Research ArticlePhosphoproteomics

Quantitative Phosphoproteomic Analysis of T Cell Receptor Signaling Reveals System-Wide Modulation of Protein-Protein Interactions

See allHide authors and affiliations

Sci. Signal.  18 Aug 2009:
Vol. 2, Issue 84, pp. ra46
DOI: 10.1126/scisignal.2000007

Abstract

Protein phosphorylation events during T cell receptor (TCR) signaling control the formation of complexes among proteins proximal to the TCR, the activation of kinase cascades, and the activation of transcription factors; however, the mode and extent of the influence of phosphorylation in coordinating the diverse phenomena associated with T cell activation are unclear. Therefore, we used the human Jurkat T cell leukemia cell line as a model system and performed large-scale quantitative phosphoproteomic analyses of TCR signaling. We identified 10,665 unique phosphorylation sites, of which 696 showed TCR-responsive changes. In addition, we analyzed broad trends in phosphorylation data sets to uncover underlying mechanisms associated with T cell activation. We found that, upon stimulation of the TCR, phosphorylation events extensively targeted protein modules involved in all of the salient phenomena associated with T cell activation: patterning of surface proteins, endocytosis of the TCR, formation of the F-actin cup, inside-out activation of integrins, polarization of microtubules, production of cytokines, and alternative splicing of messenger RNA. Further, case-by-case analysis of TCR-responsive phosphorylation sites on proteins belonging to relevant functional modules together with network analysis allowed us to deduce that serine-threonine (S-T) phosphorylation modulated protein-protein interactions (PPIs) in a system-wide fashion. We also provide experimental support for this inference by showing that phosphorylation of tubulin on six distinct serine residues abrogated PPIs during the assembly of microtubules. We propose that modulation of PPIs by stimulus-dependent changes in S-T phosphorylation state is a widespread phenomenon applicable to many other signaling systems.

Introduction

T cell receptor (TCR) signaling is central to multiple aspects of the adaptive immune response, including initial clonal expansion, the ensuing effector functions of activating specific B cells and macrophages, and the effector functions of cytotoxic T cells (1). TCR signaling also controls the maturation of T cells in the thymus. According to the current paradigm, protein phosphorylation events drive TCR signaling (2, 3). Initial tyrosine phosphorylation events mediate the activation of tyrosine kinases and protein-protein associations among receptor-proximal proteins through Src homology 2 (SH2) domains. This is followed by changes in phosphoinositide metabolism, Ca2+ influx, and the activation of serine-threonine (S-T) kinase cascades, which culminate in the activation of transcription factors, leading to the synthesis and secretion of cytokines for autocrine and paracrine action on other immune cell types (1). T cell activation is also accompanied by several other cellular phenomena, such as patterning of cell-surface proteins, endocytosis of the TCR, formation of the F-actin cup, inside-out activation of integrins, polarization of microtubules, and alternative splicing of messenger RNAs (mRNAs) (47). These cellular phenomena have important immunological consequences. For example, the polarization of microtubules toward the antigen-presenting cell (APC) ensures that exocytosis of effector proteins occurs at the interface between the T cell and the APC and thereby maintains the specificity of immune responses. Inside-out activation of integrins enables stable adherence and interaction between the T cell and the APC. Actin cytoskeleton dynamics during TCR signaling is important for other events, such as the patterning of cell-surface proteins, endocytosis of the TCR, activation of integrins, and the influx of Ca2+ (7, 8).

Substantial progress has been made over the last two decades in discovering and elucidating the roles of numerous TCR-responsive tyrosine phosphorylation events in T cell activation. Emphasis on these early events in TCR signaling has been facilitated by the availability of antibodies that specifically recognize phosphotyrosine (pTyr) residues and by the use of flow cytometric procedures to detect the presence of cell-surface markers of activation and to monitor defects in Ca2+ influx (9). Functional characterization of candidate tyrosine phosphorylation sites has been less daunting because of the low frequency of tyrosine residues in proteins. Furthermore, functional characterization has been aided by the knowledge that most tyrosine phosphorylation events result in the recruitment of SH2 domain–containing proteins in a sequence-restricted manner (10, 11). In contrast, relatively fewer S-T phosphorylation events have been discovered and characterized in the context of TCR signaling despite their molar predominance over tyrosine phosphorylation events. Most of the known S-T phosphorylation events in TCR signaling occur on S-T kinases and on transcription factors responsible for the expression of cytokine-encoding genes. However, no evidence exists to support a widespread role for S-T phosphorylation in other aforementioned phenomena associated with T cell activation. If these phenomena are indeed controlled by widespread phosphorylation, it is likely that S-T phosphorylation events will be predominantly responsible, because few proteins involved in these processes contain SH2 domains. Whether S-T phosphorylation is directly involved in these processes and, if so, how it mechanistically coordinates them are currently unclear.

Mass spectrometry–based techniques enable the identification of sites on proteins phosphorylated in vivo and the quantification of stimulus-dependent changes in phosphorylation events in a hypothesis-free manner. Furthermore, recent advances have presented an opportunity to monitor phosphorylation events at the proteome level (1214). We used large-scale quantitative phosphoproteomic methods to identify TCR-responsive phosphorylation sites and to gain system-wide insights into the role of phosphorylation during TCR signaling. Overall, we identified 10,665 unique phosphorylation sites, of which 696 showed TCR-responsive changes. Apart from 63 known phosphorylation sites, we identified 60 previously unknown TCR-responsive phosphorylation sites on proteins that have previously characterized roles in TCR signaling. Detailed in-depth analysis of all of the TCR-responsive phosphorylation sites resulted in the assignment of roles for 38.7% of these sites. Our main finding is that the scope of phosphorylation in response to TCR stimulation is widespread and that it extensively targets proteins involved in all of the salient phenomena associated with T cell activation. Detailed analysis of TCR-responsive phosphorylation sites on proteins belonging to relevant functional modules revealed that many of the TCR-responsive sites are within the mapped region of protein-protein interactions (PPIs). As further corroboration, we observed that phosphoproteins tended to participate in more PPIs than did nonphosphorylated proteins, and that proteins that contained TCR-responsive S-T phosphorylation sites formed a more extensive network than they would have been expected to form by chance. Finally, with microtubule assembly as a test system, we showed that phosphorylation of tubulin on six distinct serine residues abrogated PPIs during the assembly of microtubules. These results enabled us to deduce that S-T phosphorylation modulates PPIs in a system-wide manner to regulate diverse stimulus-dependent processes.

Results

Compilation of the Jurkat cell phosphoproteome data set

We used the human Jurkat T cell leukemia cell line for our analysis because it is a preferred model system for studying TCR signaling (9). Approximately 100 million cells were subjected to trypsinization, phosphopeptide enrichment by strong cation exchange (SCX) fractionation or immunoprecipitation with antibodies specific for pTyr residues, and immobilized metal-affinity chromatography (13, 15). Finally, the phosphopeptide mixtures were analyzed by tandem mass spectrometry (MS/MS). In an attempt to assemble a comprehensive Jurkat phosphoproteome data set, we identified phosphopeptides from multiple replicates, as well as from control and activated Jurkat cells (tables S2A and S3A). From a total of 65 such samples, we identified 11,708 unique phosphopeptides that mapped to 10,665 unique phosphorylation sites on 3084 proteins (fig. S5 and table S1). To ensure high confidence, the estimated false discovery rate (FDR) of phosphopeptide identification was kept below 0.15% in all the MS/MS analyses with composite, reversed-sequence database hits as a guide (16). A probabilistic score to estimate the certainty in the localization of phosphorylation sites was also calculated (16). Of the 10,665 phosphorylation sites, 66% had >95% certainty in their localization (fig. S6B). Further details of the experiments, enrichment of phosphopeptides, acquisition of mass spectrometry data, and data processing are provided in the Supplementary Materials. A summary flowchart of data processing steps is also provided (fig. S1).

