Critical role of WNK1 in MYC-dependent early thymocyte development

WNK1, a kinase that controls kidney salt homeostasis, also regulates adhesion and migration in CD4+ T cells. Wnk1 is highly expressed in thymocytes, and since migration is important for thymocyte maturation, we investigated a role for WNK1 in thymocyte development. We find that WNK1 is required for the transition of double negative (DN) thymocytes through the β-selection checkpoint and subsequent proliferation and differentiation into double positive (DP) thymocytes. Furthermore, we show that WNK1 negatively regulates LFA1-mediated adhesion and positively regulates CXCL12-induced migration in DN thymocytes. Despite this, migration defects of WNK1-deficient thymocytes do not account for the developmental arrest. Instead, we show that in DN thymocytes WNK1 transduces pre-TCR signals via OXSR1 and STK39 kinases and the SLC12A2 ion co-transporter that are required for post-transcriptional upregulation of MYC and subsequent proliferation and differentiation into DP thymocytes. Thus, a pathway regulating ion homeostasis is a critical regulator of thymocyte development.


Introduction
With No Lysine kinase 1 (WNK1) is a member of a family of four mammalian serine/threonine-specific protein kinases that are broadly conserved across evolution and are found in all multicellular organisms (Verissimo and Jordan, 2001). Mutations in WNK1 and WNK4 result in familial hypertension due to altered salt reabsorption in the kidney, because they regulate ion transport in kidney epithelial cells (Wilson et al., 2001). WNK kinases phosphorylate and activate the related OXSR1 and STK39 kinases, which in turn phosphorylate and activate the Na + K + Clco-transporters SLC12A1 and SLC12A2 and the Na + Clco-transporter SLC12A3 (Rafiqi et al., 2010;Thastrup et al., 2012), allowing Na + , K + and Clions to enter the cell. Furthermore, they phosphorylate and inhibit the K + Clco-transporters SLC12A4, SLC12A5, SLC12A6, SLC12A7 (Mercado et al., 2016), allowing K + and Clto leave the cell. Thus, the net effect of WNK kinase signaling is to promote movement of Na + , K + and Clions into the cell. Beyond its role in ion homeostasis, WNK1 has been proposed to regulate vesicular trafficking, proliferation and cell volume (de Los Heros et al., 2018;McCormick and Ellison, 2011).
Unexpectedly, we recently showed that signaling from both the T cell antigen receptor (TCR) and from the CCR7 chemokine receptor in CD4 + T cells lead to activation of WNK1 (Köchl et al., 2016). Furthermore, we found that WNK1 is a negative regulator of TCR-or CCR7-induced adhesion to ICAM1 mediated by LFA1. Conversely, WNK1 is a positive regulator of chemokine-induced migration through OXSR1, STK39 and SLC12A2. As a result, WNK1-deficient T cells home less efficiently to lymphoid organs and migrate more slowly through them. Thus, a pathway that regulates salt homeostasis in the kidney, also controls T cell adhesion and migration.
To generate T cells that have successfully rearranged both TCRa and TCRb genes and express an abTCR that is both restricted by self-MHC and self-tolerant, thymocytes need to pass three checkpoints (Carpenter and Bosselut, 2010). Thymic progenitors migrate from the bone marrow via the blood, entering the thymus at the cortico-medullary junction as DN1 cells. DN2 and DN3 thymocytes begin to re-arrange TCRb genes and migrate to the sub-capsular zone of the cortex. If rearrangement is successful, TCRb protein binds to pre-Ta and together with CD3g, CD3d, CD3e and CD3z forms the pre-TCR (Yamasaki and Saito, 2007). Signals from the pre-TCR are required for the survival and proliferative expansion of DN4 thymocytes, and subsequent differentiation into ISP and DP cells, a checkpoint termed b-selection (Kreslavsky et al., 2012;Mingueneau et al., 2013). Progression through this checkpoint is also supported by signals from CXCR4 and NOTCH1 (Janas et al., 2010;Maillard et al., 2006;Trampont et al., 2010).
Subsequently, DP thymocytes, now in the cortex, rearrange TCRa genes. If they successfully make an abTCR that can bind weakly to self-peptide-MHC complexes, the cells undergo positive selection into 4SP and 8SP cells and migrate into the medulla (Klein et al., 2014). Alternatively, if the abTCR binds strongly to self-peptide-MHC, auto-reactive T cells are eliminated by negative selection.
In view of the important role of migration during thymocyte development, the high expression of Wnk1 in the thymus, and our previous work showing that WNK1 regulates adhesion and migration of CD4 + T cells, we investigated a potential role for WNK1 during thymocyte development. We find that WNK1 is required for DN thymocytes to progress past the b-selection checkpoint. We show that in DN thymocytes, WNK1 is a negative regulator of LFA1-mediated adhesion and a positive regulator of chemokine-induced migration. However, changes in migration are unlikely to cause the developmental arrest in WNK1-deficient thymocytes. Instead, WNK1, acting through OXSR1, STK39 and SLC12A2, is required for pre-TCR-induced post-transcriptional upregulation of MYC, and subsequent proliferation of DN4 thymocytes and differentiation into DP cells. Thus, a pathway that regulates ion homeostasis is a critical and previously unknown regulator of early thymocyte development.

