Combinatorial Proteomic Analysis of Intercellular Signaling Applied to the CD28 T-Cell Costimulatory Receptor

Combinatorial proteomic analysis of intercellular signaling applied to the CD28 T-cell costimulatory receptor

Ruijun Tiana,b,c,1,2, Haopeng Wangd,1, Gerald D. Gisha, Evangelia Petsalakia, Adrian Pasculescua, Yu Shie, Marianne Mollenauerd, Richard D. Bagshawa, Nir Yoseff, Tony Huntere, Anne-Claude Gingrasa,g, Arthur Weissd,h,2, and Tony Pawsona,g,3

aLunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, ON, Canada M5G 1X5; bDepartment of Chemistry and cShenzhen Key Laboratory of Cell Microenvironment, South University of Science and Technology of China, Shenzhen 518055, China; dDivision of Rheumatology, Department of Medicine, Rosalind Russell and Ephraim P. Engleman Rheumatology Research Center, and hHoward Hughes Medical Institute, University of California, San Francisco, CA 94143; eMolecular and Cell Biology Laboratory, Salk Institute for Biological Studies, La Jolla, CA 92037; fDepartment of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA 94720; and gDepartment of Molecular Genetics, University of Toronto, Toronto, ON, Canada M5S 1A8

Contributed by Arthur Weiss, February 20, 2015 (sent for review October 15, 2014) Systematic characterization of intercellular signaling approxi- than taking a more global approach. Moreover, cell–cell inter- mating the physiological conditions of stimulation that involve actions are often typified by the interaction of multiple trans- direct cell–cell contact is challenging. We describe a proteomic membrane receptor-ligand pairs, frequently involving membrane strategy to analyze physiological signaling mediated by the T-cell proteins that are not receptor kinases. Short-peptide motifs from costimulatory receptor CD28. We identified signaling pathways such receptors that serve as potential ligands for downstream activated by CD28 during direct cell–cell contact by global anal- targets can be used as affinity probes for binding partners, but ysis of protein phosphorylation. To define immediate CD28 tar- these experiments suffer from high false-positive and false-neg- gets, we used phosphorylated forms of the CD28 cytoplasmic ative rates (7, 8). Importantly, existing techniques are especially region to obtain the CD28 interactome. The interaction profiles unsuitable for investigating the common situation typified by of selected CD28-interacting proteins were further characterized cells of the immune system, in which antigen-dependent signal- in vivo for amplifying the CD28 interactome. The combination of ing is induced by direct cell–cell contact, rather than by soluble the global phosphorylation and interactome analyses revealed secreted factors. broad regulation of CD28 and its interactome by phosphoryla- This latter issue is exemplified by the observation that com- tion. Among the cellular phosphoproteins influenced by CD28 munication between two distinct cell types often can be con- signaling, CapZ-interacting protein (CapZIP), a regulator of the trolled by the interaction of two transmembrane or membrane- actin cytoskeleton, was implicated by functional studies. The tethered proteins (9). Under these circumstances, investigating combinatorial approach applied herein is widely applicable for signaling events requires that cells expressing an activating characterizing signaling networks associated with membrane membrane-associated ligand are cocultured with cells displaying receptors with short cytoplasmic tails. Significance intercellular signaling | proteomics | T cells | phosphorylation | signal transduction Intracellular signaling during complex cell–cell interactions, such as between immune cells, provides essential cues leading uring some forms of intercellular communication, soluble to cell responses. Global characterization of these signaling Dgrowth factors are secreted by one cell type and bind specific events is critical for systematically exploring and understanding transmembrane receptor tyrosine kinases (RTKs) present on how they eventually control cell fate. However, proteome-wide neighboring cells. Activated RTKs then undergo autophos- characterization of intercellular signaling under physiologically phorylation and also phosphorylate associated scaffold pro- relevant conditions involving multiple interacting receptors teins, thereby creating docking sites for effector proteins with during cell–cell interactions remains challenging. We developed phosphotyrosine (pTyr)-recognition modules such as Src ho- an integrated proteomic strategy for quantitatively profiling mology 2 (SH2) and pTyr-binding (PTB) domains (1). These intercellular-signaling events mediated by protein phosphoryla- immediate targets of growth-factor receptors control a variety tion and protein–protein interaction. We applied this approach of intracellular responses, typified by the activation of serine/ to determine the influence of a single receptor-ligand pair dur- threonine protein kinases such as those involved in the ERK ing T-cell stimulation by blocking the interaction of the CD28 MAP kinase pathway. Growth-factor signaling therefore leads costimulatory receptor with its ligand. This approach is gen- to extensive changes in both tyrosine and serine/threonine phos- erally applicable to other transmembrane receptors involved in phorylation (2). signaling during complex cell interactions. The alterations in protein–protein interactions and post-

translational modifications elicited by transmembrane receptors Author contributions: R.T., H.W., A.W., and T.P. designed research; R.T., H.W., G.D.G., E.P., can in principle be characterized by mass spectrometry (MS)- A.P., Y.S., M.M., and R.D.B. performed research; N.Y., T.H., and A.-C.G. contributed new based proteomics (3–6). However, under circumstances that in- reagents/analytic tools; R.T., H.W., G.D.G., E.P., A.P., Y.S., M.M., R.D.B., A.W., and T.P. analyzed data; and R.T., H.W., A.W., and T.P. wrote the paper. volve complex cell–cell interactions, identifying physiologically The authors declare no conflict of interest. relevant signals using MS-based proteomics can be challenging. 1R.T. and H.W. contributed equally to this work. For example, obtaining a sufficiently robust response for MS 2To whom correspondence may be addressed. Email: [email protected] or tian. analysis has often required the use of cells that overexpress [email protected]. a receptor of interest and are subjected to a nonphysiological 3Deceased August 7, 2013. stimulus. For this reason, experiments have often focused on This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. specific, often stereotypic, aspects of the cellular response, rather 1073/pnas.1503286112/-/DCSupplemental.