To our knowledge, this data set of 10,665 unique phosphorylation sites on 3084 proteins is the largest from a hematopoietic and lymphocyte cell type. Seventy-four percent of these phosphorylation sites are newly identified, based on comparison to the phospho.ELM database (table S1) (17). We also defined T cell–specific proteins on the basis of large-scale mRNA expression data from the NCI60 panel of cell lines, and found 394 phosphorylation sites on 79 such proteins (table S12B). In addition, we found that classes of proteins implicated in hematopoiesis, lymphoma, or leukemia showed statistically significant overrepresentation in the total data set upon querying the Ingenuity Knowledgebase (table S12E). Our data set also includes 797 phosphorylation sites on 190 proteins implicated in the above pathophysiological processes (table S12A). Both Wnt and Notch proteins provide cues for hematopoietic lineage decisions, and we identified 127 phosphorylation sites on 38 proteins that are involved in Wnt signaling, Notch signaling, or both (18, 19). Overall, these observations emphasize the unique features and utility of our phosphorylation data set for future studies on proteins with restricted expression patterns and functions in adaptive immunity, hematopoietic lineage commitment, and leukemic and lymphomic transformation.

Identification of TCR-responsive phosphorylation sites by quantitative phosphoproteomics

We next conducted quantitative phosphoproteomic analyses on the basis of the stable isotope labeling of cells in culture (SILAC) method to identify previously unknown TCR-responsive phosphorylation sites and to gain system-wide insights into the role of phosphorylation during TCR signaling (12) (fig. S21). CD3, the signaling co-receptor complex of the TCR, was cross-linked with OKT3, an antibody specific for CD3ɛ, to initiate TCR signaling. Tyrosine phosphorylation of TCR-proximal proteins constitutes the first wave of events after TCR stimulation. Because pTyr residues constitute less than 1% of the total phosphorylated amino acid content, we specifically enriched pTyr-containing peptides with a pool of antibodies specific for pTyr residues (15) to identify TCR-responsive tyrosine phosphorylation sites in samples 5 and 15 min after stimulation of the TCR. A global phosphopeptide-enrichment approach based on cation-exchange chromatography (13) was also used for samples 15 and 60 min after stimulation of the TCR. A change in the abundance of phosphopeptides was quantified by calculating the relative intensities of light and heavy-isotope ion pairs. Phosphopeptides showing more than a 1.85-fold change in abundance after treatment with OKT3 were designated as containing TCR-responsive phosphorylation sites. A fold change of >1.85 represented a >99% certainty in fold change after stimulation, based on the variability in the quantification of isotope peak pairs (fig. S7A). Quantification was obtained from two technical replicates for each time point, and 93% of the TCR-responsive phosphopeptides that were identified in both the technical replicates had less than a 50% discrepancy in the fold changes determined (fig. S7B). In total, we identified 696 TCR-responsive phosphorylation sites on 453 proteins (table S4H), of which 70 were pTyr sites. Details on the calculation of fold changes in phosphopeptide abundance and the delineation of TCR-responsive phosphorylation sites are provided in the Supplementary Materials.

Validation of TCR-responsive phosphorylation sites

We first chose to validate the set of TCR-responsive phosphorylation sites by searching the literature, because TCR signaling is one of the most extensively investigated mammalian signaling systems and because Jurkat cells have been a widely used model for TCR signaling over the last 20 years (9). We captured 63 of the previously characterized TCR-responsive phosphorylation sites (Fig. 1, A and B, and table S8A). In addition, we also identified 60 previously uncharacterized TCR-responsive phosphorylation sites on proteins that play defined roles in TCR signaling (Fig. 1, A and B, and table S8D). However, we failed to detect phosphopeptides corresponding to 13 previously characterized sites that were amenable for identification by the MS/MS methods used. An additional 46 characterized TCR-responsive phosphorylation sites that we failed to capture were within tryptic peptides that were either shorter than 9 residues or longer than 29 residues and therefore not readily amenable for identification by collision-induced dissociation on ion trap mass spectrometers (table S8C) (20). Most prominent among these were eight sites on nuclear factor of activated T cells 1 (NFAT1) that are dephosphorylated by the phosphatase calcineurin (21). This emphasizes the need for the complementary use of other proteolytic enzymes and other ion-fragmentation methods to enable comprehensive discovery of in vivo phosphorylation sites (22). Overall, literature searches suggested that our large-scale phosphoproteomic experiments were reasonably successful in identifying TCR-responsive phosphorylation sites. It is also notable that most (~62%) of the already known and characterized TCR-responsive phosphorylation sites were on tyrosine residues, whereas ~86.5% of the new phosphorylation sites identified in this study were on S-T residues, thus underscoring the value of hypothesis-free phosphoproteomic approaches. As mentioned before, most of the known and characterized TCR-responsive phosphorylation sites are on receptor-proximal, upstream signaling proteins, proteins involved in mitogen-activated protein kinase (MAPK) cascades, and transcription factors (Fig. 1, A and B). In contrast, previously unidentified TCR-responsive sites were on proteins with roles as diverse as integrin activation, cytoskeletal remodeling, endocytosis, and alternative splicing of mRNA.

Fig. 1

Validation of TCR-responsive phosphorylation sites. Maps of TCR-responsive tyrosine phosphorylation sites (A) and S-T phosphorylation sites (B) identified on proteins involved in TCR signaling (see table S8 for references). The classes of phosphorylation sites are represented as colored dots: known and characterized TCR-responsive phosphorylation sites that showed quantitative changes in our study (red), known and characterized TCR-responsive phosphorylation sites that were not detected (gray), known and characterized TCR-responsive phosphorylation sites that were missed likely because of the presence of cysteine residues (brown), known and characterized TCR-responsive phosphorylation sites that were missed likely because of incompatibility with trypsin (blue), and previously unknown sites on proteins involved in TCR signaling that showed quantitative changes (in green). Details on these phosphorylation sites and their classification are given in table S8. Plasma membrane proteins, membrane-associated proteins, and nuclear transcription factors are depicted accordingly. Components of CD3 are shown at the centre of the plasma membrane, and inhibitory molecules are toward the left side of the map. Proteins that are excluded from the IS are placed at the extreme left of the map. Receptor-proximal signaling proteins and cytoskeleton-associated proteins are toward the right side of the map. Proteins that are part of MAPK cascades are at the bottom left part of the map. (C) Validation of six newly identified TCR-responsive phosphorylation sites by immunoprecipitation and targeted MS/MS. Names of proteins and the phosphorylation sites are given on the x axis. The fold change in the extent of phosphorylation as measured by targeted MS/MS of phosphorylated and corresponding nonphosphorylated proteins is compared with SILAC data. Product ions chosen for plotting the chromatograms are given in table S13.

Next, we quantified a handful of previously unknown TCR-inducible phosphorylation sites by targeted MS/MS. SH2 domain–containing tyrosine phosphatase 2 (SHP2), KRAB-associated protein 1 (KAP1), guanine nucleotide exchange factor H1 (GEF-H1), and B cell lymphoma 11B (BCL11B) were immunoprecipitated from control and OKT3-treated (15 min) cells. Phosphorylated and nonphosphorylated versions of the tryptic peptides containing the phosphorylation sites were subjected to targeted MS/MS (fig. S8). The extent of phosphorylation in both cell populations was estimated with these ion chromatogram pairs (23), and the fold change in the extent of phosphorylation was plotted (Fig. 1C). Overall, the quantification from targeted mass spectrometry was in agreement with the results obtained from SILAC experiments. Thus, the results from this independent biological replicate indicated that the SILAC quantifications obtained for previously unknown TCR-responsive sites were valid.

Label-free spectral count data provide a semiquantitative index of abundance. Statistical analysis of replicate data sets has typically been used for detecting differences in protein abundance (24). In this study, we evaluated the utility of replicate phosphopeptide spectral count data sets to discern changes in the extent of phosphorylation by comparing them with the SILAC data (fig. S9A). Statistical analysis with a confidence cutoff of 0.8 yielded a set comprising ~30% of the phosphopeptides that harbored TCR-responsive phosphorylation sites. Among these, more than 90% showed trends in spectral count differences that agreed with the SILAC data (fig. S9, B and C).

