Protein kinase D regulates positive selection of CD4+ thymocytes through phosphorylation of SHP-1

Thymic selection shapes an appropriate T cell antigen receptor (TCR) repertoire during T cell development. Here, we show that a serine/threonine kinase, protein kinase D (PKD), is crucial for thymocyte positive selection. In T cell-specific PKD-deficient (PKD2/PKD3 double-deficient) mice, the generation of CD4 single positive thymocytes is abrogated. This defect is likely caused by attenuated TCR signalling during positive selection and incomplete CD4 lineage specification in PKD-deficient thymocytes; however, TCR-proximal tyrosine phosphorylation is not affected. PKD is activated in CD4+CD8+ double positive (DP) thymocytes on stimulation with positively selecting peptides. By phosphoproteomic analysis, we identify SH2-containing protein tyrosine phosphatase-1 (SHP-1) as a direct substrate of PKD. Substitution of wild-type SHP-1 by phosphorylation-defective mutant (SHP-1S557A) impairs generation of CD4+ thymocytes. These results suggest that the PKD–SHP-1 axis positively regulates TCR signalling to promote CD4+ T cell development.

A n appropriate ab T cell receptor (TCR) repertoire is shaped in the thymus through multiple selection steps. In this process, transduction of signals through the TCR in CD4 þ CD8 þ double positive (DP) thymocytes determines CD4/CD8 lineage specification and generates CD4 þ CD8 À and CD4 À CD8 þ single positive (SP) thymocytes. At the DP stage, interaction of the TCR with self-peptides on major histocompatibility complex (MHC) molecules generates positive-selecting signals. DP thymocytes that undergo positive selection increase their surface expression of CD5, CD69 and TCR and differentiate into a CD4 þ CD8 int transitional stage. At this stage, persistent TCR signalling promotes CD4 lineage specification through a series of transcriptional programs 1 . However, the molecular mechanisms by which the signal duration is translated into specific responses have yet to be fully defined.
It is well established that sequential tyrosine phosphorylation events triggered by protein tyrosine kinases (PTKs), such as Src-, Syk-and Tec-family PTKs, orchestrate TCR signalling during T cell development 2 . Serine/threonine kinases also regulate T cell development by controlling transcriptional and metabolic programs 3 . Until now, loss of several serine/threonine kinases has been reported to result in defective T cell development [4][5][6][7][8][9][10] . However, the crosstalk between serine/threonine kinase and tyrosine phosphorylation cascade is not clearly understood.
PKD, initially called PKCm, is a serine/threonine kinase family now classified within the CaMK group and separated from the AGC group (named for PKA, PKG and PKC) 11 . The three isoforms, PKD1, PKD2 and PKD3, are encoded by different genes, prkd1, prkd2 and prkd3, respectively. All PKD isoforms have a highly conserved amino acid sequence with a diacylglycerol (DAG)-binding domain, pleckstrin homology (PH) domain and kinase domain 11 . Among them, PKD2 is abundantly expressed in T cells 12,13 and is phosphorylated on TCR stimulation 14 . In addition, constitutively active PKD2 drives T cell development 15 . However, thymic selection was not grossly disturbed in PKD2-deficient mice 13 . Until now, no studies have addressed the role of PKD in thymocyte development using mice lacking all PKD isoforms, and the downstream substrates of PKD in thymocytes have not fully been elucidated.
Inhibitory signalling molecules, such as protein tyrosine phosphatases (PTPs), modulate TCR signalling and thus contribute to setting the threshold for thymocyte selection 16,17 . In particular, SH2-containing protein tyrosine phosphatase-1 (SHP-1) is reported to regulate thymic selection through a variety of proposed mechanisms [18][19][20][21] . The inhibitory activity of SHP-1 is regulated by binding to the phosphorylated targets via SH2 domains and by phosphorylation of its C-terminal tyrosines (Try 536 and Tyr 564 ) 17,22 . However, the role of serine/threonine phosphorylation in the function of SHP-1 remains unclear 23 .
In this study, we established mice lacking all PKD isoforms in T cells. The generation of CD4 SP thymocytes is impaired in these T cell-specific PKD-deficient mice. Using a phosphoproteomic approach, we identify SHP-1 as a direct substrate for PKD. The phosphorylation of SHP-1 by PKD positively regulates TCR signalling. Collectively, we demonstrate that the PKD-SHP-1 axis is required for CD4 þ T cell development.

