking routines (A3725) Rulikowska A, Kilcawley KN, Doolan IA, Alonso‐gomez M, Nongonierma AB, Hannon JA, Wilkinson MG (2013)

The impact of reduced sodium chloride content on Cheddar cheese quality Int Dairy J 28:45‐55 Saboya L (2000) Current developments of microfiltration technology in the dairy industry Lait 80:541 Salaün F, Mietton B, Gaucheron F (2005) Buffering capacity of dairy products Int Dairy J 15:95‐109 Sameh A (2007) Effect of sodium chloride and pH on the rennet coagulation and gel firmness LWT 40:220‐224 Sandra S, Ho M, Alexander M, Corredig M (2012) Effect of soluble calcium on the renneting properties of casein

micelles as measured by rheology and diffusing wave spectroscopy J Dairy Sci 95:75‐82 Santos MV, Ma Y, Caplan Z, Barbano DM (2003) Sensory threshold of off‐flavors caused by proteolysis and

lipolysis in milk J Dairy Sci 86:1601‐1607 Scheidegger D et al. (2013) Evaluation of milk powder quality by protein oxidative modifications J Dairy Sci

96:3414‐3423 Schwendel BH, Wester TJ, Morel PC, Tavendale MH, Deadman C, Shadbolt NM (2015) Invited review: organic

and conventionally produced milk‐an evaluation of factors influencing milk composition American Dairy Science Association 98:721‐746

Sharma SK, Hill AR, Mittal GS (1993) Effect of milk concentration, pH and temperature on aggregation kinetics and coagulation properties of ultrafiltered (UF) milk Food Res Int 26:81‐87

Shennan DB, Peaker M (2000) Transport of Milk Constituents by the Mammary Gland Physiological Reviews 80:926‐951

Singh H (2007) Interactions of milk proteins during the manufacture of milk powders. Lait 87:413‐423 Slack AW, Amundson CH, Hill Jr CG, Jorgensen NA (1982) On‐Farm Ultrafiltration of Milk: Part 1 ‐ Technical

Feasibility Studies Process Biochemistry:6‐11 Smithers GW (2008) Whey and whey proteins — From ‘ gutter‐to‐gold ’ Int Dairy J 18:695‐704 Sorensen A, Muir DD, Knight CH (2008) Extended lactation in dairy cows : effects of milking frequency , calving

season and nutrition on lactation persistency and milk quality J Dairy Res 75:90‐97 Stapelfeldt H, Nielsen BR, Skibsted LH (1997) Effect of heat treatment, water activity and storage temperature

on the oxidative stability of whole milk powder. Int Dairy J 7:331‐339 Stepaniak L (2004) Dairy enzymology International Journal of Dairy Technology 57:153‐171 Sundheim G (1988) Spontaneous Lipolysis in Bovine Milk: Combined Effects of Cream, Skim Milk, and

Lipoprotein Lipase Activity J Dairy Sci 71:620‐626 Thomas MEC, Scher J, Desobry‐Banon S, Desobry S (2004) Milk Powders Ageing: Effect on Physical and

Functional Properties. Crit Rev Food Sci Nutr 44:297‐322 Van Renterghem R, De Block J (1996) Furosine in consumption milk and milk powders. Int Dairy J 6:371‐382 Velthuis AGJ, Asseldonk MAPMV (2011) Process audits versus product quality monitoring of bulk milk J Dairy

Sci 94:235‐249 Walstra P (1999) Casein sub‐micelles: do they exist? Int Dairy J 9:189‐192 Walstra P, Wouters JTM, Geurts TJ (2006a) Dairy Science and Technology. 2nd edn. Taylor & Francis Group,

Boca Raton, FL, U.S. Walstra P, Wouters JTM, Geurts TJ (2006b) Dairy Science and Technology, 2. edn. Taylor and Francis Group,

Boca Raton

51

Westergaard V (2004) Milk powder technology evaporation and spray drying. GEA Niro, Copenhagen, Denmark Whitaker JR (2002) Proteolytic enzymes. In: Whitaker JR, Voragen AGJ, Wong DWS (eds) Handbook of Food

Enzymology. CRC Press, pp 993‐1009 Wiking L, Bertram HC, Björck L, Nielsen JH (2005) Evaluation of cooling strategies for pumping of milk ‐ impact

of fatty acid composition on free fatty acid levels J Dairy Res 72:476‐481 Wiking L, Bjerring M (2010) Bedre mælkekvalitet fra økologiske køer ICROFS nyt 1. sep. 2010 Wiking L, Björck L, Nielsen JH (2003) Influence of feed composition on stability of fat globules during pumping

of raw milk Int Dairy J 13:797‐803 Wiking L, Frøst MB, Larsen LB, Nielsen JH (2002) Effects of storage conditions on lipolysis, proteolysis and

sensory attributes in high quality raw milk. Milchwissenschaft 57:797‐803 Wiking L, Stagsted J, Björck L, Nielsen JH (2004) Milk fat globule size is affected by fat production in dairy cows

Int Dairy J 14:909‐913 Zall RR (1987a) On‐farm ultrafiltration of milk: The California experience Milchwissenschaft 42:3‐7 Zall RR (1987b) Accumulation and quantification of on‐farm ultrafiltrated milk: The California experience

Milchwissenschaft 42:98‐100 Zisu B, Schleyer M, Chandrapala J (2013) Application of ultrasound to reduce viscosity and control the rate of

age thickening of concentrated skim milk Int Dairy J 31:41‐43 Zunin P, Boggia R, Turrini F, Leardi R (2015) Total and "free" lipids in commercial infant formulas: Fatty acid

composition and their stability to oxidation. Food Chem 173:332‐338

Paper I

Chemical Quality of Raw Milk Retentate processed by Ultra-filtration or Reverse Osmosis

at the Dairy Farm

Ida Sørensen, Søren Jensen, Niels Ottosen, Tommas Neve & Lars Wiking

International Journal of Dairy technology, February 2016, Volume 69, Issue 1, Page 31-37

Errata

The second paragraph beginning on page 34, first column, should have said:

The concentration of FFA in nonfiltrated milk was 1.19 mmol/100 g fat, and the ultra-filtrated milk

ranged from 0.98 mmol/100 g fat to 2.12 mmol/100 g FFA (Table 2), which was an overall

nonsignificant difference.

ORIGINALRESEARCH Chemical Quality of Raw Milk Retentate processed by

Ultra-filtration or Reverse Osmosis at the Dairy Farm

IDA SØRENSEN,1 SØREN JENSEN,2 NIELS OTTOSEN,2 TOMMASNEVE3 and LARS WIKING1*1Department of Food Science, Aarhus University, Blichers All�e 20, DK-8830 Tjele, 2Arla Foods Ingredients GroupP/S, Sønderupvej 26, DK-6920 Videbæk, and 3Arla Foods, Arla Strategic Innovation Centre, Roerdrumvej 10,DK-8820 Brabrand, Denmark

Concentrating raw milk at the dairy farms – rather than at the dairy – reduces energy consumptionand CO2 emission, due to less road transportation of the milk. This study demonstrates whether itis possible to use either reverse osmosis or ultra-filtration for milk concentration at the farm with-out harming the milk quality, regarding lipolysis and proteolysis. Filtration at low temperature(4 °C) secures a good milk quality. Despite reverse osmosis operating at much higher feed pres-sures, the effect on lipolysis is small, which makes this technology the most applicable, as lactose iswithhold in retentate.

Keywords Lipolysis, Free fatty acids, Proteolysis, Milk fat globule, milking system, mechanicalstress.

INTRODUCTION

During the production of several dairy productssuch as milk powder, cheese and yoghurts, aconcentration test for the milk is conducted atthe dairy, to increase the protein content of themilk. It would be beneficial if part of this con-centration can occur at the farm, so transportand cooling as well as CO2 emission can bereduced. Additionally, the water withdrawn fromthe milk may be reused at the farm. However,the impact on the milk quality is needed to bestudied, as it is yet unknown how the processinfluences the milk quality.To overcome the hydraulic resistance of the

membrane and possible fouling cake layer, acertain feed pressure is needed. The retentateflux is a result of a combination of membranespacer size, product viscosity – due to, forexample, concentration factor and temperature –and feed pressure. The retentate flux may, how-ever, decrease over time as fouling layer buildsup (Kessler 2002; Walstra et al. 2006). Mem-brane filtration at farms is known to be used inthe United States, Australia and New Zealand,and several early studies conclude that analysisof concentration of raw milk is possible, but thismight impact the milk quality. The exact mecha-

nisms are, however, not well documented (Zall1984; Kelly 1987; Garcia and Medina 1988).More effort has generally been to study and pre-vent membrane fouling (Slack et al. 1982;Brans et al. 2004), and fractionation of the vari-ous milk compounds (Tolkach and Kulozik2005; Piry et al. 2008; Toro-Sierra et al. 2011).de Boer and Nooy (1980) arrived at some ideasregarding the methods of concentration of milkat the farm. They mention that reverse osmosis(RO) might be a better option compared toultra-filtration (UF), as UF changes the composi-tion of the milk and hereby makes use of theretentate less versatile in the dairy industry –even though RO processing requires moreenergy. They tested whether temperature andstorage would influence the level of free fattyacids in the retentate and found that processingtemperatures below 7.5 °C are most preserving.The challenges related to concentrating raw milkis to maintain a high quality, while running themembrane filtration plant at a pressure, highenough to prevent fouling. Kelly (1987)expressed his concern in this issue stating thatthe fouling effect would be greater in raw milkthan in skim milk, causing a higher pressuredrop through the membrane and a decreasedflow. High-pressure pumping and general

*Author forcorrespondence. E-mail:[email protected]

© 2015 Society ofDairy Technology

Vol 69, No 1 February 2016 International Journal of Dairy Technology 31

doi: 10.1111/1471-0307.12296

mechanical treatment of the milk may very well cause dam-age to the milk fat globules (MFGs) (Wiking et al. 2003;Fuc�a et al. 2013).The disruption of the milk fat globule membrane

(MFGM) is known to increase the level of lipolysis, as thetriglycerides inside of the MFG are more reachable to thenative lipase present in milk. High levels of free fatty acidwill cause off flavour, and this fact is regarded as a qualitydefect. The enzymes and bacteria present in the milk will beconcentrated as well, so the risk of compromising the milkquality is likely. Plasmin and other proteolytic enzymesmay react differently in the concentrated milk, resulting in amore unpredictable cleaving pattern and kinetics. Thehydrolysis of b-casein and as2-casein by plasmin will causechanges in texture and flavour, together with loss in cheeseyield (Bastian and Brown 1996). However, the extent andmechanisms of the mechanical stress and enzymatic damageof the raw milk related to concentration, as well as theimpact on further processing of the retentate into, for exam-ple, milk powder, are not known.The aim of this study was to examine whether concentrat-

ing raw milk at the farm could be performed without qualitydeteriorations. It was tested whether membrane type (UF orRO), membrane spacer thickness, feed pressure, pressuredrop, volume concentration factor and temperature affectedthe concentration of free fatty acids, MFG size distributionand extent of proteolysis. The hypothesis is that the changein viscosity and distance between particles and enzymes inthe concentrated raw milk will change the kinetics andcleavage patterns of the native lipases and proteases. Thismay influence the shelf life of the milk as well at the pro-cessing ability.

MATERIALS AND METHODS

Raw material and equipmentThe raw milk was produced at the Danish Cattle ResearchCentre (Aarhus University – Foulum, Tjele, Denmark).Mainly milk from Danish Holstein cows was used, but for

the UF experiments, some Jersey milk had been mixed in.The membrane filtration plant was developed by GEA Pro-cess Engineering (Skanderborg, Denmark) and designed, soit would be appropriate for daily production on a farm of250 cows. The cows were milked by automatic milking sys-tem (AMS). The membrane filtration plant was designed asan inline process, where milk was added to the filtrationplant between the milking robot and the bulk tank. Milkfrom one robotic milking line was collected in a mobilecontainer and poured into the balance tank. There was nocooling of the milk prior to the plate heat exchanger of thefiltration plant; however, the milk was transferred directly tothe filtration plant. The process flowchart is shown in Fig-ure 1. Additionally, RO experiments were conducted onmilk from a farm with herringbone parlour milking (HPM)system, where the milk was collected from a bulk tank.The membranes of both UF and RO were 3.8″ pHt spiral

wound with either 30 mil or 48 mil spacer and were pro-duced by Alfa Laval (Lund, Sweden). The total membranesurface area per module for the 30 mil membrane was 6.3and 4.7 m2 for the 48 mil membrane.

Experimental set-upThe experiments were conducted, so many of the combina-tions of settings on the filtration plant have been includedand repeated. The settings were feed pressure, pressure dropacross the membrane, temperature, volume concentration fac-tor and membrane spacer size. Combinations of settings isshown in Table 1. The order of the various experiments waschosen mainly from a practical aspect and was therefore notcompletely random. For each day, with new experiments, 2–3 different setting combinations were conducted. For eachexperiment, samples of both retentate and raw milk were col-lected. During UF, samples of permeate were likewise col-lected to monitor the process. For the HPM bulk milk,pressure drop was at about 0.08 MPa, volume concentrationfactor (VCF) was two times the concentration, and the mem-brane spacer was 48 mil, so only feed pressure and tempera-ture varied. The temperature was either 4 °C or 7 °C.

Figure 1 Flowchart of the filtration plant setup connected to the automatic milking system. The balance tank, plate heat exchanger and membranemodule are assembled into the filtration plant unit. All the changes of settings related to this study were done through pumps and valves, on the filtra-tion plant unit.

32 © 2015 Society of Dairy Technology

Vol 69, No 1 February 2016

Milk compositionThe overall composition of the retentate and unfiltrated milkwas measured by FT-IR (MilkoScan FT2, Foss, Hillerød,Denmark). Both samples of retentate and raw milk weremeasured, together with UF permeate. The componentsmeasured by FT-IR were dry matter, fat, protein, lactose,casein and solids’ nonfat.

Analysis of free fatty acids (FFA)To determine the concentration of FFA, samples of bothretentate and raw milk were sent to Eurofins Laboratories(Vejen, Denmark), which used the Bureau of Dairy Industries(BDI) method (IDF 1991). The samples were left at 4 °C for24 h after production, in order for the lipolysis activity to behigh (Wiking et al. 2002). Afterwards, the samples were pas-teurised within the 500-ml plastic bottles where the sampleswere collected in. The pasteurisation process was conductedby applying the bottled samples to a 71 °C water bath for40 min. Immediately after pasteurisation, the samples werestored at �18 °C, until analysed at Eurofins.