Analysis of global trends in the phosphorylation data set

Next, we looked for global trends in our phosphorylation data sets (fig. S10). Protein kinases represented the functional class of proteins that showed the highest enrichment in the Jurkat phosphoproteome when compared to the human genome database (fig. S10A). This is consistent with the trend observed in the yeast phosphoproteome and with numerous examples of the regulation of kinase activity by phosphorylation (22). Interestingly, the class of nonmotor actin-binding proteins showed the highest overrepresentation among proteins with TCR-responsive phosphorylation sites. This observation is elaborated upon later.

We classified the phosphorylation sites on the basis of neighboring residues and therefore by the putative upstream kinase recognition motifs (fig. S10B). The overall distribution pattern for the Jurkat phosphoproteome was similar to what was observed for the mouse liver phosphoproteome data set (13). Because extracellular signal–regulated kinase (ERK), p38 MAPK, and c-Jun N-terminal kinase (JNK) are all activated during TCR signaling, sites with the MAPK substrate motif expectedly showed the highest overrepresentation among the TCR-inducible phosphorylation sites. We also noticed that fewer sites increased and more sites decreased in phosphorylation status after 60 min than after 15 min of treatment with OKT3 (fig. S7C).

Putative substrates of ERK in TCR signaling

Defining the substrates of a kinase has often led to elucidation of its functions in a signaling pathway (2527). Despite the collective interest in identifying kinase substrates, progress has been slow (27). Here, we chose to identify putative substrates of ERK, which is essential for T cell activation (3). The activity of ERK increases during TCR signaling and is necessary for transcription of genes encoding cytokines such as interleukin-2 (IL-2) (28). To our knowledge, 90-kD ribosomal protein S6 kinase 1 (RSK1), MAPK signal–integrating kinases 1 and 2 (MNK1 and MNK2), ETS domain–containing protein (ELK1), and stathmin are the only established substrates of ERKs in the context of TCR signaling. We used U0126, a specific chemical inhibitor of the upstream kinase mitogen-activated or extracellular signal–regulated protein kinase kinase (MEK) together with OKT3 treatment to identify putative substrate sites of ERK during TCR signaling (fig. S12B). Frequency plots of residues surrounding the 54 putative ERK substrate sites defined in this study were similar to those of 143 known ERK substrate sites (fig. S12) (17). The newly identified putative ERK substrate sites belonged to proteins of different functional classes, indicating diverse roles for ERK in TCR signaling (table S6C). It is also notable that seven proteins with previously defined roles in TCR signaling and an expression pattern mostly restricted to T lymphocytes had putative ERK substrate sites (fig. S13A). Among these seven, we further validated that Thr260 on BCL11B was a substrate site for ERK by means of transient expression of activation-deficient and constitutively active MEK (fig. S13, B and C).

ζ chain–associated protein kinase of 70 kD (Zap70) is an essential upstream tyrosine kinase that relays TCR stimulation by phosphorylating its substrates. To our knowledge, 12 Zap70 substrate sites are known, but neither a consensus recognition motif nor the preferred neighboring residues have been defined. We developed a subtractive motif-driven strategy and predicted 43 putative Zap70 substrate sites, of which 13 showed an increase in phosphorylation upon treatment of Jurkat cells with OKT3 (fig. S14 and table S6E). Further experiments with reagents such as analog-sensitive Zap70 alleles will be needed to confirm these predicted substrates (29).

Modules of proteins with TCR-responsive phosphorylation sites

Next, we considered the proteins and their TCR-responsive phosphorylation sites on a case-by-case basis. We examined the functional roles of these proteins both in general and specifically in TCR signaling, if previously studied. We reasoned that the above approach might provide emergent themes on the potential role of phosphorylation in TCR signaling, as well as coherent hypotheses that could be experimentally tested. We associated classes of proteins with TCR-responsive phosphorylation sites to all of the salient phenomena related to T cell activation. The following section discusses these protein modules, their associated phenomena, and the plausible role of phosphorylation in coordinating these phenomena.

Transcriptional activation upon stimulation of the TCR

We first examined well-known cases of transcription factors involved in the expression of genes that encode cytokines upon activation of T lymphocytes, wherein the consequences of phosphorylation have been characterized (table S8). Phosphorylation of ELK1 on Ser383, Ser389, and Ser422 mediates its enhanced association with the histone acetyltransferases cAMP (adenosine 3′,5′-monophosphate) response element–binding protein (CREB)–binding protein (CBP) and p300, which recruit basal transcription factors to the target promoter (Fig. 2A and table S10). Similarly, phosphorylation of c-Jun on Ser63 and Ser73 mediates its enhanced association with CBP. In both cases, the phosphorylation sites were within interaction domains, and phosphorylation enhances protein-protein association, which results in transcriptional activation. Phosphorylation of serum response factor (SRF) on Ser103, which is proximal to the promoter recognition domain, results in its increased binding affinity to the serum response DNA element (table S8C).

In contrast, phosphorylation of Fos on Ser362, Thr325, and Thr331 leads to an intramolecular rearrangement and an increase in the half-life of the protein (Fig. 2A and table S10). Similarly, phosphorylation of activating transcription factor 2 (ATF2) on Thr51 and Thr53, which are within the N-terminal transactivation domain, results in the release of the transactivation domain from the C-terminal bZIP domain, exposure of the intrinsic histone acetylase domain, and an increase in the half-life of the protein. In the case of the retinoblastoma gene product RB, phosphorylation of Thr821 and Thr826 results in the intramolecular association of the phosphorylated RbC core region with the B-pocket region, thereby disrupting its interaction with ELF1 (Fig. 2A and table S10). Unlike the above examples, dephosphorylation of 13 serine and threonine residues in the regulatory domain of NFAT1 unmasks the nuclear localization signal in the regulatory domain that is required for its nuclear translocation (21). Thus, in each of these well-known cases, phosphorylation within the interaction region regulates intra- or intermolecular PPIs or protein-DNA interactions, resulting in net transcriptional activation (Fig. 2A and table S10).

Fig. 2

TCR-responsive phosphorylation sites in a module of transcription factors and corepressors. (A) Illustration of diverse modes of transcription factor activation by phosphorylation. Transcription factors activated during TCR signaling are shown (in blue) with red dots representing phosphorylation sites. The characterized phosphorylation sites and the consequences of phosphorylation are depicted. (B) Depiction of transcription factors (in blue) and their corepressors (in gray) involved in the transcription of Fos and Il2. (C) A model representing the likely role of TCR-responsive phosphorylation sites in the derepression of five transcription factors, which leads to the expression of Fos and Il2. Phosphorylation of Ser103 in SRF is intractable due to the presence of tryptic ends in the immediate neighborhood. Information on previously uncharacterized TCR-responsive sites is given. (D) Network of physical interactions among transcription factors and corepressors. Only those nodes representing proteins with TCR-responsive phosphorylation sites identified from this study are colored to indicate their functions. Interactions that are modulated by TCR-responsive phosphorylation changes are indicated by red edges. Interactions suggested to be modulated based on the presence of the newly identified TCR-responsive site(s) within or in close proximity to the region of protein binding are indicated by green edges. Information on such TCR-responsive sites is provided close to the node that represents the protein harboring the TCR-responsive site and along the edge that represents the interaction (see table S10). Information on most of the interactions was obtained from HPRD (www.hprd.org/).

With the above observation in mind, we examined TCR-responsive phosphorylation events on a large number of transcriptional regulators. Initially, we selected transcription factors known to bind to Fos and Il2 promoters, because these two genes are robustly transcribed after stimulation of the TCR (2). Among the transcription factors known to activate expression of Fos, IL2, or both, we found TCR-responsive phosphorylation of ELK1, Fos, c-Jun, BCL11B, and ELF1 (Fig. 2C and table S8D). As explained before, TCR-responsive phosphorylation sites on ELK1, Fos, and c-Jun increase their transcriptional activity; the previously unknown TCR-responsive phosphorylation sites that we observed on BCL11B and ELF1 may have a similar function.