Results
PKD is phosphorylated in preselection DP thymocytes. We first investigated whether PKD is activated on TCR engagement in thymocytes by evaluating phosphorylation of its activation loop. To address this, 'unengaged' DP thymocytes that have never encountered self-peptides were prepared. To obtain preselection thymocytes bearing antigen-specific TCR, bone marrow (BM) cells from OT-I TCR Tg mice were transferred into transporter associated with antigen processing (TAP)-deficient mice, which lack functional MHC class I. The preselection thymocytes were stimulated with OVA peptides that differ in affinity such as OVA 257-264 (SIINFEKL) (strong agonist), T4 (intermediate agonist) and G4 (weak agonist). To evaluate PKD activation, we established anti-phospho-PKD1/2/3 Abs that recognize phosphorylated activation loops containing a common sequence shared by all PKDs (Fig. 1a). Substantial PKD activation, as  Reactivities of anti-pPKDs were evaluated by western blot analysis of lysates from HEK293 cells overexpressing WT PKD2 and mutants. Note that overexpressed PKD2 is constitutively phosphorylated. Anti-PKD1/2/3 (anti-PKDs) blotting was also performed as a control. (b) Preselection OT-I DP thymocytes, that are obtained from TAP-deficient mice transferred with OT-I Tg BM cells, were stimulated with a variety of OVA peptides (10 mM) for the indicated times and phosphorylation of PKD and Erk was analysed. PKD, Erk and b-actin were used as loading controls. (c) RT-PCR analysis of expression of PKD isoforms in the thymus, spleen, ovary and lung from C57BL/6 mice. mRNA expression of b-actin was analysed as a control. Threefold serial dilutions of cDNAs were used as templates. (d) RT-PCR analysis of PKD isoforms expressions in various cell subsets sorted by flow cytometry from WT thymocytes. Results are presented as expression relative to b-actin. ND, not detected. (e) mRNA copy number was evaluated by molecular indexing assay. The amount of RNA used in the experiment is indicated on the x axis. Data are representative of two independent experiments (a-c). Data are presented as mean ± s.d. of triplicate assays and representative of two independent experiments (d,e). assessed by anti-phospho-PKDs Abs, was induced by low-affinity peptides (Fig. 1b), implying a role for PKD in positive selection.
We next examined the PKD isoforms expressed in thymocytes. Reverse transcription-PCR (RT-PCR) analyses confirmed that thymocytes express abundant PKD2, less PKD3 and undetectable PKD1 mRNA (Fig. 1c,d), as previously reported 13 . The lack of PKD1 expression in thymocytes was further verified by evaluating mRNA copy number using a molecular indexing assay (Fig. 1e) 24 . To generate T cells lacking all PKD isoforms, we established T cell-specific PKD2/PKD3 double-deficient mice using Lck-Cre Tg mice (PKD2/3 DT ) (Fig. 2a,b). We confirmed that none of the three PKD isoforms was expressed in PKD2/3 DT thymocytes using Abs reacting with PKD1, PKD2 and PKD3 (Fig. 2c,d).
Defective T cell development in PKD2/3 DT mice. In PKD2/3 DT mice, the proportion and number of CD4 SP thymocytes were markedly lower compared with wild-type (WT) mice, whereas those from single-deficient (PKD2 DT or PKD3 DT ) mice were not altered (Fig. 3a,b). These trends were more prominent in the TCR hi -gated mature CD4 SP population (Fig. 3a, lower panels).
Consistent with the reduction of CD4 SP cells, PKD2/3 DT mice had small and fragmented thymic medulla (Fig. 3c). The above results might be due to a possible survival defect of CD4 SP cells ( Fig. 3d; Supplementary Fig. 1), as Bcl-2 expression was reduced by half in PKD2/3 DT cells (Fig. 3e). However, this possibility seems unlikely as promotion of survival by a Bcl-2 transgene in PKD2/3 DT mice (Fig. 3f) failed to restore the proportion and number of CD4 SP thymocytes (Fig. 3g,h). Thus, PKD is likely to regulate CD4 SP cell generation beyond any effects on survival. A small number of CD4 SP thymocytes in PKD2/3 DT mice expressed lower levels of TCR and CD5 on the cell surface (Fig. 3i). This phenomenon was also observed in CD8 SP cells, suggesting that TCR signalling is impaired during positive selection in PKD2/3 DT mice.
Role of PKD in positive and negative selection. To examine the role of PKD in positive selection, we crossed PKD2/3 DT mice with MHC class II-restricted OT-II TCR transgenic (Tg) mice. In OT-II Â PKD2/3 DT mice, the percentage of CD4 SP thymocytes was reduced to one-tenth that of PKD-sufficient OT-II mice (Fig. 4a), demonstrating that PKD is critically involved in positive selection for the CD4 lineage. In the MHC class I-restricted OT-I TCR background, the proportion of mature CD8 SP thymocytes in PKD2/3 DT mice was reduced compared with that in control OT-I Tg mice (Fig. 4b), albeit less severe than in OT-II background. Furthermore, the expression of Tg-TCR was lower in PKD2/3 DT mice (Fig. 4a,b, lower panels). Analysis of female H-Y TCR Tg mice supported this finding, as CD8 SP development was impaired in H-Y Â PKD2/3 DT mice (Fig. 4c). Thus, positive selection for both the CD4 and CD8 lineages is impaired in PKD2/3 DT mice, particularly when the abTCR is fixed by transgenes.
