Bioresource Technology 101 (2010) 2287–2293

Contents lists available at ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Double CO2 fixation in photosynthesis–fermentation model enhances algallipid synthesis for biodiesel production

Wei Xiong, Chunfang Gao, Dong Yan, Chao Wu, Qingyu Wu *

Department of Biological Sciences and Biotechnology, Tsinghua University, Beijing 100084, PR China

a r t i c l e i n f o a b s t r a c t

Article history:Received 25 July 2009Received in revised form 7 November 2009Accepted 9 November 2009Available online 5 December 2009

Keywords:Carbon fixationElementary flux modeMicroalgal biodieselPhotosynthesis–fermentation modelRubisco

0960-8524/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.biortech.2009.11.041

* Corresponding author. Tel./fax: +86 10 62781825E-mail address: qingyu@tsinghua.edu.cn (Q. Wu).

In this study, a photosynthesis–fermentation model was proposed to merge the positive aspects of auto-trophs and heterotrophs. Microalga Chlorella protothecoides was grown autotrophically for CO2 fixationand then metabolized heterotrophically for oil accumulation. Compared to typical heterotrophic metab-olism, 69% higher lipid yield on glucose was achieved at the fermentation stage in the photosynthesis–fermentation model. An elementary flux mode study suggested that the enzyme Rubisco-catalyzedCO2 re-fixation, enhancing carbon efficiency from sugar to oil. This result may explain the higher lipidyield. In this new model, 61.5% less CO2 was released compared with typical heterotrophic metabolism.Immunoblotting and activity assay further showed that Rubisco functioned in sugar-bleaching cells at thefermentation stage. Overall, the photosynthesis–fermentation model with double CO2 fixation in bothphotosynthesis and fermentation stages, enhances carbon conversion ratio of sugar to oil and thus pro-vides an efficient approach for the production of algal lipid.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Lipids including animal fats and plant oils are the main feedstockfor biofuel (biodiesel) production. Animals and most microorgan-isms are heterotrophs. They are able to efficiently synthesize a com-pact storage of energy-fat whilst releasing a certain amount of CO2.Plants, including algae, are autotrophs and they function with bulkystorage of energy-starch. Oils (lipids) are generally formed in plantseeds and constitute a very small amount of the whole plant. How-ever, absorbance of CO2 is one of the main advantages of autotro-phy, benefiting both the environment and the economy viabiomass production.

The study of algae-for-fuel has become a hot topic in recentyears with energy prices fluctuating widely and green house gasemissions increasingly becoming a cause for concern (Gouveiaand Oliveira, 2009; Jorquera et al., 2009; Pruvost et al., 2009; Yooet al., 2009). Microalgae are regarded as a good source of biofuel(especially biodiesel) that has the potential to completely displacefossil fuels because of its rapid biomass production, high photosyn-thetic efficiency and, in some species such as Botryococcus, high li-pid content (Haag, 2007). However, cultivation of autotrophicmicroalgae for biodiesel production still faces some technical chal-lenges. For example, in a photosynthesis growth model (PM), rap-idly growing cells contain lower amounts of lipids (<20% of dryweight), whereas algal cells accumulating high lipid contents

ll rights reserved.

.

(40–50% of dry weight) exhibit little growth. Heterotrophic fer-mentation of Chlorella protothecoides provides an alternative wayto solve these problems (Miao and Wu, 2004, 2006; Xu et al.,2006). The cell density and lipid content achieved in a 5-L bioreac-tor was up to 51.2 g/L and 50.3% of dry cell weight (DCW) (Xionget al., 2008). However, the fermentation growth model (FM) con-sumes organic carbon (sugar or starch) and is associated with moreCO2 release than the PM.

To develop an integrated strategy for cost-effective and environ-mentally-friendly production of microalgal biofuels, we adopted aphotosynthesis–fermentation model (PFM) for algal cultivation.This model involves the photosynthetic growth of C. protothecoidesto increase biomass and subsequent heterotrophic fermentation tomaximize cell density and lipid accumulation. In the PFM, not onlywas CO2 used for biomass production in the photosynthesis stage,but lipid biosynthesis was also enhanced in the fermentation stagecompared with the FM. A theoretical analysis suggested that theCO2 re-fixation in fermentation stage resulted in enhancing lipidsynthesis. This conclusion has been supported by further experi-mental data, confirming that the PFM is a novel approach for moreefficient biodiesel production from microalgae.

2. Methods

2.1. Cell strains and culture medium

Microalga C. protothecoides strain 0710 originally obtained fromthe Culture Collection of Alga at the University of Texas (Austin,

2288 W. Xiong et al. / Bioresource Technology 101 (2010) 2287–2293

Texas, USA) was screened in the Microalgal Fermentation and Bio-energy Laboratory at Tsinghua University, Beijing, China. Basicmedia composition was: KH2PO4 0.7 g/L, K2HPO4 0.3 g/L,MgSO4�7H2O 0.3 g/L, FeSO4�7H2O 3 mg/L, glycine 0.1 g/L, vitaminB1 0.01 mg/L, A5 trace mineral solution 1 mL/L. Glycine (5 g/L)was added to phototrophic culture broth as a nitrogen source;45 g/L glucose was used as an organic carbon source for heterotro-phic growth and glycine, limited up to 2 g/L, was used as a nitrogensource for lipid biosynthesis.