MFG size distributionMilk fat globule size distributions were determined by laserlight scattering (Mastersizer2000, Malvern Instruments) asdescribed by Wiking et al. (2003), immediately after thesamples were collected.

Analysis for proteolysisSamples of both RO retentate and unconcentrated milk werestores at 5 °C for 3 days before the start of the analysis.This was performed to ensure that the plasmin activitywould be at its highest. The assay is based on a reactionbetween free N-terminals of amino acids or peptides andfluorescamine, which will become fluorescent and detectableon a fluorometer. The exact experimental procedure isdescribed in the study by Wiking et al. (2002). Measure-ments were conducted in triplicate on a multiplate reader(BioTek Synergy 2, Holm & Halby, Brøndby, Denmark),and the results were obtained by Gen5 1.07.5 software (Bio-Tek Instruments, Winooski, VT, USA).

StatisticsAll the statistical calculations have been made with the free-ware program R 3.0.1 (R Foundation for Statistical Comput-ing, Vienna, Austria). The applied analyses were t-test,paired t-test and analysis of variance.

RESULTS

Ultra-filtration experimentsUsing ultra-filtration membranes, larger particles such asmicro-organisms, fat and proteins are retained in the retentatephase, whereas lactose, minerals and water distribute evenlybetween retentate and permeate. The milk was added into the

Table 1 Combinations of membrane filtration process settings usedfor the UF and RO experiments

UF of milk from AMS

Membranespacer, mil

Feedpressure, MPa

Temperature,°C VCF

Pressuredrop, MPa n

30 0.15 4 1.5 0.12 20.08 2

10 2 0.12 21.5 0.08 1

0.35 4 1.5 0.12 12 0.12 2

10 2 0.12 215 0.12 2

48 0.15 4 1.5 0.12 10.08 1

2 0.12 110 1.5 0.12 1

0.08 10.35 4 1.5 0.12 1

RO of milk from herringbone parlour milking system

48 2 4 2 0.08 17 1

3 4 37 3

RO of milk from AMS

30 2 4 1.5 0.12 20.08 2

2 0.12 20.08 2

10 1.5 0.12 10.08 1

2 0.12 20.08 1

3 4 1.5 0.12 10.08 1

2 0.12 20.08 1

48 2 4 1.5 0.12 30.08 3

2 0.12 210 1.5 0.12 2

0.08 22 0.12 2

0.08 23 4 1.5 0.12 2

0.08 22 0.12 2

0.08 210 1.5 0.12 1

0.08 22 0.12 1

0.08 2

n, number of replicates; R, reverse osmosis; UF, ultra-filtration.

© 2015 Society of Dairy Technology 33


filtration plant directly from the automatic milking, as aninline system. The process settings – feed pressure, processtemperature, concentration factor, pressure drop across themembrane and membrane spacer thickness – were tested forinfluence on the retentate condition. The concentration ofFFA in nonfiltrated milk was 1.38 mmol/100 g fat, and theultra-filtrated milk ranged from 0.98 mmol/100 g fat to2.12 mmol/100 g FFA (Table 2), which was an overall non-significant difference. Likewise, no effect on the UF processparameters VCF and pressure drop was observed (Table 2).An effect on processing temperature was found, that is noeffect between 5 and 10 °C was observed, but processing at15 °C produced retentate with a large concentration of FFA(only a single experiment was conducted at the high tempera-ture). Retentate produced with a feed pressure of 0.35 MPacontained a significant greater concentration of FFA(1.48 mmol/100 g fat) than the retentate produced at 0.15MPa(1.25 mmol/100 g fat). MFG size distribution showed no sig-nificant difference between retentate and nonfiltrated milk, andnone of the process settings had an impact (Table 2).

Reverse osmosis experiments

Reverse osmosis filtration in combination with automaticmilking systemsIn RO membranes, only water and small minerals, such asurea can be pressured through the membrane as permeate.Similar to the UF experiment, the filtration system was aninline process from AMS. By increasing the feed pressurefrom 2 to 3 MPa, FFA significantly decreased by0.22 mmol/100 g fat (Table 3). Also, processing tempera-ture significant influenced the concentration of FFA. Anincrease in temperature from 4 to 10 °C increased the FFAconcentration by 0.24 mmol/100 g fat. None of the otherfiltration plant settings influenced the concentration of FFAin the retentate. The size distribution of the MFG was notchanged significantly (P > 0.05) during RO processing(Table 3). About 72 h after milking, the level of proteolysis

in the retentate showed a significant influence on feed pres-sure. By increasing the feed pressure from 2 to 3 MPa,proteolysis increased by 13.4%. Membrane spacer thicknesshad a significant (P = 0.003) influence on proteolysis, andby increasing the membrane spacer thickness from 30 to 48mm, the proteolysis increased by 23.4%. Concentration fac-tor did not have a significant influence; however, a tendencywas observed towards higher level of proteolysis at twotimes the concentration factor compared to 1.5 times theconcentration factor. Taking the protein concentration inretentate into consideration gives slightly changed results, asthis still showed interaction between feed pressure and tem-perature, although feed pressure was not significant in itself.Spacer thickness still had a significant impact (Table 3).

Reverse osmosis filtration in combination with conventionalmilking systemsRO experiments were also conducted with bulk milk from aHPM system, as this is known to produce milk with lowerFFA content. Feed pressure and temperature were the onlyfiltration process settings that we varied during the experi-ments. Table 4 is showing the results from these experiments.Effects of neither process parameter nor overall levelbetween retentate and nonfiltrated milk were found on FFA,MFG size distribution and proteolysis.

DISCUSSION

Filtrations techniques influence on milk fat globulestabilityRO did not lead to more formation of FFA compared toUF, although the pressure needed for RO is 10 times larger.It was expected that the lower pressure during UF would bemore gentile to the milk, thereby retaining MFG and pro-teins at a more native state. But it appears that the ROmembrane technology is compensating, so the mechanicalstress across the membrane is limited at these low tempera-tures.

Table 2 Statistical effect of processing parameters on ultra-filtration retentate based on milk from automatic milking system. The measured fac-tors are as follows: volume weighted diameter [D (4.3)], lower 10% fractil [d (0.1)] and upper 10% fractie [d (0.9)] of the milk fat globules(MFG) and free fatty acids (FFA). The results are compared to the average level found in the retentate and raw nonfiltrated milk. Different lettersindicate significant different values between retentate and nonfiltrated milk

Processing parameter

MFG size distribution

FFA [mmol/100 g fat]D (4.3) d (0.1) d (0.9)

Feed pressure NS NS NS P = 0.0295Temperature NS NS NS P = 0.02241

Concentration factor NS NS NS NSPressure drop NS NS NS NSRetentate level 3.740 � 0.287a 0.976 � 0.200a 6.795 � 0.400a 1.377 � 0.308a

Nonfiltrated milk level 3.864 � 0.336a 1.025 � 0.200a 7.002 � 0.520a 1.19 � 0.209a

1Significant effect seen for temperatures above 10 °C.



For both UF and RO membrane filtration, no change inthe size distribution of MFG was observed compared to non-filtrated milk. This indicates that the MFG are not severelydisrupted, which has been the initial concern of implement-ing this technology at farms. Wiking et al. (2003) found anincrease in the MFG size after subjecting raw milk to shearstress, which also led to an increase in FFA concentration.Meanwhile, the results from this experiment show no signifi-cant difference in size distribution between concentrated andraw milk, and it appears that the treatment does not severelydamage the MFG. Therefore, the risk of increased FFA con-centration is thought to be less likely, as the MFGM can stillact as a barrier towards native milk lipase.The results of ultra-filtrating raw milk showed that no

damage was applied to the milk, except at processing tem-peratures above 10 °C (only based on a single measure-ment). It indicates that production temperatures higher than10 °C provide a risk for retentate with elevated levels ofFFA. Changes with regard to the increasing temperature ofmilk can cause a number of conformation mechanismsaround the MFG and MFGM. The MFGM composition and

structure can change and in turn can cause the membrane toeither weaken or strengthen, depending on the exact mem-brane molecules that are affected. This is, however, notassociated with much evidence, as the mechanism is highlycomplex (Evers 2004; Dickow et al. 2011). The temperatureis also observed to shift the ratio between solid and liquidfat inside the MFG. A higher level of solid fat is known tohave a stabilising effect on the MFG, which leads to lessdisruption and hence less formation of FFA. Wiking et al.(2003) showed a clear relation between milk at 20 °C beingmore sensitive towards mechanical stress through pumpingand milk at 5 °C. This supports the tendency found in thisstudy, where the sample at 15 °C yielded a higher level ofFFA. This indicates that the mechanical stress applied to theMFG during filtration might be more pronounced at highertemperatures, thereby giving lipase better access to thetriglycerides likewise. de Boer and Nooy (1980) showedthat cooling the milk below 7 °C before subjecting to ROwould prevent damage of the MFG.The content of FFA in the native nonfiltrated milk

from AMS had a large day-to-day variance as well as large

Table 3 Statistical effect of processing parameters on reverse osmosis retentate based on milk from automatic milking system. The measured fac-tors are as follows: volume weighted diameter [D (4.3)] and upper 10% fractile [d (0.9)] of the milk fat globules (MFG), free fatty acids (FFA)and proteolysis level. The results are compared to the average level found in the retentate and raw nonfiltrated milk. Different letters indicate sig-nificant different values between retentate and nonfiltrated milk



FFA [mmol/100 g fat] 1Leu-Equivalents [mM] Leu-Equivalents [mM/protein %]D (4.3) d (0.9)

Feed pressure NS NS P = 0.0195 2P = 0.04225 3NSTemperature NS NS P = 0.0198 2NS 3NSConcentration factor NS NS NS NS NSPressure drop NS NS NS NS NSMembrane spacer NS NS NS P = 0.00286 P = 0.003Retentate level 4.111 � 0.200a 6.607 � 0.426a 1.139 � 0.302a 1.795 � 0.355b 31.209 � 5.717b

Non-filtrated milk level 4.124 � 0.260a 6.621 � 0.543a 1.07 � 0.302a 0.96 � 0.1635a 27.639 � 4.977a

1Significant influence of retentate protein content.2Interaction between feed pressure and temperature P < 0.001.3Interaction between feed pressure and temperature P < 0.001.

Table 4 Statistical effect of processing parameters on reverse osmosis retentate based on milk from herring bone parlour milking system



FFA [mmol/100 g fat] Leu-Equivalents [mM] Leu-Equivalents [mM/protein %]D (4.3) d (0.9)

Feed pressure NS NS NS NS NSTemperature NS NS NS NS NSRetentate level 4.416 � 0.244a 7.328 � 0.469a 0.620 � 0.039a 2.165 � 0.266a 35.05 � 4.225a

Nonfiltrated milk level 4.415 � 0.548a 7.361 � 0.840a 0.580 � 0.064a 1.069 � 0.236a 31.42 � 7.086a

The measured factors are as follows: milk fat globule (MFG) size distribution, free fatty acids (FFA) and proteolysis level. The results are com-

pared to the average level found in the retentate and raw nonfiltrated milk. Different letters indicate significant different values between retentate

and nonfiltrated milk.



variance during the continual sampling from the milk line,compared to the effect from the filtration itself. This is indi-cated by the large relative standard deviation found in bothnonfiltrated milk and retentate. The natural variation in FFAcontent from individual cows is due to differences in stateof lactation, udder health, milking frequency and some dif-ferences in feeding, as the samples of raw milk by our set-up origin from single or few cows in a heard of mixed ages,state of lactation and diets. The average FFA concentrationin the AMS experiments was around 1.25 mmol/100 g fat,which is considered to be quite high. It is well known thatthe FFA content in milk from AMS is greater than that fromother milking systems, for example milking carousel or par-lous, due to increased milking frequencies, higher air intakeand more temperature fluctuations in the bulk tank whichcaused damage to the MFG and thereby enhanced lipolysis(Rasmussen et al. 2006; Wiking 2011). Lipolysis is inhib-ited by its own product (Bengtsson 1980), and this couldhave biased the results with the AMS milk. Thus, a fewexperiments with herring bone parlour milking system(HPM) were also conducted as this is milk inherently lessdamaged (see nonfiltrated milk in Tables 3 and 4). Besidesthe potential effect from product inhibition, the oppositeeffect could also be theorised, that is predamage MFG couldbe more sensitive to the filtration treatments.The experiment with the milk from HPM only tested the

effect on feed pressure and processing temperature, as theprevious results were found to have a significant effect onAMS retentate. However, no significant effect of feed pres-sure and processing temperature on milk FFA was found(Table 4). This indicates that the reason behind the differentimpacts on RO between AMS and HPM might be due tothe harsh treatment from AMS, making the milk predis-posed for further disruption during filtration. Also, the rela-tive standard deviation in both retentate and nonfiltratedmilk from HPM is lower than that seen in the milk fromAMS (which is expected, as the AMS milk was collectedcow by cow and HPM was from a bulk tank). This supportsthe suggestion that the membrane filtration in itself does notcause severe changes to the milk quality, as the HPM milk,with less indigenous variation, showed no significant differ-ence in FFA level between retentate and nonfiltrated milk.

Influence of reverse osmosis on proteolysisPlasmin is the enzyme in milk that causes the majority of theproteolytic activity. Its activity is balanced between a systemof activators, inhibitors and autolysis, which is, for example,affects the temperature and citrate concentration (Cruddenet al. 2005; Kelly et al. 2006; Ismail and Nielsen 2010). Theeffect of milk processing on proteolysis is presumed to below. However, the concentration of plasmin and its substratescould have kinetic effects on the plasmin activity.Protein content had a significant influence on proteolysis in

the AMS retentate, and the tendency of increased proteolysis

was observed according to VCF. So to compensate for thiseffect, and better focus on the mechanical influence of theprocess, a relative level of proteolysis was calculated.Increasing the RO membrane spacer thickness was the onlymechanical parameter that significantly increased the freeamino group concentration. This could perhaps be due toless mechanical pressure inside the membrane. The relativelevel of proteolysis in the retentate from AMS showed thatfeed pressure and temperature have a tendency of the sameeffect on increased proteolysis. So the direct effect ofmechanical stress on protease activity has to be furtherinvestigated. In the retentate from HPM, no significanteffect of feed pressure and temperature on proteolysis, evenin total concentration, was found. Therefore, the HPM milkproved to be less prone to both lipolysis and proteolysisthan the AMS milk. This leads to the fact that the type ofmilking system has a larger effect on the milk than themembrane filtration processing. However, the nonfiltratedmilk from HMP has a slightly higher average level of pro-teolysis than the milk from AMS. This could be due to dif-ferences between the herds e. g. by the average number oflactations (Bastian and Brown 1996).The trend of increased proteolysis in milk from both

milking systems upon higher processing temperature,together with the statistical interaction between feed pres-sure and temperature, ascribes higher plasmin activity athigher temperature (Crudden et al. 2005). Overall, theresults are very encouraging with regard to future imple-mentation of the technology at larger dairy farms. It appearsthat within regular production settings on the equipment, nosevere mechanical disruption and enzymatic degradation arefacilitated.As the loss of lactose is of great disadvantage, especially

today, where lactose in itself holds great value, the extraenergy consumption that RO requires [UF consumed in thisstudy 1.96 kWh/100 l permeate (at 0.35 MPa feed pres-sure), whereas RO consumed 2.54 kWh/100 l permeate (at3 MPa feed pressure)] is low compared to the benefit ofretaining the native milk composition.