We next investigated the function of PKD in negative selection using two different models. In male H-Y Tg Â PKD2/3 DT mice, the percentage and number of DP thymocytes were significantly higher than in control male H-Y Tg mice, suggesting inefficient negative selection in PKD-deficient thymocytes (Fig. 4d). Strikingly, male H-Y Tg Â PKD2/3 DT mice had CD8 SP thymocytes bearing high levels of the H-Y TCR (Fig. 4e). Furthermore, the peripheral T cells in male H-Y Tg Â PKD2/3 DT mice expressed normal levels of CD8, unlike low-level expression in WT H-Y controls (Fig. 4f). These results suggest that the loss of PKD results in a shift of the normal positive selection window towards high self-reactivity due to impaired TCR signalling in thymocytes.
For the second model, we used a clonal deletion model triggered by endogenous superantigens. Thymocytes bearing Vb5 þ and Vb11 þ TCR are deleted in the presence of mtv-8 and mtv-9 on I-E (Fig. 4g, H-2 25 . This deletion persisted in PKD single-deficient mice; however, it was significantly blocked in PKD2/3 DT (H-2 b/d ) mice (Fig. 4g, left panels). Irrelevant Vb6 þ cells were not affected by the absence of PKD (Fig. 4g, right panel). Thus, these two models reveal that PKD is involved in negative selection.
Collectively, our results indicate that PKD is critical for positive selection to the CD4 lineage and is also required for optimal positive selection to the CD8 lineage and negative selection.
Impaired CD4 þ CD8 int thymocyte generation by loss of PKD. DP thymocytes stimulated with positive-selecting peptides express CD69 and then differentiate into CD4 þ CD8 int intermediate cells, which are the precursors to both CD4 and CD8 SP thymocytes 1 . Although the number of CD69 þ DP cells was not altered in PKD2/3 DT mice, the CD4 þ CD8 int subpopulation was significantly decreased (Fig. 5a). The impaired generation of the CD4 þ CD8 int population was also evident in OT-I and OT-II TCR Tg backgrounds (Fig. 5b), indicating that PKD is required for the transition from the DP to the CD4 þ CD8 int stage. Consistent with this, the CD4 þ CD8 int population in PKD2/3 DT mice showed lower expression levels of CD5 than in WT mice, suggesting that loss of PKD led to the attenuation of TCR signalling during differentiation of DP to CD4 þ CD8 int stage (Fig. 5c).
To confirm this assumption, we conducted two-step differentiation assay in vitro. During recovery culture after crosslinking of TCR and co-receptor, DP thymocytes differentiate into CD4 þ CD8 int cells (Fig. 5d), as previously reported 26 . However, PKD2/3 DT DP thymocytes generated a much lower percentage of CD4 þ CD8 int cells than WT thymocytes after this treatment (Fig. 5d). Collectively, these results suggest that PKD regulates the TCR signals required for transition of DP to CD4 þ CD8 int thymocytes during positive selection.
PKD mediates signals triggered by weak-affinity peptides. To evaluate the strength of TCR signals, we performed quantitative assays using OVA peptide variants to stimulate thymocytes from TCR Tg mice. Ag peptide-induced co-receptor downregulation is a molecular indicator of the strength of TCR signalling in DP thymocytes 27 . Indeed, stimulation of thymocytes from OT-II TCR Tg mice resulted in co-receptor downregulation in accordance with the peptide affinity as seen using OVA 323-339 (strong agonist), R9 (weak agonist) and F9 (very weak agonist) (Fig. 5e, open circles) 28 . However, the responses were suppressed in PKD-deficient OT-II thymocytes, even on stimulation with the R9 variant (Fig. 5e, closed circles). Likewise, upregulation of CD69 and CD5 was also reduced in the absence of PKD (Fig. 5f). To further investigate, we stimulated thymocytes bearing OT-I TCR. Again, the co-receptor downregulation faithfully reflected the peptide affinity: namely, OVA 257-264 is the most potent, followed by T4 and G4 (Fig. 5g, open circles). We found that in the absence of PKD, the response was significantly attenuated when stimulated with a low-affinity variant, G4, which is known as a positive-selecting ligand (Fig. 5g, closed circles) 29 . These findings suggest that PKD is required for TCR signalling triggered by weak-affinity ligands, which is consistent with the above observation that positive selection is severely affected in PKD2/3 DT mice.
PKD is involved in CD4 lineage specification. As the generation of CD4 SP cells was more severely affected than CD8 SP cells in PKD2/3 DT mice (Fig. 3a,    should give rise to CD8 SP cells even in the absence of MHC class I (in an MHC class II-dependent manner), which is called 'CD8-redirection' (refs 30,31). To address this possibility, we transferred BM cells from WT or PKD2/3 DT mice to TAP-deficient mice. PKD-deficient BM cells, but not WT BM cells, gave rise to mature CD8 SP thymocytes even in TAP-deficient hosts lacking functional MHC class I (Fig. 6a), suggesting that PKD2/3 DT cells are defective in lineage commitment to CD4 SP thymocytes.