2.2. Culture conditions for the three models

2.2.1. PMPlate grown green C. protothecoides was used to inoculate

500 mL media and exposed to light at 100 lmol/(m2 s). The incu-bation temperature was 28 �C. Aseptic air was supplied through apipeline located centrally at the bottom of the column. Log-phasegrowth was then transferred to a new tube or photo-bioreactor(Infors, Switzerland) at a 1:100 (V/V) dilution. The parameters ofthe photo-bioreactor were set as: pH 6.5, gas mixture comprisingair and CO2 at a flow rate of 250–300 L/h with a CO2 concentrationof 2%.

2.2.2. FMThe heterotrophic culture was incubated at 28 �C in either shak-

ing flasks at a rate of 200 rpm or a 5-L stirring tank (Infors, Switzer-land). The flow rate of aseptic air and stirring speed in thefermenter were set as 240 L/h and 250 rpm, respectively. This al-lowed the dissolved oxygen (DO) value to be kept over 20% air sat-uration. pH of the medium was automatically controlled at6.3 ± 0.1.

2.2.3. PFMThe green cells of C. protothecoides were obtained by growing

the alga autotrophically as described above. At the end of log-phase growth, green algal cells were left to sediment overnight.The supernatant was then discarded and cell pellet was re-sus-pended in the defined heterotrophic medium. Subsequent proce-dures for heterotrophic incubation were as detailed above.

2.3. Analytical techniques

Cell growth was monitored by optical density measurements at540 nm using a UV/Visible spectrophotometer (Pharmacia BiotechUltrospec 2000). Samples were diluted to an appropriate concen-tration to keep the OD540 value between 0.2 and 0.8. At the endof every batch, wet algal cells were freeze-dried and weighed.The DCW corresponded to OD540 value by a regression equation:y = 0.4155x (R2 = 0.9933, P < 0.05), where y (g/L) is the DCW, x isthe absorbance of the suspension at 540 nm. Glucose consumptionwas analyzed using an enzymatic bio-analyzer (SBA-40C, Shan-dong Academy of Sciences). The concentration of CO2 in the in-let-gas/off-gas was determined with an online tandem gasanalyzer (Milligan Instrument, UK). Chlorophyll content was deter-mined from the absorbance of the methanol extracts at 666 nm(MacKinney, 1941). The composition of the lipid was analyzed byGC–MS according to our previous protocols (Xiong et al., 2008).The cellular lipid content was measured by gas chromatography:cells were harvested by centrifugation, washed twice with distilledwater, then lyophilized. The resulting dry algal powder was sus-pended in a mixture of 2 mL methanol acidified with 3% sulfuricacid and 2 mL chloroform containing 2.5 g/L capric acid (the inter-nal standard to correct transesterification and injection volume er-rors). The mixture was then heated in a sealed tube at 100 �C for4 h. After cooling, 1 mL of distilled water was added and thesample was vortexed for 20 s. After separation of phases, 1 lL

was injected into the gas chromatograph (HP 6890, USA), using a0.32-mm diameter column, 30 m in length. Nitrogen was used asthe carrier gas at a flow rate of 1 mL/min. Measurements startedat 80 �C for 1.5 min, then the temperature was increased to140 �C at a rate of 30 �C/min, it then reached 300 �C at a rate of20 �C/min, and was kept stable for 1 min before the analysis wasterminated. Retention times were 4.4 min for capric acid methylester, 7.4 min for hexadecanoic acid and 8.5 min for octadecanoicacid methyl ester.

2.4. Electron microscopy

Sample pretreatment for autotrophic and heterotrophic cellswas performed by standard protocols (Glauert and Lewis, 1998).In short, the procedures included fixation, dehydration, embed-ding, sectioning and staining. Cells were observed and photo-graphed using a JEM-123O transmission electron microscope(Hitachi, Japan).

2.5. Quantitation and activity assay of Rubisco

Immunoblotting was carried out to measure Rubisco in C. prot-othecoides during the phototrophic/heterotrophic process. Cellswere disrupted using an ultrasonic cell pulverizer (JY92-2D, Xinzhi,Ningbo, China) for 40 min. After centrifugation at 15,000g for10 min, proteins were extracted from the precipitates using thephenol/SDS method (Nagai et al., 2008). Protein powder was dis-solved in the rehydration solution (7 M urea, 2 M thiourea, 4%CHAPS, 65 mM DTT) and protein concentration was determinedby the method of Bradford (Bradford, 1976) with BSA as the stan-dard. Protein (20 lg) was separated by 9% SDS–PAGE and electro-phoretically transferred to polyvinylidene difluoride (PVDF)membranes. The protein blots were blocked with 5% skimmed milkin TBST buffer (20 mM Tris–HCl, pH 7.6, 0.8% NaCl, 0.05% Tween20) and incubated for 1 h at room temperature with a rabbit anti-body (provided by G. Chen, dilution 1:1000) against the tobaccoholoenzyme of Rubisco. The hydroxyperoxidase-labelled goatanti-rabbit IgG (Bio-Rad, dilution 1:5000) was used as a secondaryantibody. After autoradiographic detection, the quantitation ofband intensities was analyzed using Image software, Gelpro4.