CONCLUSION

The results of this study show no advantage with regard tothe FFA concentration of using UF compared to RO. Adecrease in RO feed pressure from 3 to 2 MPa, as well asan increase in temperature from 4 to 10 °C, increased theconcentration of FFA. In the milk from AMS, the MFG sizedistribution did not change by RO processing. The 48 mmmembrane spacer caused a higher level of proteolysis thanthe 30 mil spacer. And contrary to FFA concentration, theproteolysis level increased at a higher feed pressure. Over-all, results show that the concentration of milk by use ofRO membrane filtration is possible, without causing severechanges to the milk quality.



REFERENCES

Bastian E D and Brown R J (1996) Plasmin in milk and dairy products:an update. International Dairy Journal 6 435–457.

Bengtsson G (1980) Lipoprotein lipase - Mechanism of product inhibi-tion. European Journal of Biochemistry 106 557–562.

de Boer R and Nooy P F C (1980) Concentration of Raw whole milk byreverse osmosis and its influence on fat globules. Desalination 35201–211.

Brans G, Schro€en C G P H, van der Sman R G M and Boom R M(2004) Membrane fractionation of milk: state of the art and chal-lenges. Journal of Membrane Science 243 263–272.

Crudden A, Patrick F F and Kelly A L (2005) Factors affecting thehydrolytic action of plasmin in milk. International Dairy Journal 15305–313.

Dickow J A, Larsen L B, Hammershoj M and Wiking L (2011) Coolingcauses changes in the distribution of lipoprotein lipase and milk fatglobule membrane proteins between the skim milk and cream phase.Journal of Dairy Science 94 646–656.

Evers J M (2004) The milkfat globule membrane—compositional andstructural changes post secretion by the mammary secretory cell. In-ternational Dairy Journal 14 661–674.

Fuc�a N, Pasta C, Impoco G, Caccamo M and Licitra G (2013) Mi-crostructural properties of milk fat globules. International DairyJournal 31 44–50.

Garcia A and Medina B (1988) On-farm membrane concentration ofmilk. American Society of Agricultural Engineers 31 274–278.

IDF (1991) Determination of Free Fatty Acids in Milk and Milk Products.Bulletin 265. Brussels, Belgium: International Dairy Federation.

Ismail B and Nielsen S S (2010) Invited review: plasmin protease inmilk: current knowledge and relevance to dairy industry. Journal ofDairy Science 93 4999–5009.

Kelly P M (1987) Application of membranes to whole milk separa-tions: a review. Irish Journal of Food Science and Technology 11153–161.

Kelly A L, O’Flaherty F and Fox P F (2006) Indigenous proteolyticenzymes in milk: a brief overview of the present state of knowledge.International Dairy Journal 16 563–572.

Kessler H G (2002) Food and Bio Process Engineering - Dairy Technol-ogy, Chapter 4: Membrane Seperation-Processes, page 56–104.5th edn. M€unchen: A. Kessler.

Piry A, Kuhnl W, Grein T, Tolkach A, Ripperger S and Kulozik U(2008) Length dependency of flux and protein permeation in cross-flow microfiltration of skimmed milk. Journal of Membrane Science325 887–894.

Rasmussen M D, Wiking L, Bjerring M and Larsen H C (2006) Influ-ence of air intake on the concentration of free fatty acids and vacuumfluctuations during automatic milking. Journal of Dairy Science 894596–4605.

Slack A W, Amundson C H, Hill C G and Jorgensen N A (1982) On-farm ultrafiltration of milk - Part 1 - technical feasibility studies. Pro-cess Biochemistry 17 6–11.

Tolkach A and Kulozik U (2005) Fractionation of whey proteins andcaseinomacropeptide by means of enzymatic crosslinking and mem-brane separation techniques. Journal of Food Engineering 67 13–20.

Toro-Sierra J, Tolkach A and Kulozik U (2011) Fractionation of a-Lac-talbumin and b-Lactoglobulin from whey protein isolate using Selec-tive Thermal Aggregation, an Optimized Membrane Separationprocedure and resolubilization techniques at pilot plant scale. Foodand Bioprocess Technology 6 1032–1043.

Walstra P, Wouters J T M and Geurts T J (2006) Membrane processes.In Dairy Science and Technology, pp 341–356. Boca Raton: CRCpress as part of Taylor and Francis Group.

Wiking L (2011) Milking and handling of raw milk – influence on freefatty acids. In Encyclopedia of Dairy Sciences, 2nd edn, pp 638–641.Fuquay J W, Fox P F and McSweeney P L H, eds. San Diego: Aca-demic Press.

Wiking L, Frøst M B, Larsen L B and Nielsen J H (2002) Effects ofstorage conditions on lipolysis, proteolysis and sensory attributes inhigh quality raw milk. Milchwissenschaft 57 190–194.

Wiking L, Bj€orck L and Nielsen J H (2003) Influence of feed composi-tion on stability of fat globules during pumping of raw milk. Interna-tional Dairy Journal 13 797–803.

Zall R R (1984) Membrane processing of milk on the dairy farm. FoodTechnology 38 38–91.



Paper II

Caseinomarcropeptide release in relation to rheological properties during rennet coagulation of

raw milk reverse osmosis retentate

Ida Sørensen, Thao T. Le, Gitte Hald Kristensen, Lotte Bach Larsen & Lars Wiking

Manuscript in preparation, intended for publication in International Dairy Journal

1

Caseinomarcropeptide release in relation to rheological properties during rennet coagulation of 1

raw milk reverse osmosis retentate 2

3

Ida Sørensen, Thao T. Le, Gitte Hald Kristensen, Lotte Bach Larsen & Lars Wiking* 4

5

Aarhus University, Department of Food Science, Blichers Allé 20, 6

Tjele, 8830, Denmark 7

8

*Corresponding author. Tel.: + 45 87157805. 9

E-mail address: [email protected] 10

2

Abstract 11

12

Studies on new technology for concentrate milk at the dairy farm by reverse osmosis (RO) have 13

previously been conducted, yielding promising results on retaining the milk quality. Cheese dairies are 14

considered obvious recipients of the concentrated milk, but before implementation, the cheese making 15

properties has to be evaluated. During this study, several combinations of concentration factor and 16

rennet concentration was examined based on caseinomacropeptide (CMP) release and rheological 17

attributes of the coagulum. Furthermore, the calcium distribution of retentate compared to raw milk 18

was included in the study, as a possible explanation of observed differences. The results showed a clear 19

influence of the ratio between sample dry matter and rennet concentration on the CMP release and 20

coagulation onset. The calcium distribution shifted to a larger fraction of colloidal calcium in the 21

retentate samples compared to the raw milk. This could be part of the explanation to the increased curd 22

firming rate. 23

3

1. Introduction 24

25

Reverse osmosis (RO) membrane filtration on the dairy farm is regarded an option for saving transport 26

costs and increase bulk tank capacity (de Boer and Nooy 1980; Garcia III and Medina 1988), and one 27

of the obvious utilizations for the retentate would be in cheese production. In modern cheese 28

production, milk is concentrated and standardized according to its protein content, in order to save 29

rennet and cheese vat capacity. Ultra-filtration (UF) is often used to concentrate milk for standardizing 30

the milk protein content in order to utilize the cheese vats in cheese making. However, the process of 31

UF changes the milk composition, by allowing e.g. lactose and ions filtered through the membrane into 32

the permeate (Brans et al. 2004). Compared to using non-concentrated milk for cheese production, it 33

could be assumed that cheeses produced from UF retentate has different properties, and further on, that 34

cheese from RO retentate differs from UF cheese making properties due to altered milk component 35

concentrations. Benfeldt (2005) reported differences in the cheese ripening pattern, lower plasmin 36

activity and less degradation of β-casein when UF milk was used instead of regular raw milk was used 37

for the cheese production. 38

39

Casein is the group of milk proteins that makes the network of the basic cheese matrix. They exist as a 40

micellar colloidal structure of αs1-, αs2-, β- and κ-caseins. The exact arrangements of the casein species 41

are still being debated. However, the consensus is that α-caseins give the “framework”, β-casein is 42

dynamically drifting through the micellar structure and κ-casein is positioned primarily in the outer 43

layer (Dalgleish and Corredig 2012). The micelles are kept suspended in the water phase through a 44

negative charged electrostatic repulsion and hydrophilic characteristics. This is especially due to the 45

properties of the outer κ-casein (Fox and McSweeney 1998; McSweeney and Fox 2013). 46

4

47

Chymosin (EC 3.4.23.4) is an aspartic protease, commonly added to induce coagulation of the milk 48

during cheese production. The coagulation of milk is normally regarded as a two phase phases process: 49

First phase is the enzymatic hydrolysis of κ-casein into soluble fraction caseinomacropeptide (CMP) 50

and insoluble one. This reduces the electrostatic repulsion, and thus enables the casein micelles to form 51

a network structure. Second phase is the actual aggregation of the remaining insoluble part of the 52

casein micelles (without CMP). The second phase normally starts when approximately 70% of the κ-53

casein has been hydrolyzed (Walstra et al. 2006). The casein aggregation is mainly due to van der 54

Waals attraction and formation of calcium bridges. 55

56

Calcium is of vast importance for the coagulation properties if milk. The core of the casein micelle is 57

held together by colloidal inorganic calcium phosphate (CCP) which consists of 3Ca3(PO4)2CaH 58

citrate; Ca9(PO4)6; or CaHPO4٠2H2O. The Ca++ both found natively in the serum phase and as added 59

during the cheese making process neutralizes the negative charges of the casein and hereby facilitates 60

the aggregation (Lucey and Fox 1993), and therefore calcium in the ionic form is crucial for the 61

aggregation process in the second phase of coagulation. In addition, micellar calcium is important for 62

the integrity of the CN micelle and its coagulation properties (Udabage et al. 2001). Calcium bridges 63

afterwards further stabilize the aggregates. Lowering the pH results both in higher calcium ion activity 64

due to higher solubility of the CCP complexes (Koutina et al. 2014). The bovine milk calcium exist as 65

an equilibrium between CCP, bound as complexes in the serum phase and as free Ca++ ions. The 66

distribution is temperature dependent and reversible when re-heating after cold storage (Malacarne et 67

al. 2013). 68

5

When changes in coagulation properties are observed, it is meaningful to determine whether the change 69

is due to first or second phase. This will allow more targeted actions. The distinction between first and 70

second phase can be made by comparing the rate of released CMP representing the first phase to the 71

rennet coagulation time and curd firming rate on the other hand, representing the second phase. Sandra 72

et al. (2012) concluded that calcium activity only plays a role during the second phase of the 73

coagulation process, and did not have any influence on the first phase. It could be hypothesized that the 74

calcium distribution and the general ion strength might change during RO filtration, and thus affect the 75

second phase of coagulation. 76

77

During cheese production, not only the rennet coagulation time is of importance, but also the final 78

firmness and the rate of which its coagulum is formed matters. Frederiksen et al. (2011a) as well as 79

Logan et al. (2015) described in their studies that the size of casein micelles and fat globules have an 80

influence on the development of curd firmness, and thus giving higher cheese yield (Jensen et al. 81

2012). The combination of large fat globules and small caseins gave the highest curd firming rate, and 82

the combination of small fat globules and large caseins gave the lowest. They ascribed this mechanism 83

to smaller casein micelles being able to pack closer together, whereas the fat might only play a 84

secondary role. These interactions could also be affected in a concentrated system, where the 85

components are closer together and the water activity lower. The aim of this study was to examine the 86

cheese making properties of RO retentate produced from raw milk at a farm. This includes the effects 87

of the calcium distribution on the chymosin reaction and further on the rheological behaviour. 88

89

90

91

6

92

93

2. Materials and methods 94

2.1 Raw materials and retentate production 95

The RO membrane filtration was conducted at Danish Cattle Research Centre (Aarhus University – 96

Foulum, Tjele, Denmark), as an inline process with a 3.8” pHt spiral wound membranes (Alfa Laval, 97

Lund, Sweden), with a total surface area of 4.7 m2. Pressure across the membranes was 30 bar, and the 98

process temperature was kept at 4 ºC. A closer description of the equipment and filtration process has 99

been published by Sørensen et al. (2016). Samples of raw milk were acquired before the start of 100

filtration. First, valves on the filtration equipment were set to 1.5 VCF and subsequently to 2 VCF. 101

Retentate samples were acquired from both 1.5 VCF and 2 VCF after 30 minutes of processing time to 102

ensure a stable process. Danish Cattle Research Centre (Aarhus University – Foulum, Tjele, Denmark) 103

supplied the raw milk (Danish Holstein breed) for the experiments. The milk was collected from the 104

bulk tank just after the morning milking, and poured into the balance tank of the filtration plant. The 105

experimental production was conducted as triplicates on separate days. 106

107

2.2 Sample composition 108

The overall composition of the retentate and raw milk measured by FT-IR (MilkoScan FT2, Foss, 109

Hillerød, Denmark), giving values of dry matter content, fat, protein, lactose, and solids non-fat. 110

After 24h cold storage, samples of raw milk, 1.5 VCF retentate and 2 VCF retentate were skimmed by 111

centrifugation (3500 rpm for 20 min at 4°C), and removal of the fat phase. 112

The milk serum phase was obtained by ultracentrifugation (Beckman-Coulter Optima 113

7

L-80XP, Beckman Coulter Inc., Brea, CA) skim milk at 100000 × g at 30 ºC for 1 hour in a T4-TI-70 114

rotor. 115

116

Total calcium of both skim milk and serum phase was measured by titrating with 117

Ethylenediaminetetraacetic acid (EDTA). The samples were acidified to pH 4.3, centrifuged at 3500 118

rpm for 5 minutes. The supernatant was collected and added 0.1 N borax buffer. The titration was 119

conducted with a Calcium electrode (sclON 6.1241.070, Metrohm, Herisau, Switzerland) and a 120

reference electrode (LL ISE Reference 6.0750.100, Metrohm, Herisau, Switzerland) on an auto-121

titration system (862 Compact Titrosampler, Metrohm, Herisau, Switzerland). Analytical replication 122

was conducted as duplicates. Further description of the method has been made by (Poulsen et al. 2017). 123