Next, we examined the TCR-responsive phosphorylation of proteins known to repress the expression of Fos or Il2 or of corepressors of the above transcription factors: nuclear receptor corepressor 2 (NCOR2), Brahma protein homolog (BRG1), metastasis-associated protein 1 (MTA1), heterochromatin protein 1 homolog (HP1α), and RB (Fig. 2B and table S9A). As mentioned before, TCR-responsive phosphorylation of RB on Thr821 and Thr826 results in the intramolecular association of the phosphorylated segment, thereby displacing ELF1. Further, we noticed that TCR-responsive phosphorylation of NCOR2, BRG1, and HP1α was within, or proximal to, their regions of interaction with their target transcription factors: Ser2261 of the corepressor NCOR2 is within the domain responsible for its physical interaction with SRF, Fos, and c-Jun; Ser1382 of BRG1 is adjacent to the LXCXE motif that binds to RB; and Ser92 of HP1α is within the BCL11B-binding region (Fig. 2D, fig. S15, and table S10). Based on the known roles of these proteins and on known cases of modulation of PPIs by phosphorylation, we propose a model of transcriptional derepression (Fig. 2, B and C). In this model, TCR-responsive phosphorylation of transcriptional activators and repressors disrupts their associations with each other, which results in robust transcription of Fos and Il2.

We next sought additional support for the model of coordinated transcriptional regulation by inducible phosphorylation within interaction regions. In total, we found nine TCR-responsive phosphorylation sites that modulate PPIs; seven of these were within the respective interaction regions (Fig. 2D and table S10). Further, we found an additional 17 plausible cases in which the TCR-responsive phosphorylation site was situated within or proximal to the region responsible for interaction with another transcriptional regulator (Fig. 2D, fig. S15, and table S10).

Patterning of T cell proteins

Imaging studies have revealed that engagement of the TCR by the APC leads to dramatic redistribution of cell-surface and cell polarity proteins and that these redistributions of receptors are important for optimal downstream signaling (6, 30). The region of contact between the T cell and APC, called the immunological synapse (IS), is marked by three segments, namely, the central supramolecular activation cluster (cSMAC), the peripheral SMAC (pSMAC), and the distal SMAC (dSMAC). These segments are defined on the basis of the accumulation of the TCR complex, the lymphocyte function-associated antigen–1 (LFA1) complex, and CD45 in the cSMAC, pSMAC, and dSMAC, respectively (Fig. 3A).

Fig. 3

TCR-responsive phosphorylation sites in proteins involved in the patterning of receptors. (A) Schematic of protein redistribution patterns after TCR stimulation and APC engagement. Organization of the IS into cSMACs, pSMACs, and dSMACs for both x-z and x-y planes is shown. (B) Schematic of surface proteins of T cells that show patterning and the responsible proteins that connect them to the actin cytoskeleton along with their TCR-responsive phosphorylation sites. TCR-responsive phosphorylations are denoted by the red dots. Phosphorylation of Thr558 on moesin is intractable due to the presence of tryptic ends in the immediate neighborhood. Similarly Thr758 of the β2 integrin subunit was not identified (table S8B). (C) Primary sequence of these proteins depicting the TCR-responsive sites and the regions that mediate PPIs. The interacting protein is shown below the blue line that indicates the region that mediates the physical interaction. TCR-responsive sites that showed increased or decreased phosphorylation status are represented in red and green, respectively. Known consequences of the characterized phosphorylation events are noted. See table S10 for details.

Intercellular adhesion molecule–3 (ICAM-3) is thought to mediate the initial interaction between the APC and the T cell and is concentrated across the interface with the APC (table S8D). Further, bulky glycoproteins such as CD43 and P-selectin glycoprotein ligand 1 (PSGL1), and polarity proteins such as DLG1 and SCRIB are excluded from the IS to mark the rear end of the engaged T cell (Fig. 3A). Multiple studies have shown that these patterning events require an intact, dynamic actin cytoskeleton and the bridging molecules that link the receptors to the actin cytoskeleton (Fig. 3B and table S8D) (31, 32). Uniquely, in the case of CD3ζ, phosphorylation on Tyr153 promotes its direct association with actin microfilaments (33). In another case, phosphorylation of the bridging molecule moesin on Thr558 exposes its C-terminal actin-binding region and promotes the exclusion of CD43 from the IS (34). Similarly, we observed previously unidentified changes in phosphorylation of the F-actin–binding region of CD2AP and talin, which link CD2 and LFA-1, respectively (Fig. 3C, table S8D, and table S10). We also observed TCR-responsive phosphorylations within or proximal to the moesin-binding region of ICAM-3, CD43, and PSGL1, of which phosphorylation of Ser525 on ICAM-3 promotes its interaction with moesin (Fig. 3C). Further, we observed changes in the phosphorylation status of DLG1 and SCRIB, which show dynamic redistribution to and from the IS (table S8D) (35). These proteins are responsible for epithelial cell polarity and promote T cell polarity upon engagement of the APC. Overall, these observations point to a role for phosphorylation changes in promoting the interaction between proteins involved in receptor patterning that is observed after engagement of the TCR by the APC.

The LFA1 complex, which segregates to the pSMAC, is composed of α4 and β2 integrins (Fig. 3B). LFA1 is also activated by inside-out mechanisms to maintain adherence to the APC and provide costimulatory signals in the T cell (36). We observed TCR-responsive phosphorylation sites on 10 proteins involved in inside-out activation of LFA1 (table S9I). Among them, phosphorylation of L-plastin on Ser5 promotes activation of LFA1 (table S8A).

Polarization of microtubules and translocation of the microtubule organizing center

Engagement of the TCR by the APC leads to polarization of the microtubule network and the movement of the microtubule organizing center (MTOC) to the vicinity of the IS (Fig. 4A) (7, 37). These changes enable the directed secretion of effector molecules to the stably interacting target cell, thus ensuring specificity of immune responses. These changes are also important for the targeting of TCRs in recycling endosomes back at the IS where they can accumulate (38). Although not extensively investigated in the context of T cell activation, studies in various model systems point to conserved means of polarization of the microtubule network and reorientation of the MTOC (Fig. 4A) (39): Localized activation of Rho family guanosine triphosphatases (GTPases) leads to the capture of dynamic microtubules through their plus-end tracking proteins (+TIPs) at the plasma membrane. These captured microtubules are further stabilized by differential binding of microtubule-associated proteins (MAPs). The stabilized microtubules are marked by posttranslational modifications. Reorientation of the MTOC is driven by the sliding of these modified microtubules at the cell periphery due to minus-ended activity of anchored dynein motor proteins.

Fig. 4

TCR-responsive phosphorylation sites in proteins involved in the polarization of microtubules. (A) Schematic of polarization of microtubules after TCR stimulation. The steps involving selective microtubule capture through +TIPs, microtubule stabilization, and translocation of the MTOC are depicted. Representations of different classes of proteins and types of microtubules are given on the right. (B) Primary sequence representation of proteins denoting TCR-responsive sites and regions that mediate PPIs. The interacting protein is shown below the blue line that indicates the region that mediates the physical interaction. TCR-responsive sites that show increased or decreased phosphorylation status are marked in red and green, respectively. The functional roles of these proteins are also indicated as for (A). Phosphorylation of Tyr451 of tubulin α1 was not quantified by isotope labeling (table S8A).