E1594–E1603 | PNAS | Published online March 17, 2015 www.pnas.org/cgi/doi/10.1073/pnas.1503286112 Downloaded by guest on September 29, 2021 its corresponding receptor. However, exploring the resulting signaling pathways regulated by a specific endogenous receptor PNAS PLUS signal transduction pathways is challenging because the identity activated by direct cell–cell contact. of the cell from which a particular protein is derived is lost after cell lysis. The T.P. laboratory developed a proteomic approach Results [termed quantitative analysis of bidirectional signaling (qBidS)] Global Phosphoproteomic Analysis of Endogenous CD28 Signaling. for profiling cell-specific tyrosine phosphorylation events re- We sought to develop a general proteomics approach to quan- sulting from contact between cells expressing either an Eph re- titatively measure global signaling events mediated by an en- ceptor tyrosine kinase or its ephrin ligand, respectively (10). In dogenous receptor activated by direct cell–cell contact. As a test these experiments, the EphB2 receptor or its ephrin-B1 ligand, biological system, we systematically explored how CD28-medi- a transmembrane protein that itself has intrinsic signaling ac- ated costimulatory signaling coordinates with TCR signaling for tivity, was stably expressed in human embryonic kidney (HEK) full activation of T cells by antigen-presenting B cells, using 293 cells so that the two cell types segregated in culture due to a combination of MS-based interaction-proteomics and phos- repulsive EphB2/ephrin-B1 signaling. The two distinct cell pop- phoproteomics. To specifically explore physiological CD28-de- ulations were then differentially labeled with stable isotopes pendent protein phosphorylation events, we used a coculture [stable isotope labeling by amino acid in cell culture (SILAC)] system involving Jurkat leukemic T cells and Raji lymphoblastoid and cocultured. This strategy not only can identify tyrosine B cells (Fig. 1A) (13, 14). Distinct combinations of SILAC la- “ phosphorylation events modulated after the interaction between beling were used to differentially label T cells with either me- ” “ ” EphB2 and ephrin-B1 but also can indicate from which cell type dium (M) or heavy (H) SILAC isotopes, and with B cells “ ” a specific phosphopeptide originates, and thus can infer path- being in an unlabeled light (L) state. After coculture of T cells ways activated in either cell type involved in EphB2/ephrin-B1 and B cells for 5 min, this differential labeling allowed us to bidirectional signaling. However, these experiments have limi- distinguish peptides and phosphorylation sites derived from B tations. One is that they depend on overexpression of receptors cells (L) or T cells (M or H), as well as those derived from T cells and ligands in a model-cell system. Another is that, once two that were fully stimulated by coculture with B cells (M) or T cells cells adhere to one another, they may interact through multiple in which the TCR (and other interacting receptor/ligand pairs that are not defined here) is engaged but CD28 is not (H) (Fig. different ligands and receptors, making it difficult to confidently A define the exclusive signaling output of a single class of receptor 1 ). Specifically, Jurkat leukemic T cells were stimulated by (9, 11). coculturing with Raji lymphoblastoid B cells presenting the Here, we overcome these issues by using a combination of superantigen staphylococcal enterotoxin E (SEE), which binds to proteomic and cell-biological approaches to systematically identify the TCR and to MHC class II molecules. Under these con- specific signaling events downstream of the endogenous CD28 ditions, the majority of the T-cell population (up to 88.2%) was efficiently activated, as indicated by potent ERK phosphoryla- costimulatory receptor on T cells. CD28 is a type I transmembrane tion, reflecting TCR signaling (Fig. 1B and SI Appendix,Fig.S1A); protein that binds through its extracellular region to B7 proteins consistent with this result, we observed T and B cells forming (CD80 and CD86), which are transmembrane proteins expressed a typical immunological synapse, with strong induction of tyrosine on the surface of antigen-presenting cells (APCs) and are up- phosphorylation at the site of synaptic contact in T cells (SI Appendix, regulated by inflammatory signals. Upon the interaction of T cells Fig. S1B). INFLAMMATION and APCs, the T-cell receptor (TCR) engages a peptide antigen- IMMUNOLOGY AND To selectively block CD28 signaling, we made use of cytotoxic bound major histocompatibility complex molecule on the APCs, T-lymphocyte antigen-4 (CTLA4), a coreceptor expressed on resulting in a primary signal. Simultaneously, costimulatory sig- activated T cells, which binds the same family of B7 ligands on B naling is established predominantly, but not exclusively, through cells as does CD28, but with more than 20-fold higher affinity the association of CD28 on T cells with B7 ligands on the surface (15). CTLA4-Ig, a CTLA4-Ig recombinant fusion protein that is of the APCs. Both TCR and CD28 signals are required for a full used in the treatment of rheumatoid arthritis, was added to T-cell response, resulting in, for example, production of the IL-2 specifically block the interaction of CD28 on Jurkat leukemic T cytokine. Although it lacks an intrinsic catalytic domain, the short cells with B7 ligands on the surface of Raji lymphoblastoid B 41 amino acid cytoplasmic tail of CD28 has highly conserved cells. Incubation with CTLA4-Ig completely inhibited CD28 tyrosine-based motifs that are phosphorylated by cytoplasmic ty- signaling in the coculture assay, even after 22 h, as suggested by rosine kinases upon T-cell stimulation and that bind targets with an efficient block in the production of the cytokine IL-2, a sig- SH2 domains in a pTyr-dependent manner, as well as proline-rich nature target of CD28-mediated costimulation (Fig. 1C). To sequences that potentially bind SH3 domains (12). It therefore can focus on proximal CD28 signaling events and minimize potential help orchestrate the response to the signal from an APC, such as secondary effects, a relative early time point after T-cell ac- aBcell. tivation was chosen. We were guided by the kinetics of ERK Because the complex cellular interactions of T cells and APCs phosphorylation after Raji-SEE stimulation was systematically stimulate multiple receptors (e.g., the integrin receptor) in ad- investigated. Maximal ERK phosphorylation was observed at dition to CD28, we have developed techniques to distinguish 5 min after T-cell stimulation (SI Appendix, Fig. S1A). Therefore, specific events uniquely dependent solely on CD28 signaling to 5 min was chosen to activate the Jurkat T cells in the phospho- undertake a comprehensive screen of tyrosine and serine/thre- proteomic analysis. Signals that are present in M-labeled T cells onine phosphorylation sites selectively regulated by CD28. We cocultured with SEE-presenting B cells, but are specifically have used a complementary approach using the full cytoplasmic blocked by treatment with CTLA4-Ig in H-labeled T cells, are tail of CD28 to identify the cytoplasmic targets with which it therefore dependent on B7-stimulated CD28. interacts. This approach involved immunoprecipitation coupled We sought to develop a sensitive MS assay to identify tyrosine to mass spectrometry (IP-MS) to identify a network of inter- and serine/threonine phosphorylation events regulated by en- acting proteins immediately proximal to CD28. We then in- dogenous CD28 on a global scale. To this end, we optimized a tegrated these data to dissect CD28-based signaling networks two-step phosphoproteomic approach involving tryptic digestion activated in T cells after coculture with APCs. We targeted one followed by either anti-pTyr IP or strong cation exchange (SCX) newly identified CD28-regulated phosphoprotein, CapZ-inter- fractionation and subsequent titanium dioxide (TiO2) enrich- acting protein (CapZIP), to demonstrate its importance in the ment (Fig. 1A). By combining these enrichment steps with the CD28-dependent functional response. Our results demonstrate use of a high-resolution Orbitrap Elite instrument with fast se- the value of a combinatorial and broad-based strategy to identify quence speed (16), we identified 936 phosphorylation sites by

Tian et al. PNAS | Published online March 17, 2015 | E1595 Downloaded by guest on September 29, 2021 Fig. 1. Exploring global phosphorylation events mediated by endogenous CD28 in direct cell–cell contact. (A) Schematic of the phosphoproteomic approach. (B) FACS analysis of ERK phosphorylation in resting Jurkat leukemic T cells (gray histogram) or Jurkat leukemic T cells stimulated by 30 ng/mL SEE presented by Raji lymphoblastoid B cells (solid line) after 5 min. (C) The ELISA analysis of cytokine IL-2 production by Jurkat leukemic T cells after their stimulation with Raji lymphoblastoid B cells and SEE at different concentrations in the presence of the indicated concentration of CTLA4-Ig. (D) Volcano plot of the global quantification of T cell-specific phosphosites. Each circle represents one site scaled by frequency of peptide detection. Three CD28 pTyr sites are indicated by an “X”.(E) The identified CD28 phosphorylation sites are shown with site position and localization probability. The relative positions of each measurement against noise distribution are indicated. The sites were labeled according to a full-length human CD28 sequence.