124

The method for ionic calcium concentration measurements was based on work by Koutina et al. (2015) 125

using a Ca2+-meter (LAQUA twin compact Ca2+ METER B-751, electrode model S050, Horiba, Kyoto, 126

Japan) directly on the sample serum phase. The obtained ionic strength (Mw) was converted to [Ca2+ ] 127

through a standard curve of CaCl2 solutions. 128

129

The distribution of calcium fractions was calculated as follows: 130

[Ca colloidial] = [Ca total, skim] – [Ca total, serum] 131

[Ca serum bound] = [Ca total, serum] – [Ca2+] 132

133

134

2.3 Rheological measurements 135

2.3.1 Reorox 136

8

The skimmed samples were pH adjusted to 6.5 with 10% (vol/vol) lactic acid and incubated at 33°C for 137

30 min in a water bath. Chymosin (ChyMax, 200 IMCU /ml, Chr. Hansen, Hørsholm, Denmark) was 138

diluted in Milli-Q ultrapure water (Millipore, Billerica, MA) such that adding 20µl to 10ml milk would 139

give 0.03, 0.04 and 0.05 international milk-clotting units (IMCU) as the final chymosin concentration. 140

Diluted chymosin was added to the sample, mixed well for 10 seconds and transferred to the rheometer 141

(AR G2, TA Instruments, New Castle, Delaware) (20 ml) and the ReoRox (ReoRox4, oscillatory 142

rheometer, MediRox AB, Nykobing, SE) (1 ml in each of 3 channels). 143

144

2.3.2 Conventional rheometer 145

The rheometer was set to run for 40 minutes at 33 ºC at an oscillatory strain of 6.89×10-5 rad at 1 Hz, 146

and data was obtained through Rheology Advantage Data Analysis V5.7.0 (TA Instruments, New 147

Castle, DE). The ReoRox method has been fully described by Frederiksen et al. (2011b). Rennet 148

coagulation time was chosen as the parameter from ReoRox to describe the second stage of 149

coagulation. The G’ max value was not reproducible, and therefore not included, as the samples with 150

firm gels would detach from the cup. 151

152

2.4 CMP determination by LC 153

The method for CMP determination was adapted from Frederiksen et al. (2011a) and (Jensen et al. 154

2015). The samples were incubated at 33°C for 30 min in a water bath prior to the start of the 155

experiment, and pH was adjusted to 6.5 with 10% (vol/vol) lactic acid. Chymosin (chymax) was diluted 156

in milli Q water such that adding 20µl to 10ml milk would give 0.03, 0.04 and 0.05 international milk-157

clotting units (IMCU) as the final chymosin concentration in each sample. One ml of milk sample 158

withdrawn at each time point: 0, 30 s, and 1, 2, 5, 10 and 20 min after addition of chymosin dilution, 159

9

and mixed with 20 µL of pepstatin stock solution [1 mg of pepstatin A (Sigma-Aldrich, St. Louis, 160

Missouri) per milliliter of 10% (vol/vol) acetic acid in methanol], placed on ice to stop the reaction. 161

The samples were carefully mixed with 100 µL of acetic acid (CH3COOH; 10% for raw milk samples, 162

15% for 1.5 VCF retentate and 20% for 2 VCF retentate) to reach pH 4.6. After 2 min of incubation on 163

ice, 100 µL of sodium acetate (CH3COONa; 1 N for raw milk samples, 1.5 N for 1.5 VCF retentate and 164

2 N for 2 VCF retentate) was added as a buffer, and the solution was mixed and centrifuged for 10 min 165

at 14000 rpm at 4°C. The supernatants containing the CMP was isolated, and stored at -18 ⁰C until the 166

further analysis of CMP by LC. One hundred µl of supernatant was mixed 300 µl of 6N GdnHCl and 6 167

µl of 1N DTE and incubated in a shaker at 37 ⁰C for 1 hour, then centrifuged at 14000 rpm at 7 ⁰C for 168

10 minutes. The supernatant was filtered through filtervails (Mini-UniPrep syringeless filter device, 0.2 169

µm pore, PTFE filter media, Whatman, Maidstone, GB) and loaded into the LC. 170

A commercial CMP standard was included to compare the results and the elution time. The CMP 171

standard was dissolved to a 5% solution, and prepared with GdnHCl and DTE. The samples were 172

loaded (injection volume of 50µl) into a LC-ESI-MS (Agilent LC 1100 series connected to an ESI-173

single-Q-MS, Agilent Technologies, Palo Alto, CA, USA) using the exact procedure described by Le et 174

al. (2016). Data analysis was conducted through LC/MSD ChemStation (Agilent Technologies, Santa 175

Clara, CA). The CMP content was calculated as the total curve area between 20 and 45 minutes 176

retention time divided by the total curve are between 20 and 75 minutes retentation time 177

(approximately total protein present) in accordance to the elution times of detected compounds 178

described by Le et al. (2016). 179

180

2.5 Statistics 181

10

The statistical analysis in this study was processed through the statistical freeware program R 3.0.1 (R 182

Foundation for Statistical Computing, Vienna, Austria), as an analysis of variance with a significance 183

level of P < 0.05. To distinguish difference between groups (raw milk, 1.5 VCF retentate and 2 VCF 184

retentate) Tukey’s honest significant difference test was applied. 185

11

3. Results 186

3.1 Composition and calcium distribution in milk and retentate 187

The raw milk used for the experiments had an average protein content of 3.71 % and a total dry matter 188

content of 9.6 % (Table 1). The retentate from 1.5 VCF concentration had an average protein content of 189

5.79 % and dry matter content of 13.68 %, and the 2 VCF retentate had 7.42 % protein and 16.9 % dry 190

matter. Thus, the actual concentration factor of 2 VCF was around 1.8-2, and the 1.5 VCF was 1.4-1.5. 191

The permeate had no measurable dry matter content by FT-IR. Concentrating the milk resulted in a 192

significant decrease in pH, from 6.68 to 6.52, measured after 24 hours of storage at 4 ⁰C. 193

194

The total calcium content increased from 1589 mg/l to 3210 mg/l between raw milk and 2 VCF. The 195

total calcium in the serum phases increased from 497 mg/l to 867 mg/l between raw milk and 2 VCF 196

retentate. Ionic calcium content did however not change between the raw milk and retentate samples. 197

The permeate had no measurable calcium content, neither bound nor ionic. 198

199

3.2 Curd formation 200

Results obtained from the conventional rheometer showed that raw milk had a shorter gelation time 201

compared to retentate concentrated by a VCF of 1.5 and 2 (Table 3a), measured as the time needed for 202

the elastic modulus to reach 1 Pa. Gelation time was negatively correlated to chymosin concentration. 203

The ratio between protein content and chymosin (Table 2) had a significantly negative impact on the 204

gelation time (P < 0.001). After 40 minutes of reaction time, both retentate samples (1.5 VCF and 2 205

VCF) reached a higher elastic modulus compared to the raw milk sample where a chymosin 206

concentrations of 0.04 IMCU and 0.05 IMCU were used (Table 3b). At 0.03 IMCU the coagulation of 207

the retentate samples, especially VCF 2 was delayed to such a degree, that here was no real gel 208

12

formation after 40 minutes. Likewise, the elasticity at 40 minutes was significantly influenced by the 209

ratio between protein content and chymosin concentration individually as well as the ratio between the 210

two (P < 0.05). An example of the elasticity during the coagulation process for the three different 211

sample type with a chymosin concentration of 0.05 IMCU can be seen in Figure 3. The rennet 212

coagulation time obtained from the ReoRox measurements were very much in line with the results 213

from the rheometer (Table 4). Figure 4 shows the relation between rennet coagulation time and the 214

protein to chymosin ratio. 215

216

3.3 CMP release 217

The CMP content was calculated as the peak area at 214 nm UV absorbance curve from 20-45 minutes 218

retention time compared to the total peak area of 20 to 75 minutes retention time. Since it was not 219

possible to make an accurate protein quantification on the LC measurements, the CMP content was not 220

converted into a mass unit, but kept as a relative value to the total peak area. Figure 1 shows 221

representative results of LC measurements of whey from a retentate sample, a raw milk sample and a 222

CMP standard. It is noticed that the retention time of CMP, match in all the different samples – 223

including the standard. The standard does however not contain peptides and proteins with a retention 224

time above 42 minutes, as there are no whey proteins. According to the study from Le et al. (2016), the 225

peak at 33 minutes represents CMP variant A and the peak at 42 minutes represent CMP variant B. The 226

areas from 20 to 30 and from 44 to 55 minutes are protein fragments. The peaks between 60 and 75 227

minutes are α-Lactalbumin and β-Lactoglobulin variants. The retentate samples generally had a larger 228

peak area of CMP compared to samples from raw milk, but the peak area from the whey proteins were 229

of relatively higher intensity compared to the CMP peaks. Thus, the relative CMP content was lower in 230

the retentate samples. CMP variant A appeared to be of slightly higher abundance than CMP variant B, 231

13

which is in accordance to the results of Jensen et al. (2015) on milk samples from Danish Holstein 232

cows. 233

234

After 20 minutes, the chymosin reaction was stopped for all samples, since excessive gelation made 235

sample withdrawal inaccurate. As can be seen on Figure 2, the relative amount of released CMP did not 236

show a consistent pattern during the first 2 minutes of reaction time, whereas the time interval between 237

2 to 10 minutes displayed a certain degree of linearity for all the combinations of chymosin 238

concentration and milk samples. The retentate samples released relatively less CMP compared to the 239

raw milk sample at a constant chymosin concentration (Figure 2). Furthermore, the relative amount of 240

CMP found after 10 minutes of reaction time is positively correlated (P<0.001) with VCF : chymosin 241

ratio. The relative CMP after 10 minutes reaction time was negatively correlated to protein content and 242

positively correlated to chymosin concentration, and the slope of the linear area between 2 and 10 243

minutes was negatively correlated to the protein content of the samples (P< 0.05), with no significant 244

influence of chymosin concentration. 245

246

247

4. Discussion 248

Several studies have dealt with coagulation and curd quality from UF retentate (Sharma et al. 1993; 249

Waungana et al. 1998; Sandra et al. 2011). The primary distinction between UF and RO retentate is the 250

ionic balance systems. Changes to the ionic balance plays an important role during milk coagulation 251

(Ferrer et al. 2008; Zhao and Corredig 2016), and especially calcium distribution is known to greatly 252

influence the coagulation properties (Klandar et al. 2007). An uptake of calcium into the casein micelle 253

is associated with a decrease in pH due to exchange between Ca2+ and H+. This will typically improve 254

14

the coagulation properties as the optimum pH for chymosin is approached, as well as a neutralizing 255

effect of the casein micelle electrostatic repulsion. In this study, the sample pH was standardized prior 256

to the rheological tests, so the contribution of calcium to the milk acidity is no longer a factor that can 257

explain the observations. The observed decrease in pH in the present study was likely a result of ionic 258

balances and acids present in the milk - such as free fatty acids and amino acids being concentrated in 259

RO process (Sørensen et al. 2016), rather than caused by microbial growth. Both protein content and 260

pH of the raw milk used in this study is comparable to the levels found by Jensen et al. (2012) in milk 261

with good coagulation properties from Danish Holstein-Friesian cows. As pH decreased due to the 262

membrane filtration, it was not possible to make a complete distinction between concentration level 263

and pH as the single factor influencing the calcium distribution. In order to do so, the pH should have 264

been adjusted prior to determine the calcium content of the various phases. 265

266

The raw milk calcium contents observed in this study were at the higher end of what has previously 267

been reported for milk that exhibit good coagulation properties (Maciel et al. 2015), and generally a 268

high calcium content has been correlated with higher curd firming rate and gel strength. The percentage 269

distributions of calcium fractions reveal that the ionic calcium decreased and the casein bound 270

increased, whereas the serum bound remained constant. Since calcium has great impact on cheese 271

making properties, it has been common to use it as a supplement to the cheese milk. Adding calcium to 272

the system does however not result in the same calcium distribution pattern as was observed in this 273

study as Philippe et al. (2003) observed an increase in serum and ionic calcium after supplementation 274

rather than calcium binding to the casein micelles. This difference in observations might be ascribed in 275

the calcium to protein ratio, which changes when calcium is added to the system rather than being 276

constant as is seen for the retentate. 277

15

278

One way to evaluate coagulating properties is to measure viscosity and elasticity through rheological 279

measurements as the coagulum is formed. This provides comparable data of both the time it takes to 280

form the coagulum (either the time it takes to change from mainly viscous to mainly elastic properties, 281

or the time to reach maximum coagulum strength) and the strength of the coagulum. These properties 282

are important, and need to be considered in a cheese production, in order to control the production 283

process (Frederiksen et al. 2011b). Based on comparison between raw milk curd and retentate curd it 284

can be seen dry matter, and hereby protein concentration, together with the chymosin concentration 285

influence both curd firming rate and coagulation onset. Similar effects on like delayed coagulation 286

onset and increased curd firming rate have been observed with UF retentate and can be explained the 287

change in coagulation properties by a stronger interaction between casein micelles when the initial 288

distance between them is shorter (Karlsson et al. 2007). The results of CMP release indicates that κ-289

casein is hydrolysed at a lower rate in concentrated milk presumably due to the decreased substrate to 290

enzyme ratio. Karlsson et al. (2007) and Caron et al. (1997) found increased gel firmness in rennet 291

coagulated milk when increasing the protein content by addition of milk powder. The results of these 292

studies are very much in line with the observations made in this current study, even by a different 293

setup. The electrostatic interactions between casein micelles are likewise affected by the ionic balance 294

of both Ca2+, Na+ and pH (Karlsson et al. 2005), all of which are influenced by the RO process. 295

296

In order to distinguish between effects on first or second phase of coagulation, the study on the 297

hydrolyzation rate of κ-casein were used to describe first phase. Rennet induced aggregation of casein 298

micelles normally happen when approximately 70 % of the κ-casein has been hydrolysed as a pseudo 299

first order reaction rate (Mellema et al. 1999). Increasing the rennet concentration does however 300

16

change the reaction rate with a higher slope of the linear phase (Lomholt and Qvist 1999; Sandra et al. 301