In fact, expression of ThPOK, a critical regulator of CD4 lineage commitment 32 , was decreased in PKD2/3 DT CD4 þ CD8 int cells, whereas Runx3, a transcription factor essential for CD8 lineage specification, was not decreased (Fig. 6b). Forced expression of ThPOK restored the percentage of CD4 SP thymocytes to nearly WT levels (Fig. 6c). However, the expression of the CD5 maturation marker in CD4 þ CD8 int cells remained low even in the presence of ThPOK transgene (Fig. 6d). Thus, these results suggest that PKD regulates the development of   CD4 SP cells through at least two mechanisms: (i) promotion of TCR signals in the DP stage that trigger transition to the CD4 þ CD8 int stage; and (ii) specification of CD4 lineage development via a ThPOK-dependent process.
SHP-1 is a PKD substrate in thymocytes. The above results demonstrate that PKD promotes TCR signalling during thymic selection. However, proximal signalling events including Ca 2 þ influx and Erk activation following TCR ligation were comparable between WT and PKD2/3 DT thymocytes (Fig. 7a,b). Furthermore, in DP cells, TCR-induced tyrosine phosphorylation of proximal signalling molecules, such as CD3z and ZAP-70, as well as MAPKs, such as Erk, p38 and JNK, was unaffected by the loss of PKD (Fig. 7c).
On the other hand, expression of a truncated PKD2 transgene lacking the kinase domain inhibited the generation of CD4 SP thymocytes (Fig. 7d), suggesting that the PKD kinase activity is required for thymocyte development.
Hence, to clarify downstream signalling events, we performed an unbiased proteomic search for novel PKD substrates in thymocytes. To identify proteins that exhibit TCR-induced phosphorylation in WT but not PKD-deficient thymocytes, phosphoproteins from unstimulated WT, stimulated WT and stimulated PKD2/3 DT thymocytes were labelled with Cy2 (blue), Cy3 (red) and Cy5 (green) fluorescent markers, respectively, and analysed by two-dimensional fluorescence difference gel electrophoresis (2D-DIGE) 33 . We detected six 'red' spots representing proteins that were phosphorylated in a stimulation-and PKD-dependent manner (Fig. 8a). These spots were identified to be pro-IL-16, SHP-1, NCK1 and Gads by mass spectrometry (MS) (Fig. 8a,b; Supplementary Data 1). SHP-1, NCK1 and Gads are reported to be involved in T cell development 17,34,35 . Phosphate-affinity SDS-PAGE (Phos-tag) western blotting 33 showed that the migration of SHP-1, NCK1 and Gads was retarded (indicating phosphorylation) in stimulated-WT thymocytes ( Fig. 8c; Supplementary Fig. 2). In PKD2/3 DT cells, the levels of phosphorylation of these proteins were reduced (Fig. 8c). In vitro kinase assays with recombinant PKD2 and PKD3 revealed that SHP-1, Gads and NCK1 are direct substrates for PKD2/3, and were phosphorylated at an efficiency similar to substrates previously identified in other tissues (Fig. 8d) 36 . We next used MS and site-directed mutagenesis, and identified the sites phosphorylated by PKD as SHP-1 Ser 557 , Gads Ser 186 and NCK1 Ser 85 (Fig. 8e, Supplementary Fig. 3 and Supplementary Data 2). Notably, these sites are similar in sequence to the PKD consensus phosphorylation motif, L/V/IxRxxpS/T (Fig. 8f) 37 .
These substrates were also phosphorylated in OT-I thymocytes on stimulation with various OVA peptides as assessed by Phos-tag western blotting. Among these substrates, the phosphorylation of SHP-1 was largely dependent on PKD2/3 (Fig. 9a). In order to further examine SHP-1 phosphorylation in vivo, tryptic phosphopeptides were enriched using TiO 2 from each  thymocyte. The liquid chromatography (LC)-MS data revealed that a phosphorylated peptide corresponding to amino acid residues 555-570 was increased by TCR stimulation in a PKD2/3-dependent manner (Fig. 9b, Supplementary Fig. 4 and Supplementary Data 3). Tandem mass spectrometry (MS/MS) analysis confirmed phosphorylation at Ser 557 of endogenous SHP-1 derived from stimulated-WT thymocytes (Fig. 9c). Finally, we developed an antibody that recognizes the phosphorylated Ser 557 in SHP-1. TCR stimulation increased the levels of phospho-Ser 557 SHP-1 in thymocytes from WT but not PKD2/3 DT mice (Fig. 9d). These results indicate that PKD mediates TCR-induced phosphorylation of Ser 557 in SHP-1. Importantly, the Ser 557 amino acid is highly conserved across mammalian species (Fig. 9e).