We found TCR-responsive phosphorylation sites on several proteins that are important for the process of microtubule polarization: tubulin isoforms α1, α2, α6, and β1; the destabilizing MAP stathmin; two stabilizing MAPs, MAP1A and MAP4; two +TIPs CLASP2 and CLIP1; and the dynein motor components, DYNC1I2 and DYNC1LI1 (Fig. 4B). Among the known TCR-responsive events, phosphorylation of stathmin on Ser25 and Ser38 abrogates its binding to microtubules, which results in stabilization of microtubules (table S8A). Similarly, phosphorylation of α1 tubulin on Tyr451 abrogates its incorporation into microtubules and is thought to promote translocation of the MTOC (37). Similar to that of Tyr451, phosphorylation of Ser172 of β1 tubulin also abrogates its incorporation into microtubules (40). Further, we have observed that phosphorylation of Ser48 and Ser439 phosphorylation in α tubulin also prevented its incorporation into microtubles (see below). Thus, extensive phosphorylation of tubulins near the IS may promote translocation of the MTOC, as previously proposed (37). Among the other phosphorylation sites with known consequences, we identified two that were TCR-responsive. Phosphorylation of MAP4 on Ser1073 and of DYNC1I2 on Ser84 abrogated their interaction with microtubules and dynactin, respectively (Fig. 4B and table S10). Previously uncharacterized TCR-responsive phosphorylations of cytoplasmic linker protein 1 (CLIP1) and CLIP-associated protein 2 (CLASP2) may promote capture of microtubules at the plasma membrane and thereby promote stabilization of microtubules (table S9C). In all of the above cases, TCR-responsive phosphorylation sites were within the regions of PPI (Fig. 4B and table S10). We also observed changes in the phosphorylation states of nine other MAPs (table S9C). Among these, DIA1 is important for translocation of the MTOC in T cells upon their engagement by APCs (table S8D). Thus, it appears that TCR-responsive phosphorylation modulates the key subprocesses associated with microtubule polarization.

Formation of the F-actin cup at the IS

Engagement of the TCR by the APC leads to selective formation of F-actin, referred as the F-actin cup, beneath the IS (Fig. 5A) (4). A dynamic actin cytoskeleton is crucial for multiple events in an engaged T cell, such as receptor patterning and endocytosis, activation of integrins for adherence to the APC, influx of Ca2+, and other downstream signaling events necessary for the production of cytokines (7, 8, 31). Therefore, it is not entirely surprising that non-motor actin cytoskeletal proteins form the most overrepresented class among the proteins with TCR-responsive phosphorylation sites (fig. S10). Our compilation revealed that proteins with a wide repertoire of regulatory properties on the actin cytoskeleton, such as bundling, cross-linking, capping, severing, and branching showed TCR-responsive changes in phosphorylation state (Fig. 5B and table S9D). The known events include phosphorylation of Tyr291 of WASP, which increases its nucleation and branching activity, phosphorylation of Ser3 of cofilin1 (CFL1), which reduces its actin-depolymerizing and G-actin–sequestering activities, and phosphorylation of Ser5 of L-plastin (LCP1), which increases its avidity for F-actin (table S8A). Eleven of the proteins displayed in the actin network have previously been implicated in TCR signaling (Fig. 5B and table S9D). In addition, we found that 11 out of 26 proteins had previously unknown TCR-responsive phosphorylation sites within the region or domain responsible for their binding to F-actin (Fig. 5B). Further, we identified 15 TCR-responsive phosphorylation sites on 11 modulators of Rho family GTPases (table S9E). These findings suggest a widespread role for phosphorylation in regulating F-actin dynamics and remodeling of the actin cytoskeleton in T cells after stimulation of the TCR.

Fig. 5

TCR-responsive phosphorylation sites in proteins involved in remodeling of the actin cytoskeleton and endocytosis of the TCR. (A) Schematic of F-actin cup formation after TCR stimulation and APC engagement. Network of physical interactions among actin cytoskeletal proteins (B) and proteins involved in receptor endocytosis (C). Nodes representing proteins with observed TCR-responsive sites are distinctly colored to denote their relevant functions and previously characterized role in phenomena associated with T cell activation. Unigene names have been provided for the proteins. Proteins with multiple defined roles are represented as concentric circles. Interactions that are modulated by TCR-responsive changes in phosphorylation are marked in red. Interactions suggested to be modulated based on the presence of the previously uncharacterized TCR-responsive site(s) within or in close proximity to the region of protein binding are marked in green. Information on such TCR-responsive phosphorylation sites is provided close to the node that represents the protein harboring the TCR-responsive site and along the edge that represents the interaction. See table S10 for details.

Endocytosis of the TCR

Stimulation of the TCR leads to an increase in the rate of its internalization and subsequent degradation by clathrin-mediated endocytosis (41, 42). Ligand-induced internalization and degradation is mediated by ubiquitination of CD3ζ and phosphorylation of Ser148 on CD3γ (table S10) (43). We detected phosphorylation changes in five proteins involved in TCR endocytosis, of which phosphorylation of Tyr1477 of clathrin heavy chain promotes its assembly (Fig. 5C and table S9F). Clathrin-mediated endocytosis progresses through a series of subprocesses, namely, receptor recognition, assembly of clathrin, formation of coated pits, invagination of the membrane, fission of vesicles, movement away from the plasma membrane, and, finally, uncoating of the endosome (44). These events are also dynamically coupled to the actin cytoskeleton (45). The set of proteins that we have observed to undergo changes in phosphorylation are involved in all of the above processes except for uncoating of clathrin from endocytosed vesicles. These proteins also show PPIs among each other, and 10 of the 19 sites of TCR-responsive phosphorylation are within or proximal to the regions that mediate PPIs (Fig. 5C and table S10). Further, we have observed changes in phosphorylation of 12 other proteins that control receptor and membrane trafficking in endosomes and in vesicles targeted for exocytosis (table S9F).

Alternative splicing of mRNAs

To date, 22 mRNAs are known to undergo alternative splicing as a consequence of TCR signaling, among which functional consequences have been studied in the case of CD44, CD45, Fas, platelet endothelial cell adhesion molecule 1 (PECAM1), and cytotoxic T lymphocyte antigen 4 (CTLA4) (46). Alternative splicing is controlled by the SR family of proteins that recognize exon-intron splice enhancer or suppressor sequence elements on pre-mRNAs (5). Furthermore, these proteins also antagonize each other’s functions (47). Phosphorylation of SR proteins regulates their interactions with other SR proteins, cotranscriptional recruitment, and their shuttling between the cytoplasm and the nucleus (47). We have observed fold changes in the phosphorylation status of five SR family proteins after treatment with OKT3 (table S9G). These proteins interact with each other and, similar to proteins from previous modules, the phosphorylation sites are in domains that are critical for protein-protein or protein-RNA interactions (fig. S16). Of these five proteins, SFRS4 (Srp75), SFRS6 (Srp55), and SFRS8 (Swap) are implicated in the exclusion of exons 4, 5, and 6 in CD45 mRNA after T cell activation (table S8 and fig. S9G). In addition, we detected changes in phosphorylation of seven other proteins that are components of spliceosome complexes. Overall, phosphorylation appears to play a role in alternative splicing of mRNAs after stimulation of the TCR by modulating the interactions between proteins that control the selection of alternative splice sites and the splicing reactions.

Nuclear pore and nucleocytoplasmic transport

We detected an increase in the phosphorylation status of seven proteins involved in nuclear transport (fig. S17 and table S9H). A handful of studies have implicated phosphorylation in modulating rates of nuclear import of proteins in a stimulus-dependent manner. Nup214 and RanGAP1 are targets of MAPKs that regulate nuclear transport (48, 49); however, the sites of phosphorylation involved in the modulation of nuclear transport rates are not known. We have identified MAPK-dependent phosphorylation sites on RanGAP1 and Nup214, as well as other nucleoporins that show increased phosphorylation after TCR stimulation (table S9H). A study proposed that widespread phosphorylation of nucleoporins controls cell cycle–dependent assembly and disassembly of the nuclear pore (50). Our unanticipated observations predict a similar role for phosphorylation in stimulus-dependent modulation of nuclear transport.

