Quantitative Proteome Analysis of Pluripotent Cells by iTRAQ Mass Tagging Reveals Post-transcriptional Regulation of Proteins Required for ES Cell Self-renewal

Abstract

Embryonic stem cells and embryonal carcinoma cells share two key characteristics: pluripotency (the ability to differentiate into endoderm, ectoderm, and mesoderm) and self-renewal (the ability to grow without change in an untransformed, euploid state). Much has been done to identify and characterize transcription factors that are necessary or sufficient to maintain these characteristics. Oct-4 and Nanog are necessary to maintain pluripotency; they are down-regulated at the mRNA level by differentiation. There may be additional regulatory genes whose mRNA levels are unchanged but whose proteins are destabilized during differentiation. We generated proteome-wide, quantitative profiles of ES and embryonal carcinoma cells during differentiation, replicating a microarray-based study by Aiba et al. (Aiba, K., Sharov, A. A., Carter, M. G., Foroni, C., Vescovi, A. L., and Ko, M. S. (2006) Defining a developmental path to neural fate by global expression profiling of mouse embryonic stem cells and adult neural stem/progenitor cells. Stem Cells 24, 889–895) who triggered differentiation by treatment with 1 μm all-trans-retinoic acid. We identified several proteins whose levels decreased during differentiation in both cell types but whose mRNA levels were unchanged. We confirmed several of these cases by RT-PCR and Western blot. Racgap1 (also known as mgcRacgap) was particularly interesting because it is required for viability of preimplantation embryos and hematopoietic stem cells, and it is also required for differentiation. To confirm our observation that RACGAP-1 declines during retinoic acid-mediated differentiation, we used multiple reaction monitoring, a targeted mass spectrometry-based quantitation method, and determined that RACGAP-1 levels decline by half during retinoic acid-mediated differentiation. We knocked down Racgap-1 mRNA levels using a panel of five shRNAs. This resulted in a loss of self-renewal that correlated with the level of knockdown. We conclude that RACGAP-1 is post-transcriptionally regulated during blastocyst development to enable differentiation by inhibiting ES cell self-renewal.

ES cells are interesting because of their pluripotent developmental potential and capability for self-renewal. Pluripotency is the ability to differentiate into cells representing or derived from any of the three primordial germ layers that make up a postimplantation embryo. This property is shared with embryonal carcinoma (EC),¹ a cancer of the testis. Self-renewal is distinct from proliferation in that the daughter cells are identical to the mother cells at both the genetic and epigenetic levels (1).

Aggregation and treatment of mouse ES and EC cells with retinoic acid (RA) has been shown to increase the proportion of ES and EC cells that differentiate into the neuroectodermal lineage (1, 2), although the rate at which these cells differentiate is quite different (3). Differentiation by RA treatment and aggregation is a standard first step in many mouse ES cell differentiation protocols and is sufficient to efficiently drive mouse EC cells of line P19 to neural differentiation (3).

The last few years have led to many advances in our understanding of the undifferentiated state of ES cells. The transcription factor OCT 4 also called POU5F1 was considered to be the key determinant and biomarker of undifferentiated cells. OCT-4 was later found to dimerize with another transcription factor, SOX-2 (4, 5). Around that time, another gene Nanog, was identified as being necessary for the undifferentiated ES cell state. More recent work from Yamanaka and co-workers (6, 7) and Thomson and co-workers (8) has identified combinations of transcription genes including Oct-4, Sox-2, Nanog, Lin-28, c-Myc, and Klf-4 that are sufficient to reprogram somatic cells into ES-like induced pluripotent stem cells.

Although the transcriptional networks that establish and maintain the pluripotent state have been extensively studied, there has been relatively little work identifying how signaling from the environment affects cells as they exit the undifferentiated state. Recent studies of early differentiation have identified CASPASE 3 as a negative regulator of NANOG and RONIN protein stability (9, 10). Loading of ribosomes in undifferentiated ES cells is subject to extensive post-transcriptional control (11), resulting in changes in protein levels during differentiation that are not observed at the mRNA levels. Our present study compared the proteome and transcriptome of pluripotent cells to identify cases where the proteins but not the mRNAs are enriched before or after differentiation.

There have been several deep proteomic analyses of human and mouse ES cells. Graumann et al. (12) identified 5,111 proteins and quantified subcellular localization of proteins in undifferentiated mouse ES cells. Stable isotope labeling by amino acids in cell culture was used in the study for a self-to-self comparison. Swaney et al. (14) described 11,995 phosphopeptides from human ES cells (see also Ref. 13). Our study identified 5,489 proteins and provided quantitative comparisons of 4,986 proteins in mouse ES cells before and after differentiation. In addition, our study compared embryonal carcinoma cells with ES cells, which share the properties of pluripotency and self-renewal but are otherwise distinct.

We designed our experiments to replicate the cell lines and differentiation protocols published by Aiba et al. (3) to compare our measurements of the proteome with their measurements of the transcriptome (3). To achieve the necessary coverage and sensitivity we used the multidimensional protein identification technology LC-ESI MS/MS method (15, 16) to identify peptides, and in vitro iTRAQ mass tag labeling to measure their relative abundance.

iTRAQ labeling was done after protein extraction and protease digestion; therefore, growth in labeling medium was not required, and >90% of the peptides from any sample were tagged. iTRAQ offers the advantage of multiplexing with four or eight different mass tags; we used four tags to enable four samples to be compared directly within the same on-line chromatography run.

We confirmed biomarkers that distinguish undifferentiated pluripotent cells as predicted by prior analysis of the transcriptome. More importantly, we report protein biomarkers that are under post-transcriptional regulation including several cases involving proteins that are essential for self-renewal. We discovered that RACGAP-1, which is enriched at the protein level but not the mRNA level in pluripotent cells, is necessary for ES cell self-renewal.

EXPERIMENTAL PROCEDURES

Cell Culture

P19 cells (ATCC number CRL-1825) were cultured as described in Aiba et al. (3) Cells were grown in DMEM supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 100 units of penicillin/streptomycin. Cells were subcultured at 1:10 every 2–3 days on 150-mm tissue culture dishes treated with 0.1% gelatin.