pTyr IP, of which 778 were pTyr sites, and 15,861 phosphoryla- Previous studies indicate that multiple signaling pathways are tion sites using the SCX/TiO2 workflow, of which 13,516 were downstream of CD28. Our pathway analysis revealed that many phospho-serines (pSers), and 2,184 were phospho-threonines proteins involved in known CD28-related signaling pathways (pThrs) (Dataset S1). Two biological replicates gave an overlap were identified by at least one phosphorylation site, and more of about 75% for sites identified by pTyr IP and 90% for the than 20 of those phosphorylation sites were significantly reduced SCX/TiO2 workflow, demonstrating reasonable reproducibility upon CD28 inhibition (Fig. 2A) (12). For example, we observed of both procedures (SI Appendix, Fig. S2A). The results obtained a remarkable reduction of phosphorylation of Y191 in the CD28 demonstrated high identification confidence and MS1 mass ac- cytoplasmic tail, which is known to recruit phosphatidylinositol curacy (SI Appendix, Fig. S2 B and C). 3-kinase (PI3K) through its SH2 domain, leading to PI3K acti- To delineate T cell-specific phosphorylation sites dependent vation. Consistent with this observation, phosphorylation of on CD28 costimulation, we applied stringent quantification fil- several proteins involved in the PI3K pathway, including PI3K, ters and show the results derived from every single MS spectrum. cAMP-binding protein 1 (CREB1), and mammalian target of This effort is mainly due to the fact that Jurkat leukemic T cells rapamycin (mTOR), changed after inhibiting CD28 signaling. and Raji lymphoblastoid B cells are completely different cell We also observed increased phosphorylation of filamin A types and that their proteome profile and expression level of the (FLNA), which interacts with CD28 and is involved in regulating same protein could be significantly different. Consequently, ac- cytoskeleton remodeling (17). Protein kinase C theta (PKCθ) curate measurement of the SILAC ratios between coeluted was recently reported to associate with CD28 and control NFκB SILAC-labeled peptides from distinct cell types could be chal- activation (18). Although we didn’t confidently detect any lenging. We strictly selected phosphopeptides that were reliably phosphorylation changes of PKCθ, the phosphorylation of its identified (Andromeda score > 42) and quantified in at least downstream signaling targets, CARMA1, as well as the tran- three MS/MS spectra from two biological replicates, and that scription factor NFkB, was reduced after CD28 blockade. We had negligible peak overlap, defined by the SILAC light peak observed dynamic CD28-dependent regulation of phosphoryla- overlapping into the medium isotope peak by less than 20%. tion-signaling events in fully activated T cells. For example, Furthermore, we used the H/L and M/L ratios (T cells/B cells) to phosphorylation of CD28-proximal signaling complexes [e.g., define and remove B cell-specific phosphopeptides. The cutoffs VAV1, phospholipase Cγ1 (PLCγ1), and SHC1] tended to be were selected based on removal of phosphopeptides from PYK2, consistently decreased whereas nuclear proteins such as NFAT which was predominantly phosphorylated in B cells as judged by transcription factors mostly exhibited increased serine/threonine manual inspection, and from six well-characterized B cell-spe- phosphorylation. Aside from known CD28-related signaling cific proteins, including CD19, LYN, BTK, CD22, BLNK, and pathways, the global phosphorylation analysis quantitatively PIK3AP1, which had significantly low H/L and M/L ratios. In identified phosphoproteins belonging to a broad spectrum of total, 8,208 T cell-specific phosphorylation sites were quantified, canonical signaling pathways with up to 66.7% coverage of all with a probability of being assigned to the correct residue of ≥0.33 pathway components (SI Appendix, Fig. S2D and Dataset S3). and where 6,267 sites have site probability of ≥0.75 (Dataset S2). Interestingly, most of the enriched signaling pathways within the The majority of phosphorylation sites exhibited a tight distribution 598 CD28-regulated phosphorylation sites were down-regulated around zero, indicating no significant change when CD28 signaling (Fig. 2B and Dataset S3). TCR signaling and a number of other was blocked, and demonstrating the high accuracy of the SILAC immune signaling pathways were down-regulated by CD28 quantification (Fig. 1D). A total of 598 sites showed significant blockade. It is noteworthy, however, that events associated with changes upon CTLA4-Ig treatment, with a P value (Wilcoxon test) the TCR signaling pathway did not dominate the down-regulated less than 0.01 and log2 ratio H/M cutoff at ±0.95 (98.5% and 1.5% events, suggesting that CD28 may influence events indepen- quartiles, respectively). dently of the TCR. These data provide a broad map of signaling

E1596 | www.pnas.org/cgi/doi/10.1073/pnas.1503286112 Tian et al. Downloaded by guest on September 29, 2021 PNAS PLUS

Fig. 2. Pathway analysis of the phosphoproteomics data. (A) Pathway map of CD28 signaling-related proteins. The proteins with phosphorylation sites significantly regulated by CD28 inhibition are indicated. The pathway map was based on a previous report (12). (B) Global canonical pathway analysis of

significantly regulated phosphoproteins. Additive log2 ratio H/M of all phosphorylation sites associated with a pathway are presented.

events specifically regulated by endogenous CD28 activated by cytoplasmic polypeptide was used to affinity purify CD28-bind- contact between Jurkat T and Raji B cells. ing proteins from T-cell lysates (Fig. 3A). The CD28 cytoplasmic tail contains only 41 amino acids, of which we synthesized a 33 Identification of a Protein-Interaction Network Involving CD28. The amino acid peptide that contains all of the highly conserved CD28-dependent pathways identified in the preceding phos- motifs, including two of the identified sites of tyrosine phosphoproteomic analysis are most likely controlled by proteins that phorylation, as well as proline-rich sequences (SI Appendix, Fig. are recruited to the evolutionarily conserved pTyr-containing or S3A) (CD28cyto). We anticipated that, using longer synthetic proline-rich motifs on CD28 (SI Appendix, Fig. S3A). Indeed, in peptides corresponding to almost the entire CD28 cytoplasmic the preceding analysis, we confidently identified three pTyr sites domain, together with appropriate modifications, would provide on the CD28 cytoplasmic tail that were significantly down- a physiologically relevant and comprehensive view of the receptor’s regulated upon CTLA4-Ig–mediated inhibition (Fig. 1E). In- immediate targets. To ensure that we identified phosphorylation- terestingly, two of these pTyr sites, Y191 (located in the motif dependent interactions, we synthesized several forms of CD28cyto

YMNM) and Y209 (located in the motif PYAPP) were relatively that were variously unmodified (YY), phosphorylated on either INFLAMMATION abundant and strongly inhibited by CTLA4-Ig treatment whereas Y191 or Y209 (pYY or YpY), doubly phosphorylated on both IMMUNOLOGY AND phosphorylation on Y218 was less abundant and more weakly Y191/Y209 (pYpY), or had alanine substitutions for Pro208 and inhibited. By IP and Western blot, we observed reduced re- Pro211 to destroy a potential PXXP SH3-binding motif (pYAA). cruitment of the p85α subunit of phosphatidylinositol 3′-kinase To examine phosphorylation-dependent protein interactions, (PI3K) to CD28 after CTLA4-Ig treatment, which corroborates the the various forms of biotin-tagged CD28cyto were immobilized on observation that phosphorylation of Y191 is down-regulated upon streptavidin agarose beads and incubated with SILAC-labeled CD28 inhibition because this residue is in a site that favors binding T-cell lysates, and associated proteins were identified by MS. For to the PI3K p85α SH2 domains (SI Appendix,Fig.S3B). We also quantitative purposes, we applied a three-channel SILAC ap- identified two phosphoserine and phosphothreonine sites on the proach that allowed us to perform two biological replicates in the CD28 cytoplasmic tail but with low site localization probability same experiment with reversed SILAC labeling (21). For the scores that don’t meet the data cutoff. Other function-related purpose of identifying pTyr-dependent interactions on both posttranslational modifications including ubiquitination would be Y191 and Y209, we did pulldowns using pYpY peptide with both possibly identified with a more dedicated workflow. SILAC light- and heavy-labeled T-cell lysates, while using YY To explore the targets immediately proximal to CD28, and peptide with SILAC medium-labeled T-cell lysate as a common thus to investigate how CD28 is connected to the spectrum of control (Fig. 3A). We thus obtained two replicate ratios of proteins whose phosphorylation is regulated by its signaling, we pYpY:YY peptide pulldowns as represented by H/M and L/M sought to systematically profile proteins that bind the major pTyr ratios (Fig. 3B and Dataset S4). This approach generated highly and Pro-rich sites on CD28 identified in the phosphoproteomics reproducible data between two biological replicates and nicely screen. The comprehensive identification of proteins associated differentiated pTyr-dependent interacting proteins from the with transmembrane receptors is technically problematic due to majority of other proteins that bound nonspecifically or to the the harsh protein extraction conditions needed to solubilize the nonphosphorylated YY peptide. Twenty-eight CD28-binding receptors for affinity purification, to which the relevant com- proteins were confidently identified, including 8 proteins pre- plexes are relatively labile (19, 20). Compared with alternative viously identified as associating with the CD28 cytoplasmic do- strategies of using short peptide motifs corresponding to receptor- main (Fig. 3C). Based on the accurate quantification afforded by docking sites, we expected that the “full-length” cytoplasmic do- SILAC, use of the various phosphorylated and mutant forms of main with physiologically relevant modifications could recruit CD28cyto identified three families of proteins that bound selec- more critical downstream proteins associated with an activated tively to either the pYMNM site (group I), the PXXPP motif receptor (7, 8). (group II), or the PpYAPP sequence (group III) of CD28 (Fig. 3 To attempt to capture a relevant view of the receptor’s im- B and C, and SI Appendix, Fig. S4 A–D). Western blots for 8 mediate targets and the role of the regulated pTyr sites in me- CD28cyto-associated proteins that contained either pTyr-binding diating interactions, we devised a proteomic approach to define SH2 domains or PXXP-binding SH3 domains also gave results the CD28 interactome in which a chemically synthesized CD28 that were consistent with the MS data (SI Appendix, Fig. S4E).