To further examine the role of the SHP-1 phosphorylation in T cell development, we turned to the DPK cell line, a pigeon cytochrome c (PCC)-specific DP cell line that retains the potential to undergo differentiation in vitro 38 . Stimulation of DPK cells with cognate peptide on I-E k resulted in downregulation of CD8 expression and differentiation into CD4 SP cells as previously reported 38 (Fig. 10a, Mock). Forced expression of a phosphorylation-defective mutant of SHP-1 (SHP-1 S557A ) reduced the generation of CD4 SP cells (Fig. 10a). Likewise, SHP-1 S557A also inhibited CD5 upregulation induced by peptide stimulation (Fig. 10b). These results suggest that SHP-1 S557A can suppress TCR signalling in a dominant negative manner and that phosphorylation of Ser 557 in SHP-1 promotes TCR signalling. We finally examined the in vivo relevance of the phosphorylation of Ser 557 in SHP-1 by establishing knock-in mice expressing mutant SHP-1 (SHP-1 S557A/S557A mice) (Fig. 10c). Thymocytes from SHP-1 S557A/S557A mice expressed comparable amount of SHP-1 to that of WT, whereas TCR-induced phosphorylation was impaired (Fig. 10d), indicating that Ser 557 is a major phosphorylation site on TCR stimulation. To address the role of this site, we evaluated the ability of SHP-1 S557A/S557A thymocytes to develop into SP cells in the context of competition with WT. CD3e-deficient (CD3e D5/D5 ) mice lacking T cells were transferred with mixed BM cells from CD45.1 þ WT and CD45.2 þ SHP-1 S557A/S557A mice and analysed after T cell reconstitution. SHP-1 S557A/S557A thymocytes showed impaired development into CD4 SP population compared with WT thymocytes (Fig. 10e), indicating that Ser 557 is required for the optimal generation of CD4 SP thymocytes. Collectively, these results suggest that the phosphorylation of Ser 557 in SHP-1 can positively regulate TCR signalling in DP thymocytes to promote T cell development.

Discussion
In this study, we established T cell-specific PKD-deficient mice to characterize the role of PKD in T cell development. PKD appears to control the generation of CD4 SP thymocytes through promoting (i) TCR signalling during positive selection, and (ii) CD4 lineage specification. Since the decreased number of CD4 þ CD8 int cells in PKD2/3 DT mice was restored by a Bcl-2 transgene (Fig. 3h), an important function of PKD during positive selection may be promoting survival that allows DP cells to differentiate into the CD4 þ CD8 int stage. On the other hand, defective CD4 lineage specification on loss of PKD is likely due to the incomplete induction of ThPOK during positive selection, as CD4/CD8 ratio was restored by ThPOK transgene, but not by Bcl-2. Optimal ThPOK induction requires 'duration' of TCR signalling during positive selection 39 . In contrast, the expression of Runx3, a transcription factor essential for CD8 lineage specification, is controlled by TCR-induced immediate upregulation of IL-7R and subsequent IL-7 signalling 40 . As Runx3 induction occurred normally in PKD-deficient mice (Fig. 6b), PKD is apparently dispensable for the transient TCR signalling required for Runx3 induction. In this respect, we speculate that another critical role of PKD may be to maintain the 'duration' of TCR/co-receptor signalling after sensing a weakaffinity ligand. Consistent with this idea, PKD was efficiently activated by stimulation with weak-affinity peptide variants in preselection DP thymocytes ( Supplementary Fig. 5).
PKD2 is abundantly expressed in thymocytes, while the amount of PKD3 is considerably lower. Nevertheless, the strikingly different phenotype of PKD2/3 DT versus PKD2 DT mice ARTICLE demonstrates that a limited amount of PKD3 is sufficient to support SP thymocyte generation. The redundant functions of PKD2 and PKD3 in T cell development suggest that their substrate specificities are likely to overlap. Indeed, all substrates tested in this study were phosphorylated by both PKD2 and PKD3 with similar efficiencies (Fig. 8d). The PKD consensus phosphorylation motif has been reported to be L/V/IxRxxS/T 37 . The new phosphorylation sites that we have identified herein are similar but not identical to this motif (Fig. 8f). This finding underscores the important contribution of unbiased phospho-proteomics 33,41 , in addition to motif-based approaches 42 , in identifying novel kinase substrates.