S-T phosphorylation modulates PPIs in a system-wide fashion

Phosphorylation of tyrosine residues and their subsequent recognition by SH2 domain–containing proteins has been widely studied in the context of several signaling pathways. We were able to tabulate 60 tyrosine phosphorylation sites that have been previously studied in the context of TCR signaling, of which 48 mediate protein-protein association through an SH2 domain (table S10). During the case-by-case consideration of TCR-responsive phosphorylation sites, we also found 34 instances wherein S-T phosphorylation had already been demonstrated to modulate physical interactions between proteins (table S10). Of these 34 S-T phosphorylation sites, 24 were within the characterized binding regions of these proteins. We also found 61 cases in which the previously unknown TCR-responsive phosphorylation sites were located within the region of interaction among proteins involved in almost all of the subprocesses of the salient phenomena associated with T cell activation (table S10). These 61 novel TCR-responsive phosphorylation sites may modulate PPIs and thereby control the phenomena associated with T cell activation. Among these, 76% of the sites and the adjoining amino acid residues that determine kinase-specificity are conserved in all the tetrapod phyla, which indicates the importance of these phosphorylation events. These results suggest a system-wide role for S-T phosphorylation in modulating PPIs, far beyond the known roles of S-T phosphorylation motifs that are recognized by specific domains (51).

To assess the extent of the influence of S-T phosphorylation on PPIs by an independent approach, we performed network analyses of extensive protein-interaction information assembled in the Human Protein Reference Database (HPRD, www.hprd.org/). We noticed that S-T phosphorylated proteins were involved in more physical interactions than were proteins selected randomly from the PPI data set (P < 10−10; table S11). S-T-phosphorylated proteins were also more likely to interact among themselves than was expected by chance (P < 10−10; table S11 and Fig. 6A). These observations have also been made for yeast phosphoproteins in studies that used interaction data sets derived from unbiased large-scale studies (22). This indicates that our observations of human phosphoproteins indeed reflect the true scenario and not the likely overrepresentation of interactions for extensively studied proteins. Overall, there is a correlation between S-T phosphorylation status and the extent of interactions of a protein.

Fig. 6

Network analysis to assess the influence of inducible serine-threonine phosphorylation on PPI. (A) Comparison between the networks of interactions from randomly chosen phosphoproteins and the network of interactions among proteins with TCR-responsive phosphorylation sites. From a set of 1935 phosphoproteins that are in the HPRD, 326 phosphoproteins were randomly selected 150 times for a direct comparison with 326 proteins that have TCR-responsive phosphorylation site(s) that are also in the HPRD and that do not interact through SH2 domains or pTyr residues. The network properties were calculated with the “tYNA” tool (65). The following properties that describe the network of interactions among the chosen proteins were tabulated: degree, clustering coefficient, eccentricity, and betweenness (see main text for definitions). These parameters were averaged over all of the nodes of a network for plotting and comparison. The distribution and the averaged parameters of the random networks are shown. Network properties for variant networks obtained from proteins with TCR-responsive phosphorylation sites are also shown. P values estimating the statistical significance of TCR-responsive phosphorylation were empirically derived from the 150 random simulations. When a combined parameter derived from taking the product of degree, eccentricity, and betweenness was used, the networks of proteins with TCR-responsive phosphorylation sites showed an accentuated difference from the randomized networks (fig. S18A). (B) A representative network of interactions between 326 randomly chosen phosphoproteins. The largest connected component is on the left, whereas other, smaller components are on the right. Not all of the 326 proteins interact with other chosen proteins in the set, and hence these are not present in the network. (C) Network of interactions between proteins with TCR-responsive phosphorylation sites that do not interact through SH2 domains or pTyr residues. Nodes representing ERK1-2, RB, and c-Jun, which have the highest number of interacting partners, are shown in orange.

Further extending the network analysis, we asked whether 326 randomly chosen phosphoproteins would form networks similar to those of 326 proteins with TCR-responsive S-T phosphorylation sites. The following properties describing the network of interactions among the chosen proteins were compared: (i) “Degree” refers to the number of edges (interactions) of a protein (node). (ii) The “clustering coefficient” refers to the extent of connectivity among the first neighbors of a node. (iii) “Eccentricity” refers to the size of the network based on the longest of all of the “shortest path lengths” to a node. (iv) “Betweenness” refers to the capacity for information transfer based on the number of node pairs whose shortest path lengths pass through a node. Shortest path length refers to the smallest number of steps that are needed to reach one node from another among all of the possible connections through the edges. These parameters were averaged over all of the nodes of a network to define its properties. In the context of the entire network database, phosphoproteins with TCR-responsive S-T phosphorylation sites did not interact with more proteins than would be expected from randomly selected sets of phosphoproteins (table S11). Among the chosen phosphoproteins, those with TCR-responsive S-T phosphorylation sites formed a network with both a higher degree and betweenness than would be expected by chance (Fig. 6A). Even when four proteins, ERK1, ERK2, Rb, and c-Jun, which have the maximum number of interactions in the TCR-responsive set, were removed, betweenness remained high. Similar results were obtained when only the largest connected component was considered. All four network properties increased when additional interactions from the literature, which have not yet been incorporated into the HPRD, were included (Fig. 6A). Proteins whose interactions may be modulated by TCR-responsive phosphorylation may not harbor an observed TCR-responsive phosphorylation site. Therefore, we included common first neighbors of proteins with TCR-responsive S-T phosphorylation sites and observed higher betweenness in the enlarged network than would be expected by chance (fig. S18B). Comparison with random choice experiments clearly highlighted that TCR-responsive phosphorylations have a biological purpose. Average betweenness of nodes in a network represents the potential for information transfer in the network and is a relevant parameter in the analysis of signal transduction networks (52, 53). Our results suggest that changes in phosphorylation upon treatment of T cells with OKT3 represent information transfer along interacting proteins by means of modulating PPIs in a widespread manner. These results also provide further support to our interpretations from the case-by-case analysis presented earlier.

Biochemical evidence for the role of S-T phosphorylation in modulating PPIs during microtubule assembly

To experimentally test our inference on the role of S-T phosphorylation in modulating PPIs in a system-wide fashion, we chose the microtubule polymerization system as a test case. This system was chosen mainly for the following three reasons. First, extensive interactions between tubulin isoforms are central to the assembly of microtubules (54, 55). Second, OKT3-dependent changes in phosphorylation were observed on multiple isoforms of tubulin (Fig. 4B). Third, structurally and functionally equivalent microtubules can be experimentally assembled from isolated cell extracts with taxol. We tested whether phosphorylated forms of tubulins were selectively incorporated into or excluded from taxol-stabilized microtubules when compared to the remaining unpolymerized soluble tubulin (fig. S22). Through Western blotting analysis, phosphoserine and phosphothreonine residues were detected only in tubulin immunoprecipitated from the initial cytosolic extract and in unpolymerized soluble tubulin, but not in tubulin from polymerized microtubules (Fig. 7A). To confirm these observations, we performed detailed quantitative targeted MS/MS. Phosphorylated forms of tubulin were considerably lower in abundance in the polymerized microtubules than in the initial cytosolic extract (Fig. 7B and fig. S19A). In total, phosphorylation of six different serine residues resulted in the exclusion of the phosphorylated forms of tubulin from polymerized microtubules (Fig. 7B). Further, exclusion of phosphorylated forms of tubulin from polymerized microtubules is reflected by the enrichment of phosphorylated tubulin in the unpolymerized, soluble tubulin fraction compared to that of the initial extract. In the case of phosphorylation of Ser439 and Ser48 on tubulin α1, we generated phosphomimetic (Ser to Glu) and phosphorylation-deficient (Ser to Ala) mutants and compared their extent of incorporation into microtubules with that of the wild-type (WT) protein (Fig. 7C). Consistent with the above results, the phosphomimetic forms of tubulin α1 were less efficiently incorporated into taxol-stabilized microtubules than was WT tubulin α1.