Tian et al. PNAS | Published online March 17, 2015 | E1597 Downloaded by guest on September 29, 2021 6 Synthetic CD28 A CD28 B PIK3CD C PIK3CA GADS cytoplasmic domain 5 PIK3R1 PIK3C1 PIK3CA 4 PIK3R2 GRB2 RASA1/VAV3 GRB2 PIK3CD PIK3C1 PLCG1/SHIP1 CSK/GADS PIK3R2 GRAP

3 PIK3R3 II YMNM

YMNM I

pY pY pY GRAP

YMNM PTPN11

P 2 STS1/ZAP70/CRKL

MNM MNM 191 MNM CIN85 VAV1 SH2D1A PIK3R1 PIK3R3 1 CD2AP LCK/THEMIS/STAT5

0 PYAPP

P SHIP1

PYAPP

P A

PYAPP -1 STS2 YA P pY

209 APP SLP76 CD28

APP P A STS1 -2 CBL -3 CSK STAT3 STAT5 Cytoplasmic domain YY pYpY pYY YpY pYAA -4 PTPN11 VAV3 III

n= 41 aa 2nd replicate) H/M (pYpY/YY Ratio -5 n= 33 aa 2 ARAP1 LCK -6 PLCG1 ZAP70 Log SH2D1A -7 -2-3-4-5-6-7 -1 0123456 VAV1 CRKL SILAC SILAC SIALC Log Ratio L/M (pYpY/YY 1st replicate) “Light” “Medium” “Heavy” 2

30 20 # of overlapped proteins

SILAC T cell lysates D -Log10 (p-value) 25 Proteins 16 pYpY YY pYpY 20 12 CD28cyto pull down 15 (p-value) 8 10 10 1:1:1 combine -Log 5 4

On-bead digestion 0 0

Mass spectrometry

Fig. 3. SILAC-based CD28cyto pulldown approach reveals phosphorylation-dependent interactome of CD28. (A) Schematic of the experimental workflow. (B) A 2D plot shows double phosphorylation-dependent interacting proteins to CD28. Positive hits are annotated. (C) Integrated map of the CD28 interactome. Proteins previously identified as interacting with CD28 are underlined. Motif-binding preference to pYMNM (I), PxxP (II), and PpYAPP motif (III) are highlighted in shaded circle. The width of the lines is scaled by the number of spectral counts. (D) The top 10 pathways enriched in the CD28 interactome and the number of overlapping proteins are indicated.

Pathway analysis revealed that more than 50% of the CD28 bound constitutively and selectively to the target, proteins whose interactors are involved in immune signaling pathways, including binding was induced by CD28 activation, and proteins that CD28 signaling in T-helper cells, demonstrating the biological bound nonspecifically. GRB2 was investigated because it is relevance of the CD28 interactome obtained by the in vitro known to interact with CD28 on both the YMNM and the CD28cyto pulldown (Fig. 3D). Our data recapitulated previous PYAPP motifs and to function as an adaptor to link receptors to observations demonstrating that the YMNM and PYAPP motifs multiple downstream signaling pathways (Fig. 4B). A group of are the two dominant motifs on the CD28 cytoplasmic tail (12), proteins that selectively, but constitutively, interacted with including the binding of PI3K family proteins as well as the GRB2 in T cells was reliably identified with high confidence; in GRB2 and GADS adaptors to the CD28 pYMNM motif and contrast, the association between GRB2 and CD28 required GRB2 SH3 domain-mediated interaction with the PYAPP motif. CD28 stimulation, consistent with an SH2–pTyr interaction. In addition, we identified seven previously unidentified inter- Among the constitutively GRB2-interacting proteins were CBL, actions to the PYAPP motif that are dependent on pTyr, the SOS1, and THEMIS (23). Suppressor of T-cell receptor signal- proline-rich sequence, or both. Because these newly identified ing 1 (STS1; gene name UBASH3B), a negative regulator of proteins contain SH2 or SH3 domains as GRB2 and GADS T-cell activation, is a potential atypical tyrosine phosphatase that adaptors, they likely bind to the CD28 cytoplasmic tail in an contains an SH3 domain that recognizes a proline-rich motif and indirect manner. However, a more dedicated experimental a UBA domain that binds ubiquitin (24, 25). The CD28cyto workflow is needed to confirm their binding modes. The PYAPP pulldown experiments have shown that STS1 binding required motif has been suggested to be indispensable for mediating T- both the PXXP and pTyr elements within the CD28 PYAPP cell functions such as proliferation and IL-2 secretion, in contrast motif. Interestingly, our IP-MS data showed that STS1 also to the YMNM motif (12). Therefore, our approach and findings interacted with GRB2 and that both of them bound to the E3 may be the basis for further experiments to study the mecha- ubiquitin ligases CBL and CBL-B (Fig. 4C). The association of nisms underlying the importance of these motifs. STS1 with CD28 was validated by immunoprecipitating endog- Like CD28, most immune receptors and ligands contain rather enous CD28 (SI Appendix, Fig. S5A). We also found that the short cytoplasmic tails (Dataset S5). Peptides of this relatively interaction between STS1 and GRB2 is enhanced upon the ac- short length, together with their posttranslational modifications, tivation of both TCR and CD28 signaling, demonstrating its can be readily synthesized using current technology, indicating potential recruitment to CD28-associated signaling complexes that the overall approach described above could have widespread through GRB2 (SI Appendix, Fig. S5B). applications. We further followed up experimentally an interaction hub that focused on CD28, GRB2, and CBL/CBL-B, which contains 14 A High-Resolution in Vivo CD28 Interactome Reveals Phosphorylation- proteins, including 8 identified by CD28cyto pulldown experi- Dependent Interaction Hubs. To validate and extend the signaling ments (Fig. 4D). By using IP-Western blot (IP-WB) and IP-MS network out from these CD28-binding partners identified by in approaches, we validated 12 pairs of protein–protein interactions vitro CD28cyto pulldown experiments, we performed IP-MS from within this interaction hub and found that 6 of these interactions Jurkat leukemic T cells (Fig. 4A). For this purpose, we used could be enhanced upon stimulation of both TCR and CD28 a piggybac transposon system to efficiently generate Jurkat leu- signaling (Fig. 4D and SI Appendix, Fig. S5B). Interestingly, we kemic T cells stably expressing specific 3xFlag-tagged CD28- found that the phosphorylation level of 14 phosphorylation sites binding proteins, such as GRB2 (22). These cells were then on nine proteins, as identified by our initial unbiased phospho- either left unstimulated or were stimulated by the anti-CD28 proteomic screen in cell lysates (Fig. 1 and Dataset S3), all antibody and lysed, and proteins associated with the Flag-tagged consistently decreased upon CD28 inhibition. Of these 14 polypeptide were identified by IP-MS. A three-channel SILAC phosphorylation sites, 10 are pTyr sites, which is consistent with approach was used to quantitatively distinguish proteins that the down-regulation of three pTyr sites on the CD28 cytoplasmic