For differentiation, P19 cells were plated on bacteriological plates in tissue culture medium supplemented with 1 μm RA (10 mm stock of all-trans-retinoic acid (ATRA) dissolved in 100% ethanol) as described in Aiba et al. (3). Medium was replaced on day 2, and EBs were collected for processing on day 4.

129/SvEv mouse ES cells from ATCC were cultured as described in Aiba et al. (3). Cells were grown in DMEM supplemented with 15% heat-inactivated FBS and 100 units of penicillin/streptomycin, 1,000 units of LIF, 1 mm sodium pyruvate, 1 mm non-essential amino acids, and 55 μm β-mercaptoethanol. Cells were subcultured at 1:5 every 2–3 days on 150-mm tissue culture dishes treated with 0.1% gelatin.

For differentiation, 129/SvEv mouse ES cells were plated on bacteriological plates in ES culture medium minus LIF as described in Aiba et al. (3). Medium was replaced on days 2, 4, and 6. Cells were treated with 1 μm RA on days 4–8.

Cell Lysis, Reduction, and Trypsin Digestion

100 μl of cell pellets were lysed in 250 μl of lysis buffer (2% (w/v) RapiGest (Waters catalogue number 186002122), 1 mm EDTA, and 50 mm Hepes buffer pH 7.2). Cysteines were reduced and alkylated using 1 mm tris(2-carboxyethyl)phosphine (Fisher catalogue number AC36383) at 95 °C for 5 min followed by 2.5 mm iodoacetamide (Fisher catalogue number AC12227) at 37 °C in the dark for 15 min. Protein concentrations were measured using the Bradford assay (Pierce). Proteins were digested with trypsin (Roche Applied Science catalog number 03 708 969 001) at 37 °C with an enzyme-to-substrate ratio (w/w) of 1:50 overnight.

iTRAQ Labeling of Peptides

For iTRAQ (Applied Biosystems, Foster City, CA) derivatization, an aliquot of each digested sample (100 μg of total protein) was treated with one tube of one of the iTRAQ reagents in 70% isopropanol at pH 7.2 for 2 h at room temperature. Labeled samples were dried down in a vacuum concentrator. 100 μl of water was added to each tube to dissolve the peptides. Samples tagged with four different iTRAQ reagents were pooled together. 1% trifluoroacetic acid (TFA) was added to a final pH of 1.4 to precipitate RapiGest. Samples were incubated at 4 °C overnight and then centrifuged at 16,100 × g for 15 min. Supernatant was collected and centrifuged through a 0.22-μm filter and was used for LC-MS/MS analysis. iTRAQ labeling efficiency was calculated by searching the MS/MS data specifying four possible iTRAQ modifications: 1) fully labeled, 2) N terminus-labeled only, 3) lysine-labeled only, and 4) non-labeled. Using the above protocol, we obtained higher than 90% iTRAQ labeling efficiency for all data sets (Table I).

Table I. Filtering criteria for autovalidation of database search results (spectrum score, spectrum SPI%).

	1+ peptide	2+ peptide	3+ peptide
MS/MS spectrum cutoff scores (score, SPI%)	>14, >50%	>12, >50%	>14, >50%

On-line Separation of Peptides by HPLC

An Agilent 1100 HPLC system (Agilent Technologies, Santa Clara, CA) delivered a flow rate of 300 nl/min to a three-phase capillary chromatography column through a splitter. Using a custom pressure cell, 5-μm Zorbax SB-C₁₈ (Agilent Technologies) was packed into fused silica capillary tubing (200-μm inner diameter, 360-μm outer diameter, 20 cm long) to form the first reverse phase column (RP1). A 5-cm-long strong cation exchange (SCX) column packed with 5-μm polysulfoethyl (PolyLC, Inc.) was connected to the RP1 using a zero dead volume 1-μm filter (Upchurch catalogue number M548) attached to the exit of the RP1. A fused silica capillary (100-μm inner diameter, 360-μm outer diameter, 20 cm long) packed with 5-μm Zorbax SB-C₁₈ (Agilent Technologies) was connected to an SCX column as the analytical column (the second reverse phase column). The electrospray tip of the fused silica tubing was pulled to a sharp tip with the inner diameter smaller than 1 μm using a laser puller (Sutter catalogue number P-2000). The peptide mixtures were loaded onto the RP1 using the custom pressure cell. Columns were not reused. Peptides were first eluted from the RP1 to the SCX column using a 0–80% acetonitrile gradient for 150 min. The peptides were fractionated by the SCX column using a series of salt gradients (from 10 mm to 1 m ammonium acetate for 20 min) followed by high resolution reverse phase separation using an acetonitrile gradient of 0–80% for 120 min. Typically, it took 4 days (38 salt fractions) for each full proteome analysis.

Tandem Mass Spectrometry Analysis

Spectra were acquired using an LTQ linear ion trap tandem mass spectrometer (Thermo Electron Corp., San Jose, CA) using automated, data-dependent acquisition. The mass spectrometer was operated in positive ion mode with a source temperature of 150 °C.

The full MS scan range of 400–2000 m/z was divided into three smaller scan ranges (400–800, 800–1050, and 1050–2000) to improve the dynamic range. Both collision-induced dissociation (CID) and pulsed Q dissociation (PQD) scans of the same parent ion were collected for protein identification and quantitation. Each MS scan was followed by four pairs of CID-PQD MS/MS scans of the most intense ions from the parent MS scan. A dynamic exclusion of 1 min was used to improve the duty cycle of MS/MS scans. About 20,000 MS/MS spectra were collected for each salt step fractionation.

Data Analysis

The raw data were extracted and searched using Spectrum Mill v3.03 (Agilent Technologies). The CID and PQD scans from the same parent ion were merged together. MS/MS spectra with a sequence tag length of 1 or less were considered to be poor spectra and were discarded. The remaining MS/MS spectra were searched against the International Protein Index (IPI) mouse database (v3.31; 56,555 protein sequences). The enzyme parameter was limited to fully tryptic peptides with a maximum miscleavage of 1. All other search parameters were set to the default settings of Spectrum Mill (carbamidomethylation of cysteines, iTRAQ modification, ±2.5 Da for precursor ions, ±0.7 Da for fragment ions, and a minimum matched percent scored peak intensity (SPI%) of 50%). A concatenated forward-reverse database was constructed to calculate the in situ false discovery rate (FDR). The total number of protein sequences in the combined database was 113,110. Cutoff scores (Table II) were dynamically assigned to each data set to maintain the false discovery rate at less than 1% at the protein level. Only proteins with two or more unique peptides were validated and selected for following quantitative analysis. Proteins that share common peptides were grouped to address the database redundancy issue. The proteins within the same group shared the same set or subset of unique peptides.