Fig. 7

Phosphorylated forms of tubulins are excluded from polymerized microtubules. (A) Evidence for the exclusion of phosphorylated tubulins from microtubules by Western blotting analysis. Immunoprecipitated tubulin from the initial cytosolic extract (IT), unpolymerized soluble tubulin (ST), and polymerized microtubule (MT) were analyzed by Western blotting for the presence of pSer and pThr residues. Total tubulin on the same membrane was also determined. Polymerized microtubules lacked detectable S-T phosphorylation despite a >5-fold higher amount of tubulin on the membrane. (B) Evidence for the exclusion of phosphorylated tubulins from microtubules by quantitative and targeted MS/MS. The extent (%) of phosphorylation in the three samples was calculated from pairs of ion-chromatographic peaks of phosphorylated and nonphosphorylated tryptic peptides (see fig. S19A for an example). Product ions used to plot the ion chromatograms are shown in table S13. Triplicate measurements for seven phosphorylation sites are shown. Reduced amounts of phosphorylation on six distinct serine residues were seen in polymerized microtubules (MT), and the resultant enrichment of phosphorylated forms in the remaining soluble tubulin (ST) is also shown. The less than expected enrichment of β-tubulin phosphorylated on Ser172 and Ser105 in the “ST” sample may be due to an inaccuracy in the instrument response near the limit of detection. Detected tryptic peptides spanning the phosphorylation sites are common to more than one tubulin isoform but only one of their names is provided here. Only the phosphorylated form of Ser439 on TUBA8 was detected in all the three samples, thus acting as an internal control for equivalent detection of phosphorylated forms in all three samples. (C) Phosphomimetic mutation of Ser439 and Ser48 on tubulin α1 decreased its ability to be incorporated into taxol-stabilized microtubules. Western blotting of endogenous tubulin and different versions of green fluorescent protein (GFP)–tagged tubulin α1 (WT, S439E, S439A) in polymerized microtubule (MT) and unpolymerized soluble tubulin (ST) samples is shown. The incorporation index was calculated from densitometry of the Western blot. Western blotting analysis for the S48E and S48A mutants is shown in the supplementary material (fig. S19B). The incorporation index is the fraction of GFP-tagged tubulin in the “MT” sample over that of the “ST” sample, which is normalized to that of endogenous tubulin in the two samples. The unpolymerized soluble tubulin (ST) in the Western blot came directly from the supernatant of the microtubule-pelleting step and not from the subsequent immunoprecipitation step, as in previous cases. (D) A model depicting the presence of phosphorylated serine residues within the region of longitudinal contacts and lateral contacts between microfilaments of microtubules. This information was gleaned from crystal structure studies of microtubules (see main text).

We further examined whether these phosphorylation sites were within regions of PPIs in the microtubule structure (54, 55). We found that Ser48 of α-tubulin and Ser115 of β-tubulin were within the regions of lateral contacts between microtubule protofilaments (Fig. 7D and table S14). Similarly, Ser168 and Ser172 of β-tubulin were within the regions of longitudinal contacts in protofilaments and within regions interacting with the guanine nucleotide, which is necessary for the formation of heterodimers (56). These phosphorylation sites map to regions containing temperature-sensitive mutations in yeast tubulins, and touch-sensitive and polymerization-defective mutations in Caenorhabditis elegans tubulins (54, 55). Indeed, a paper reported the exclusion of Ser172-phosphorylated β-tubulin from microtubules during mitosis upon phosphorylation by cyclin-dependent kinase 1 (Cdk1) (40). Although Ser439 of α-tubulin is part of the C-terminal region that is disordered and projected outward from the microtubule, phosphorylation of Ser439 reduces incorporation of α-tubulin into microtubules for all but one isoform. This may be due to the disruption of the interaction between the C terminus of α-tubulin and the +TIPs, which chaperone the loading of tubulin onto growing microtubules (57). For example, the EB1 family of +TIPs have calponin homology domains that interact with microtubules through the C terminus of tubulins (58). These observations indicate that all modes of PPIs during the assembly of microtubules are abrogated by phosphorylation of serine residues in the region responsible for each of the specific modes of interaction (Fig. 7D). These observations also corroborate our inferences from the case-by-case analysis of functional modules of proteins and from our network analysis. Overall, these results provide evidence for the involvement of multiple serine phosphorylation events in modulating PPIs during the assembly of microtubules.

Discussion

We conducted large-scale phosphoproteomic experiments to gain system-wide insights into TCR signaling. We identified 10,665 phosphorylation sites, of which 696 showed TCR-responsive changes. These included 60 previously unknown TCR-responsive phosphorylation sites on proteins that have defined roles in TCR signaling, ~86.5% of which are S-T sites, underscoring the utility of large-scale, hypothesis-free experiments (table S8). The exact roles of these phosphorylations can now be investigated by targeted, hypothesis-based experiments. As detailed earlier, our main finding is that the scope of phosphorylation in TCR signaling is widespread and that it extensively targets proteins involved in all of the salient phenomena associated with T cell activation. Changes in the phosphorylation status of proteins that regulate the transcription of cytokine-encoding genes are suggestive of a model of coordinated disruption of interactions among transcription factors and their corepressors to induce robust transcription (Fig. 2C), whereas changes in the phosphorylation status of proteins that regulate the patterning of cell-surface receptors appear to promote their interactions with each other (Fig. 3B). Spatially restricted phosphorylation of tyrosine residues in the C terminus of α1 tubulin directly beneath the IS is proposed to promote translocation of the MTOC (37). It is thought that the reduced abundance of polymerized microtubules or increased frequency of microtubule catastrophes selectively in the cell-cell contact area act as dynamic guides for translocation of the MTOC. Similarly, we have observed that TCR-inducible phosphorylation of serine residues on multiple tubulin isoforms abrogates their incorporation into microtubules (Fig. 7). Thus, serine phosphorylation on tubulins may also contribute to translocation of the MTOC.

Overall, it was difficult to propose compelling and coherent hypotheses about protein modules because of multiple features of complexity, such as the dynamic nature of phosphorylation, spatial aspects of the processes, and the involvement of proteins in multiple steps of the processes. Nonetheless, we have highlighted nearly 150 proteins with previously unknown TCR-responsive phosphorylation sites that are part of diverse functional modules (table S9). The overriding, unifying, and emergent theme across all of the protein modules is the plausible influence of inducible phosphorylation on PPIs that control the diverse aspects and subprocesses of all of the phenomena associated with T cell activation. Analysis of the protein modules provided a list of 34 TCR-responsive S-T phosphorylations known to modulate PPIs, of which 24 sites were within interaction regions (table S10). We also found that 61 previously unknown TCR-responsive S-T phosphorylation sites were situated within or proximal to the region of interaction among the proteins belonging to relevant modules. We provide experimental evidence for the involvement of multiple S-T phosphorylations in modulating PPIs during the assembly of microtubules (Fig. 7). Phosphorylation of six serine residues within regions of longitudinal and lateral protofilament contacts and interaction with +TIPs results in the exclusion of phosphorylated tubulins from the microtubules. Further, we found that proteins with S-T phosphorylation sites interacted with more proteins and also tended to interact more among themselves than did proteins selected irrespective of known phosphorylation sites (Fig. 6 and table S11). From these analyses, we deduced that S-T phosphorylation modulates PPIs in a system-wide fashion.