E1598 | www.pnas.org/cgi/doi/10.1073/pnas.1503286112 Tian et al. Downloaded by guest on September 29, 2021 PNAS PLUS

Fig. 4. Expanding the CD28 interactome in vivo by both IP-MS and IP-WB approaches. (A) Schematic of the experimental workflow for the IP-MS approach. Triple Flag-tagged GRB2 (B) and STS1 (C) were stably transfected into Jurkat leukemic T cells and SILAC-labeled for pulling down interacting proteins. Stimulation-dependent interactions are labeled in red color. Bait proteins are labeled in bold. (D) Validated interaction map of a protein interaction hub

focused on CD28, GRB2, and CBL/CBLB. INFLAMMATION IMMUNOLOGY AND

tail. Docking protein 1 (DOK1) has been previously indicated to CD28 interaction network showed that immune signaling, in- be tyrosine-phosphorylated upon CD28 activation by antigen- cluding TCR signaling, is overrepresented and additively down- presenting cells (26). We characterized three phosphorylation regulated by CD28 inhibition (SI Appendix, Fig. S6B and Dataset sites with a decreased phosphorylation level on DOK1 (Y409, S3). The extended CD28 interaction network forms clear phos- Y449, and T406) and identified a stimulation-dependent in- phorylation-dependent interaction hubs around proteins such as teraction of DOK1 with Crk-like protein (CRKL), which po- GRB2, the PI3K family, the STAT family, CD2AP and CIN85, tentially links it to the CD28-mediated interaction network and CBL. Interestingly, two well-characterized CD28-interacting (SI Appendix, Fig. S5B). ADAP is a well-characterized adaptor proteins are most notable; GRB2 has extensive connections to 37 protein that positively regulates TCR signaling (27). We identi- newly recruited phosphoproteins whereas PI3K p85α (PIK3R1) fied a decreased phosphosite at Y813 of ADAP (Dataset S2) has broad associations with multiple components in the CD28 and showed that ADAP constitutively associated with CRKL interactome. These observations might explain a functional im- (SI Appendix, Fig. S5B). ODIN is a newly characterized protein portance of GRB2 as a key adaptor for regulating critical CD28- containing ankyrin repeats, a SAM domain, and a PTB domain associated downstream signaling. (28). Odin constitutively bound both GRB2 and CD2-associated protein (CD2AP) (SI Appendix, Fig. S5B). The phosphorylation Costimulation-Regulated Phosphorylation of CapZIP and Its Function level of ODIN Y445 strongly decreased upon CD28 inhibition, in Regulation of IL-2 Production. Our analysis of the phosphoryla- with log2 ratio H/M of −2.2 (Dataset S3). These results validate tion-dependent CD28 interaction network suggested that one of tight and consistent phosphorylation-dependent regulation of the main functions of CD28 signaling is to regulate actin dy- the CD28 interactome, demonstrating the usefulness of the namics (Fig. 5B). Consistent with this observation, it was recently combinatorial analysis of the phosphoproteome and interactome reported that Rltpr, an actin-uncapping protein, plays an es- for dissecting the CD28-mediated signaling network. sential role in CD28-mediated signaling (29). Rltpr contains To obtain a comprehensive view of the phosphorylation- a capping protein (CP) interaction (CPI) motif that interacts dependent CD28 interaction network, we queried our 28 CD28- with the CapZ protein, thereby inhibiting actin filament capping interacting proteins for their interactors in the STRING and and promoting actin polymerization. In our global phosphopro- BioGRID databases (Fig. 5A). Ninety-three proteins were teomics study, no significant change of RLTPR phosphorylation obtained that have regulated phosphorylation sites as identified was observed. However, the phosphorylation pattern of CapZIP, by the phosphoproteomic screen (Dataset S3). These proteins another member of the CPI motif-containing protein family, are highly enriched for protein functions such as intracellular- was significantly changed upon blockage of CD28 costimu- signaling cascades and actin-cytoskeleton organization (Fig. 5B lation (Dataset S2). CapZIP, highly expressed in immune and SI Appendix, Fig. S6A). Pathway analysis of the extended cells and skeletal muscle, can be phosphorylated by multiple

Tian et al. PNAS | Published online March 17, 2015 | E1599 Downloaded by guest on September 29, 2021 Fig. 5. Phosphorylation-dependent CD28 interactome. (A) An interaction map of the extended CD28 interaction network. CD28 and its 28 interacting

proteins are displayed in the outer ring whereas newly recruited interacting proteins are presented inside. The additive log2 ratio H/M of all regulated phosphorylation sites associated with a protein is presented. (B) Enriched Gene Ontology (GO) term analysis of the newly recruited phosphorylation- dependent CD28 interactome (see SI Appendix, Fig. S6A for an extended version). Representative enriched biological processes are presented.