WNK1 pathway genes expressed throughout thymic development
To extend the previous observation that Wnk1 is expressed at a high level in the thymus (Shekarabi et al., 2013), we re-analyzed RNAseq data from multiple stages of thymic development (Hu et al., 2013), focusing on expression of WNK-family kinases and proteins in the WNK pathway. We found that Wnk1 is the only member of the WNK-family expressed in thymocytes, with high expression at all stages of development (Supplementary Figure 1). Substantially more Oxsr1 than Stk39 is expressed at all stages, except for 4SP thymocytes where the expression of Stk39 exceeds levels of Oxsr1. Of the Na + K + Cland Na + Clco-transporters, only Slc12a2 is expressed in the thymus. Finally, there is substantial expression of the Slc12a6 and Slc12a7 K + Clco-transporters. Thus, multiple genes in the WNK1 pathway are expressed throughout thymic development.

Essential role for WNK1 in thymocyte development
To test whether WNK1 is required during T cell development, we used the CD2Cre transgene to delete loxP-flanked (floxed) alleles of Wnk1 (Wnk1 fl ) starting at the DN2 stage of development. We compared Wnk1 fl/fl /CD2Cre mice with control Wnk1 fl/+ /CD2Cre mice in which thymocytes have homozygous or heterozygous loss of Wnk1 respectively. We found that in the absence of WNK1, there was a progressive loss of cells from the DN4 stage onwards, with a 92% reduction in numbers of ISP thymocytes and a 98-99% reduction in DP, 4SP and 8SP thymocytes ( Figure 1A-C).
Analysis of Wnk1 mRNA levels showed that while Wnk1 expression was lost in DN and ISP populations in thymi from Wnk1 fl/fl /CD2Cre mice, from the DP stage onwards the expression levels were similar to or higher in comparison to control animals ( Figure   1D), demonstrating that the few remaining DP cells are those that have failed to delete Wnk1. Hence WNK1 is absolutely required for the development of DP thymocytes.
Next, we evaluated whether the kinase activity of WNK1 is required for thymic development. We generated compound heterozygous mice with both a kinase-inactive allele (Wnk1 D368A ) and a floxed allele of Wnk1 and a tamoxifen-inducible Cre in the ROSA26 locus (ROSA26-CreERT2, RCE) and generated bone marrow chimeras by reconstituting irradiated RAG1-deficient mice with marrow from either Wnk1 fl/D368A /RCE mice or Wnk1 fl/+ /RCE mice as a control. Following tamoxifen treatment of these chimeras, lymphocytes are the only cell type in which all cells delete the floxed Wnk1 allele, leaving them with either a wild-type or D368A Wnk1 allele. We found that after 7d of tamoxifen treatment, development of thymocytes expressing only WNK1-D368A was again very strongly blocked from the DN4 stage onwards ( Figure 1E, F), demonstrating that WNK1 kinase activity is required for thymocyte development from the DN to DP stages.