Table II. Summary of whole proteome data sets, total validated spectra identified, and iTRAQ labeling efficiency.

	Total validated spectra	Unmodified	Fully iTRAQ-modified	N terminus iTRAQ-modified only	Lysine iTRAQ-modified only
ES cells
No. spectra	55,638	861	52,262	1,927	588
Total intensity	1.52e+10	1.17e+8	1.52e+10	4.7e+8	1.07e+8
EC cells
No. spectra	94,399	7,529	60,811	23,528	2,531
Total intensity	4.67e+10	2.42e+9	3.13e+10	1.2e+10	1.06e+9
EC nuclei
No. spectra	49,746	1,139	41,641	6,601	365
Total intensity	7.32e+9	7.18e+7	6.41e+9	8.01e+8	3.66e+7

Data Access

All mass spectra used in this study are publicly available at the Proteome Commons Tranche web site (https://proteomecommons.org/tranche/).

Multiple Reaction Monitoring (MRM) Analysis of RACGAP-1

Undifferentiated mES 129/SvEv cells and differentiated mEB samples were lysed and digested by the protocol described above. MRM experiments were performed on an Agilent 6410 triple quadrupole mass spectrometer equipped with an Agilent 1200 nanoflow LC system, Agilent 1200 microautosampler, and HPLC-Chip Cube MS interface. Two synthetic peptides from RACGAP-1 (VSLLGPVTTPEFQLVK and TDTDNLGTPQNTGGMR) were purchased from Eton Bioscience, Inc. One synthetic peptide (VAPEEHPVLLTEAPLNPK) from actin and one from GAPDH (VPTPNVSVVDLTCR) were also purchased for the purpose of data normalization. Method optimization was performed to determine the optimal transitions and collision energies for each protein. Synthetic peptide standards were spiked into samples at a high level (1pmol of peptide/μg of sample) for method optimization. Each sample was analyzed both with and without spiked-in standards. The RACGAP-1 peptide standards were spiked in at 2fmol/μg of total peptide, which is close to the endogenous RACGAP-1 level. There were four total samples (mES no spike-in, mES with spike-in, mEB no spike-in, and mEB with spike-in). Each sample was analyzed six times. A total of 24 LC/MRM runs were performed, and the order of injections was randomized to minimized systematic errors. 1 μg of digested peptides was injected by Agilent 1200 autosampler and subjected to a 40-min reverse phase separation. The intensities from the best transitions (RACGAP-1: 864.5 → 1258.7; collision energy; 28, dwell time; 50 ms; β-actin: 977.5 → 1291.8; collision energy, 31; dwell time, 50 ms) were used to calculate the absolute amount of RACGAP-1.

Quantification by iTRAQ Mass Tagging Reagent

Protein iTRAQ intensities were calculated by summing the peptide iTRAQ intensities from each protein group. Peptides shared among different protein groups were removed before quantitation. A minimal total iTRAQ reporter ion intensity (sum of all four channels) of 100 was used to filter out low intensity spectra. Isotope impurities of iTRAQ reagents were corrected using correction factors provided by the manufacturer (Applied Biosystems). Median normalization was performed to normalize the protein iTRAQ reporter intensities in which the log ratios between different iTRAQ tags (115/114, 116/114, and 117/114) are adjusted globally such that the median log ratio is 0.

Quantitative analysis was performed on the normalized protein iTRAQ intensities. Protein ratios between undifferentiated and differentiated cells were calculated by taking the ratios of the total iTRAQ intensities from the corresponding iTRAQ reporters. t test (two-tailed, paired) was used to calculate the p values. Proteins with more than 50% change and p values less than 0.05 were considered significantly changed after differentiation.

Reanalysis of Microarray Data

Microarray data from Aiba et al. (3) were obtained from PubMed Gene Expression Omnibus (GEO) data sets. Data were processed by normalizing expression values to reference data and generating undifferentiated-to-differentiated ratios of mRNA measurements from each replicate of the experiment. Ratios were averaged in Microsoft Excel using the geomean function, and Student's t test was performed to obtain p values. Protein and mRNA data were combined in Microsoft Access.

RT-PCR

RNA was isolated from undifferentiated and differentiated cells using TRIzol (Invitrogen catalogue number 15596-026) according to the manufacturer's instructions. RNA quality was checked by agarose gel electrophoresis, and 2 μg of total RNA was treated with DNase Turbo (Ambion catalogue numberAM2238) according to the manufacturer's instructions. After DNase inactivation, cDNA was synthesized by reverse transcription with/without Superscript II (Invitrogen) and 10 μm random hexamer primers. RT-PCRs were performed using the following primer sequences: β-actin, F-GATCTGGCACCACACCTTCTACAATG and R-CGTACATGGCTGGGGTGTTGAAG; Oct-4, F-CTCCCGAGGAGTCCCAGGACAT and R-GATGGTGGTCTGGCTGAACACCT; Pdlim-7, F-GCACTCAGGAGCAGGCACGATGG and R-CCTCCGGGCGTGAGCCG; Dpysl-2, F-GTGACGCCCAAGACGGTGAC and R-ATGTTGTCGTCAATCTGAGCACCAG; H2afy, F-GGAGAAGAAGGGCGGCAAGG and R-GGCCTGCACTAATAGCAGCTC; Utf-1, F-GGTTCGCCGCCGCTCTACTG and R-GCAGGGGCAGGTTCGTCATTTTC; Racgap-1, F-TCCTTATGATCCACfFCTACAGAGAGTG and R-GCGCTCCACCACCTTG; and Sall-4, F-GGAGAGAAGCCTTTCGTGTG and R-CTCTATGGCCAGCTTCCTTC. All RT-PCRs were performed at the minimal cycle numbers necessary to observe bands (25 cycles for Utf-1 and actin; 27 for others) avoid saturation of the PCR products.