stress-activated protein kinases (SAPKs) when cells are exposed phosphorylation changes that were affected by blocking CD28 to cellular stresses (30). A previous study suggested that phos- engagement was more extensive than anticipated. The global phorylation of CapZIP may cause the dissociation of CapZIP phosphorylation analysis using cell-based stimulation showed some from CapZ and thereby regulate actin-filament assembly (30). intersection of CD28-mediated signaling with TCR signaling but Among 18 CapZIP phosphorylation sites identified by our phos- also involved a number of other signaling pathways, some of which phoproteomic screening, three phosphoserine sites (S132, S135, have been associated with other immune receptors. and S333) were significantly changed upon CD28 inhibition, with Studying signal transduction events mediated by endogenous S132 and 135 showing significant down-regulation and S333 receptors under physiologically relevant conditions is crucial for showing significant up-regulation, suggesting that CapZIP might understanding fine-tuned signaling outputs and their cellular be involved in CD28-mediated costimulation (Dataset S2). To effects. Toward this end, we have designed our SILAC approach examine the role of CapZIP in the CD28-signaling pathway, we with the following considerations. We studied endogenously generated CapZIP-deficient cells from the Jurkat T-cell lines by expressed CD28 receptors during a physiologically relevant in- using CRISPR/Cas9 technology. The double-nicking strategy was tercellular interaction. We used CTLA4-Ig to specifically in- used to minimize off-target mutagenesis (Fig. 6A) (31). We suc- terrupt CD28 receptor activation, leaving signaling by other ceeded in generating four single-cell clones of CapZIP-deficient receptors, including the TCR, intact. This physiologically rele- −/− (CapZIP ) Jurkat cells (Fig. 6B). The deficiency of CapZIP did vant activation condition maintains the complexity of the natural not affect surface expression of CD3 and CD45 (Fig. 6C). Al- cell–cell communication. It has been well-characterized that −/− though two of four CapZIP single-cell clones had reduced a broad spectrum of different costimulatory and coinhibitory −/− surface CD28 expression level by up to 60%, the other CapZIP receptors and their ligands are engaged and activated during clones had CD28 expression that was indistinguishable from the TCR-dependent T-cell activation (9). The assumption of our parental Jurkat line. The reduction of CD28 expression was not study is that, in terms of T-cell activation, CD28 has some key, statistically significant (Fig. 6D). After stimulation with SEE pre- unique functions that either work independently or are in- −/− sented by Raji B cells, WT and CapZIP Jurkat cells up-regu- tegrated with other signals generated by TCR or other molecules lated comparable amounts of CD69, indicative of intact TCR involved in cell–cell interaction. In our experimental setting, the −/− signaling in CapZIP cells (Fig. 6E). CD69, a major activation only difference between two activation conditions is CD28 marker downstream of the Ras/MEK/ERK signaling pathway, can blockade. We acknowledge that other receptor-mediated signals be induced by TCR stimulation alone. However, the IL-2 may rely on or influence the CD28 signals. Our experiments −/− production in SEE-stimulated CapZIP Jurkat cells was nearly indeed aim to determine which of these pathways are CD28- −/− abolished (Fig. 6F). Notably, CapZIP Jurkat cells were able to dependent. produce similar levels of IL-2 compared with WT Jurkat cells, To explore the multidimensional nature of receptor signaling, when cells were stimulated by PMA plus ionomycin, which bypass a combination of different unbiased screening approaches on the TCR/CD28 proximal signaling (SI Appendix,Fig.S7). Col- a global scale is required. Our combinatorial proteomic ap- lectively, these data suggest that CapZIP is required for CD28 proach allowed the identification of signaling events associated costimulation-dependent IL-2 production. with CD28 costimulation through both a global phosphorylation analysis and a protein-interactome screen. The phosphopro- Discussion teome analysis provides a global view of CD28 receptor activa- Although the importance of CD28-mediated costimulation for tion across a broad spectrum of different downstream signaling T-cell activation has been explored for more than 29 y, the pathways. The synthetic cytoplasmic-domain pulldown approach downstream signaling pathways that are regulated by CD28 are captured CD28-associated protein complexes without protein still poorly understood. Current models derived from targeted solubilization issues. The combination of these phosphoproteo- experimental studies have yielded controversial views, especially mic approaches facilitated the study and recognition of phos- for the functional role of the PI3K pathway in CD28 signaling phorylation-dependent recruitment of protein complexes. (12). Our combinatorial proteomic analysis provides a window The CD28 interactome analysis was characterized by a spec- into the physiologically relevant CD28-regulated protein phos- trum of associated signaling proteins that are shared with other phorylation and its interaction networks. The extent of protein important immune signaling pathways. Prominent among these

E1600 | www.pnas.org/cgi/doi/10.1073/pnas.1503286112 Tian et al. Downloaded by guest on September 29, 2021 PNAS PLUS INFLAMMATION IMMUNOLOGY AND

Fig. 6. Generation and characterization of CapZIP-deficient cells. (A) Schematic of targeting human CapZIP exon 2 using Cas9 double-nicking strategy. The target regions of each sgRNA are labeled in blue, and PAM sequences are highlighted in red. (B) Immunoblot analysis with a CapZIP-specific or GAPDH-specific antibody in total cell lysates. Single-cell clones (clones 1–7) were generated from cells transfected with Cas9-nickase and sgRNAs targeting CapZIP. CapZIP protein level remained largely unchanged in clones 1–3 and was completely deleted in clones 4–7. (C) Representative FACS analysis of surface expression of CD3, CD45, and CD28 on resting CapZIP-sufficient cells (blue, clone 2) and CapZIP-deficient cells (red, clone 5). Gray-filled histogram, unstained control. (D) Bar chart showing CD28 surface expression level on resting CapZIP-sufficient cells (blue, WT Jurkat and clones 1–3) and CapZIP-deficient cells (red, clone 4–7). (E) Jurkat cells were stimulated by SEE (30 ng/mL) presented by Raji B cells. CD69 up-regulation on CapZIP-sufficient cells (blue, clone 2) and CapZIP-deficient cells (red, clone 5) was measured by FACS 22 h after stimulation. Gray-filled histogram, untreated control. (F) The ELISA analysis of cytokine IL-2 production by Jurkat cells 22 h after stimulation with Raji B cells and SEE at indicated concentrations. **P < 0.01, ***P < 0.001, n = 4, unpaired t test. The CapZIP-sufficient samples (WT Jurkat and clone 1–3) are labeled in blue, and the − − CapZIP / samples (clone 4–7) are labeled in red.

pathways are those that involve remodeling the actin cytoskele- proteome will provide an important starting point for many other ton and the PI3K pathway. As a preliminary effort to explore this studies aimed at understanding CD28 costimulatory signaling. dataset, we focused on the actin cytoskeleton because our recent Importantly, this SILAC-based phosphoproteomic approach is studies (32) and those of others (29) suggested an important role generally applicable for the study of other endogenous receptors for CD28 in regulating the actin cytoskeleton. The genetic study activated by direct cell–cell contact because blocking reagents of Liang et al. (29), which identified an important role of Rlptr, are available for most of the well-characterized receptors. With led us to focus on CapZIP. Ablation of CapZIP expression led to advances in synthetic biology, the synthetic cytoplasmic-domain impaired IL-2 production in our Jurkat-SEE/Raji system, which pulldown approach could, in principle, be applied to study a mimicked the effects of CD28 blockade but had no effect on broad spectrum of transmembrane receptors. As such, this com- TCR signaling, based on CD69 expression. Although the precise binatorial approach provides a widely applicable workflow to role of CapZIP is not clear and further work is required to un- systematically understand signaling events mediated by mem- derstand its role and that of its CD28-regulated phosphorylation, brane receptors with short cytoplasmic tails in more physiologi- these studies do support an important role for CD28 signaling in cally relevant contexts. actin-cytoskeleton regulation and have identified a previously Materials and Methods unappreciated role for CapZIP downstream of CD28. Our fur- Cell–Cell Stimulation and Sample Preparation for Phosphoproteomics. Medium ther systems-wide interactome analysis and experimental vali- and heavy SILAC-labeled Jurkat leukemic T cells were starved in SILAC media dation revealed a prominent GRB2-proximal interaction hub without dialyzed FBS for 12 h whereas two sets of equal amounts of un- that is widely regulated by CD28-dependent phosphorylation labeled Raji lymphoblastoid B cells were kept in normal culture. After that, and merits further exploration. Clearly, our CD28-regulated 1 × 108 medium and heavy SILAC-labeled Jurkat leukemic T cells were used