2007). The same effect was not observed in this present study, as protein content of the samples 302

appeared to be the dominant factor. Sandra et al. (2011) found a significant relation between protein 303

content in UF retentate and gel strength, with no changes to the amount of κ-casein that needed to be 304

hydrolysed. As in accordance to our study, Lomholt and Qvist (1997) found that higher rennet 305

concentration contributed to a higher gel-firming rate, even after all κ-casein had been hydrolysed, and 306

thus speculated rennet to contribute to the final gel strength beside the hydrolytic process. Since the 307

substrate to enzyme ratio was lower in the retentate compared to raw milk, this effect has likely not 308

been a contributing factor during this present study. This confirms the theory that the changes in the 309

rennet coagulation properties of RO retentate was not directly influenced by the first phase; only due to 310

the ratio between substrate and enzyme. 311

312

5. Conclusion 313

Concentrating raw milk by reverse osmosis caused a shift in calcium distribution towards more 314

colloidal calcium and less ionic serum calcium followed by, and a decrease in pH. A difference in 315

coagulation properties between RO retentate and raw milk was observed, leading to longer rennet 316

coagulation time but a higher curd firming rate. By decreasing the concentration of chymosin, the 317

retentate samples showed a decrease in curd firming rate and longer rennet coagulation time. The 318

results are comparable to findings by Caron et al. (1997) and Karlsson et al. (2007), who assigned the 319

observations to chymosin to protein ratio and changes in casein micelle interactions. These properties 320

were highly ascribed to the ratio between rennet and sample protein concentration. This statement 321

supports the results in this present study, where also the rennet reaction rate was highly dependent on 322

the sample protein content and rennet concentration. Thus, it appears that mechanical process of RO 323

17

does not contribute to changes in coagulation properties, as it is primarily an effect of substrate to 324

enzyme ratio, which could explain the delay of rennet coagulation time. However, the increased curd-325

firming rate could be related to the calcium distribution and ionic balances, and this would require 326

further studies to confirm. 327

328

6. References 329

Benfeldt C (2005) Ultrafiltration of cheese milk: Effect on plasmin activity and proteolysis during cheese 330 ripening Int Dairy J 16:600‐608 331

Brans G, Schroën CGPH, van der Sman RGM, Boom RM (2004) Membrane fractionation of milk: state of the art 332 and challenges Journal of Membrane Science 243:263‐272 333

Caron A, St‐Gelais D, Pouliot Y (1997) Coagulation of Milk Enriched with Ultrafiltered or Diafiltered 334 Microfiltered Milk Retentate Powders International Dairy Journal 7:445‐451 335

Dalgleish DG, Corredig M (2012) The Structure of the Casein Micelle of Milk and Its Changes During Processing 336 Annu Rev Food Sci Technol 3:449‐467 337

de Boer R, Nooy PFC (1980) Concentration of raw whole milk by reverse osmosis and its influence on fat 338 gloubles Desalination 35:201‐211 339

Ferrer MA, Hill AR, Corredig M (2008) Rheological Properties of Rennet Gels Containing Milk Protein 340 Concentrates J Dairy Sci 91:959‐969 341

Fox PF, McSweeney PLH (1998) Lactose. In: Dairy chemistry and biochemistry. Blackie Academic & profetional, 342 London, UK, pp 21‐66 343

Frederiksen PD et al. (2011a) Composition and effect of blending of noncoagulating, poorly coagulating, and 344 well‐coagulating bovine milk from individual Danish Holstein cows J Dairy Sci 94:4787‐4799 345

Frederiksen PD, Hammershøj M, Bakman M, Andersen PN, Andersen JB, Qvist KB, Larsen LB (2011b) Variations 346 in coagulation properties of cheese milk from three Danish dairy breeds as determined by a new free 347 oscillation rheometry‐based method Dairy Science and Technology 91:309‐321 348

Garcia III A, Medina B (1988) On‐farm Membrane Concentration of Milk American Society of Agricultural 349 Engineers 31:274‐278 350

Jensen HB, Pedersen KS, Johansen LB, Poulsen NA, Bakman M, Chatterton DEW, Larsen LB (2015) Genetic 351 variation and posttranslational modification of bovine kappa ‐casein: effects on caseino‐macropeptide 352 release during renneting J Dairy Sci 98:747‐758 353

Jensen HB, Poulsen NA, Andersen KK, Hammershøj M, Poulsen HD, Larsen LB (2012) Distinct composition of 354 bovine milk from Jersey and Holstein‐Friesian cows with good , poor , or noncoagulation properties as 355 reflected in protein genetic variants and isoforms J Dairy Sci 95:6905‐6917 356

Karlsson AO, Ipsen R, Ardö Y (2007) Rheological properties and microstructure during rennet induced 357 coagulation of UF concentrated skim milk Internatioal Dairy Journal 17:674‐682 358

Karlsson AO, Ipsen R, Schrader K, Ardö Y (2005) Relationship between physical properties of casein micelles and 359 rheology of skim milk concentrate J Dairy Sci 88:3784‐3797 360

Klandar AH, Lagaude A, Chevalier‐Lucia D (2007) Assessment of the rennet coagulation of skim milk: A 361 comparison of methods Int Dairy J 17:1151‐1160 362

18

Koutina G, Knudsen JC, Andersen U, Skibsted LH (2014) Temperature effect on calcium and phosphorus 363 equilibria in relation to gel formation during acidification of skim milk Int Dairy J 36:65‐73 364

Koutina G, Knudsen JC, Andersen U, Skibsted LH (2015) Influence of colloidal calcium phosphate level on the 365 microstructure and rheological properties of rennet‐induced skim milk gels LWT ‐ Food Science and 366 Technology 63:654‐659 367

Le TT et al. (2016) Using proteomics to characterise storage‐induced aggregates in acidic whey protein isolate 368 drinks Int Dairy J 60:39‐46 369

Logan A, Leis A, Day L, Øiseth SK, Puvanenthiran A, Augustin M (2015) Rennet gelation properties of milk: 370 Influence of natural variation in milk fat globule size and casein micelle size Int Dairy J 46:71‐77 371

Lomholt SB, Qvist KB (1997) Relationship between rheological properties and degree of κ‐casein proteolysis 372 during renneting of milk J Dairy Res 64:541‐549 373

Lomholt SB, Qvist KB (1999) Gel firming rate of rennet curd as a function of rennet concentration Int Dairy J 374 9:417‐418 375

Lucey JA, Fox PF (1993) Importance of calcium and phosphate in cheese manufacture ‐ a reviewI J Dairy Sci 376 76:1714‐1724 377

Maciel GM, Hammershøj M, Frederiksen PD, Sørensen J, Bakman M, Poulsen NA, Larsen LB (2015) Dairy 378 processing and cold storage affect the milk coagulation properties in relation to cheese production 379 Dairy Science and Technology 95:101‐114 380

Malacarne M et al. (2013) Effects of storage conditions on physico‐chemical characteristics, salt equilibria, 381 processing properties and microbial development of raw milk Int Dairy J 29:36‐41 382

McSweeney PLH, Fox PF (2013) Casein Micelle structure, Functions and Inteactions. In: Advanced Dairy 383 Chemistry. 4 edn. Springer Science and Buisness Media, New York, U.S, 384

Mellema M, Leermakers FAM, De Kruif CG (1999) Molecular Mechanism of the Renneting Process of Casein 385 Micelles in Skim Milk, Examined by Viscosity and Light‐Scattering Experiments and Simulated by Model 386 SCF Calculations Langmuir : the ACS journal of surfaces and colloids 15:6304‐6313 387

Philippe M, Gaucheron F, Le Graet Y, Michel F, Garem A (2003) Physicochemical characterization of calcium‐388 supplemented skim milk Lait 83:45‐59 389

Poulsen NA et al. (2017) Novel genetic variation associated to CSN3 strongly affects rennet‐induced milk 390 coagulation Int Dairy J 71:122‐130 391

Sandra S, Alexander M, Dalgleish DG (2007) The rennet coagulation mechanism of skim milk as observed by 392 transmission diffusing wave spectroscopy Journal of colloid and interface science 308:364‐373 393

Sandra S, Cooper C, Alexander M, Corredig M (2011) Coagulation properties of ultrafiltered milk retentates 394 measured using rheology and diffusing wave spectroscopy Food Res Int 44:951‐956 395

Sandra S, Ho M, Alexander M, Corredig M (2012) Effect of soluble calcium on the renneting properties of casein 396 micelles as measured by rheology and diffusing wave spectroscopy J Dairy Sci 95:75‐82 397

Sharma SK, Hill AR, Mittal GS (1993) Effect of milk concentration, pH and temperature on aggregation kinetics 398 and coagulation properties of ultrafiltered (UF) milk Food Res Int 26:81‐87 399

Sørensen I, Jensen S, Ottosen N, Neve T, Wiking L (2016) Chemical Quality of Raw Milk Retentate processed by 400 Ultra‐filtration or Reverse Osmosis at the Dairy Farm. Int J Dairy Sci 69:31‐37 401

Udabage P, McKinnon IR, Augustin MA (2001) Effects of mineral salts and calcium chelating agents on the 402 gelation of renneted skim milk J Dairy Sci 84:1569‐1575 403

Walstra P, Wouters JTM, Geurts TJ (2006) Dairy Science and Technology. 2nd edn. Taylor & Francis Group, Boca 404 Raton, FL, U.S. 405

Waungana A, Singh H, Bennett RJ (1998) Rennet coagulation properties of skim milk concentrated by 406 ultrafiltration: Effects of heat treatment and pH adjustment Food Res Int 31:645‐651 407

19

Zhao Z, Corredig M (2016) Influence of sodium chloride on the colloidal and rennet coagulation properties of 408 concentrated casein micelle suspensions J Dairy Sci 99:6036‐6045 409

410

Table 1: composition of milk samples including calcium content and distribution between stages (bound to casein micelle, bound calcium in the serum phase and ionic calcium). Differences between sample types are indicated with letters and the significance level.

Raw 1.5VCF 2VCF Significance

level

DM 9.60±0.055a 13.68±0.381b 16.90±0.221c p<0.001

Protein 3.71±0.078a 5.79±0.324b 7.42±0.276c p<0.001

pH 6.68±0.026c 6.59±0.017b 6.52±0.026a p=0.01

Total calcium skim (mg/l) 1589±32a 2470±86b 3210±49c p<0.001

Total calcium serum

(mg/l) 496±24a 684±14b 867±24c p<0.001

Ionic calcium (mg/l) 100.61±3.550 107.27±4.613 106.16±1.463 NS

Casein bound calcium % 68.7 72.2 73.0

Serum bound calcium % 24.9 23.4 23.7

Ionic calcium % 6.33 4.35 3.31

Table 2: Ratio between average protein content of the samples and chymosin concentration.

IMCU Raw 1.5 VCF 2 VCF

0.03 0.008 0.005 0.004

0.04 0.011 0.007 0.005

0.05 0.013 0.008 0.007

Table 3a and 3 b: (3a) Gelation time, as the time needed to reach a G’ of 1 Pa, of the different sample types with increasing rennet concentration. (3b) The G’ meassured 40 minutes after addition of rennet to the samples with increasing rennet concentration.

Table 4: Rennet coagulation time of the different sample types with increasing rennet concentration, measured on ReoRox.

RCT / min


0.03 24.5±2.83 30.6±3.37 36.6±1.05

0.04 18.2±1.77 20.0±0.35 27.1±1.77

0.05 14.5±1.44 17.8±0.73 21.8±1.41

Gelation time (Pa=1) min


0.03 24.7±4.74 31.9±3.52 38.9±2.60

0.04 17.7±2.12 22.2±3.38 27.6±3.34

0.05 13.9±1.65 16.9±2.15 21.8±2.21

G' at 40 min


0.03 27.7±0.25 27.7±0.30 4.4±0.09

0.04 50.5±0.22 73.3±0.73 101.1±0.92

0.05 66.9±0.24 177.2±0.66 210.9±0.91

Figure 1: LC absorbnace spectrum of whey from a 2 VCF RO retentate sample and a raw milk sample reacted for 20 minutes with 0.05 IMCU rennet concentration - compared to a commercial G-CMP standard.

Figure 2: CMP content, relativ to total whey proteins, at certain timepoints during the renneting reaction on raw milk, 1.5 VCF retentate and 2 VCF retentate with 0.05 IMCU rennet concentration.

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.140

0.160

0 200 400 600 800 1000 1200 1400

Relative CMP content

time, s

Raw

1.5 VCF

2 VCF

Figure 3: Elasticity curve formed during the rennet coagulation process of raw milk, 1.5 VCF retentate and 2 VCF retentate with a rennet concentration of 0.05 IMCU.

Figure 4: Rennet coagulation time as a function of sample substrate to enzyme ratio, measured on ReoRox.

0.001

0.01

0.1

1

10

100

1000

0 5 10 15 20 25 30 35 40 45 50

log G' , Pa (elasticity)

Time, minutes

Raw

1.5 VCF

2 VCF

0

5

10

15

20

25

30

35

40

45

0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016

RCT / min

Chymosin to protein ratio

Paper III

Storage stability of whole milk powder produced from raw milk reverse osmosis retentate

Ida Sørensen, Tommas Neve, Niels Ottosen, Lotte Bach Larsen, Trine Kastrup Dalsgaard & Lars

Wiking

Dairy Science & Technology, 2017, Volume 96, Page 873–886

ORIGINAL PAPER

Storage stability of whole milk powder producedfrom raw milk reverse osmosis retentate

Ida Sørensen1& Tommas Neve2 & Niels Ottosen3

&

Lotte Bach Larsen1& Trine Kastrup Dalsgaard1

&

Lars Wiking1

Received: 13 June 2016 /Revised: 30 November 2016 /Accepted: 2 December 2016 /Published online: 19 December 2016# INRA and Springer-Verlag France 2016

Abstract Implementation of reverse osmosis filtration at the dairy farm will reduce thevolume of milk, which has to be transported, and thereby potentially reduce energyconsumption and CO2 emission. The aim of this study was to examine the quality ofwhole milk powder produced from reverse osmosis retentate concentrated at the farm.Whole milk powder prepared from reverse osmosis retentate, with a volume concen-tration factor of 2, was compared to powder from non-concentrated milk, as well as to arange of commercial whole milk powders. A storage experiment of the stability ofretentate powder for up to 12 months at room temperature was conducted and evaluatedfor quality parameters, including proteolysis, oxidation, furosine and colour. The resultsshowed that concentration of the oxidation products hexanal, heptanal and nonanalincreased during storage of both retentate powder and powder from non-concentratedmilk, but not to a higher extent than found in commercial powder of similar storageconditions. Detectable furosine was higher in powder prepared from non-concentratedmilk than that in powder from pre-concentrated milk, and further no changes in colourwas found during storage. However, high variation in powder composition betweenproduced powders, especially with regard to moisture content, could have affectedsome quality parameters. In conclusion, pre-concentrating milk by reverse osmosis atthe farm did not have significant effects on the overall quality of the produced milkpowders in this study.