WNK1 is required for thymocytes to develop past the pre-TCR checkpoint
The block in T cell development between the DN3 and DN4 and subsequent DP stages in the absence of WNK1 coincides with the b-selection checkpoint, where DN3 cells that have productively rearranged the TCRb locus and express TCRb protein assemble the pre-TCR. Subsequent pre-TCR signals result in survival and proliferative expansion of DN4 cells, and differentiation into DP cells. To determine if this checkpoint might be compromised in WNK1-deficient thymocytes because of a failure to assemble the pre-TCR, we assessed expression of pre-TCR components. Flow cytometric analysis of DN2 and DN3 thymocytes showed that a similar fraction (~10%) of both control and WNK1-deficient cells expressed intracellular TCRb, corresponding to cells that have successfully rearranged the TCRb locus, indicating that loss of WNK1 does not impair TCRb locus rearrangement (Figure 2A). A similar analysis of DN4 cells showed that most control cells expressed intracellular TCRb, as would be expected for cells successfully undergoing b-selection. In contrast, only ~10% of WNK1-deficient DN4 cells express intracellular TCRb, similar to the percentage in DN3 cells ( Figure   2A). Thus, in the absence of WNK1, there is no selective expansion of TCRb + cells.
Cell surface levels of CD3e and pre-Ta in WNK1-deficient DN2 and DN3 cells were comparable to controls, while CD3e levels were increased in WNK1-deficient DN4 cells ( Figure 2B). Taken together, these results show that the developmental block in WNK1-deficient DN thymocytes is unlikely to be due to a failure to make pre-TCR subunits. Instead, the phenotype may be caused by failure of pre-TCR signaling.
To directly test this, we made use of an approach in which anti-CD3e antibodies are injected into RAG1-deficient mice that have arrested thymocyte development at the DN3 stage because of a failure to rearrange TCR genes. RAG1-deficient DN3 thymocytes express small amounts of the CD3 proteins on the cell surface and binding of anti-CD3e mimics a pre-TCR signal, resulting in differentiation of the thymocytes into DP cells (Levelt et al., 1993;Shinkai and Alt, 1994). As expected, injection of anti-CD3e into control Wnk1 fl/+ /CD2Cre/Rag1 -/mice resulted in the generation of large numbers of DP thymocytes. By stark contrast, DP thymocytes were completely absent in anti-CD3e treated Wnk1 fl/fl /CD2Cre/Rag1 -/mice ( Figure 2C), indicating that WNK1 is essential for pre-TCR signaling that controls the DN-DP transition in thymocytes.

RNAseq of WNK1-deficient thymocytes reveals significant changes in 5 distinct processes
To gain further insight into the role of WNK1 in this developmental transition, we used RNA sequencing (RNAseq) to analyze the transcriptome of DN3 and DN4 cells sorted from the thymi of Wnk1 fl/+ /CD2Cre/Rag1 -/and Wnk1 fl/fl /CD2Cre/Rag1 -/mice, harvested 0, 3, 24 and 48 h after injection with anti-CD3e ( Figure 3A, B, Supplementary Table 1). Principal component analysis (PCA) of the results revealed that the biggest differences between the samples were caused by anti-CD3e stimulation and not by genotype ( Figure 3C). Both, WNK1-deficient thymocytes and control cells followed a similar trajectory on the PCA plot as a function of time after anti-CD3e injection, suggesting that pre-TCR signaling is able to induce similar transcriptional changes in control and WNK1-deficient thymocytes. Nevertheless, comparing the transcriptomes of control and WNK1-deficient thymocytes, we identified a large number of statistically significant differentially expressed genes (DEGs) at each time point, including in unstimulated cells (Figure 3D, Supplementary Tables 2, 3).
To further characterize these changes, we carried out a process enrichment analysis taking into account all DEGs at each time point and found that most of the identified processes fit into 5 functional groups: genes associated with TCR signaling, protein translation, cytoskeleton, adhesion and cell cycle ( Figure 3E). The TCR signaling and translation signatures were most enriched at the 0 h time point, cytoskeleton and adhesion at 48 h, and cell cycle at 24 h. Thus, WNK1 deficiency causes changes in several key biochemical pathways, any or all of which could be causing the developmental arrest.

WNK1 deficiency does not perturb early pre-TCR-induced signaling and protein translation
We investigated the potential contribution of each of the perturbed processes to the developmental block caused by loss of WNK1. Inspection of the TCR signaling signature showed that WNK1-deficient thymocytes had decreased levels of mRNA for Cd3d, Cd3e, Cd3g, Lck and Syk, coding for CD3 proteins and the LCK and SYK tyrosine kinases and an increase in Lcp2 mRNA, which codes for the SLP76 adapter protein ( Figure 4A), suggesting that pre-TCR signaling may be perturbed, though we note that at least for CD3e, reduced mRNA level in WNK1-deficient cells did not result in reduced cell surface protein level ( Figure 2B). However, analysis of transcriptional changes driven by pre-TCR stimulation showed that both control and WNK1-deficient cells upregulated Nur77 and downregulated Il2ra, Hes1, Rag2, and Ptcra to a similar extent ( Figure 4B). Furthermore, anti-CD3e-induced downregulation of surface levels of CD25 (encoded by Il2ra) also occurred normally in WNK1-deficient cells ( Figure 3B, 4C). These results imply that loss of WNK1 does not have a major effect on pre-TCR signaling, at least for the first 24 h after stimulation.
The most highly enriched processes in the RNAseq analysis were associated with protein translation, with many of the genes in these processes being downregulated in WNK1-deficient cells, particularly at the 0 h time point ( Figure 3E, 4D). We hypothesized that this could lead to lower overall protein synthesis and might explain the developmental block. To test this, we measured levels of phosphorylated S6 (pS6), a hallmark of active translation, in DN3 and DN4 thymocytes from mice 24 h after injection of anti-CD3e, and also directly tested the capacity of the cells to synthesize proteins using a puromycin incorporation assay. We found that pS6 levels in WNK1deficient cells were similar to those in control cells before stimulation and higher than in controls 24 h after anti-CD3e stimulation ( Figure 4E). Furthermore, rates of protein synthesis were increased to a similar extent by anti-CD3e in both control and mutant thymocytes ( Figure 4F). Thus, despite the reduced expression of genes associated with translation, the absence of WNK1 does not cause substantial changes in rates of protein synthesis, and these are unlikely to explain the developmental block.