For activity assay, cells from 250 mL cultures were suspended in10 mL extraction buffer (pH 7.8) that contained 100 mM Tris–HCl,20 mM KCl, 1 mM EDTA and disrupted by ultrasonic cell pulverizer(JY92-2D, Xinzhi, Ningbo, China) for 40 min. The crude preparationwas clarified by centrifugation at 15,000g for 10 min at 4 �C. Car-boxylase activity of Rubisco was assayed by measuring [14C]-3-phosphoglyceric acid. Briefly, 20 lL of crude preparation that con-tained 2 mg/mL protein were pre-incubated for 15 min in 460 mLassay buffer (100 mM Tris–HCl [pH 8.2], 20 mM MgCl2, 1 mMDTT, 10 mM NaH14CO3 [0.2 lCi/lmol]) to fully activate the en-zyme. Reactions were initiated by addition of 20 lL RuBP[10 mM, pH 6.5] and were terminated by the addition of 200 lLHCl (2 M). Radiolabel was determined by liquid scintillation count-ing to calculate activity of Rubisco.

2.6. Elementary flux mode analysis

The network of biochemical reactions in C. protothecoides cellswas assembled based on specific genomic knowledge of Chlorella(http://genome.jgi-psf.org/ChlNC64A_1/ChlNC64A_1.info.html andhttp://genome.jgi-psf.org/Chlvu1/Chlvu1.home.html) and otherphotosynthetic microorganisms (Shastri and Morgan, 2005; Yanget al., 2000, 2002). For network construction and elementary modescomputation, Cellnetanalyzer 9.1, a MATLAB package for structuraland functional analysis of biochemical networks was utilized(Klamt et al., 2007).

W. Xiong et al. / Bioresource Technology 101 (2010) 2287–2293 2289

3. Results and discussion

3.1. Growth of C. protothecoides in the PFM

Three different modes (PM, FM and PFM) are designed for culti-vating C. protothecoides. The PM was performed by growing algaunder illumination in the absence of organic carbon. Cells alwayskept green in the PM and the average cell growth rate was about0.286 g/L/d. The FM using ‘‘glucose-adapted” algal cells was alsocarried out as control. A ‘‘glucose-adapted” cell line was obtainedby the sequential transfer of C. protothecoides to glucose-enrichedmedium until chlorophyll was no longer detectable. In this case,broths were yellow as cells were continually cultivated under het-erotrophic conditions. In the PFM, green C. protothecoides cellswere grown phototrophically for 120 h (photosynthesis stage)and then sedimented, washed and re-suspended in a glucose-con-taining medium for another 120 h incubation (fermentation stage).The cell growth and oil yield in the PM and the PFM were com-pared in Section 3.2.

During cultivation, the green color of the broth in the fermenta-tion stage of the PFM was observed to gradually fade. Changes inpigment were quantitatively compared (Fig. 1) and chlorophyllcontents decreased from 0.45 to 0.029 mg/g DCW during a 120-hincubation, while the chlorophyll content of cells in the PM re-mained unchanged at about 0.45 mg/g DCW. The de-greening phe-nomenon is due to the biodegradation of chlorophyll (Engel et al.,1991; Hortensteiner et al., 2000). Electron microscopy (EM)showed that chloroplasts, the high electron density photosyntheticorganelles, were clearly visible in photosynthetic cells. Membraneswere abundantly accumulated in these chloroplasts and a numberof starch granules could also be seen. In contrast, thylakoid mem-branes rapidly disappeared within 48 h after cells had undergoneheterotrophic metabolism, suggesting degeneration of chloro-plasts. Instead, the cytoplasm was almost totally filled with largelipid droplets, over 1 lm in diameter (images not shown). Bio-chemical and ultrastructural experiments together suggested thatthe occurrence of chlorophyll breakdown and chloroplast degener-ation was associated with lipogenesis during the fermentationstage of the PFM.