It has been suggested that SHP-1 dephosphorylates several key molecules involved in TCR signalling mainly through interactions with ITIM-containing transmembrane proteins 17 . Recently, Themis, a molecule critical for positive selection, was reported to recruit SHP-1 to antagonize TCR signalling, thereby preventing the apoptosis of DP cells that encountered weak agonists during positive selection 43 , although some debate exists 44 . Given that loss of Themis or PKD cause similar defects in T cell development, one might speculate that PKD also antagonizes TCR signalling to ensure positive selection. However, this possibility seems less likely, as the expression levels of CD5, CD69 (Fig. 5f) and Nur77 ( Supplementary Fig. 6), which represent the summation of TCR signalling intensity 45,46 , were reduced in PKD-deficient cells. This finding is consistent with the fact that active PKD induces upregulation of CD5 in thymocytes 15 . In addition, a 'selection shift' from negative selection to positive selection has been reported in mice with defective TCR signalling, such as ZAP-70-mutated mice or CD3z À / À mice 47,48 , similar to what we observe herein for PKD2/3 DT mice (Fig. 4e,f). In contrast, in Themis À / À mice, proximal TCR signalling is enhanced rather than attenuated 43 . Therefore, although PKD and Themis can both act on SHP-1, they appear to regulate TCR signalling in opposite directions. The generation of Themis À / À Â PKD2/3 DT triple-deficient mice will clarify this issue. What is the molecular basis for the positive signalling resulting from the phosphorylation of Ser 557 in SHP-1? There are at least three possibilities: First, the phosphorylation by PKD may constrain the inhibitory function, such as phosphatase activity, of SHP-1. Alternatively, the phosphorylation might alter the proper cellular localization of SHP-1 away from the area where efficient inhibition of TCR signalling takes place. Indeed, a 6-mer motif including Ser 557 in SHP-1 is reported to regulate its subcellular localization 49 . Second, we cannot rule out the possibility that serine-phosphorylated SHP-1 dephosphorylates proteins containing tyrosines that on phosphorylation, negatively regulate TCR signalling. For example, SHP-1 dephosphorylates the C-terminal regulatory tyrosine of Src family PTKs in osteoblasts to activate their kinase activities 50 . Third, SHP-1 may function as an adaptor rather than a phosphatase through its C-terminal phospho-Ser 557 , and may recruit positive regulators into the vicinity of the TCR. In every case, it would   be extremely intriguing to identify the protein(s) that selectively bind to Ser 557 -phosphrylated SHP-1. Johnson et al. 51 recently reported that T cell-specific SHP-1 deletion using Cd4-Cre Tg mice had no obvious effect on T cell development. This may be explained by the loss of both positive and negative regulation via SHP-1 in thymocytes lacking SHP-1 protein.
The observation that the defects found in SHP-1 S557A/S557A mice are milder than that of PKD2/3 DT mice could be explained by a possible contribution of other PKD substrates, such as Gads and NCK1. Indeed, co-transfection of phosphorylation-defective forms of three other PKD substrates (Gads, NCK1 and pro-IL-16) together with SHP-1 S557A resulted in more severe impairment in CD5 upregulation (Supplementary Fig. 7). Thus, we cannot exclude the possibility that PKD might control T cell development by modulating multiple TCR signalling pathways through various substrates in addition to SHP-1.
Histone deacetylase 7 (HDAC7) is also reported to be a substrate for PKD 52 and we confirmed this in in vitro kinase assay (Fig. 8d), whereas we could not detect PKD-dependent phosphorylation of HDAC7 in thymocytes by 2D-DIGE   ( Fig. 8a,b). HDAC7-deficient mice show poor generation of SP thymocytes 53 . However, the phenotype of PKD-deficient mice is unlikely to be attributed to HDAC7, as loss of HDAC7 results in unique phenotypes, such as limited usage of TCR Ja segments, which are not observed in PKD2/3 DT mice. In addition, in HDAC7-deficient mice, the introduction of TCR transgenes (OT-I or OT-II) reduced rather than exacerbated the defect in thymocyte development 53 , which contrasts with the phenotypes seen in PKD-deficient mice.
Currently, the pathway upstream of PKD during thymic selection is not clearly understood. PKD is activated by binding to DAG and, more importantly, by phosphorylation of serine residues in the activation loop (Ser 707/711 in PKD2 and Ser 731/735 in PKD3) 3 . In T cells, classical as well as novel PKCs, including PKCa, d, e, Z and y, are implicated in the phosphorylation of these residues 14 . Among them, PKCy and PKCZ are abundantly expressed in DP thymocytes, and recently, PKCy À / À PKCZ À / À double-deficient mice were demonstrated to have a decreased number of SP thymocytes compared with single-deficient mice 54 .
In this regard, one can speculate that PKDy and PKCZ act as upstream regulators of PKD after TCR engagement in thymocytes. This hypothesis warrants further investigation.
In conclusion, we propose that the PKD represents one of the molecular switches that converts weak TCR engagement to positive selection and CD4 lineage specification. The critical roles of PTKs and PTPs in thymic selection have been clearly established. The present study brings to light the additional importance of serine/threonine kinases as key modulators of TCR signalling cascade regulated by tyrosine phosphorylation during thymic selection.
Cell survival assay. Thymocytes were cultured in medium at 37°C with 5% CO 2 .
After incubation for different times, cells were stained with APC-conjugated anti-CD4, PE-conjugated anti-CD8, FITC-conjugated Annexin V at the concentration described in the 'Cells and reagents' section and propidium iodide at room temperature for 15 min, and immediately analysed by flow cytometry.