Keywords Membrane filtration . Oxidation . Proteolysis . Maillard reaction . Furosine

Dairy Sci. & Technol. (2017) 96:873–886DOI 10.1007/s13594-016-0309-y

* Lars [email protected]

1 Department of Food Science, Aarhus University, Blichers Allé 20, 8830 Tjele, DK, Denmark2 Arla Strategic Innovation Centre, Arla Foods, Roerdrumvej 10, 8820 Brabrand, DK, Denmark3 Arla Foods Ingredients Group P/S, Sønderupvej 26, 6920 Videbæk, DK, Denmark

http://crossmark.crossref.org/dialog/?doi=10.1007/s13594-016-0309-y&domain=pdf

1 Introduction

Milk powder quality is commonly evaluated in two categories: physical and chemical.The physical quality is related to handling the powder during production (flowability)during shipment (bulk density) and by the consumer (wettability). These attributes areoften associated with production-related factors such as atomization and dry matter ofthe evaporated milk and thereby size distribution and surface composition of thepowder particles (Kim et al. 2009; Murrieta-Pazos et al. 2012). However, much ofthe powder quality depends on the raw material. A high level of proteolytic andlipolytic activities in the milk, due to either bacterial growth or somatic cells, willtransfer these qualities to the milk powder and perhaps advance the reactions found inthe powder during storage (Celestino et al. 1997). Sert et al. (2016) has in a recent studyfound a correlation between elevated somatic cell count and loss of functional proper-ties such as solubility, wettability and dispersability of whole milk powder. Thisresulted in poor texture when the powder was reconstituted into yoghurt.

In relation to chemical quality, nutritional value and flavour may be compromisedthrough alteration of fat and proteins and by oxidation. Therefore, it is of absoluteimportance to ensure the product quality when implementing new technologies. Heattreatments of the milk prior to drying might denature whey proteins which can theninteract with the casein micelles, and thereby alter the functionality (Singh 2007). TheMaillard reaction is also catalysed by high temperature, moisture content and pH andwill influence the appearance, colour, flavour, odour and digestibility due to essentialamino acids being less accessible after lactosylation (Thomas et al. 2004; Dalsgaardet al. 2007). Furosine is formed as a further stage of lactosylation, when fructoselysine,generated from lactulosyl-lysine, is hydrolysed in acidic conditions during analysis.This rather early stage of Maillard reaction is a good indicator of heat damage, sincethese products are often found in freshly produced milk powder and UHT milk, andstudies have correlated the formation of furosine with especially heat modifications ofβ-lactoglobulin (Van Renterghem and De Block 1996). Humidity plays a significantrole in Maillard reaction during storage of the powder. Even at low storage temperature,where furosine is not normally formed, humid conditions can accelerate the process(Van Renterghem and De Block 1996). Maillard reaction that reaches some of the endproducts will often go by the nomination of non-enzymatic browning and will yield apowder with a lower lightness value (L*) and more yellowness (b*) and redness (a*)(Thomas et al. 2004). Oxidation is another aspect that will give unpleasant flavour tothe powder and is often associated with surface composition of the particles and storageconditions (Pisecky 1997; Romeu-Nadal et al. 2007). Secondary lipid oxidation prod-ucts, such as hexanal, heptanal and nonanal, have aromatic properties linked to grassyand floral/citrus notes (Mahajan et al. 2004). Lipid oxidation in milk powder is assumedto be mainly from autoxidation, since enzymatic oxidation in powder is insignificantdue to the low water activity (Parkin 2008).

We previously showed that concentrating milk at the farm using reverse osmosis(RO) did not induce any change of free fatty acid concentration and proteolytic activity,provided that raw milk of good quality was used (Sørensen et al. 2016). It was thereforeassumed that the products made from the RO retentate, such as powder, would be ofsimilar quality compared to the products made from raw milk. The aim of this studywas thus to compare the quality of whole milk powders obtained either from RO

874 Sørensen I. et al.

retentate or from raw milk. In this aim, we produced milk powders from the same rawmilk, either pre-concentrated or not on the same spray drying pilot plant, and wecharacterized the properties of these powers and their storage stability, in comparison toreal-scale commercial whole milk powders.

2 Materials and methods

2.1 Production of retentate

Danish Cattle Research Centre (Aarhus University—Foulum, Tjele, Denmark) provid-ed raw milk for the RO filtration studies. From the bulk tank, 800 L of milk wastransferred to a smaller bulk tank with cooling and stirring system. The milk was amixture from Danish Holstein and Jersey cows (2:1 ratio). Membrane filtration wasconducted at the Danish Cattle Research Centre as a batch process, where the milkcirculated between the small bulk tank and the filtration plant (pilot filtration plantproduced by GEA, Skanderborg, Denmark), until the desired concentration factor wasreached. The filtration was carried out by RO with two 3.8″ pHt spiral woundmembranes produced by Alfa Laval (Lund, Sweden), with a total surface area of2 × 4.7 m2. Pressure across the membranes was 30 bar, and the process temperaturewas kept a 4 °C. To produce 400 L of permeate (with a volume concentration factor of2), a process time of 9 h was required. All the batches of raw milk had a dry matter of13.7%, and the RO filtration resulted in retentate with a dry matter of 25.0 and 24.7%.The morning after concentrating the milk, the retentate and 400 L of fresh raw milkfrom the same heard was transported to GEA Niro (GEA Process Engineering, Søborg,Demnark) and stored for 1 day at 4 °C, before further processing. This experimentalproduction was conducted as duplicates of both RO retentate and fresh raw milk on twosubsequent days. The samples, ‘Non-conc 1’ and ‘Conc 1’ referring to non-concentrated raw milk and pre-concentrated raw milk, were dried on the same day—on the second day after the production of ‘Conc 1’ retentate. The procedure was thenrepeated for ‘Non-conc 2’ and ‘Conc 2’.

2.2 Powder manufacture and storage

Before powder production, both the retentate and raw milk batches were thermised at67 °C for 90 s and afterwards evaporated to approximately the same level of dry mattercontent (41.5–46.8%), followed by homogenization. The powder was produced as amultistage drying process, with an inlet temperature of 180 °C and outlet of 75 °C. Asystem of three-stage fluidized beds was used, and powder fines retrieved from theoutlet air were recycled into the drying chamber close to the atomizer nozzle. Aschematic overview of the milk treatment and powder production is shown in Fig. 1.

The powders were stored in air and light sealed bags at room temperature ofapproximately 20 °C for 3, 6 and 12 months. The atmosphere was not modified, andno vacuum applied to the bags. Aliquots for furosine and oxidation measurements weretaken at the indicated time points and stored at −80 °C until analysed. The referencepowders were acquired from Arla Arinco (Arla Foods, Videbæk, Denmark) and ArlaAkafa (Arla Foods, Svenstrup, Denmark) and had been stored at the dairies in bulk

Quality of powder from raw milk RO retentate 875

bags at room temperature. Two reference samples had been stored in sealed packagewith alternated atmosphere. All the reference samples were acquired and analysed atthe same time at the end of the storage experiment. Thus, all the reference samples wereproduced at different times and as different batches.

2.3 Powder composition

2.3.1 Protein content

Protein content was estimated by the Kjeldahl method (AOAC 2005), using KjeltecInstruments (Kjeltec System Autosampler 8460 and Kjeltec System Autoburette 8400Analyser Unit, Foss, Hillerød, Denmark). The protein content was determined bydissolving a 0.2 g powder sample in 12 mL 98% sulfuric acid and 5 mL pure hydrogenperoxide. Two Kjeldahl taps (copper sulphate and potassium sulphate, Foss, Hillerød,Denmark) were added to the solution. The samples were then set to react at 420 °C for1 h and 20 min and added sodium hydroxide 27.5% to convert the ammonium sulphateinto ammonia gas, before being titrated with hydrochloric acid 0.1 M. The nitrogencontent was multiplied by a factor of 6.38, to calculate the estimated protein content ofthe sample.

2.3.2 Fat content

Total fat content was measured by the Rose-Gottlieb method (AOAC 2000). A 0.3 gpowder sample was dissolved in 10 mL demineralized water overnight. Afterwards,1 mL ammonia was added to the tubes together with 10 mL ethanol 96% and congo red

Two batches of both concentrated and non-concentrated 1 day of storage

Thermised

Retentate production

Homogenized

Evaporated

Transported/ stored at GEA

Spray dried

67ºC 90 s

Inlet temp. 180ºC Outlet temp. 75ºC Water content 1.9-3.6

Dry matter 41.5% - 46.8%

Fig. 1 Process diagram of powder production on pilot spray drier. Included is the pre-concentration step


indicator solution. The fat phase was then dissolved in 25 mL diethyl ether and 25 mLpetroleum ether. After 30 min, the fat phase was withdrawn and the solvents evaporateso the fat was left at the bottom of the tube.

2.3.3 Insoluble particles

Insoluble particles were measured by dissolving 10 g of sample in 100 mL of waterusing a Solubility Index Mixer (type AC, Labinco BV, Breda, Nederland) for 1.5 min.The samples then rested for 15 min before being stirred and poured into 50-mLcentrifuge tubes and centrifuged for 5 min at 800 rpm at room temperature. The top45 mL was poured off and replaced by water. The samples were stirred beforecentrifugation was repeated. The amount of sediment was then regarded the insolubilityindex.

2.3.4 Surface free fat

Surface free fat analysis was conducted by Arla Arinco (Videbæk, Denmark). Thesamples were washed with petroleum ether, filtrated, and the solvent evaporated, soonly the surface fat of the particles would be left in the beaker.

2.3.5 Particle size distribution

Particle size distribution was determined by sieving 50 g of sample through a sievetower with a 630 μm and a 400 μm metal sieve with an amplitude of 60 for 3 min. Thedistribution of powder between the sieves was then measured.

2.4 Proteolysis

The powder was reconstituted in demineralized water to approximately the sameconcentration as raw milk and kept for 3 days at 4 °C. The composition of thereconstituted milk was measured by FT-IR (MilkoScan FT2, Foss, Hillerød, Denmark).The level of proteolysis was measured by the reaction between N-terminals of aminogroups and fluorescamine and quantified by a standard row of leucine. Thus, the resultswere measured as Leucine equivalents (mmol·L−1) divided by the total protein content.A description of the fluorescamine assay can be found in the study by Wiking et al.(2002). Results were obtained by analysis on a multi plate reader (BioTek Synergy 2,Holm & Halby, Brøndby, Denmark) with the Gen5 1.07.5 software (BioTek Instru-ments, Winooski, VT, United States). The concentration of free N-terminals was thendivided by the protein content in the reconstituted milk, to find the relative proteolysisin the sample. The measurements were conducted in triplicates.

2.5 Oxidation

The method for measuring oxidation products was adapted from Jensen et al. (2011),on the same equipment, with a few modifications to the procedure. A sample size of200 mg powder was put into a HPLC tube with 1 mL water and 5 μL 0.01 mg·mL−1

internal standard (hexanal D12). For subtraction of the volatile compounds into the GC-


MS, a grey SPME fibre (50/30 um DVB/CAR/PDMS, stableflex 2 cm, Grey-notched)from Supleco (Bellefonte PA, USA) was incubated at 50 °C for 30 min and injectedinto a GC-MS MSD 5975 from Agilent Technologies (Waldbronn, Germany) with aninlet temperature of 275 °C. A mixed external standard was used for quantification ofthe oxidation products found: hexanal (targeted ion 56 m/z and qualifier ions 82 and72 m/z), heptanal (targeted ion 70 m/z and qualifier ions 55, 81 and 86 m/z) andnonanal (targeted ion 57 m/z and qualifier ions 98, 82 and 70 m/z) using hexanal D12as an internal standard according to Wold et al. (2015) using a target ion of 64 m/z andqualifier ions 80 and 92 m/z.

2.6 Colour

Colour was measured by colorimeter (Konica Minolta portable spectrophotometer, To-kyo, Japan), using the parameters L* (lightness), a* (red/green) and b* (yellow/blue). Themeasurements were conducted as triplicates through a thin transparent plastic bag.

2.7 Furosine

The method for furosine measurement was adapted from Jansson et al. (2014). Samplesof whole milk powder (0.15 g) were hydrolysed in 10 mL 8 mol·L−1 hydrochloric acidfor 20 h at 110 °C and filtered after cooling. The filtrated hydrolysate was diluted 1:4 in3 mol·L−1 hydrochloric acid, and 500 μL was transferred to HPLC filter vails. Furosineconcentration was determined through ion-pair RP-HPLC using a Spherisorb ODS25 μm column (250 × 4.6 mm i.d.) (Grace Davison, Australia), with 0.06 mol·L−1

acetate buffer and a flow rate of 0.5 mL·min−1. The amount of furosine was calculatedbased on a standard curve with a concentration of 0.3–10 μg·mL−1, made from a stocksolution of 0.604 mg·mL−1 furosine dihydrochloride (99.4% purity) from thePolyPeptide Group (Strasbourg, France).

2.8 Statistics

The statistical analysis of variance was processed through the statistical freewareprogram R 3.0.1 (R Foundation for Statistical Computing, Vienna, Austria). The effectsof powder fat content, storage time and protein content, respectively, on surface free fat,oxidation and furosine were analysed by the following model: γi = α + βχ(i) + ei,i = 1,…,11; where γ was the value of the dependant variable and χ was the value of theindependent variable in samples 1 to 11 and ei as the residual error. α and β were theintercept and slope estimated for the linear model. The effects of powder type onproteolysis and colour were tested by the model: γi = T(i) + ei; i = sample 1,…,11,where γ was the value of the dependant variable and Twas the effect of type in samples1 to 11 and ei as the residual error. Tukey’s HSD test (R package agricolae, version 1.2–4) was applied for evaluation of treatment differences among powder types. Oxidationproducts as dependant on interaction between storage time and powder type were of themodel: γij = SP(ij) + S(i) + P(j) + eij,; where γ refers to the value of the specific oxidationproduct as the dependant variable, Si = effect of storage time (i = 0, 3, 6, 12 months)and P = effect of powder type (j = conc, non-conc and reference). P < 0.05 was used asthe significant threshold in all the models.