3.2. Biomass and lipid yield of heterotrophic cells in the PFM

After being grown in medium with 45 g/L glucose, the biomassand lipid producing performance of C. protothecoides cells were

Fig. 1. Chlorophyll content in three cell growth models. Relative chlorophyllcontent of C. protothecoides cells (mg/g DWC) in the PM (closed circle), the PFM(open circle) and the FM (open square). Data represent the mean ± standarddeviation (SD) of three independent measurements.

evaluated. Incubation of ‘‘glucose-adapted” cells in the FM wasperformed as a control experiment. The batch time profiles(Fig. 2) surprisingly showed that C. protothecoides cells synthesizedmore biomass and lipids in the PFM than in the FM. Correspond-ingly, sugars were completely consumed at 84 h in the PFM,whereas ‘‘glucose-adapted” yellow algal cells in the FM left about2 g/L of glucose unused. An important indicator for biofuel feed-stock production is the conversion ratio of glucose to biomassand glucose to oil (Ybiomass/glc, Yoil/glc). A biomass yield of 0.33 g/gof glucose and a lipid yield of 0.176 g/g of glucose were finally ob-tained at 156 h in the FM, the carbon ratio of glucose to oil (Coil/Cglc) was 27.8%. Meanwhile, bleaching Chlorella cells in the fermen-tation stage of the PFM exhibited improved phenotypes. A higherbiomass yield of 0.617 g/g of glucose and a lipid yield of 0.298 g/g of glucose, were respectively obtained at the end of incubationresulting in a carbon ratio of glucose to oil (Coil/Cglc) as 47.0%.The PFM batches achieved 69.32% higher lipid yield than thatachieved in the equivalent FM. It meant that consumption of glu-cose was largely reduced for the synthesis of same amount of lipid.These results are of potentially great consequence for efficient bio-fuel production.

3.3. Metabolic pathway analysis

To address the question why the conversion ratio of glucose tooil (Yoil/glc) in the PFM fermentation stage was so much higher thanthat in the FM, analysis of elementary flux modes (Schuster et al.,1999; Schwender et al., 2004) was carried out within the Chlorellametabolic network. Through computation with the Matlab packageCellnetanalyzer (Klamt et al., 2007) (detailed in Supplementarymaterial), 18 elementary modes (EMs) were obtained. Amongthem, 12 EMs generate stearic acid (C18:0) as an end-product. TheseEMs can be further divided into phototrophic, heterotrophic andin-between modes.

In the photosynthetic routes (modes 3, 7 and 9 in Fig. E1), fattyacids are formed from the Calvin cycle product glyceraldehyde-3-phosphate (GAP) by the following reaction sequence: transforma-tion of GAP to pyruvate through glycolysis, then conversion to fattyacids via acetyl-CoA (3CO2 + 6NADPH + 9ATP ? glyceraldehyde-3-phosphate; Glyceraldehyde-3-phosphate ? acetyl-CoA + CO2 + 2-NADH + 2ATP; 9Acetyl-CoA + 8ATP + 16 NADPH ? C18:0). The netcarbon stoichiometry shows that generation of one molecule ofstearic acid (C18:0) requires 71 molecules of ATP and 52 moleculesof reductant, all of which originate from absorbed solar energy viaphotoreaction (The stoichiometric equation was shown in TableE1). This result suggested CO2-initiated oil synthesis consumedgreat amounts of cofactors.

In comparison, typical carbon metabolism in the heterotrophicmode requires far fewer cofactors (ATP and reducing equivalents)for oil synthesis than that in the autotrophic mode (mode 16, de-tailed in Supplementary material). In this case, ATP and reductantare almost balanced between oil biosynthesis and glucose oxida-tion (The stoichiometric equation was shown in Table E1). Hexoseis first degraded to pyruvate via glycolysis and then transformed toacetyl-CoA, the precursor for fatty acid synthesis. The net stoichi-ometry of this process is:

9Glucose! 18CO2 þ 2C18:0

In this metabolism, the decarboxylation step is catalyzed bypyruvate dehydrogenase (PDH) and results in a total loss of onethird of the organic carbon, i.e. generation of one molecule oftwo-carbon acetyl-CoA from three-carbon pyruvate, releases onemolecule of CO2. Therefore, the maximum carbon conversion ratioof glucose to oil (Coil/Cglc) is limited to 66.67%. In practice, extra CO2

could be released in the oxidative pentose phosphate pathway and

Fig. 2. Comparison cell growth and oil synthesis between the FM and the fermentation stage of the PFM. Concentration of glucose (circle), DCW (diamond) and oil (triangle),in bleaching cells during incubation in the PFM (left panel) and in chlorophyll-less cells in the FM (right panel). Data represent the mean (SD < 0.05) of three independentmeasurements in a single-batch run of duplicate experiments.

2290 W. Xiong et al. / Bioresource Technology 101 (2010) 2287–2293

citrate cycle for providing essential reductant or energy. Thus, theCoil/Cglc in real heterotrophic cells might be lower than 66.67%.