In vivo negative selection model. C57BL/6 (H-2 b ) mice were backcrossed onto B10.D2 (H-2 d ) mice and analysed for the expression of Vb5, Vb6 and Vb11. Thymocytes expressing Vb5 or Vb11 are deleted by binding of Mtv-8 and Mtv-9 superantigens only when I-E is also expressed. Thus, deletion occurs on H-2 b/d but not H-2 b/b backgrounds. Thymocytes expressing Vb6 are not deleted by these superantigens.
Two-step differentiation assay. In vitro differentiation of CD4 þ CD8 int thymocytes was performed using purified DP thymocyets 26 . Thymocytes in RPMI1640 medium supplemented with 10% (v/v) charcoal-and dextran-treated FBS were plated on 96-well plates at 5 Â 10 5 cells per well and stimulated for 20 h in wells coated with 5 mg ml À 1 anti-TCRb and 10 mg ml À 1 anti-CD2. Cells were washed and immediately analysed for CD4 and CD8 expressions by flow cytometry (stimulatory culture) or transferred to a new plate, washed extensively, and cultured for another 20 h in same medium, followed by the analysis for the expression of CD4 and CD8 by flow cytometry (recovery culture).
Co-receptor downregulation assay (Dulling assay). Co-receptor downregulation was examined using OT-II thymocytes or preselection OT-I thymocytes 27 . Overall, 1 Â 10 5 thioglycollate-induced peritoneal cells were used as APC and pulsed with peptides before the addition of thymocytes (1.5 Â 10 5 -5.0 Â 10 5 ) for stimulation of OT-II thymocytes. To stimulate preselection OT-I thymocytes, peptides were loaded on MHC class I expressed on thymocytes themselves. The expression of CD4 and CD8 was analysed by flow cytometry after 16-20 h.
Ca 2 þ mobilization. Thymocytes were freshly isolated and loaded with 1.2 mM Fluo-4 (Molecular Probes) in phenol red-free RPMI1640 containing 5% FBS for 30 min at 30°C. After washing, cells were stained with anti-CD3, phycoerythrinconjugated anti-CD8 and PerCP-Cy5.5 -conjugated anti-CD4 at the concentration described in the 'Cells and reagents' section for 30 min on ice. Cells were washed, suspended in PBS containing 5% FBS, prewarmed to 37°C before analysis and were kept at 37°C during event collection on FACSCalibur (Becton-Dickinson). For cell stimulation, various concentrations of goat anti-hamster IgG were added to crosslink anti-CD3 Abs. Maximum Ca 2 þ influx was obtained by the addition of 10 mM ionomycin (Calbiochem). Mean fluorescence ratio was calculated by FlowJo software (Tree Star).
Intracellular staining. For intracellular staining of phosphorylated proteins, thymocytes were stained with anti-CD3, anti-CD4 and anti-CD8 at the concentration described in the 'Cells and reagents' section for 30 min on ice and cross-linked by goat anti-hamster IgG as described above. Cells were washed, fixed and permeabilized using cytofix/cytoperm solution (BD Biosciences) for pCD3z, pErk, pp38 and pJNK staining or fixation/permeabilization buffer (eBioscience) for pZAP70 staining. Data were analysed by FACSCalibur flow cytometer and CellQuest software (BD Biosciences).
2D-DIGE. Thymocytes were freshly isolated and cultured for 16 h to upregulate TCR. Cells were then incubated with 10 mg ml À 1 anti-CD3 mAb for 30 min on ice and stimulated by cross-linking by 100 mg ml À 1 goat anti-hamster IgG Ab for 2 min at 37°C. The unstimulated WT, stimulated WT and stimulated PKD2/3 DT cells were lysed and phosphoproteins were enriched using PhosphoProtein Purification Kit (Qiagen). The interfering non-protein materials were subsequently removed by 2-D Clean Up Kit (GE Healthcare). Overall, 30 mg of the resultant phosphoproteins were minimally labelled with 400 pmol Cy2, Cy3 or Cy5 fluorescent dye (CyDye DIGE Fluors, GE Healthcare) for 30 min on ice. After quenching the labelling reaction with 10 nmol lysine, differentially labelled samples were mixed and subjected to first-dimension isoelectric focusing on immobilized pH gradient strips (24 cm; pH4-7) using an Ettan IPGphor II system. The second-dimensional SDS-PAGE was carried out on 10% acrylamide gel. The Cy2, Cy3 and Cy5 signals were individually acquired with a Typhoon 9400 scanner (GE Healthcare).
Phos-tag western blotting. Lysates from thymocytes were subjected to PAGE clean up kit (Nacalai tesque) before phosphate-affinity SDS-PAGE is performed using 6-7.5% acrylamide gels containing 25 or 50 mM Phos-tag acrylamide (Wako chemicals) and 50 or 100 mM MnCl 2 . After electrophoresis, gels were washed in Trans-Blot Turbo transfer buffer (Bio-Rad) added with 1 mM EDTA for 15 min and then incubated in the buffer without EDTA for 15 min. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes using Trans-Blot Turbo Transfer System (Bio-Rad) with the program for HIGH molecular weight proteins (1.3A, 25 V, 10 min) and analysed by immunoblotting. For l protein phosphatase (l PP) treatment, lysates were treated with l PP for 30 min at 30°C before purification using PAGE clean up kit.