Proteins with TCR-responsive phosphorylation sites not only had higher connectivity but also higher betweenness than would be expected by chance (Fig. 6A). This suggests that changes in phosphorylation represent information transfer along the interacting proteins and hence potential routes for signaling pathways. As an illustration, one can consider “protein A,” which upon phosphorylation gets released from a complex with “protein B.” Protein B now presents an exposed site, which upon phosphorylation allows “protein C” to bind to it. This allows for information transfer from protein A to protein C by changes in phosphorylation status and PPI. The above illustration is a simplified case and diverse variations can be expected in the transfer of intracellular information. For example, phosphorylation in a distant but spatially neighboring site may initially cause an intramolecular rearrangement needed for the subsequent binding or release of an interaction partner, as is the case with Rb and cortactin (59, 60). The actions of phosphatases may be relevant, as occurs with WASP (61). Kinase activity may be passive; that is, its basal activity is sufficient and phosphorylation is contingent on the exposure of the target site by a prior event, as with WASP. Multisite phosphorylations can add further complexity to the mode of interaction, as is the case with Cdc25 (62). Overall, phosphorylation and PPIs can influence each other and operate “hand in hand,” providing an elegant mode of information transfer at a systems level.

The extent of the influence of S-T phosphorylation on PPIs as deduced by our analysis is substantially more than would be expected from the current collective knowledge. We have corroborated the findings from our case-by-case analysis of TCR-responsive phosphorylation sites by performing network analyses of randomly chosen phosphoproteins (Fig. 6). This was important for many reasons. First, most of the TCR-responsive phosphorylation sites that we analyzed were in domains or large regions that contain tens or hundreds of residues that are responsible for PPIs. There is a large probability of finding an inducible phosphorylation site within a domain of interest by chance alone when the domain makes up such a large portion of the protein. Second, the spatial dimension was not considered in our analyses, albeit because of lack of information. Although the inducible phosphorylation site is within the domain of interest, it may not be a part of the interaction surface. Similarly, residues outside the domain may also contribute to the interaction surface. Third, systematically elucidating the role of a few additional phosphorylation sites does not substantially strengthen or weaken a proposal made at the systems level. Therefore, we chose to perform comparative analyses of networks with randomly chosen phosphoproteins.

As noted before, we observed TCR-responsive phosphorylation sites on more than 100 proteins that have defined roles in TCR signaling or on proteins controlling diverse processes that follow TCR stimulation. Network analysis suggests that changes in phosphorylation of proteins observed after CD3 cross-linking represent information transfer along the interacting proteins and, hence, potential routes for signaling pathways. Thus, results from our study along with other reports underscore the promise of phosphoproteomic approaches in providing impetus for studies of poorly characterized signaling systems (12, 63). Finally, system-wide modulation of PPIs by stimulus-dependent changes in S-T phosphorylation status may be a general phenomenon applicable to many other signaling systems.

The role of protein phosphorylation in modulating different stimulus-dependent cellular processes is emphasized in this report; however, it is likely that phosphorylation also plays an important role in basal, housekeeping processes. Most intracellular processes such as transcription, splicing, nucleocytoplasmic transport, and endocytosis can be thought of as cyclical events with many proteins participating in discrete biochemical steps, especially from the perspective of the protein machinery that carry out these processes. Alternating phosphorylation and dephosphorylation of the protein machinery can be used to drive these cyclical processes. This may provide an explanation for the existence of hundreds of phosphorylation sites in proteins of multimeric complexes such as the nuclear pore or the spliceosome: 141 phosphorylation sites on 20 nuclear pore proteins and 253 phosphorylation sites on 51 proteins involved in splicing (table S12F). Finally, our report may seem to suggest that every phosphorylation event has an intracellular consequence. But it is likely that inconsequential phosphorylation events may occur as bystander responses to interaction with other proteins, promoter elements, or second messenger molecules. In our view, these inconsequential phosphorylation events can be likened to silent mutations. Similarly, the extent of conservation of phosphorylation sites and their adjoining residues may indicate whether the phosphorylation event has a dominant consequence or not. Among the newly identified TCR-responsive phosphorylation sites that are implicated in modulating PPIs, 76% of the sites and their adjoining kinase-specificity determining residues are conserved in all the tetrapod phyla, which indicates the importance of these phosphorylation events (table S10). Nevertheless, even if a particular phosphorylation event is inconsequential, a change in the phosphorylation status of a protein recorded in a large-scale experiment is still likely to be indicative of involvement of that protein in the signaling pathway.

In addition to TCR-responsive phosphorylation sites, we identified a total of 10,665 phosphorylation sites. Particularly interesting is a subset of more than 1000 phosphorylation sites on proteins with restricted expression, function, or both in adaptive immunity, hematopoietic lineage commitment, and leukemic and lymphomic transformation. It is well established that deregulation of proteins controlling commitment of cells to a hematopoietic lineage causes cellular transformation (64). Querying the Ingenuity Knowledgebase brought to our attention that proteins implicated in hematopoiesis, leukemia, and lymphoma extensively interact with each other and regulate each other in multiple ways (fig. S20). Phosphorylation sites on these proteins may function to modulate interactions between proteins involved in hematopoiesis and influence the transformation processes, analogous to the TCR-responsive phosphorylation sites.

Materials and Methods

Details of all of the materials and methods used in this study can be found in the Supplementary Materials.

Acknowledgments

V.M. and D.K.H. designed research. V.M. conducted the experiments and processed and analyzed the data. D.H.L. and V.M. wrote new software routines needed for data processing and analysis. D.H.L. and J.K.E. provided support for the computational infrastructure. K.R., S.I.H., and L.W. acquired data sets that contributed ~2000 phosphorylation sites in the final Jurkat phosphoproteome data set of 10,665 phosphorylation sites. V.R. provided expertise and reagents for experiments on microtubules. V.M. and D.K.H. wrote the paper. D.H. dedicates this work to the memory of Richard D. Berlin. V.M. performed this work in partial fulfillment of the requirements for a Doctor of Philosophy degree from the Department of Cell Biology, University of Connecticut School of Medicine, Farmington, CT. We thank B. J. Mayor and T. Hunter for useful comments on the manuscript. We thank A. Alpert, J. Blenis, and I. Brodskiy for SCX chromatography columns, Mek constructs, and α1 tubulin constructs, respectively. We also thank S. P. Gygi, J. Villen, and S. A. Beausoleil for help with Ascore analysis. V.M. thanks his thesis committee members L. M. Loew, B. J. Mayer, K. P. Claffey, and G.-H. Fong for useful discussions. This work was supported by NIH Grants RO1 HL 67569 and PO1 HL 70694.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/2/84/ra46/DC1

Materials and Methods (incorporating Figs. S1 to S4)

Fig. S5. Accumulation of unique phosphopeptides in the Jurkat phosphoproteome.

Fig. S6. Distribution of the number of sites in phosphopeptides and the certainty in their localization.

Fig. S7. Summary of TCR-responsive fold changes in the abundance of phosphopeptides as determined by SILAC experiments.

Fig. S8. Illustration of the quantification of fold change by targeted MS/MS.

Fig. S9. Comparison of phosphopeptide spectral count data with SILAC data.

Fig. S10. Global trends in the phosphorylation data set.

Fig. S11. Abundance and subcellular localization of phosphoproteins based on protein spectral count data.

Fig. S12. Putative substrates of ERK during TCR signaling.

Fig. S13. Validation of Thr260 of BCL11B as a target of ERK.

Fig. S14. Predicted substrates of Zap70.

Fig. S15. Primary sequence representation of transcriptional regulators with TCR-responsive phosphorylation sites.

Fig. S16. Interactions among proteins involved in alternative splicing of mRNA that have TCR-responsive phosphorylation sites.

Fig. S17. TCR-responsive phosphorylation events in nuclear pore components.

Fig. S18. Network analysis to assess the influence of inducible phosphorylation of Ser or Thr residues on PPIs.

Fig. S19. Typical product ion chromatogram showing the lack of phosphorylation in polymerized microtubules.

Fig. S20. Network representation of nuclear proteins involved in hematopoiesis, lymphoma, and leukemia for which phosphorylation sites were identified in the current study.

Fig. S21. Schematic of the experimental work flow used to identify TCR-responsive phosphorylation sites by SILAC.

Fig. S22. Phosphorylated forms of tubulins are excluded from polymerized microtubules.

Descriptions of supplementary tables.

References

Tables S1 to S14

References and Notes

View Abstract

Navigate This Article