Besides conventional heterotrophic and autotrophic carbonmetabolism, it was found that 5 EMs among 12 oil-generatingEMs, allowed higher efficiency of carbon conversion (80–100%, de-tailed in Supplementary material). The simplest example of theseEMs was shown in Fig. E1 (mode 2 and mode 4), revealing a max-imum carbon ratio of glucose to oil of 80%. The metabolic routecould be described as follows: 1 mol glucose is first metabolizedinto 1.2 mol ribulose-1,5-bisphosphate through the non-oxidativepentose phosphate pathway without carbon loss. Then conversionof ribulose-1,5-bisphosphate and CO2 to 3-phosphoglyceric acid(PGA) occurs, allowing fixation of 1.2 mol CO2 and production of2.4 mol PGA. These reactions yield 20% more PGA than the typicalheterotrophic mode. In following reactions, PGA is converted topyruvate and a debt of 2.4 mol CO2 was incurred by PDH frompyruvate, resulting in a total loss of 20% carbon as CO2(CCO2 /Cglucose

= (2.4 mol C–1.2 mol C)/1 mol � 6 C), such that remaining 80% ofcarbon in glucose was converted into oil. Since 33.33% of carbonin glucose is lost in the heterotrophic mode and 20% is lost in thisexample EM, this mode theoretically releases 66.67% less CO2 thanthe typical heterotrophic mode (mode 16). Consequently, morefatty acids could be synthesized from these saved carbons. In addi-tion, it should be noted that among 5 EMs, there are also modesthat have even higher carbon conversion ratio (83.33–100%)(modes 6, 15 and 17, shown in Table E1 and Fig. E1); these modesallow for more fatty acid production as well as less CO2 release.

Carefully checking of EMs that have higher carbon efficiency re-vealed that CO2 re-fixation by Rubisco provided more precursorsfor extra lipid synthesis. This might explain why the conversion ra-tio of glucose to oil (Yoil/glc) in the fermentation stage of the PFMwas so much higher than that of the FM. In the PFM, it is reason-able that the apparatus for CO2 assimilation remains available inChlorella cells after 120 h of photosynthetic growth. Thus, theopportunity is created to incorporate the autotrophic and hetero-trophic metabolic pathways in the fermentation stage of the PFMfor efficient lipid production as well as low CO2 emission.

3.4. CO2 release in the FM and the fermentation stage of the PFM

According to above theoretical analysis, higher lipid yield wouldallow lower CO2 loss in the fermentation stage of the PFM. Wetherefore monitored CO2 release levels during oil accumulation inboth the FM and the fermentation stage of the PFM by infraredgas analysis. CO2 concentrations in the inlet-gas and off-gas ofthe 5-L-bioreactor under specific air pressure were measured. Fac-tors including temperature, pH, stirring speed and air flow were

maintained constant and CO2 evolution rate (CER) could thus becalculated via integration of the net volume percentage of CO2. Un-der exponential growth, the CERs in the FM and the fermentationstage of PFM were measured as 16.07 mmol/L/h and 10.71 mmol/L/h. Corresponding oil production rates were 0.68 mmol/L/h and1.23 mmol/L/h, respectively. It meant 61.5% less CO2 was liberatedin the fermentation stage of the PFM than in typical heterotrophicmetabolism (control group) upon the same yield of oil. Theseexperiments suggested that carbon loss indeed reduced in thePFM. According to the glucose consumption rates of 5.95 mmol/L/h in the FM and 6.94 mmol/L/h in the fermentation stage of thePFM, the distribution of carbon from glucose were also obtained.It further showed that in the PFM, 25.8% of carbon in glucosewas released as CO2 whereas 53.2% was observed to be stored inthe form of oil. In contrast, 34.3% of glucose was converted intooil in the FM with a 45% carbon loss as CO2. This experimental dataindicated that much higher carbon efficiency as well as lower CO2

emission was achieved in the fermentation stage of the PFM thanin the FM. This result provided direct evidence to support theoret-ical pathway analysis in Section 3.3.

3.5. Rubisco in the fermentation stage of the PFM

The existence of active Rubisco, the most important enzymecatalyzing the addition of CO2 to ribulose bisphosphate (RuBP),helps to support our findings. We first investigated the fate ofRubisco during the process of ‘‘glucose-bleaching” in the PFM.Immunoblot analysis was performed to measure Rubisco expres-sion levels in the three culture models. After separation of the sol-uble proteins in the cell extracts by SDS–PAGE, immunoblottingusing an anti-Rubisco antibody showed that this enzyme was dom-inant in the phototrophic green cells of the PM. Its large subunit(RbsL) accounted for about 5% of the total soluble protein. How-ever, it was undetectable in ‘‘glucose-adapted” yellow cells in theFM, implying that algal cells exposed to long-term heterotrophicgrowth did not express Rubisco. In comparison, trophic conversionin the PFM allowed a decline in the relative content of Rubisco intotal proteins. Nevertheless, both large and small subunits of Rubi-sco were clearly detectable at the fourth day of heterotrophicgrowth in the fermentation stage (Fig. 3). After the photosyntheticalgal cells were transferred to heterotrophic conditions, Rubiscoactivity was also measured. Cultures grown in the presence of glu-cose showed a gradual increase in carboxylase activity from5.2 nmol C/mg protein/min to 8.5 nmol C/mg protein/min at theearly stage. It was probably due to the rising concentration ofCO2 upon glucose input, which saturated Rubisco carboxylaseactivity and prevented its oxygenase activity. The increase of Rubi-

Fig. 3. Rubisco in the fermentation stage of the PFM. Algal proteins isolated fromthe PM, FM and PFM at the indicated times were loaded onto 9% SDS–PAGE andanalyzed by immunoblotting with an anti-Rubisco antibody. RbcL, large subunit;RbcS, small subunit of Rubisco.

