PROTEORED Multicentric Study

QUANTITATIVE PROTEOMICS METHODSPECTRAL COUNT

2010

UNIVERSITY OF BARCELONA

Salamanca 16th March

Objectives:

•Test each laboratory abilities to perform quantitative proteomic analysis.

•Comparison of methodologies for relative quantitative analysis of proteomes. The study should provide data to assess and compare performance of different methodologies and intra- and inter-lab reproducibility of these.

•Evaluation of data reporting and data sharing tools (MIAPE documents, standard formats, public repositories).

Samples:

•Each participant laboratory will receive two protein mixture samples, labeled A and B, containing each 100 µg of total protein.

•100 micrograms of each protein mixture A and B dissolved in 6M Urea /1% CHAPS, at 6 micrograms/microliter concentration.

Samples contain:

A mixture of around 150 E. Coli proteins (identical in each sample). This mixture has been prepared by fractionation of the cytoplasmatic proteome of E.Coli. It contains soluble proteins, of a wide range of pI and Mw.

Four spiked mammalian proteins:

•CYC_HORSE (Cytochrome C, Mw 12362), added at the ~ 1 pmol/ 1 mg total protein level.

•MYG_HORSE (Apomyoglobin, Mw 16952), at ~ 200 fmol / 1 mg total protein

•ALDOA_RABIT (Aldolase, Mw 39212), at ~ 25 fmol / 1 mg total protein

•ALBU_BOVIN (Serum albumin, Mw 66430), at ~ 1 fmol / 1 mg total protein

These four proteins have been spiked in different amounts in samples A and B, with ratios ranging from 1.5:1 to 5:1 between the two samples.

Purpose of the analysis:

•The intended purpose of the analysis is to measure the ratios between samples A and B for the four spiked proteins. The “matrix” E. Coli proteins, which should be unchanged, will provide a measure of dispersion for the method used.

•The samples can be also used to test methods for absolute quantitation, if desired.

•In order to evaluate reproducibility in an homogenous dataset, we ask to perform a minimum of 4 replicate analysis of the samples. (Depending on the method of choice this would demand a maximum of 4 + 4 LC-MS runs).

Methods:

•Sample complexity has been chosen to allow for the analysis of the mixture on single LC-MS runs. In principle, there is no need for pre-fractionation. A long enough gradient (90-120 min) gradient is suggested, but this of course will strongly depend on the MS instrument available for analysis.

•1-2 micrograms of total protein per run should be enough to cover the range of abundances of the spiked proteins in the samples. Again, this will depend a lot on the instrument used, and should be adjusted by each Lab. according to their expertise.

•The sample is primarily intended to test non-targeted relative quantitation methodologies. Both label-based methods (ICPL, iTRAQ, TMT, O18,...) and label-free methods (based on spectral counts, Hi3, “LCMS Image analysis”...) can be performed and tested to analyze the samples. Some of them will require 4 + 4 LC-MS runs, while others (i.e. 8-plex iTRAQ) could require a single run to provide comparable measurements of reproducibility.Try to choose the number of replicate analysis in a way that 4 independent measurements of each A:B ratio are obtained, so that comparable statistics can be calculated.

•The sample can be also used if desired to test targeted methods, such MRM methods for relative or absolute quantitation. The concentration of the spiked proteins is probably too high to provide a real challenge for those methods, but it can still be useful for test purposes (one can test accuracy, sensitivity on serial dilutions of the sample...)

•The amount of sample provided, as well as the concentration of the spiked proteins, should allow also a 2D-DIGE analysis of the samples, although this is not the main purpose of the experiment.

Quantitative Proteomic Approaches• Label free– Spectral counting– Ion current based (Extracted ion chromatograms)– Other• Stable isotope labeling– Stable isotope label reagent as ICAT and ITRAQ– Metabolic labeling (SILAC, 15N)– Others

Shotgun Proteomics• Digestion of proteins and separation of peptides– Extensive chromatographic separation (one or mutlipledimensional separations, columns,..)• Data acquisition– Data-dependent acquisition (Automated acquisition of MS/MS spectra from as many precursor ions as possible)• Data analysis– Automated interpretation of the MS/MS spectra (DB search)

Spectral Counting Summary• Spectral count correlates well with protein abundance• Fold change can be calculated and statistically evaluated• Simple and straightforward implementation• Sensitive to protein abundance changes – for abundant proteins 2 fold change easily detected with high confidence

Limitations• The response to increasing protein amount is saturable• Noisy data at low spectral counts – large difference in spectral count necessary to determine significant change