Tian et al. PNAS | Published online March 17, 2015 | E1601 Downloaded by guest on September 29, 2021 for the following steps. Two sets of 1 × 108 Raji lymphoblastoid B cells were spectrometers (Thermo). The obtained raw data were processed by Max- washed and resuspended into 1 mL of SILAC RPMI 1640. After adding 20 Quant (version 1.2.2.5) coupled with Andromeda for database searching ng/mL SEE or 20 ng/mL SEE plus 10 μg/mL CTLA4-Ig (final concentration), the against the International Protein Index (IPI) human database (version 3.79; Raji lymphoblastoid B cells were incubated at 37 °C for 30 min. Both Jurkat 91464 entries). The assignment of phosphorylation sites to tryptic peptides leukemic T cells and Raji lymphoblastoid B cells were then transferred to ice was performed automatically in the MaxQuant software environment with for 10 min, mixed together with equal volumes, quickly centrifuged at 500 × proper evaluation of the site localization probabilities (36). Evidence tables g for 5 min at 4 °C to promote cell–cell contact, and stimulated at 37 °C for containing all of the quantified peptide information were used to generate 5 min without resuspending the cell pellet, to promote cell–cell contact. After the final identification information as shown in Dataset S1. The R language that, the cells were lysed in lysis buffer [50 mM Tris·HCl, pH 7.4, 150 mM package was used to perform quantitative analysis of the global phos- NaCl, 1% Nonidet P-40, 0.1% sodium deoxycholate, 1 mM sodium ortho- phorylation dataset with data obtained from the Maxquant evidence tables. vanadate, protease inhibitors mixture (Complete mini; Roche), phosphatase The minimum Andromeda score was set as 42. In addition, the minimum site inhibitor mixture (PhosSTOP; Roche)], and the two sets of cell lysates were localization probability was 0.33. The pathway analysis was performed using then mixed together. The soluble proteins were centrifuged at 4 °C and gene annotations from the Kyoto Encyclopedia of Genes and Genomes were precipitated with four volumes of acetone at −20 °C overnight. The (KEGG) database (37) (P value cutoff 0.1; Fisher’s exact test) and Ingenuity protein precipitate was harvested by centrifugation at 12,000 × g,4°C,and Pathways Analysis (P value cutoff 0.05; Fisher’s exact test) (Fig. 3D). Further solubilized in 8 M urea sample buffer (8 M urea, 20 mM Hepes, pH 8, 1 mM interactome analysis related to Fig. 4D was performed based on STRING v9.1 sodium orthovanadate, 2.5 mM sodium pyrophosphate, 1 mM β-glycer- (38) (score cutoff 805) and BIOGRID (39). ophosphate) with mild sonication. The pellet fraction was solubilized in the For CD28 cytoplasmic domain pulldowns and IP-MS experiments, only same 8 M urea sample buffer containing benzonase. Total protein concen- proteins identified and quantified with at least two “Unique + Razor Pep- tration was measured by the BCA assay (Pierce). The obtained proteins were tides” were considered. Only proteins identified and quantified in at least reduced with DTT, alkylated with iodoacetamide, and digested with trypsin two out of three experiments as shown in Fig. 3B and SI Appendix, Fig. S4 A

as described (10). The obtained peptide sample was split for either immu- and B were considered as positive hits. The log2 ratio cutoff was set as 1. noprecipitation or strong-cation exchange (SCX) fractionation following by pYMNM motif- and PpYAPP motif-dependent interactions were defined titanium dioxide (TiO2) enrichment. For tyrosine-phosphorylated peptide by SILAC quantification with three-peptide pulldown combinations (SI enrichment, the peptide sample was dissolved in 100 mM Tris·HCl, pH 7.4, Appendix,Fig.S4A and B), which were cross-validated by YpY peptide 100 mM NaCl, and 0.3% Nonidet P-40 and enriched by anti-phosphotyrosine pulldown (SI Appendix, Fig. S4C). The result, as shown in SI Appendix, Fig. antibody 4G10-conjugated beads (Millipore). The rest of the sample was S4D, was used to define PxxPP motif-dependent interactions, and two se- fractionated into 24 fractions by SCX chromatography as described (33) and lected proteins (i.e., CIN85 and STS1) were further validated by Western blot enriched by a TiO2 kit (GL Sciences). (SI Appendix, Fig. S4E).

Flow-Cytometry Analysis and ELISA. Flow-cytometry analysis of phosphory- Western Blots. For IP-WB analysis of GRB2, CRKL, DOK1, CD2AP, CIN85, ODIN, lated ERK was performed as described (34). Briefly, after stimulation, cells and ADAP, 50 × 106 cells were starved in PBS for 20 min at 37 °C, and cells were were fixed in 4% (wt/vol) formaldehyde (Polysciences Inc.) and permeabilized stimulated with anti-TCR or anti-TCR plus anti-CD28 antibody for 5 min at 37 °C. in 95% (vol/vol) ice-cold methanol. Permeabilized cells were stained with The cells were lysed in Nonidet P-40 lysis buffer and incubated with anti-Flag a pERK antibody (Cell Signaling) followed by a PE-conjugated antibody M2 beads (Sigma) or anti-mouse IgG beads (Sigma) overnight. The beads were against rabbit IgG and APC-conjugated anti-CD3. The level of pERK in gated washed and boiled in 2× SDS/PAGE loading buffer for Western blot analysis. + CD3 populations was measured by an LSRFortessa cell analyzer. For the Western blot quantitation was performed with ImageJ software (version 1.47v). ELISA, the supernatants of the coculture system were collected 22 h after Protein complexes pulled down by conjugated CD28 cytoplasmic domain stimulation, and IL-2 production was measured by using a Human IL-2 ELISA or antibodies were boiled in 2× SDS/PAGE loading buffer, resolved on 10% Kit (BD Biosciences) according to the manufacturer’s instructions. SDS/PAGE gels, transferred to nitrocellulose, and analyzed.

CD28cyto Pulldown and Immunoprecipitation Assays. The CD28 cytoplasmic Generation of CapZIP-Deficient Jurkat Cell Lines. Two single-guide RNAs domain with biotin tag and various modifications was synthesized in-house, (sgRNAs) targeting the CapZIP gene were cloned into pX335 (Addgene). HPLC-purified, and conjugated on streptavidin agarose beads. Jurkat leu- Jurkat cells were cotransfected with these two sgRNA plasmids and a GFP + kemic T cells were SILAC-labeled with light, medium, or heavy lysine and reporter plasmid. At 24 h posttransfection, GFP live Jurkat cells were single 6 arginine. Then, 50 × 10 SILAC-labeled cells were lysed in Nonidet P-40 lysis cell sorted into a 96-well plate. Three weeks later, single cell clones were buffer [25 mM Tris·HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 1 mM so- expanded upon confluency. dium orthovanadate, protease inhibitors mixture (Complete minbi; Roche), phosphatase inhibitor mixture (PhosSTOP; Roche)] and incubated with dif- MS Data Storage. The raw MS data files associated with this study can be ferent types of beads overnight at 4 °C. The beads were washed four times downloaded from the Mass Spectrometry Interactive Virtual Environment · with washing buffer (25 mM Tris HCl, pH 7.4, 150 mM NaCl, 0.1% Nonidet (MassIVE) (massive.ucsd.edu/ProteoSAFe) via the following ftp access: ad- P-40) and two times with 20 mM ammonium bicarbonate at 4 °C. The as- dress, massive.ucsd.edu/ProteoSAFe; user name, CD28; password, CD28RAJI. sociated proteins were digested on the beads by trypsin as described (35) and combined for MS analysis. ACKNOWLEDGMENTS. We acknowledge K. Colwill, L. Taylor, B. Larsen, For IP-MS analysis of GRB2 and STS1, WT and triple Flag-tagged Jurkat C. Zhang, and M. Tucholska for technical help and J. Jin, Y. Zheng, C. Chen, leukemic T cells were SILAC-labeled as indicated in Fig. 2A. Then, 1 × 108 cells G. M. Findlay, and M. Kofler for helpful discussions. We thank A. Salomon were starved in PBS for 20 min at 37 °C, and the heavy SILAC-labeled and N. Krogan for constructive suggestions after their reading of this cells were stimulated with an anti-CD28 antibody for 5 min at 37 °C. The cells manuscript. We thank J. Bluestone, who provided the CTLA4-Ig. This work were lysed in Nonidet P-40 lysis buffer, precleared with mouse IgG beads was supported by grants from Ontario Research Fund GL2 (to T.P. and (Sigma), and incubated with anti-Flag M2 beads (Sigma) overnight. The A.-C.G.), the Canadian Institutes of Health Research (CIHR) (Grant MOP- 84314) (to A.-C.G.), the Howard Hughes Medical Institute (to A.W.), the NIH/ beads were washed and digested as described above. National Cancer Institute (Grant CA82683) (to T.H.), and a Shenzhen grant (ZDSYS20140509142721429) (to R.T.). R.T. is the recipient of a fellowship MS Analysis and Data Analysis. Samples were analyzed on LTQ-Orbitrap Elite from the CIHR. H.W. is the recipient of a postdoctoral fellowship from the US (for phosphoproteomics) or Classic (for part of CD28cyto pulldown) mass Arthritis Foundation.