W. Xiong et al. / Bioresource Technology 101 (2010) 2287–2293 2291

sco activity lasted about 18 h after transfer to the new media. Sub-sequently, the loss of carbon-fixing capacity commenced, indicat-ing that the synthesis of Rubisco ceased in this period. Based onthese results, we concluded that although the expression levels de-creased in the PFM, the remaining Rubisco probably proceeded CO2

fixing that might contribute to the enhanced efficiency of carbonuse. Does the fixed CO2 by Rubisco target into the metabolic path-way for lipid synthesis? The answer was provided by Mistuo andShigetoh who used NaH14CO3 as a tracer to investigate carbonmetabolism in the process of ‘‘glucose-bleaching” in C. prototheco-ides (Matsuka and Miyachi, 1974). They observed that radioactivityin the lipid fraction increased remarkably after suspending thegreen cells in glucose-enriched media in dark, suggesting that car-boxylase was indeed active and responsible for the yield of fattyacids from CO2. Overall, above experimental results indicated thatthere were twice CO2 fixations in the PFM, one occurred in photo-synthesis stage the other in fermentation stage.

3.6. Optimization of the PFM in bioreactor

After providing a reasonable explanation to the experimentalresults, another aim of the study was to realize the full potentialof the PFM. An improved scheme for process control and sugarfeeding for cell growth and oil production was therefore devel-oped in bioreactors. In the first stage (photosynthesis stage), cellswere cultivated under illumination in 5-L photo-bioreactors for

Fig. 4. Profiles for the improved PFM of C. protothecoides cultured in a 5-L bioreactor. (aDCW (diamond) and oil (triangle) during incubation. Data represent the mean of three

120 h without a sugar supply, and autotrophic cells growing inthe ‘‘photosynthesis” stage were concentrated to a higher celldensity at the beginning of heterotrophic incubation. The aim ofthis operation was to increase the contribution ratio of CO2 tothe final biomass (oil) and increase the number of producing cellsfor oil synthesis in the next fermentation stage. We thus inocu-lated 20 g/L phototrophic cells in sugar-enriched and nitrogen-deficient medium for lipid production. In the second fermentationstage, culture conditions (Fig. 4a) were finely controlled, enablinga good micro-climate for cell growth and lipid accumulation. Theamount of dissolved oxygen (DO), which reflected the real-timestate of cell growth and glucose uptake in a given agitation andair flow, was recorded. The DO value increased sharply at about23, 43.5, 58.5, 64.5, 86 and 94 h when the carbon source was ex-hausted. In response, glucose solution (100 g/L) was fed into theculture medium at the indicated times to satisfy the demand fora carbon source. The stirring rate was increased gradually tomaintain a DO value of over 20%. Under such conditions, exponen-tial growth of Chlorella in the culture media emerged at the ninthhour and lasted approximately 35 h, after which cells entered intostationary phase reaching a maximum cell density of 123 g/L(Fig. 4b). The average rate of cell growth was shown as 23.9 g/L/d during the whole ‘‘fermentation” stage. After optimization, thehighest oil content of 58.4% DCW was achieved. In comparison,the highest cell density and oil content reported previously were51.2 g/L and 50.3%, respectively in glucose-fed batch experimentsin a 5-L bioreactor (Xiong et al., 2008). Our novel approach istherefore promising for industrial application. According to GC–MS analysis, the composition of fatty acids obtained in the PFMwas identical with those in conventional heterotrophic growth(data not shown). Corresponding properties of biodiesel werecomparable to conventional diesel fuel and comply with US stan-dard for biodiesel (Xu et al., 2006). This is advantageous since noadditional refinery procedure would be needed for the generationof fuel grade lipids.

3.7. Comparison of the PFM with the PM, FM and mixotrophic model

Atmospheric CO2, the chief greenhouse gas implicated in cli-mate change, is rising towards 390 ppm at an accelerating rate.Therefore, achieving energy production with carbon neutrality isurgent. In this context, most studies of alga-based biofuel produc-tion focus on a photosynthetic cultivation model. However, boththe cell growth rate and oil content under phototrophic conditionsare extremely low. The analysis of elementary flux modes suggeststhat production of fatty acids in the PM requires substantial

) Fermentation parameters during incubation. (b) Concentration of glucose (circle),independent measurements (SD < 0.05).