Western Blotting

Cells were collected in radioimmune precipitation assay buffer, and proteins were quantified by the Bradford assay. Protein samples were prepared in 1× SDS loading buffer containing β-mercaptoethanol, sonicated, run on 10% Tris-glycine protein gels, and transferred to PVDF membranes. Membranes were blocked with 5% (w/v) nonfat dry milk in PBS and 0.05% Tween and probed with the appropriate antibodies at concentrations from 1:100 to 1:1,000. Specific antibodies used were as follows: αUTF-1 (rabbit 1:1,000; Chemicon), αSALL-4 (1:1,000; Santa Cruz Biotechnology), PDLIM-7 (rabbit 1:1,000; Chemicon), H2AFY (1:100; Santa Cruz Biotechnology), and DPYSL-2 (1:1,000; Abnova). β-ACTIN (mouse 1:1000; Santa Cruz Biotechnology) was used as loading control in all cases.

Cellular Immunofluorescence

Undifferentiated cells and differentiated cells were plated on coverslips coated with poly-d-lysine and cultured overnight. Cells were fixed in 4% paraformaldehyde, permeabilized with PBS and 0.2% Triton, blocked with 10% BSA in PBS and 0.1% Triton for 10 h, and stained with the appropriate pair of primary antibodies diluted 1:1,000 (OCT-4 and NEUN or UTF-1 and βIII-TUBULIN) in blocking buffer or left in blocking buffer for secondary antibody alone controls. Cells were washed three times for 15 min in PBS and 0.1% Triton and incubated with secondary antibodies from Molecular Probes diluted 1:1,000 (anti-mouse 534 and anti-rabbit 488). Slides were washed three times and mounted in VECTASHIELD (Vector Laboratories) with DAPI.

Knockdown of Racgap-1

10⁵ undifferentiated ES cells were transfected via lipofection with Lipofectamine 2000 (1 μl in 50 μl of Opti-MEM and 1 μg of plasmid DNA). Each well was transfected with one of five Racgap1-targeting shRNA vectors from Open Biosystems, control plasmid (pCAG), or mock (no DNA). Cells were grown in undifferentiated conditions in the absence of penicillin/streptomycin for 24 h and then selected in 1 μg/ml puromycin for 24 h. After this time, cells were passaged onto 6-well plates, puromycin was removed, and the cells were allowed to form colonies for 3 days. Cells were fixed in 4% paraformaldehyde, blocked in 10% BSA, and immunostained with a rabbit anti-Oct4 antibody from Santa Cruz Biotechnology (H-134) diluted 1:100 in 10% BSA. Cells were then stained with a Pierce 3,3′-diaminobenzidine staining kit (catalogue number 36000). Colonies were counted, and standard deviation, mean, and Student's t test were quantified.

RESULTS

Differentiation of ES and EC Cells Results in Loss of Pluripotency Biomarkers

To better understand the protein makeup of pluripotent cells and to identify differences between the transcriptome and the proteome, we replicated the experimental design of a transcriptome study from Aiba et al. (3). The experimental design consisted of differentiating ES or EC cells by aggregation and retinoic acid treatment, resulting in formation of embryoid bodies that are composed of cells that have lost expression of pluripotency markers and are in various stages of entering the neuroectodermal lineage. Specifics of the two protocols are illustrated in Fig. 1.

Fig. 1.

Overview of experimental design and sample collection. A, ES cells of line 129/SvEv from ATCC were grown in undifferentiated conditions in the presence of LIF as described in Aiba et al. (3) or differentiated by aggregation in bacteriological plates in the absence of 1 μm ATRA for 4 days and then in the presence of 1 μm ATRA for 4 days as described under “Experimental Procedures.” Medium was refreshed at differentiation day 2, and cells were collected at the indicated time points by treatment with Versene, washed in 10 mm Hepes-buffered saline (pH 7.4), and frozen at −80 °C. Cells were dissolved in 2% RapiGest, reduced, trypsinized, and labeled with the iTRAQ mass tagging reagent. B, P19 cells from ATCC were grown in undifferentiated conditions as described in Aiba et al. (3) or differentiated by aggregation in bacteriological plates in the presence of 1 μm ATRA for 4 days as described under “Experimental Procedures.” Medium was refreshed at differentiation day 2, and cells were collected at the indicated time points by treatment with Versene, washed in 10 mm Hepes-buffered saline (pH 7.4), and frozen at −80 °C. Cells were dissolved in 2% RapiGest, reduced, trypsinized, and labeled with the iTRAQ mass tagging reagent. mESC, mouse ES cells; mEC, mouse EC cells.

To verify that undifferentiated cells expressed pluripotency biomarkers and that differentiated cells had lost these markers, we used immunofluorescence to assay key biomarkers (supplemental Fig. S1). Undifferentiated ES cells strongly expressed OCT-4 and UTF-1 but did not express the neuroectodermal marker βIII-TUBULIN or the neuron-specific marker NEUN (supplemental Fig. S1A). Differentiated ES cells lost expression of OCT-4 and UTF-1, had no up-regulation of βIII, and weakly expressed NEUN (supplemental Fig. S1B). In contrast, undifferentiated EC cells strongly expressed βIII along with OCT-4 and UTF-1 (supplemental Fig. S1C). Differentiated EC cells also maintained residual levels of UTF-1 protein (supplemental Fig. S1D).

Validated Mass Spectra

We recorded ∼6.8 million MS/MS spectra. After merging and filtering, we validated a total of 386,374 spectra using the cutoff scores listed in Table I. Among the validated spectra, 580 were from the decoy database, indicating an FDR of 0.15% at the spectrum level. There were a total of 36,794 unique peptides from the validated spectra; among them, 132 were from the decoy database, resulting in an FDR of 0.36% at the peptide level. Different charge states from the same peptide were considered as one peptide.