Preparation of GST fusion proteins. SHP-1, Gads and NCK1 cDNAs were cloned into pGEX-6P-1 vector (GE Healthcare). For the mutant GST fusion proteins, Ser 557 and Ser 591 of SHP-1, Ser 186 of Gads and Ser 85 of NCK1 were changed to alanine by site-directed mutagenesis. The vectors were transformed into E. coli BL21 and GST fusion proteins were purified using glutathione sepharose (GE Healthcare).
In vitro kinase assay. For in vitro kinase assay using [g-32 P]ATP, 2 mg GST-SHP-1, GST-Gads and GST-NCK1 were incubated with 50 ng active PKD2 or PKD3 (Carna Bioscience) and 100 mM [g-32 P]ATP (5 mCi) in 20 ml kinase buffer (5 mM MgCl 2 and 20 mM Tris-HCl, pH7.5) for 30 min at 30°C. The reaction was stopped by addition of sample buffer followed by heating at 95°C for 5 min. Half of the sample was subjected to 5-20% gradient SDS-PAGE and CBB staining. Phosphorylated proteins were visualized by autoradiography.
Identification of proteins and peptides by LC-MS/MS analysis. The 2D-DIGE gel was silver-stained (Thermo Scientific) and protein spots were excised and digested with trypsin (Promega). The obtained peptides were analysed by a capillary liquid chromatography system (Waters/Micromass) connected to a Q-TOF Ultima mass spectrometer (Waters/Micromass). Raw data were acquired and processed using MassLynx version 4.0. (Waters/Micromass) to generate a peak list file for MS/MS ion search. The peak list files were searched against the NCBI non-redundant protein database restricted Mus musculus using the MS/MS ion search on Mascot search engine (Matrix Science). To identify in vitro phosphorylation sites, GST fusion proteins phosphorylated by active PKD2 or PKD3 were subjected to SDS-PAGE. The gel was stained with Coomassie Brilliant Blue and protein bands were excised and digested with trypsin. The resultant peptides were analysed by an ADVANCE UHPLC (Michrom Bioresources) connected to an Orbitrap Velos Pro mass spectrometer (for GST-Gads and GST-NCK1) or by an EASY-nLC1000 UHPLC (Thermo Fisher Scientific) connected to an Orbitrap Fusion Tribrid mass spectrometer (for GST-SHP-1).
Identification of phosphopeptides by LC-MS/MS analysis. The unstimulated WT, stimulated WT and stimulated PKD2/3 DT thymocytes were lysed in buffer containing 50 mM Tris-HCl, pH 8.0, 8 M urea, 1 mM DTT and PhosSTOP phosphatase inhibitor mixture (Roche). The lysates were sonicated and cleared by centrifugation at 20,000g at 4°C for 15 min. Proteins (each 50 mg) were reduced with 5 mM DTT for 30 min and alkylated with 27.5 mM iodoacetamide for 30 min in the dark. The urea concentration was then reduced to 1 M before addition of 0.5 mg trypsin. After 4 h at 37°C, the digestion was quenched by acidification with 0.5% TFA. The digests were desalted using InertSep RP-1 (GL Sciences), concentrated and adjusted to binding conditions for TiO 2 -based phosphopeptide enrichment (80% acetonitrile, 0.5% TFA and 300 mg ml À 1 lactic acid). Phosphopeptides were enriched using 50 mg of Titansphere Phos-TiO column (GL Sciences) in accordance with the manufacturer's instructions. Peptides were eluted with 200 ml of 5% ammonia and then with 200 ml of 5% pyrrolidine. The collected eluates were acidified with TFA and desalted using GL-Tip SDB (GL Sciences). LC-MS/MS analysis of enriched phosphopeptides was performed on an EASY-nLC 1200 UHPLC (Thermo Fisher Scientific) connected to a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific). The peptides were separated on an EASY-Spray column (15 cm Â 75 mm, PepMap C18, 3 mm) with a linear gradient from 4 to 28% ACN in 0-100 min followed by increase to 80% ACN in 100-110 min. The mass spectrometer was operated in a data-dependent acquisition mode with a top 10 MS/MS method. Raw data were converted to the mgf files using Proteome Discoverer version 2.0 (Thermo Fisher Scientific), which were then searched against the SwissProt database restricted to M. musculus using the MS/MS ion search on Mascot search engine. The false discovery rate was set to be o0.01. To generate extracted ion chromatograms, the raw data were processed using Xcalibur software (Thermo Fisher Scientific).
Data availability. The mass spectrometry proteomics have been deposited in the jPOST repository (https://repository.jpostdb.org) under the accession codes JPST000031, JPST000032 and JPST000033. All other data supporting the findings of this study are available within the article and Supplementary Information, and can also be obtained from the corresponding author upon request.