2292 W. Xiong et al. / Bioresource Technology 101 (2010) 2287–2293

amounts of ATP and reductant. Moreover, there are also problems,such as light deficiency and night biomass loss (Ogbonna and Tana-ka, 1996) that negatively affect oil productivity. According to a re-cently published report, the optimal rate of oil synthesis forphotosynthetic microalgae in outdoor photo-bioreactors was nohigher than 0.2 g/L/d (Rodolfi et al., 2009). To challenge the prob-lem, the PFM merging the beneficial features of the PM and FMwas designed. It was thought that cell growth rate and oil contentin the fermentation stage in the PFM should be higher than that inthe PM and lower than that in the FM. However, by using the PFM,both cell growth and lipid production were substantially enhanced(Fig. 2). The average rates of cell growth and lipid productionreached as high as 23.9 g/L/d and 11.8 g/L/d, respectively. Themuch higher conversion ratio of glucose to oil in the fermentationstage of the PFM than in the FM was beyond our expectations. Toexplain those findings, we provide a theoretical insight that Rubi-sco-catalyzed CO2 fixation coupled with the sugar metabolic path-way enhances net carbon efficiency (Supplementary material). Ingeneral, production of acetyl-CoA, the precursor for fatty acids inconventional sugar metabolism, was accompanied with equivalentcarbon loss as CO2 through the decarboxylation of pyruvate. Thesereactions limited the maximum Coil/Cglc to 66.7%. In the PFM, CO2

fixation by Rubisco was continuing after feeding glucose. Thus,the theoretical limitation of maximum Coil/Cglc could be higherthan 66.7% because released CO2 might be re-fixed and then targetinto lipid biosynthetic pathway (Supplementary material). Theexistence of this process was confirmed by lowered CO2 releaseand remained active Rubisco after sugars are provided (Fig. 3). Itsuggests that the enhancement of carbon efficiency in the fermen-tation stage of PFM is therefore a result of Rubisco-involved sugarmetabolism.

Besides the PFM, mixotrophic culture is an alternative way tocombine heterotrophic and autotrophic metabolism. There aretwo reasons for us to adopt integrated PFM strategy rather thanuse a simplified mixotrophic approach. First, the effect of ‘‘glu-cose-bleaching” leads to biodegradation of chlorophyll and se-verely inhibits photosynthesis. Second, the assembly andmaintenance of a photosystem requires large quantities of pig-ments and proteins that are produced from nitrogen source,whereas accumulation of neutral lipids tends to occur in nitro-gen-limited environments (Richardson et al., 1969). We thereforeseparated the culture process into two independent but sequentialparts: the ‘‘green” stage (nitrogen-sufficient photosynthetic culti-vation) and the ‘‘yellow” stage (nitrogen-deficient heterotrophicgrowth for lipid accumulation). This design was proved more effi-cient than mixotrophic cultivation.

Double utilization of CO2 in both the photosynthesis and fer-mentation stages not only benefits to reducing the emission ofgreenhouse gas but also lowers the consumption of sugar sub-strates to a great extent, thereby also providing a cost-save ap-proach for alga-based biodiesel production. Based onoptimization of the PFM in bioreactor (Fig. 4), the PFM couldbe scaled up to an industrial level for commercial biodiesel pro-duction in the near future. The integrated system for manufac-turing microalgal biodiesel could be centered around largepower plants, where sufficient flue gas containing a high concen-tration of CO2 is available. Light, water and inorganic nutrientsare provided for microalgal farming during the ‘‘green stage”.By feeding fermentable sugars hydrolyzed from Jerusalem arti-choke, sweet sorghum, cassava or other non-food crops to reducethe cost in sugar substrate, the resultant phototrophic algal cellscontaining active Rubisco can be incubated in stirred tanks for oilproduction and subsequent biodiesel refinery. This integrated ap-proach may enhance the efficiency of carbon utilization and ex-pands the pathways available for industrial-scale microalgalbiofuel production.

4. Conclusions

In summary, the integrated strategy of the PFM merges the ben-eficial features of both autotrophs and heterotrophs. By means ofdouble CO2 fixation in both photosynthesis and fermentationstages, it simultaneously achieved the enhancement of carbon effi-ciency for biofuel synthesis and the reduction of greenhouse gasemission. This strategy lowers the consumption of sugar substrateslargely, thereby opening a door for cost-effective biodiesel produc-tion from microalgae.

Acknowledgements

Antibody for Rubisco immunoblotting provided by professorDabing Zhang at Jiaotong University (Shanghai, China) is gratefullyacknowledged. This study was funded by the NSF Guangdong jointproject U0633009, NSF project 30670476 and 30970224, the Na-tional High Technology Research and Development Program of Chi-na (863 Program) 2007AA05Z400, MOST overseas cooperationproject 20070574.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.biortech.2009.11.041.

References

Bradford, M.M., 1976. Rapid and sensitive method for quantitation of microgramquantities of protein utilizing principle of protein–dye binding. AnalyticalBiochemistry 72, 248–254.

Engel, N., Jenny, T.A., Mooser, V., Gossauer, A., 1991. Chlorophyll catabolism inChlorella protothecoides – isolation and structure elucidation of a red bilinderivative. FEBS Letters 293, 131–133.