Efficiency of iTRAQ Mass Tagging

iTRAQ (17) can be used for multiplexed peptide profiling of up to four different samples. This approach labels samples with four independent reagents of the same mass that, upon fragmentation in MS/MS, give rise to four unique reporter ions (m/z = 114–117) that are subsequently used to quantify the four different samples, respectively.

iTRAQ reagents originally were not usable in ion trap instruments because of the “one-third rule.” To illustrate, fragment ions of an m/z 900 parent will not be detected below m/z 300, and this would normally prevent the detection of the iTRAQ reporter ions. To overcome this limit, a new collision-activated fragmentation technique, called PQD, was invented that enables routine and reliable measurement of ions down to 50 m/z (18). We have developed and successfully applied a combined CID-PQD scan approach for peptide identification and quantitation (19, 20).

To determine the iTRAQ labeling efficiency, MS/MS spectra were searched against a concatenated forward and reverse mouse protein database, searching for four possible iTRAQ modifications: 1) unmodified, 2) fully modified, 3) modified on the N terminus, and 4) modified on lysine side chains only. In all cases, labeling of tryptic peptides with the iTRAQ reagent was at least 92% efficient. The iTRAQ labeling efficiency of ES cell samples was 98.5%, whereas EC whole cell and ES nuclear samples were labeled at an efficiency of 92 and 97.7%, respectively (Table II).

Managing Protein Redundancy

A problem that any large scale proteomics study must deal with is the redundancy in protein databases. In many cases, tryptic peptides can be shared between different isoforms of a protein derived from the same locus, homologous proteins, or even unrelated proteins.

We addressed this redundancy by assigning proteins sharing the same peptides to a “protein group.” Each protein group has at least one unique peptide. Proteins within the group may contain additional peptides. If two proteins share peptides and contain peptides not present in the other, they will be assigned to separate protein groups. A “group leader” heads protein groups that contain more than one protein accession number. The group leader is the protein with the highest identification score (the sum of the identification scores of the identified peptides composing the protein), which usually is the protein containing the largest number of peptides. In the remaining text, the term “protein” should be taken to mean “protein group leader.” For iTRAQ quantitation, peptides that are shared between protein groups were removed, and only peptides that are unique to each group were used.

iTRAQ Mass Tags Enable Relative Quantitation of Proteins

Analysis of 129/SvEv ES cells before and after retinoic acid-mediated differentiation identified 4,053 proteins with an FDR of 0.56% (23 proteins from the decoy database of 4,053 proteins identified) (Table III and supplemental Table S6, all data). Removing all single peptide hits lowered the FDR to 0.1% (four proteins from the decoy database of 3,613 proteins with >1 peptide identified; supplemental Fig. S2). After removing proteins with low iTRAQ reporter ion intensities (sum of all channels below 100), we calculated the relative abundance of 3,566 proteins before and after differentiation. Similarly, we calculated the ratios of 3,801 proteins from EC cells with an FDR of 0.2% (nine proteins from the decoy database; Table III). The distribution of -fold changes in all three data sets is illustrated in Fig. 2.

Table III. Summary of whole proteome data sets, replicates, peptides identified, and total protein groups identified.

	Biological replicates	Total proteins	Reverse database proteins	Forward database proteins	FDR	iTRAQ labeling efficiency (N terminus, lysine, or both)	Proteins with total iTRAQ reporter intensity >100
					%	%
129/SvEV ES cells	3	4,053	23	4,030	0.56	98.5	3,566
Whole EC cells	2	4,501	9	4,492	0.20	92	3,801
EC nuclei	3	4,046	16	4,030	0.39	97.7	3,569
Total	8	5,523	34	5,489	0.62	95.2	4,986

Fig. 2.

Protein quantification. Distributions of -fold change values for each data set are shown. The histogram shows a normal distribution of -fold changes of proteins before and after differentiation.

To increase the depth of our proteome coverage, we purified nuclei from cells before and after differentiation using a fractionation kit (Pierce catalogue number 78833). This worked well for EC cells, but fractionations of ES cells and ES-derived EBs failed because of the tough, fibrous nature of the ES-derived EBs. Analysis of nuclei from EC cells enabled 3,669 protein ratios to be measured using the same criteria described above.

iTRAQ Mass Tag Ratios Reveal Conserved Protein Biomarkers of Pluripotency

To identify proteins in pluripotent cell types, we first tried a two-tailed paired Student's t test with a p value cutoff of 0.05 and a -fold change cutoff of 1.5-fold. Supplemental Table S1 summarizes the results. Only four proteins were classified as specific to the pluripotent cells across the three data sets (Fig. 3A); 21 proteins were specific for differentiated cells across all data sets (Fig. 3B). The proteins present before differentiation in all three data sets include the adhesion protein CDH-1 (E-cadherin) and the nuclear importin KPNA-2. Of the 21 proteins enriched across all three RA-differentiated cells, six are metabolic enzymes.

Fig. 3.

Overview of proteins enriched before and after differentiation using various criteria. A, to identify proteins generally enriched in pluripotent cells, we performed a Student's t test on proteins >50% enriched in undifferentiated cells as described under “Experimental Procedures.” Proteins with a p value <0.05 and -fold change of 50% were considered significantly enriched. The Venn diagram in A represents the overlap of proteins that make the statistical and -fold change cutoff. B, to identify proteins generally enriched in cells differentiated with RA, we performed a Student's t test on proteins >50% enriched in differentiated cells as described under “Experimental Procedures.” Proteins with a p value <0.05 and -fold change of 50% were considered significantly enriched. The Venn diagram in B represents the overlap of proteins that make the statistical and -fold change cutoff. C, to increase the number of proteins identified as generally enriched in pluripotent cells, we ignored statistical significance and used 50% enrichment as a -fold change cutoff. The Venn diagram in C represents the overlap of proteins that make the -fold change cutoff. These proteins include known factors such as UTF-1, TCF-3, and OCT-4 (POU5F1). D, to increase the number of proteins identified as generally enriched in RA-treated cells, we ignored statistical significance and used 50% enrichment as a -fold change cutoff. The Venn diagram in D represents the overlap of proteins that make the -fold change cutoff. These proteins include known RA response factors such as CRABP-1 and CRABP-2 as well as the neural differentiation-associated proteins HOXB-6 and N-CADHERIN.