3 Results

3.1 Powder composition and characteristics

Prior to powder production, the raw milk had a fat content of 4.5–4.9% and aprotein content of 3.6%; while the pre-concentrated milk had a fat content of8.3–8.6% and a protein content of 6.8–6.9%. After powder production, bothbatches of powder made from non-concentrated milk had a fat content of 17.8–21.9% and a protein content of 29.3–30.8% (Table 1). The powders from pre-concentrated milk had a fat content of 29.9–31.0% and a protein content of26.3–26.5%. Variation in the powder composition between pre-concentrated andnon-concentrated was presumably a result of handling at the spray dryingfacility. The water content of all the samples was between 1.8–3.5%. Thesurface free fat varies between 0.17 g/100 g fat to 2.59 g/100 g fat, and thehighest amounts were found in the powder from pre-concentrated milk. Therewas a significant (P < 0.05; R2 = 0.89) correlation between the total fat contentin the powder and the level of surface free fat. All of the powders made fromraw milk and pre-concentrated milk had similar particle size distribution, withabout 98% of the particle mass having a diameter of 400 μm or smaller. Theinstant reference powders had a larger portion of particles in the 630 to 400 μmrange, and the regular reference powders were more or less equally distributedbetween the fraction of 630 to 400 μm and 400 μm to smaller.

3.2 Proteolysis and oxidation during storage

Proteolysis was measured as level of free N-terminals in the newly produced powdersand subsequently again after 3, 6 and 12 months of storage and compared to commer-cial powders of different ages as references relative to protein content (Fig. 2). No trendtowards increased proteolysis during storage was observed, and only the powder originhad a significant influence on the level of free N-terminals. Overall, the commercialreference powders had a higher level of proteolysis relative to protein content comparedto the small-scale produced powder form raw and pre-concentrated milk. The powderfrom raw milk had the lowest level of proteolysis in spite of being the powder with thehighest protein content.

During storage, an increase of the secondary lipid oxidation products hexanal,heptantal and nonanal was found. Interaction between storage time and powder type(powder from RO retentate and non-concentrated milk) was significant (Fig. 3). Thehexanal content increased from below quantification limit of 10 ng/100 mg sample infreshly produced powder to about 98 ng/100 mg sample in 12-month-old powder fromraw milk and 66 ng/100 mg sample in powder from pre-concentrated milk. Only thereference powder stored in a bulk bag for 12 months showed detectable concentrationsof oxidation products (Ref. regular 12 month in Table 1), with a hexanal concentrationof 144 ng/100 mg sample (Fig. 3a). Even with a significant influence of powder originon hexanal development during storage, the specific powder composition did not haveany impact on the result. There is a significant effect (P = 0.0015) of interactionbetween storage time and moisture content on the hexanal concentration. Heptanalincreased from below quantification limit of 5 ng/100 mg sample for both raw and pre-


concentrated milk powder to about 21 ng/100 mg sample in the raw sample and 15 ng/100 mg sample in the pre-conc. sample (Fig. 3b). The 12-month-old reference sample(Ref. regular 12 month in Table 1) was comparable to the raw sample on heptanalcontent with 24 ng/100 mg sample. Like hexanal, there is a significant (P = 0.002)interaction between storage time and moisture content. Nonanal was below the detec-tion limit of 5 ng/100 mg sample in all the freshly produced samples and increased to9 ng/100 mg in the raw milk powders and 10 ng/100 mg in the pre-conc. powders after12 months of storage (Fig. 3c). The increase of nonanal during storage was significantfor all of them. After 6 months of storage, the raw milk powder had reached an averagelevel of nonanal comparable to what was also found after 12 months of storage. Theconcentration of nonanal powder from pre-concentrated milk increased significantlyfrom 6 to 12 months of storage and in the end exceeded the level found in the raw milkpowder samples, and a significant interaction between powder type and storage timewas found. However, the fat and protein composition of the powders did not have adirect impact on the nonanal development. The 12-month-old reference sample(Ref. regular 12 month in table 1) had a nonanal content of 6 ng/100 mgsample and is there for lower than both powder from raw milk and pre-concentrated milk.

Table 1 Composition and physical characterization of small-scale powders produced from non-concentrated(non-conc) raw milk and raw milk pre-concentrated (conc) at the farm from reverse osmosis, on twosubsequent days (1 and 2) of production, and several commercial reference samples (Ref.)—both instantand regular stored in bulk bags for up to 12 months and two reference powders stored in sealed bags withaltered air composition

Composition Particle size distribution

Fat%

Protein%

Water%

Surface fatg/100 g

fat

Insolubleparticles

≥630 μm 630–400 μm ≤400 μm

Non-conc 1 17.81 30.78 3.55 0.17 0.2 0.20 1.20 98.60

Conc 1 29.89 26.47 2.17 1.57 0.2 0.90 2.20 96.90

Non-conc 2 21.89 29.29 2.81 0.85 0.1 0.58 1.00 98.42

Conc 2 31.00 26.26 1.87 2.59 0.2 0.26 1.82 97.92

Ref. regular fresh 28.24 23.63 3.17 0.99 0.1 0.30 53.19 46.52

Ref. regular12 month

27.97 23.45 3.38 0.83 0.1 0.16 37.36 62.48

Ref. instant6 month

26.47 26.89 3.31 1.14 0.1 4.49 64.22 31.29

Ref. instant fresh 27.91 23.97 2.71 1.49 0.1 2.30 83.21 14.49

Ref. regular3 month

25.76 24.17 3.04 0.88 0.1 0.20 44.10 55.70

Ref instant sealed12 month

28.36 23.54 2.85 1.70 0.1 0.98 68.98 30.04

Ref regularsealed12 month

26.34 23.99 3.22 1.64 0.1 0.52 6.13 93.35


3.3 Colour changes and furosine formation during storage of powder

No change of colour was observed during storage, and the L*, a* and b* variables wereonly significantly affected by powder type (from non-concentrated milk, pre-conc. milkand various commercial whole milk powders) as seen in Fig. 4a–c. The commercialsamples were on average lighter but with more colour—a,* indicating greenish colourand b*, indicating yellowish colour. The powders made from raw milk were thedarkest—with the lowest L* value—and had almost the same level of a* and b* asthe commercial powders. The powders from pre-concentrated milk were the lowest incolour—a* and b* and a lightness as the reference samples.


0 2 4 6 8 10 12 14

Leu

-Equ

ival

ents

[m

mol

· L

-1]

/ pro

tein

%

18

20

22

24

26

28

30

32

34

36

38

Conc 1

Conc 2

Non-conc 1

Non-conc 2

Reference

Fig. 2 Proteolysis expressed as Leucine equivalents found in whole milk powders produced from non-concentrated (non-conc—batches 1 and 2) raw milk and pre-concentrated (conc—batches 1 and 2) raw milkmanufactured on a pilot-scale spray drier and several commercial reference powders, as dependent on storagefor 0, 3, 6 or 12 months at room temperature of approximately 20 °C

00

50

100

150

200

0 3 6 12

Hex

anal

ng/

100m

g sa

mpl

e


a

00

05

10

15

20

25

30

0 3 6 12

Hep

tana

l ng

/100

mg

sam

ple


b

Non-conc Conc Ref

00

02

04

06

08

10

12

0 3 6 12

Non

anal

ng/

100m

g sa

mpl

e


c

Fig. 3 Oxidation products found in powders produced from non-concentrated (non-conc) raw milk and pre-concentrated (conc) raw milk manufactured on a pilot-scale spray drier, compared to a commercial reference(Ref) sample. a hexanal concentration during storage. b Heptanal concentration during storage. c nonanalconcentration during storage. The error bars show the standard deviation found between thesamples of same type


Furosine increased significantly (P = 0.03) upon storage of the powder samplesproduced from non-concentrated and pre-concentrated milk. The raw milk powder hadthe highest level of furosine, both before and after storage, and it was the powder typewith the highest amount of protein. Furosine was positively correlated with both proteincontent (P = 0.004; R2 = 0.7) and storage time, but no interaction between those factorswas found by statistical analysis. The linear relation is not very evident, so a largersample size might provide a model with better fit. When adding all the commercialreference samples, a more robust statistical model could be formed on the influence ofprotein and storage time on the concentration of furosine (Fig. 5). There was nocorrelation found between colour and furosine concentration, with the only relationbeing the powder composition that affects both colour and furosine.

4 Discussion

4.1 Impact of powder composition and pre-concentrating milk at the farm

In our recent study, we found that the technology of concentrating milk at the farm didnot affect the milk quality regarding lipolysis and proteolysis activity (Sørensen et al.2016) and thus theorized that the concentrated milk is still of a quality suitable for high-quality milk powder production.

The powder composition had a significant effect on many of the quality parametersanalysed in this study, i.e. surface free fat, proteolysis, colour and as a co-influence on thefurosine concentration. The surface free fat was highly correlated with fat content as to beexpected. Fitzpatrick et al. (2004) showed that it is not only the total fat content that isresponsible for the final content of free fat, and Koc et al. (2003) listed a number ofproduction factors during the spray drying process that might contribute to the final level offree fat, such as process temperature and shear stress. Surface free fat onwholemilk powdermight however not affect important parameters such as flow ability and hydrophobicity,since it is inevitable to have the particle surface covered in free fat, and thickness of the free

95.0

95.5

96.0

96.5

97.0

97.5

98.0

Non-conc Conc Ref

L*

-5.0

-4.5

-4.0

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0Non-conc Conc Ref

a*

10

11

12

13

14

15

16

17

Non-conc Conc Ref

b*

a

b

b

a

b

c

a

b

c

Fig. 4 Colour L* (lightness), a* (greenness) and b* (yellowness) as an average during storage of powdersproduced from non-concentrated (non-conc) raw milk and pre-concentrated (conc) raw milk manufactured ona pilot-scale spray drier, compared to several commercial reference samples (Ref)—both instant and regular, ofvarious ages. The lower case letters (a, b and c) indicate significant difference between groups on the same plot(P < 0.05). The error bars represent the standard error of mean found between the samples of same typethroughout the entire storage period


fat layer is not correlated to quality loss (Nijdam and Langrish 2006; Kim et al. 2009). Ascan be seen from the results in this study, the powders made from pre-concentrated milkhad the highest amount of fat and surface free fat, but did not show any sign of poorerquality compared to the powder made from raw milk with less fat and surface free fat.

The powders from retentate and non-concentrated milk made on the small-scaleindustrial spray drier were characterized by most of the particles being smaller than400 μm. The particle size is highly dependent on production method, including initialconcentration, atomization, fluidized bed and return of fines into the drying chamber.Larger particles are associated with better wettability, flowability and lower bulkdensity and are therefore one of the parameters that defines instant powder (Pisecky1997). So the particle size distribution of the powders from the small-scale productionis to be expected, and the results correlate well with the study of Jin and Chen (2009),that reported particles in the size range of 100–450 μm.

Celestino et al. (1997) described how certain proteolytic enzymes, especially ofbacterial origin, might be resilient to the temperatures and conditions during spraydrying of the milk. The proteolytic activity during storage of the powder might howeverbe so low, that the proteolysis in the raw milk during storage before drying is still themajor contributor to the total concentration of proteolytic products. Thus, the activityduring powder storage may be negligible. These thoughts support the findings in thisstudy, where no effect of storage time has been observed, neither in the powdersproduced from raw and pre-concentrated milk nor any of the reference samples.

Generally, the colour of milk powder is often considered in relation to Maillardreaction (Le et al. 2011). In this study, no correlation between neither storage time nordevelopment of furosine was found, indicating that the Maillard reaction had not

Fig. 5 Furosine formation found in whole milk powders produced from non-concentrated (non-conc—batches 1 and 2) raw milk and pre-concentrated (conc—batches 1 and 2) raw milk manufactured ona pilot-scale spray drier and several commercial reference powders, as dependent on protein content of thepowders and storage time. Blue freshly produced, green 3 months, yellow 6 months and red 12 months


reached a level that would result in browning. It could be expected that powders withhigher protein content would be more vulnerable to the heat treatment during powdermanufacture Rozycki et al. (2007) since colour development in whole milk powder isfaster during heat treatment, at a given temperature, with increased protein content, andthis was especially the case at pH 6–7. Thus, this emphasizes the importance of rawmaterial quality and production process for the final product.

Overall, the composition of the produced powders varied considerably. This madethe interpretations of the influence of composition of the milk (normal or concentrated)on the resulting powder quality difficult. Even though it can be considered as a strengthof the present study that an industrial low scale powder production was applied, it wasalso evident that there were challenges with reproducibility, due to the variability inhow the concentrated and raw milk was handled before drying. It is therefore concludedthat the variation between trial days and batches were larger than the contribution fromvariation in the milk used for spray drying.

4.2 Storage stability in relation to oxidation and furosine formation

In this present study, the oxidation products hexanal, heptanal and nonanal were foundto increase during storage and with interaction of powder type. These oxidationcomponents are often associated with oxidation of whole milk powders and infantformulas during storage, and they all have a rather low odour threshold (Fenaille et al.2003; Romeu-Nadal et al. 2007). Fat and protein contents of the powder did not have asignificant influence on oxidation, and the powder made from raw milk generally had ahigher level of oxidation despite the lower fat content and lower surface free fat,compared to the pre-concentrated milk from the same heard. Likewise, Zunin et al.(2015) found that there is no correlation between free fat and level of oxidation. It iswell known that water activity accelerates the process (Nielsen et al. 1997; Stapelfeldtet al. 1997), and in the present study, moisture content and storage interacted on theformation of hexanal and heptanal. Therefore, even though neither of the powdersamples in this study have considerably high moisture content, the differences are stillenough to have measurable impact on the powder quality.

Furosine is a so-called artificial amino acid that is formed by the acid hydrolysisduring the analysis of the first stage Maillard reaction products. The disadvantage ofusing furosine as an indicator of Maillard reaction is that during the acid hydrolysis,lysine is formed together with furosine, and if the Maillard reaction is on a moreadvanced stage, where fructoselysine has been further degraded, furosine will not beformed. The advantages are, on the other side, that even though lysine is formed by theacid hydrolysis during analysis, the furosine formation is still consistent. So whenevaluating furosine results, it is important to keep in mind whether the Maillard reactioncould be on a more advanced stage (Thomas et al. 2004). In this study, no changes incolour was found during storage, indicating that the Maillard reaction was still at aninitial stage and thus furosine analysis was considered to be a reliable marker. Maillardreaction is known to be accelerated by water activity (Van Renterghem and De Block1996; Thomsen et al. 2005). Nevertheless, the results of this study indicated nocorrelation between moisture content of the powder and furosine formed, in contrastto the oxidation results. Protein content was, however, found to be associated with afurosine formation, in accordance with the study of Morgan et al. (2005). This supports


the findings in our study, where a combined effect of protein content and storage timeinfluenced the furosine formation, especially when including various commercialwhole milk powder as described by the model in Fig. 5. It could be argued that non-concentrated milk was subjected to more evaporation than pre-concentrated prior tospray drying. According to Oldfield et al. (2005), the preheat treatment is more harmfulto the whey proteins than the evaporation process. Thus, it does not appear to be of anydisadvantage to concentrate the milk through RO—even if it is conducted at the farm.Taken together, it seems like a promising method that does not have negative influenceon milk quality like proteolysis, in spite of the more concentrated milk matrix andcloser presence of e.g., milk enzymes and its milk substrates. However, an eventualimplementation will depend on economic calculations on the feasibility and willpotentially be of higher benefit in countries with long distances and extensive, butlarge farms.