1. Seet BT, Dikic I, Zhou MM, Pawson T (2006) Reading protein modifications with in- 6. Huttlin EL, et al. (2010) A tissue-specific atlas of mouse protein phosphorylation and teraction domains. Nat Rev Mol Cell Biol 7(7):473–483. expression. Cell 143(7):1174–1189. 2. Scott JD, Pawson T (2009) Cell signaling in space and time: Where proteins come 7. Schulze WX, Deng L, Mann M (2005) Phosphotyrosine interactome of the ErbB- together and when they’re apart. Science 326(5957):1220–1224. receptor kinase family. Mol Syst Biol 1:2005.0008. 3. Olsen JV, et al. (2006) Global, in vivo, and site-specific phosphorylation dynamics in 8. Jones RB, Gordus A, Krall JA, MacBeath G (2006) A quantitative protein interaction signaling networks. Cell 127(3):635–648. network for the ErbB receptors using protein microarrays. Nature 439(7073):168–174. 4. Bisson N, et al. (2011) Selected reaction monitoring mass spectrometry reveals the 9. Zhu Y, Yao S, Chen L (2011) Cell surface signaling molecules in the control of immune dynamics of signaling through the GRB2 adaptor. Nat Biotechnol 29(7):653–658. responses: A tide model. Immunity 34(4):466–478. 5. Gingras AC, Gstaiger M, Raught B, Aebersold R (2007) Analysis of protein complexes 10. Jørgensen C, et al. (2009) Cell-specific information processing in segregating pop- using mass spectrometry. Nat Rev Mol Cell Biol 8(8):645–654. ulations of Eph receptor ephrin-expressing cells. Science 326(5959):1502–1509.

E1602 | www.pnas.org/cgi/doi/10.1073/pnas.1503286112 Tian et al. Downloaded by guest on September 29, 2021 11. Acuto O, Michel F (2003) CD28-mediated co-stimulation: A quantitative support for 26. Michel F, Attal-Bonnefoy G, Mangino G, Mise-Omata S, Acuto O (2001) CD28 as PNAS PLUS TCR signalling. Nat Rev Immunol 3(12):939–951. a molecular amplifier extending TCR ligation and signaling capabilities. Immunity 12. Boomer JS, Green JM (2010) An enigmatic tail of CD28 signaling. Cold Spring Harb 15(6):935–945. Perspect Biol 2(8):a002436. 27. Griffiths EK, et al. (2001) Positive regulation of T cell activation and integrin adhesion 13. Fraser JD, Newton ME, Weiss A (1992) CD28 and T cell antigen receptor signal by the adapter Fyb/Slap. Science 293(5538):2260–2263. transduction coordinately regulate interleukin 2 gene expression in response to su- 28. Pandey A, et al. (2002) Cloning of a novel phosphotyrosine binding domain con- perantigen stimulation. J Exp Med 175(4):1131–1134. taining molecule, Odin, involved in signaling by receptor tyrosine kinases. Oncogene 14. Abraham RT, Weiss A (2004) Jurkat T cells and development of the T-cell receptor 21(52):8029–8036. signalling paradigm. Nat Rev Immunol 4(4):301–308. 29. Liang Y, et al. (2013) The lymphoid lineage-specific actin-uncapping protein Rltpr is 15. Scalapino KJ, Daikh DI (2008) CTLA-4: A key regulatory point in the control of au- essential for costimulation via CD28 and the development of regulatory T cells. Nat toimmune disease. Immunol Rev 223:143–155. Immunol 14(8):858–866. 16. Michalski A, et al. (2012) Ultra high resolution linear ion trap Orbitrap mass spec- 30. Eyers CE, et al. (2005) The phosphorylation of CapZ-interacting protein (CapZIP) trometer (Orbitrap Elite) facilitates top down LC MS/MS and versatile peptide frag- by stress-activated protein kinases triggers its dissociation from CapZ. Biochem J mentation modes. Mol Cell Proteomics 11(3):O111.013698. 389(Pt 1):127–135. 17. Tavano R, et al. (2006) CD28 interaction with filamin-A controls lipid raft accumula- 31. Ran FA, et al. (2013) Double nicking by RNA-guided CRISPR Cas9 for enhanced ge- tion at the T-cell immunological synapse. Nat Cell Biol 8(11):1270–1276. nome editing specificity. Cell 154(6):1380–1389. 18. Kong KF, et al. (2011) A motif in the V3 domain of the kinase PKC-θ determines its 32. Tan YX, et al. (2014) Inhibition of the kinase Csk in thymocytes reveals a requirement localization in the immunological synapse and functions in T cells via association with for actin remodeling in the initiation of full TCR signaling. Nat Immunol 15(2): CD28. Nat Immunol 12(11):1105–1112. 186–194. 19. Babu M, et al. (2012) Interaction landscape of membrane-protein complexes in Sac- 33. Villén J, Gygi SP (2008) The SCX/IMAC enrichment approach for global phosphoryla- charomyces cerevisiae. Nature 489(7417):585–589. tion analysis by mass spectrometry. Nat Protoc 3(10):1630–1638. 20. Humphries JD, et al. (2009) Proteomic analysis of integrin-associated complexes 34. Wang H, et al. (2010) Tonic ubiquitylation controls T-cell receptor:CD3 complex ex- identifies RCC2 as a dual regulator of Rac1 and Arf6. Sci Signal 2(87):ra51. pression during T-cell development. EMBO J 29(7):1285–1298. 21. Song C, et al. (2011) Improvement of the quantification accuracy and throughput for 35. Breitkreutz A, et al. (2010) A global protein kinase and phosphatase interaction phosphoproteome analysis by a pseudo triplex stable isotope dimethyl labeling ap- network in yeast. Science 328(5981):1043–1046. proach. Anal Chem 83(20):7755–7762. 36. Cox J, Mann M (2008) MaxQuant enables high peptide identification rates, in- 22. Kahlig KM, et al. (2010) Multiplexed transposon-mediated stable gene transfer in dividualized p.p.b.-range mass accuracies and proteome-wide protein quantification. human cells. Proc Natl Acad Sci USA 107(4):1343–1348. Nat Biotechnol 26(12):1367–1372. 23. Paster W, et al. (2013) GRB2-mediated recruitment of THEMIS to LAT is essential for 37. Kanehisa M, Goto S, Sato Y, Furumichi M, Tanabe M (2012) KEGG for integration and thymocyte development. J Immunol 190(7):3749–3756. interpretation of large-scale molecular data sets. Nucleic Acids Res 40(Database issue): 24. Kowanetz K, et al. (2004) Suppressors of T-cell receptor signaling Sts-1 and Sts-2 bind D109–D114. to Cbl and inhibit endocytosis of receptor tyrosine kinases. J Biol Chem 279(31): 38. Franceschini A, et al. (2013) STRING v9.1: Protein-protein interaction networks, with 32786–32795. increased coverage and integration. Nucleic Acids Res 41(Database issue):D808–D815. 25. Carpino N, et al. (2004) Regulation of ZAP-70 activation and TCR signaling by two 39. Stark C, et al. (2006) BioGRID: A general repository for interaction datasets. Nucleic related proteins, Sts-1 and Sts-2. Immunity 20(1):37–46. Acids Res 34(Database issue):D535–D539. INFLAMMATION IMMUNOLOGY AND

Tian et al. PNAS | Published online March 17, 2015 | E1603 Downloaded by guest on September 29, 2021