This list excludes a large number of proteins known to be enriched in pluripotent cells such as OCT-4 (21–28), UTF-1 (29–34), and DNMT-3B (35–38). These and other known pluripotency-specific proteins were identified in our profiles. They met the threshold of >1.5-fold enrichment before differentiation in all replicates across all three data sets (supplemental Table S2). We found that the t test was not an adequate statistical measure of reproducibility of enrichment. Variability in the dynamic range of quantitative measurements across experiments resulted in poor p values in cases where enrichment was reproducibly observed across all three replicates. For this reason, we decided that the t test is too stringent for identifying proteins with real -fold differences. Instead we proceeded with a -fold change cutoff to identify proteins with two or more unique peptides that reproducibly fall into the category of >50% enriched before or after differentiation across all three biological replicates.

Using -fold change alone, we identified 101 proteins >1.5-fold enriched before differentiation in both ES and EC cells in all three data sets (Fig. 3C). These include known pluripotency-associated proteins OCT-4, UTF-1, DNMT-3B, CDH-1 (39–47) (embryonic cadherin), and TCF-3 (48). -Fold change also identified 175 proteins enriched after retinoic acid-mediated differentiation in all three experiments (Fig. 3D and supplemental Table S3). These proteins include the retinoic acid response proteins retinol-binding protein 1 (RBP-1) and cellular retinoic acid-binding proteins 1 and 2 (CRABP-1 and CRABP-2) (49–56); Cadherin-2 (CDH-2; neural cadherin) (57–62); the embryonic patterning protein HOXB6 (63–67); and metabolic proteins. Thus, identification of proteins with >1.5-fold change during differentiation of two pluripotent cell types across three experiments reveals a group of proteins that includes many already known to be associated with pluripotency or differentiation and additional proteins not previously associated with pluripotency.

Biological Processes Enriched before and after Differentiation

To better understand the proteins enriched in cells before or after differentiation, we used the DAVID Bioinformatics Resource to identify gene ontology (GO) terms associated with the undifferentiated or differentiated states (http://david.abcc.ncifcrf.gov/home.jsp) (68, 69). The GO terms relating to biological processes enriched before differentiation in all three data sets included ribosome biogenesis, one-carbon metabolism, and amine biosynthesis (supplemental Fig. S4A). The GO terms relating to biological processes enriched after RA- and aggregation-mediated differentiation all related to oxidative phosphorylation of glucose via the TCA cycle (supplemental Fig. S4B).

Comparison of GO Terms Enriched in Proteome and Transcriptome Data

To compare the insights provided by genome-wide mRNA and protein profiles, we used the DAVID Bioinformatics Resource to identify GO terms (http://david.abcc.ncifcrf.gov/home.jsp) (68, 69) associated with our protein data set and with the published mRNA data from Aiba et al. (3). In all cases, mRNA data produced more GO terms than the protein data because of the greater depth of coverage. GO annotations of proteins were very similar to the annotations of mRNAs, especially in undifferentiated ES cells (supplemental Fig. S4). There were more differences between the mRNA and protein data of differentiated ES cells as well as undifferentiated and differentiated EC cells. However, in all cases, the annotations assigned to the proteins were >50% shared with the annotations assigned to the transcripts, indicating that most pathways observed were not subject to extensive post-transcriptional regulation.

Identification of Post-transcriptional Regulation in Pluripotent Cells

Our experiments were modeled on the global transcriptome study of Aiba et al. (3) using cell lines and differentiation conditions that replicated their study. We reanalyzed the data from Aiba et al. (3) and compared the resulting mRNA measurements directly with our proteome measurements. Direct comparisons of mRNA and protein ratios for ES and EC cells are illustrated as scatter plots in Fig. 4, A and B, respectively. Of the 276 proteins enriched before or after differentiation, 196 had corresponding data from the microarray; when these were compared, 97 of 196 (49.49%; supplemental Table S4) agreed with our protein data. These gene products are therefore putative markers of pluripotency observed in independent experiments at both the mRNA and protein levels. On the other hand, 43 of 196 genes (21.94%) were enriched at the protein level but not at the mRNA level (supplemental Tables S5 and S7). These genes may be subject to post-transcriptional regulation during RA-mediated differentiation of pluripotent cells. Mechanisms such as translational repression by miRNAs and destabilization by ubiquitylation or other mechanisms of proteolysis could explain the discrepancies between changes in protein and mRNA levels.

Fig. 4.

Protein data to confirms some, but not all, transcript-level observations from Aiba et al. (3). A, scatter plot of 129/SvEv ES gene product ratios measured at the mRNA and protein levels. Only gene products observed at the protein and mRNA levels were included in this graph. Pearson's correlation coefficient was calculated as 0.362. Axes represent log₂ scale ratios with positive numbers representing proteins enriched before differentiation and negative numbers representing proteins enriched after differentiation. Points falling in the first and third quadrants represent cases of agreement between protein and mRNA data. Points falling in the second and fourth quadrants represent cases of disagreement and putative cases of post-transcriptional regulation. Points falling on axes represent proteins or transcripts whose levels are unchanged in the corresponding data set. B, scatter plot of EC gene product ratios measured at the mRNA and protein levels. Only gene products observed at the protein and mRNA levels were included in this graph. Pearson's correlation coefficient was calculated as 0.409. Axes represent log₂ scale ratios with positive numbers representing proteins enriched before differentiation and negative numbers representing proteins enriched after differentiation. Points falling in the first and third quadrants represent cases of agreement between protein and mRNA data. Points falling in the second and fourth quadrants represent cases of disagreement and putative cases of post-transcriptional regulation. Points falling on axes represent proteins or transcripts whose levels are unchanged in the corresponding data set.

To confirm the discrepancy between protein and mRNA changes, we tested four selected cases (Pdlim-7, H2afy, Dpysl-2, and Sall-4) by RT-PCR and by Western blot (supplemental Fig. S5). These genes were chosen from candidate genes whose protein levels were changed but whose mRNA levels were unchanged and that had commercial antibodies available. In each of the four cases, the results of the Western blots confirmed our iTRAQ data: SALL-4 and PDLIM-7 were enriched before differentiation, whereas DPYSL-2 and H2AFY were enriched after differentiation. The RT-PCR assays did not, in every case, confirm the observations from the microarray studies: Sall-4, Dpysl-2, and H2afy all changed at the mRNA level in correspondence with the observed protein changes, whereas Pdlim-7 mRNA was unchanged before and after differentiation, confirming that PDLIM-7 is subject to post-transcriptional regulation during RA-mediated differentiation.