Glauert, A.M., Lewis, P.R., 1998. Biological Specimen Preparation for TransmissionElectron Microscopy. Princeton University Press, Princeton, NJ.

Gouveia, L., Oliveira, A.C., 2009. Microalgae as a raw material for biofuelsproduction. Journal of Industrial Microbiology and Biotechnology 36, 269–274.

Haag, A.L., 2007. Algae bloom again. Nature 447, 520–521.Hortensteiner, S., Chinner, J., Matile, P., Thomas, H., Donnison, I.S., 2000. Chlorophyll

breakdown in Chlorella protothecoides: characterization of degreening andcloning of degreening-related genes. Plant Molecular Biology 42, 439–450.

Jorquera, O., Kiperstok, A., Sales, E.A., Embirucu, M., Ghirardi, M.L., 2009.Comparative energy life-cycle analyses of microalgal biomass production inopen ponds and photobioreactors. Bioresource Technology.

Klamt, S., Saez-Rodriguez, J., Gilles, E.D., 2007. Structural and functional analysis ofcellular networks with CellNetAnalyzer. Bmc Systems Biology.

MacKinney, G., 1941. Absorption of light by chlorophyll solutions. Journal ofBiological Chemistry 140, 315–322.

Matsuka, M., Miyachi, S., 1974. Photosynthetic metabolism of C-14O2 in process ofglucose-bleaching of Chlorella protothecoides. Plant and Cell Physiology 15, 919–926.

Miao, X.L., Wu, Q.Y., 2004. High yield bio-oil production from fast pyrolysis bymetabolic controlling of Chlorella protothecoides. Journal of Biotechnology 110,85–93.

Miao, X.L., Wu, Q.Y., 2006. Biodiesel production from heterotrophic microalgal oil.Bioresource Technology 97, 841–846.

Nagai, K., Yotsukura, N., Ikegami, H., Kimura, H., Morimoto, K., 2008. Proteinextraction for 2-DE from the lamina of Ecklonia kurome (laminariales):recalcitrant tissue containing high levels of viscous polysaccharides.Electrophoresis 29, 672–681.

Ogbonna, J.C., Tanaka, H., 1996. Night biomass loss and changes in biochemicalcomposition of cells during light/dark cyclic culture of Chlorella pyrenoidosa.Journal of Fermentation and Bioengineering 82, 558–564.

Pruvost, J., Van Vooren, G., Cogne, G., Legrand, J., 2009. Investigation of biomass andlipids production with Neochloris oleoabundans in photobioreactor. BioresourceTechnology 100, 5988–5995.

Richardson, B., Orcutt, D.M., Schwertner, H.A., Martinez, C.L., Wickline, H.E., 1969.Effects of nitrogen limitation on the growth and composition of unicellularalgae in continuous culture. Applied Microbiology 18, 245–250.

Rodolfi, L., Zittelli, G.C., Bassi, N., Padovani, G., Biondi, N., Bonini, G., Tredici, M.R.,2009. Microalgae for oil: strain selection, induction of lipid synthesis andoutdoor mass cultivation in a low-cost photobioreactor. Biotechnology andBioengineering 102, 100–112.

Schuster, S., Dandekar, T., Fell, D.A., 1999. Detection of elementary flux modes inbiochemical networks: a promising tool for pathway analysis and metabolicengineering. Trends in Biotechnology 17, 53–60.

W. Xiong et al. / Bioresource Technology 101 (2010) 2287–2293 2293

Schwender, J., Goffman, F., Ohlrogge, J.B., Shachar-Hill, Y., 2004. Rubisco without theCalvin cycle improves the carbon efficiency of developing green seeds. Nature432, 779–782.

Shastri, A.A., Morgan, J.A., 2005. Flux balance analysis of photoautotrophicmetabolism. Biotechnology Progress 21, 1617–1626.

Xiong, W., Li, X.F., Xiang, J.Y., Wu, Q.Y., 2008. High-density fermentation ofmicroalga Chlorella protothecoides in bioreactor for microbio-diesel production.Applied Microbiology and Biotechnology 78, 29–36.

Xu, H., Miao, X.L., Wu, Q.Y., 2006. High quality biodiesel production from amicroalga Chlorella protothecoides by heterotrophic growth in fermenters.Journal of Biotechnology 126, 499–507.

Yang, C., Hua, Q., Shimizu, K., 2000. Energetics and carbon metabolism duringgrowth of microalgal cells under photoautotrophic, mixotrophic and cycliclight-autotrophic/dark-heterotrophic conditions. Biochemical EngineeringJournal 6, 87–102.

Yang, C., Hua, Q., Shimizu, K., 2002. Metabolic flux analysis in Synechocystis usingisotope distribution from C-13-labeled glucose. Metabolic Engineering 4, 202–216.

Yoo, C., Jun, S.Y., Lee, J.Y., Ahn, C.Y., Oh, H.M., 2009. Selection of microalgae for lipidproduction under high levels carbon dioxide. Bioresource Technology 101(Suppl. 1), S71–S74.