Fu et al, 2006

Spectral count reflects relative abundance of a protein (r2 ≥0.99)Issues to address:- Variability of Spectral counts- Sensitivity of Spectral count to protein abundance changes- How to determine relative changes between two samples

Variability of Spectral countingLCMSMS analysis of replicate SCX fractions of K562 cell lysates, G-testOld W. et al, MCP 2005

How to determine relative changes between two samples Fold change determinationOld W. et al, MCP 2005• Practical issue – no peptides found in one of the compared samples• Data discontinuity (spectral count – integers) – not amenable to Student t-test• Differences in sampling depth

Fold change determination.RSC = log2[(n2 + f)/(n1 + f)] + log2[(t1 - n1 + f)/(t2 - n2 + f)]n1, n2 - spectral counts for sample 1 and 2t1, t2 – total spectral count (sampling depth) for samples 1 and 2f – correction factor 1.25 (Beissbarth et al – Bioinformatics 2004)Observed RSC correlates well with expected RSC for standard proteins spiked into complex samples (Old W. et al, MCP 2005)

• 100 micrograms of each protein mixture A and B are dissolved • in 6M Urea /1% CHAPS, at 6 micrograms/microliter concentration.

Samples were kept at -20ºC .

Precipitation with TCA/ACETONE

Re suspended in 100 uL 0.3 % SDS/50 mM Tris HCl pH 8.0/200 mM DTT

5 uL(5ug) Sample digested with trypsin O/N at 1/100 ratio

Separate with nanoHPLC (4 replicas 1uL)

MS/MS LTQ VelosOrbitrap

Analysis Spectra analyzedProteored A1 11.976Proteored A2 12.090Proteored A3 12.567Proteored A4 14.889

Proteored B1 14.444Proteored B2 14.936Proteored B3 15.115Proteored B4 15.852

TOTAL 111.869

SEQUEST PARAMSpeptide_mass_tolerance = 0.07fragment_ion_tolerance = 0.6diff_search_options = 15.9949 M 0.000 C 0.000 X

Item LC-MS run A-1 A-2 A-3 A-4 B-1 B-2 B-3 B-4

Total Sample A-B (Combined AB1-

4 runs)**

                     

1 Number of MS/MS spectra acquired 11976 12090 12567 14889 14444 14936 15115 15852 13984

2 Number of total assigned peptides id. 1724 1702 1705 2660 2136 2388 2440 2576 2166

3 Number of unique peptides id. 1226 1183 1190 1559 1389 1447 1499 1660 1394

4 Number of E Coli proteins id. (total) 209 213 211 266 223 250 244 266 235

5 Number of E Coli single hit- proteins id 19 31 28 28 18 20 18 30 24

6 Number of Spiked proteins id. 4 3 3 3 4 4 4 4 3.6

7 FDR* 0 0 0 0 0 0 0 0.0038 0.0005

                     

8 Total Number of proteins quantitated 5               

9Number of proteins quantitated > 3 peptides 2               

10Number of proteins quantitated > 2 peptides 3               

11 Number of proteins quantitated 1 peptide                  

                    A/B ratio

12 Average of A/B ratios for E Coli proteins                

13 Standard deviation A/B ratios                

14 % CV A/B ratios E Coli proteins                

                     

                     

                     

The Normalized Spectrum Counts bar chart shows a protein's relative abundance across different samples.  The y-axis is the normalized count of the spectra matching any of the peptides in the protein. This count depends upon the protein, peptide, required mods and search filters set on the Samples page. Each bar along the x-axis is for a different biological sample. The bars are color coded. Each sample category is colored a different color. The bar chart can be used as a visual confirmation of a differential expression flagged by the Quantitative Analysis in the Samples view.

FDR= 0 242 Proteins

SAMPLE A1

SAMPLE A2

SAMPLE A3

SAMPLE A4

SAMPLE B1

SAMPLE B2

SAMPLE B3

SAMPLE B4 Av A STD A Av B STD B A/B B/A

ALDOA_RABIT 16 18 21 18 10 12 10 10 18.25 2.06 10.50 1.00 1.74 0.58ALBU_BOVIN 2 1 0 1 5 1 1 1 1.00 0.82 2.00 2.00 0.50 2.00MYG_HORSE 17 21 15 19 9 14 14 12 18.00 2.58 12.25 2.36 1.47 0.68CIC_HORSE 13 14 12 21 22 27 27 19 15.00 4.08 23.75 3.95 0.63 1.58