5 Conclusion

The present study show that the concentration of proteolysis products depended on thepowder origin and especially the commercial powders had a higher level of proteolysis.Concentration of proteolysis products did not increase during storage of any of thepowders. Surface free fat was significantly correlated with the fat content of thepowders. Storage time influenced the concentration of the oxidation products hexanal,heptanal and nonanal for both powder from RO retentate and from non-concentratedmilk, and after 12 months, the oxidation was still in the same range as commercialwhole milk powder. Furthermore, hexanal and heptanal were influenced by the mois-ture content. The formation of furosine was dependent on both storage time and finalpowder composition, but the Maillard reaction was at an early stage and thus notreflected through colour measurements. Colour did not change during storage. The rawmaterial handling at the pilot-scale spray drier derived some compositional effect e.g.,protein and fat % in the powder, which could have affected some quality parameters,epically the moisture content of the produce powders was low. However, this is a well-known challenge when using pilot-scale instead of a real production scale. Overall,concentrating the milk at the farm prior to powder production did not affect the powderquality, compared to powder from non-concentrated milk, and thus seemed a promisingprocedure to avoid transport of large volumes prior to processing.

Acknowledgements We thank Søren Skjølstrup Jensen, Simon Andersen (Arla Food Ingredients, Videbæk,Denmark) and Nils Mørk (GEA Process Engineering, Skanderborg, Denmark) for the knowledge andtechnical assistance; Mette Krogh Larsen (Arla Arinco, Videbæk, Denmark) for acquiring reference samplesand laboratory facilities; and Rita Albrechtsen and Gitte Hald Kristiansen (Aarhus University, Deparment ofFood Science, Tjele, Danmark) for assisting in laboratory work.

References

AOAC (2000) Official Methods of Analysis of AOAC International. Method No. 905.02. 17th edAOAC (2005) Official Methods of Analysis of AOAC International. Method No. 930.29. 18th ed


Celestino EL, Iyer M, Roginski H (1997) The effects of refrigerated storage of raw milk on the quality ofwhole milk powder stored for different periods. Int Dairy J 7:119–127

Dalsgaard TK, Nielsen JH, Larsen LB (2007) Proteolysis of milk proteins Lactosylated in model systems. MolNutr Food Res 51:404–414

Fenaille F, Visani P, Fumeaux R, Milo C, Guy PA (2003) Comparison of mass spectrometry-based electronicnose and solid phase microextraction gas chromatography-mass spectrometry technique to assess infantformula oxidation. J Agric Food Chem 51:2790–2796

Fitzpatrick JJ, Iqbal T, Delaney C, Twomey T, Keogh MK (2004) Effect of powder properties and storageconditions on the flowability of milk powders with different fat contents. J Food Eng 64:435–444

Jansson T et al (2014) Lactose-hydrolyzed milk is more prone to chemical changes during storage thanconventional UHT milk. J Agric Food Chem 62:7886–7896

Jensen BM, Nielsen JH, Sørensen J, Mortensen G, Dalsgaard TK (2011) Changing oxidation in whey fatconcentrate upon addition of green tea extract. Eur Food Res Technol 233:631–636

Jin Y, Chen XD (2009) Numerical study of the drying process of different sized particles in an industrial-scalespray dryer. Dry Technol 27:371–381

Kim EH-J, Chen XD, Pearce D (2009) Surface composition of industrial spray-dried milk powders. 2. Effectsof spray drying conditions on the surface composition. J Food Eng 94:169–181

Koc AB, Heinemann PH, Ziegler GR (2003) A process for increasing the free fat content of spray-dried wholemilk powder. J Food Sci 68:210–216

Le TT, Bhandari B, Deeth HC (2011) Chemical and physical changes in milk protein concentrate (MPC80)powder during storage. J Agric Food Chem 59:5465–5473

Mahajan SS, Goddik L, QianMC (2004) Aroma compounds in sweet whey powder. J Dairy Sci 87:4057–4063Morgan F, Nouzille CA, Baechler R, Vuataz G, Raemy A (2005) Lactose crystallisation and early maillard

reaction in skim milk powder and whey protein concentrates. Lait 85:315–323Murrieta-Pazos I, Gaiani C, Galet L, Calvet R, Cuq B, Scher J (2012) Food powders: surface and form

characterization revisited. J Food Eng 112:1–21Nielsen BR, Stapelfeldt H, Skibsted LH (1997) Differentiation between 15 whole milk powders in relation to

oxidative stability during accelerated storage: analysis of variance and canonical variable analysis. IntDairy J 7:589–599

Nijdam JJ, Langrish TAG (2006) The effect of surface composition on the functional properties of milkpowders. J Food Eng 77:919–925

Oldfield DJ, Taylor MW, Singh H (2005) Effect of preheating and other process parameters on whey proteinreactions during skim milk powder manufacture. Int Dairy J 15:501–511

Parkin KL (2008) Enzymes. In: Damodaran S, Parkin KL, Fennema OR (eds) Fennema's food chemistry,Fourth edn. CRC Press, Boca Raton, pp. 331–435

Pisecky J (1997) Hb Milk Pow Manuf. Handbook of Milk Powder Manufacture. Niro A/S, Copenhagen.Romeu-Nadal M, Chávez-Servín JL, Castellote AI, Rivero M, López-Sabater MC (2007) Oxidation stability

of the lipid fraction in milk powder formulas. Food Chem 100:756–763Rozycki SD, Pauletti MS, Costa SC, Piagentini AM, Buera MP (2007) The kinetics of colour and fluorescence

development in concentrated milk systems. Int Dairy J 17:907–915Sert D, Mercan E, Aydemir S, Civelek M (2016) Effects of milk somatic cell counts on some physicochemical

and functional characteristics of skim and whole milk powders. J Dairy Sci 99:In pressSingh H (2007) Interactions of milk proteins during the manufacture of milk powders. Lait 87:413–423Stapelfeldt H, Nielsen BR, Skibsted LH (1997) Effect of heat treatment, water activity and storage temperature

on the oxidative stability of whole milk powder. Int Dairy J 7:331–339Sørensen I, Jensen S, Ottosen N, Neve T, Wiking L (2016) Chemical quality of raw milk retentate processed

by ultra-filtration or reverse osmosis at the dairy farm. Int J Dairy Sci 69:31–37Thomas MEC, Scher J, Desobry-Banon S, Desobry S (2004) Milk powders ageing: effect on physical and

functional properties. Crit Rev Food Sci Nutr 44:297–322Thomsen MK, Lauridsen L, Skibsted LH, Risbo J (2005) Temperature effect on lactose crystallization,

maillard reactions, and lipid oxidation in whole milk powder. J Agric Food Chem 53:7082–7090Van RenterghemR, De Block J (1996) Furosine in consumption milk and milk powders. Int Dairy J 6:371–382Wiking L, Frøst MB, Larsen LB, Nielsen JH (2002) Effects of storage conditions on lipolysis, proteolysis and

sensory attributes in high quality raw milk. Milchwissenschaft 57:797–803Wold JP, Skaret J, Dalsgaard TK (2015) The action spectrum for photooxidation versus the absorption

spectrum of full fat bovine milk. Food Chem 179:68–75Zunin P, Boggia R, Turrini F, Leardi R (2015) Total and “free” lipids in commercial infant formulas: fatty acid

composition and their stability to oxidation. Food Chem 173:332–338


Appendix 1

Somatic cell count in reverse osmosis retentate – unpublished data

Introduction: In order to consider the impact from SSC on results of e.g. proteolysis in the raw milk

retentate, it was decided to acquire data on SSC before and after RO membrane filtration. Due to very

low SSC in the raw milk, and no significant correlation to other results of the studies, it was decided

not to conduct further experiments on the topic and not publish the data.

Materials and methods: During two subsequent days of experiment, 30 ml of retentate and raw milk

sample was collected as duplicates in labelled tubes with a preservative solution, and sent to Eurofins

Laboratories (Vejen, Denmark) for analysis. The samples were measured by flow cytometry on a

Fossomatic (Foss, Hillerød, Denmark).

Results: The raw milk had an average of 148.000 cells /ml, and the RO retentate of 1.5 VCF and 2 VCF

had an average of 125.000 cells/ml and 47.000 cells/ml respectively. The results had a very consistent

tendency of decreasing SSC during the filtration process.

Table 4: Average somatic cell count as cells pr ml sample of raw milk, reverse osmosis (RO) retentate of 1.5 voulme concentration factor (VCF) and RO retentate of 2 VCF.

Raw 1.5 VCF 2.0 VCF

148.000 125.000 47.000

Appendix 2

Colony forming units found in reverse osmosis retentate – unpublished data

Introduction: As microbial growth was a major concern during the initiation of the project, several

bacteological samples were collected in the entire duration of the studies. It was concluded that the

filtration process did not cause severe microbial problems, and thus the subject was not further studied.

Materials and methods: Samples of raw bulk milk and RO retentate were collected as duplicates by

using sterile syringes. The samples were put on ice in a thermo box and transported to Eurofins

Laboratories (Vejen, Denmark) for analysis. The analysis were conducted by flow cytometry on a

BactoScan (Foss, Hillerød, Denmark).

Results: the results showed that the filtration equipment had contamination issues from time to time,

and thus the CFU results were very inconsistent. After becoming aware of the issue, and making an

effort to properly avoid contamination, the CFU count was significantly reduced. The results in table 5

shows the average CFU/ml found in raw milk and RO retentate after implementation of a better

hygienic routine.

Table 5: Average CFU/ml in raw milk samples and RO retentate of 2 VCF.

Raw 2.0 VCF

3000 4500

Appendix 3

Viscosity of reverse osmosis retentate during storage – unpublished data

Introduction

After evaporation of the raw milk and RO retentate prior to drying, it was noticed that the RO retentate

was far more viscous than the raw milk at same dry matter content. It was unfortunately not possible to

conduct rheological measurements at that given time. So in order to assess these possible changes in

viscosity of RO retentate during storage, experiments were set up, so the viscosity vas measured during

storage at 4°C for both raw milk and RO retentate of various dry matter content. The gelation of heated

evaporated milk is mostly linked to the denaturation of whey proteins and rearrangement of three-

dimensional protein network (Bienvenue et al. 2003).

Materials and methods

Samples of raw milk and retentate of two different dry matter content were stored at 4°C, and viscosity

measured every day up to 3 days of storage. This was done on both samples with the full fat content

and samples skimmed directly after production. The viscosity was measured on a rheometer through a

bob cup system, with a cooling unit of 4°C.

Results and discussion

Figure 11 shows the collected results of viscosity. No significant changes have been observed during

storage of either raw milk or RO retentate, even though a tendency of viscosity increase is seen in the

full fat 2 VCF sample. The viscosity is significantly (p<0.001) dependent on dry matter content, and

thus also concentration factor. This is a known relation, so the results are as expected (Carr 1999).

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

Day 0 Day 1 Day 2 Day 3

Viscosity Pa*

s

Storage time

Whole milk

Raw

1.5 VCF

2 VCF

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0 10 20 30 40Viscosity Pa*

sDry matter %

Whole milk

0

0.005

0.01

0.015

0.02

0.025

0.03

Day 0 Day 1 Day 2 Day 3

Viscosity Pa*

s

Storage time

Skim

Raw

1.5 VCF

2 VCF

0

0.005

0.01

0.015

0.02

0.025

0 5 10 15 20 25

Viscosity Pa*

s

Dry matter %

Skim

Figure 11: Viscosity of retentate and raw milk samples during storage

Appendix 4

Cream separation from reverse osmosis retentate – unpublished data

Introduction

It is essential to ensure that the retentate has the proper ability to be separated into cream and skim

milk, without too high loss of fat into the skim milk, in order for the milk to be further processed into

high quality products. According to Fjaervoll (1968), the ability to clean skim the skim milk fraction is

highly dependent on temperature and viscosity. Also factors such as fat content and fat globules size

will affect the process. Due to the higher viscosity found in the retentate, it could potentially be harder

to achieve a pure separation between skim milk and cream. Therefore, it was decided to include

separation experiments in the project.

As a result of equipment incapability, it was very difficult to control temperature during separation at

the pilot scale centrifuge. So the results have been sparse, but enough to give an indication of the

properties.

Materials and methods

The separation experiments were conducted preliminary on a lab centrifuge, prior to the pilot plant

centrifuge experiments. The lab tests were done on both raw milk and 2 VCF RO retentate at 4000 rpm

for 5 min at 4°C, 10°C, 20°C, 30°C and 40°C.

Assistance from GEA Westfalia, Germany, came to perform the pilot plant centrifugation trials, at the

AU Foulum facilities. First, the centrifuge capacity was adjusted on water, then a pre-experiment on

raw milk, before the experiment on 1.5 VCF and 2 VCF RO retentate. It was intended to perform the

separation on a range of temperatures in order to fully evaluate the separation properties of the

retentate. However, the temperature was not possible to fully control, so only a very limited range of

samples were collected.

The milk composition was measured on a Foss MilkoScan through FTIR, but the skim milk from

retentate has values exceeding the limits of calibration. Thus, samples were sent to Eurofins

Laboratories for chemical analysis of fat and protein content.

Results and discussion

The results from the lab scale experiments showed that separation of retentate is possible, and

especially at 40°C the separation was very clear (Figure 12).

Figure 12: Pictures of samples of 2 VCF retentate and raw milk, centrifuged in the lab at 40°C. A

separation between cream and skim milk is seen.

The results from the pilot plant centrifuge (table 6) show that it is possible for both 1.5 VCF and 2 VCF

retentate to be skimmed to a fat level below 0.1 %, and still have a rather high fat content in the cream.

A higher loss of protein into the cream was observed, compared to what would normally be the case for

skimming of raw milk.

Table 6: Fat and protein content of 1.5 VCF and 2 VCF retentate before centrifugation, and the fat and

protein content in the skim milk and cream at different processing temperatures.

Processing

temperature

Fat content

(g/100g)

Protein content

(g/100g)

2 VCF feed

7.1 6.73

2 VCF skim 48 0.05 5.66

2 VCF skim 44 0.1 7.04

2 VCF skim 43 0.09 7.13

2 VCF cream 40 32.43 4.64

1.5 VCF feed

6.57 5.34

1.5 VCF skim 39 0.07 5.81

1.5 VCF

cream 43 22.05 4.23

quality and utilization of raw milk retentate produced at the dairy...

Documents