Racgap-1 Is Necessary for Self-renewal and Is Degraded in Differentiating Cells

A fifth gene Racgap-1, was selected for further study. Antibodies were not available for Western blots so we developed an MRM assay to confirm the iTRAQ data.

We first confirmed that Racgap-1 mRNA is unchanged during differentiation by RT-PCR (Fig. 5A) and then used MRM to measure transitions from several peptides as described under “Experimental Procedures.” As shown in Fig. 5B, MRM confirmed that RACGAP-1 is enriched at the protein level before the differentiation of ES cells.

Fig. 5.

RACGAP-1 is a post-transcriptionally regulated protein necessary for mES cell self-renewal. A, to confirm that Racgap-1 levels are unchanged during differentiation, we performed RT-PCR analysis on ES cells (ESC) and ES cells differentiated as in the initial experiment (aggregated in the absence of LIF for 4 days and grown in suspension in the presence of 1 μm ATRA for 4 more days). B, we used MRM to confirm that RACGAP-1 is significantly (***, p = 0.00167) enriched in ES cells before differentiation. RT-PCR image colors are inverted for clarity. C, knockdown of Racgap-1 results in significant reduction of OCT-4⁺ ES colonies. Experiments were performed in triplicate, and Student's t test was used to calculate significance using values from control cells and cells transfected with an ineffective shRNA (sh1). p values are listed in the table under the chart (Error bars are coefficients of variation.) N/A, not applicable.

We tested whether Racgap-1 is necessary for ES cell self-renewal using a library of five shRNAs from Open Biosystems (www.openbiosystems.com). We transfected three replicates of ES cells with the vectors, selected for transfected cells after 24 h, passaged the cells onto 6-well plates, and allowed them to grow. After 3 days, plates were immunostained with rabbit anti-OCT-4 (1:100; Santa Cruz Biotechnology) and 3,3′-diaminobenzidine-stained (Pierce catalogue number 36000) to quantify OCT-4⁺ colonies (Fig. 5C). Cells transfected with the most effective vectors (shRNAs 2, 3, and 5) resulted in the fewest colonies, significantly less than either the control (p = 0.00077, 0.00037, and 0.00305, respectively) or the ineffective shRNA 1 (p = 0.022, 0.024, and 0.048, respectively). A separate experiment using HEK293 cells (with no mouse Racgap-1 target mRNA sequence) confirmed that the plasmids were able to confer similar puromycin resistance on the cells without disrupting their growth (data not shown). We conclude that the reduction in ES cell colony number is due to a decrease of RACGAP-1 protein.

DISCUSSION

Proteins Associated with the Undifferentiated State

Proteins enriched in undifferentiated cells in all three data sets included proteins known to be associated with pluripotency (OCT-4, UTF1, TCF-3, and DNMT-3B). We did not detect NANOG or RONIN (THAP-11) likely because of their low expression levels and lack of tryptic cleavage sites. To better understand and categorize the types of proteins enriched before differentiation in all three data sets, we used the Gene Functional Classification tool from DAVID (http://david.abcc.ncifcrf.gov/home.jsp) (68, 69) to characterize the types of proteins enriched before differentiation (supplemental Fig. S3). The proteins fell into general classes: rRNA processing and ribosome assembly, general RNA binding, transcriptional regulators, ATP-dependent helicases, and serine/threonine kinases (all involved in the cell cycle).

Proteins Associated with RA-mediated Differentiation

In addition to RA-responsive genes such as Crabp-1 and -2, we used the Gene Functional Classification tool from DAVID to identify classes of proteins enriched after RA-mediated differentiation in all three data sets. These clusters included proteins involved in the TCA cycle and general catabolism/metabolism; a family of dihydropyrimidinase-like proteins 2, 3, 4, and 5 involved in axon guidance and dendrite projection; cytoskeletal proteins; Ca²⁺-binding proteins; chromatin structural proteins of the H1/H5 family; and small GTPases involved in signal transduction.

Comparison between Proteome and Transcriptome of Pluripotent and Differentiated Cells

The results of this study suggest that the transcriptome is not a reliable predictor of the proteome. This is in agreement with several other proteome analyses in multiple systems using multiple techniques that have concluded that the proteome and transcriptome correlate only weakly if at all (70–73). Our data confirmed that products from the core transcription factors associated with pluripotency (Oct-4, Sox-2, Utf-1, Rex-1, and Sall-4) change similarly at the protein and mRNA levels.

Putative Cases of Post-transcriptional Regulation

A potential benefit of coupled proteome and transcriptome measurements is the ability to identify post-transcriptional regulation. We replicated an experimental design from Aiba et al. (3), and by comparing our quantitative iTRAQ data with their quantitative Agilent microarray data, we were able to identify cases where our data sets disagreed. We chose four of these genes to confirm by RT-PCR and Western blot. In all cases, Western blots confirmed our proteome data, but in three of the cases, changes in protein levels were accompanied by changes in mRNA levels, contradicting the microarray-based measurements. The one exception was Pdlim-7, which was indeed unchanged at the mRNA level during the differentiation of EC cells but was down-regulated at the protein level during differentiation of EC cells. PDLIM-7 is thought to be a protein scaffold that assembles signaling molecules including protein kinase C (PKC) on the actin cytoskeleton to mediate signaling during development.

Sampath et al. (11) predicted that a 14-3-3 protein, YWHAB, would be more abundant in differentiated ES cells even though its mRNA levels were unchanged. Our iTRAQ data and the reanalyzed Aiba et al. (3) data confirmed their prediction, making it likely that Ywhab is post-transcriptionally regulated.

Role of Racgap-1 in ES Cell Self-renewal

Previous studies have determined that Racgap-1 is necessary for division in some cell types. It is phosphorylated by AURORA KINASE B and becomes a RHOA GAP rather than a RAC-1 or CDC-42 GAP (74–77). There seem to be several other mechanisms to limit its activity toward CDC-42 and RAC (78). During mouse development, Racgap-1 is expressed in the brain/ectoderm. Racgap-1 is necessary for normal development of the preimplantation embryo. Embryos that lack Racgap-1 develop past the zygote stage but fail to form an inner cell mass. Conditional knock-out of the gene in hematopoietic stem cells or B cells blocks proliferation and differentiation and leads to apoptosis, but this is apparently independent of the GAP activity of the protein (79, 80).