SAMPLE A1

SAMPLE A2

SAMPLE A3

SAMPLE A4

SAMPLE B1

SAMPLE B2

SAMPLE B3

SAMPLE B4

Av ASTD A

Av BSTD B

0

5

10

15

20

25

30ALDOA_RABIT

MYG_HORSE

CIC_HORSE

ALBU_BOVIN

ALBUMIN_BOVIN SAMPLE A2 Xcorr 0.88 DeltaCn 0.46

SAMPLE A1 Xcorr 3.16 DeltaCn 0.43

SAMPLE A1 Xcorr 3.23 DeltaCn 0.64SAMPLE B2 Xcorr 2.9 DeltaCn 0.57

P-value=0.52

ALDOA_RABIT SAMPLE A2 Xcorr 2.12 DeltaCn 0.5

SAMPLE A3 Xcorr 5.33 DeltaCn 0.78

SAMPLE B3 Xcorr 5.35 DeltaCn 0.77

P-value=0.0.00053

CYC_HORSE SAMPLE A1 Xcorr 4.78 DeltaCn 0.66

SAMPLE B3 Xcorr 5.46 DeltaCn 0.66

P-value=0.0.018

MYG_HORSE SAMPLE B1 Xcorr 3.87 DeltaCn 0.55

516.2?

635.3?653.4?

a11-H2O-H2O+2H

y10+2Hy11+2H

y13-H2O-H2O+2H

y13-NH3+2H+1

y13+2H

b14-NH3-H2O+2H

y7+1y8+1

y9+1

y10+1

b3

y3

b4

y4

b5 b6y5b7

y6

y7

b9

y8

b10

y9

y10

b11y11b12

b13

b14

V E A D I A G H G Q E V L I RR I L V E Q G H G A I D A E V

m/z

Rel

ati

ve

Inte

ns

ity

0%

20%

40%

60%

80%

100%

0 250 500 750 1000 1250 1500

1,605.85 AMU, +2 H (Parent Error: 3.7 ppm)

b7+1

y15-NH3-NH3+2H

y15+2H

b8-NH3-H2O

y9+1

y3 y4

b4

b5y5

b6

y6

y7

b7

y8 b8

y9

b9

y10

b10b11

y11

b12

y12

y13

b13

y14

b14

G H H E A E L K P L A Q S H A T K

K T A H S Q A L P K L E A E H H G

m/z

Rel

ati

ve

Inte

ns

ity

0%

20%

40%

60%

80%

100%

0 250 500 750 1000 1250 1500 1750

1,852.96 AMU, +2 H (Parent Error: 1.3 ppm)

SAMPLE A3 Xcorr 4.66 DeltaCn 0.72

P-value=0.000019

FDR= 0 219

AV STD R A A STD R B AV A/B0

0.2

0.4

0.6

0.8

1

1.2

0.30

0.12

0.95

ECOLI

ECOLI

CIC_HORSE

MYG_HORSE

ALDOA_RABIT

Conclusions:

• Spectral count can be an easy way to try to perform quantitative proteome analysis, but :

• Needs the ability to perform different LC runs with very low dispersion.• The response to increasing protein amount is saturable.• Noisy data at low spectral counts – large difference in spectral count necessary to determine significant change.

M José FidalgoEva OlmedoFrancisco FernándezJosep M Estanyol

Oriol Bachs

Proteomic Facility University of Barcelona

SAMPLE A1

SAMPLE A2

SAMPLE A3

SAMPLE A4

SAMPLE B1

SAMPLE B2

SAMPLE B3

SAMPLE B4

MITJANA A

DES A

MITJANA B

DES B0

5

10

15

20

25

30

35ALDOA_RABIT

MYG_HORSE

CIC_HORSE

ALBU_BOVIN

SAMPLE A1

SAMPLE A2

SAMPLE A3

SAMPLE A4

SAMPLE B1

SAMPLE B2

SAMPLE B3

SAMPLE B4 M A DES A M B DES B A/B B/A

ALDOA_RABIT 23 27 26 29 18 17 15 15 26.25 2.50 16.25 1.50 1.62 0.62ALBU_BOVIN 2 1 0 1 5 1 1 2 1.00 0.82 2.25 1.89 0.44 2.25MYG_HORSE 31 28 31 31 18 17 15 18 30.25 1.50 17.00 1.41 1.78 0.56CIC_HORSE 22 18 17 26 28 33 31 27 20.75 4.11 29.75 2.75 0.70 1.43

fmol/microgram E. Coli protein

  MW A B B/A A/BALDOA_RABIT 39212 50 25 0.50 2.00BSA_BOVIN 66430 1 5 5.00 0.20MYG_HORSE 16952 520 200 0.38 2.60CYC_HORSE 12362 1000 1500 1.50 0.67