We observed that Racgap-1 mRNA is expressed at the same level in undifferentiated and differentiated ES cells, but our protein measurements showed 2-fold enrichment in undifferentiated cells. Knockdown of Racgap-1 by shRNA caused a significant loss of OCT-4⁺ colony formation. This phenotype may be due to the fact that the gene is necessary for proliferation of the inner cell mass. The mechanism by which RACGAP-1 is post-transcriptionally regulated is unknown. A search of the microRNA target database (found at www.microRNA.org) (81) identified 57 putative miRNA binding sites, but none of the 10 top miRNAs are expressed in the neural lineage. Therefore, it is unlikely that the predicted miRNAs are responsible for the post-transcriptional regulation of RACGAP-1.

Protein Level Changes in Pathways during Differentiation of Pluripotent Cells

Ribosome Assembly Complexes

The functional annotation clusters from undifferentiated cells include ribosome biogenesis and rRNA-binding proteins. These proteins are enriched in undifferentiated ES and EC cells. Initially, we hypothesized that the cells are rapidly proliferating and thus need to produce many ribosomes. However, the Sampath et al. (11) study determined that ES cells contain fewer ribosomes per cell than cells from EBs (as measured by rRNA content normalized to genomic DNA content), and the ribosomes that are present in ES cells are less loaded with mRNA than cells from EBs (11). Why then do we see an enrichment of proteins associated with rRNA synthesis and regulation before differentiation?

Many of these proteins, apart from roles in rRNA synthesis and ribosome assembly, also regulate cell proliferation. RPS19BP-1 (82), GNL-3 (83, 84), and NOLC-1 (84) all inhibit p53-mediated growth arrest, whereas BOP-1 inhibits growth of non-pluripotent 3T3 cells via the G₁/S checkpoint, which is absent in ES cells (85). NPM-3 (86) and NOC-3l (87, 88) regulate chromatin structure/replication and are associated with several rapidly proliferating cell types. NOLA-3 (NOP-10) is also a member of the telomerase complex, which is essential for extensive proliferation (89). These alternate functions suggest that the ribosome protein enrichment is due in part to their recruitment to perform other tasks in proliferating cells in addition to increased ribosome synthesis activity.

Metabolism

The most striking biological process up-regulated in differentiating ES and EC cells is an increase in oxidative glucose metabolism. This response might be counterintuitive; however, ES cells and many rapidly dividing cancer cell types use glycolysis (90–92) as their primary energy source (Warburg effect), whereas oxidative phosphorylation of glucose via the TCA cycle is increasingly recognized to be associated with quiescent cell types such as neurons (93–95).

Retinoic Acid Signaling

ES and EC cells were aggregated and treated with retinoic acid, leading to an increase in the level of RA response proteins. ES and EC cells up-regulated CRABP-2, CRABP-1, and RBP-1. RA signaling leads to increases in TGFB-1 signaling (2, 96, 97). Although we did not see TGFB-1, we did see changes in its receptor (enriched in differentiated EC cells) and in downstream targets of TGFB-1 such as SMAD-5 (98) (enriched in the nucleus of differentiated EC cells). We also saw an increase in SMAD-5 in the nuclear fraction of differentiated cells, suggesting that the cells are indeed subject to increased TGFβ signaling in response to retinoic acid.

Signaling Pathways Activated during Differentiation

There were several signaling molecules enriched before differentiation. These include AURORA KINASE A, AURORA KINASE B, BUB-1, and POLO-LIKE KINASE 1. These proteins are all involved in cell division (99–107) and likely reflect the rapid proliferation associated with pluripotent cells.

Differentiated cells are enriched with several types of signaling proteins including the 14-3-3 proteins YWHAB, YWHAZ, and YWHAE. These proteins bind phosphoproteins and mediate their translocation, activity, or degradation, causing changes in development, apoptosis, and metabolism (108, 109). Interestingly, these proteins have been shown to interact with two proteins that are enriched before differentiation in all three data sets, FOXO-1 and TJP-2, suggesting that the 14-3-3 proteins may be regulating these proteins during differentiation. Differentiated cells have up-regulated GSK3B protein levels; this kinase is involved in neuronal differentiation (110) and is a negative regulator of the propluripotency AKT signaling pathway (111, 112).

Adhesion

Another group of proteins that changed during differentiation is involved in cell adhesion. Undifferentiated ES and EC cells express higher levels of the embryonic cadherin (CDH-1) as well as tight junction-associated protein TJP-2 and junction plakoglobin JUP. During differentiation, cells begin to express N-cadherin (CDH-2) and down-regulate JUP and TJP-2.

Interestingly, undifferentiated EC cells, but not ES cells, express three laminin subunits that make up the Laminin-511 complex. Recent publications have identified exogenous Laminin-511 as a potent positive regulator of pluripotency in mouse and human ES cells (113). Mouse ES cells grown on a matrix of Laminin-511 no longer require external LIF to remain undifferentiated. EC cells may maintain their undifferentiated state in the absence of external signals such as LIF due to autocrine Laminin-511 signaling.

Transcriptional Regulators and Chromatin Remodeling

Aggregation and RA treatment of ES and EC cells resulted in down-regulation of several transcription factors and chromatin-modifying proteins. These include the well known transcription factors OCT-4, SALL-4, UTF-1, and TCF-3 plus the chromatin-modifying protein DNMT3B. Aggregation and RA treatment caused up-regulation of histone family proteins H1/H5 and of transcription factors HMGB-3, DACH-1, and HOXB-6.

Both UTF-1 and H1/H5 have been implicated in epigenetic transcriptional regulation by remodeling and stabilizing condensed chromatin (114–117). Changes in these factors likely reflect changes in chromatin state during differentiation, resulting in loss of expression of undifferentiated state-specific proteins such as OCT-4, SALL-4, and UTF-1 and induction of neural differentiation-specific proteins such as CRABP-1, CRABP-2, NESTIN, HOXB-6, and N-CADHERIN.