WNK1 regulates thymocyte migration and adhesion
In the process enrichment analysis of the RNAseq data we identified cytoskeleton and adhesion signatures. At the 24 h time point WNK1-deficient thymocytes had increased expression of Itgb2 and Itgal, the subunits of LFA1, as well as Rap1a and Rap2a that encode GTPases required for integrin signaling, and several chemokine receptors (Ccr2, Ccr5 and Cxcr5), although Cxcr4 expression was reduced ( Figure 5A); some of the same changes were also visible 48 h after stimulation. In agreement with this, cellsurface LFA1 was increased and CXCR4 was decreased on both WNK1-deficient DN2+DN3 and DN4 thymocytes ( Figure 5B). We have previously shown that WNK1 is a negative regulator of LFA1-mediated adhesion and a positive regulator of chemokine-induced migration in CD4 + T cells (Köchl et al., 2016). These functions, together with the transcriptional changes seen here, suggest that WNK1 may also regulate adhesion and migration of DN thymocytes, which could account for the developmental block.
To investigate if CXCR4 can activate the WNK1 pathway in thymocytes, we stimulated primary mouse DN3 cells from Rag1 -/mice with CXCL12, the ligand for CXCR4, in the presence or absence of the WNK1 inhibitor WNK463 and analyzed phosphorylation of OXSR1 on Ser325, a WNK1 substrate (Vitari et al., 2005). We found that CXCL12 resulted in the phosphorylation of OXSR1, which was abrogated in WNK1 inhibitortreated cells ( Figure 5C). Thus, CXCR4 transduces signals through WNK1 leading to the phosphorylation and, presumably, activation of OXSR1.
These results suggest that, as in CD4 + T cells, WNK1 may also regulate CXCR4induced adhesion and migration in DN thymocytes, possibilities we tested directly. We found that in response to stimulation through either CD3e or CXCR4, or with Mn 2+ which directly activates LFA1, WNK1-deficient DN3 thymocytes showed increased binding of ICAM1, indicating increased LFA1-mediated adhesion ( Figure 5D). Furthermore, WNK1-deficient DN3 cells had reduced migration in response to CXCL12 ( Figure 5E). Thus, WNK1 is a negative regulator of LFA1-mediated adhesion and a positive regulator of CXCL12-induced migration in DN3 thymocytes.
Loss of CXCR4 results in defective thymocyte development in part because CXCR4induced migration is required for DN thymocytes to migrate from the cortico-medullary junction where they first arrive in the thymus, to the sub-cortical zone where as DN3 cells they undergo b-selection (Trampont et al., 2010). Thus, the block in development of WNK1-deficient thymocytes may be due to defects in this stereotypical CXCR4induced migration. To address this, we generated mixed bone marrow chimeras by reconstituting irradiated B6.SJL mice (CD45.1 + CD45.2 -) with an 80:20 mixture of marrow from B6.SJL mice and from either Wnk1 fl/+ /CD2Cre/Rag1 -/or Wnk1 fl/fl /CD2Cre/Rag1 -/mice (CD45.1 -CD45.2 + ). In the thymi of these animals, the RAG1-deficient hematopoietic cells can only develop as far as the DN3 stage, and we were able to analyze their distribution within normal thymic architecture generated by the excess of wild-type B6.SJL thymocytes. We found that WNK1-deficient DN3 cells accumulated in the sub-cortical zone with a similar frequency to WNK1-expressing cells and showed no excess accumulation at the cortico-medullary junction ( Figure   5F). Thus, despite an in vitro migration defect, WNK1-deficient DN3 thymocytes migrate to the correct anatomical location in vivo, even when in competition with WT counterparts. Collectively these data indicate that the developmental block seen in WNK-1 deficient thymocytes is unlikely to be caused by defective localization within intra-thymic microenvironments.
To extend this, we made use of an in vitro thymocyte development assay in which DN thymocytes are cultured on OP9-DL1 cells. In this system, over the course of 7d, thymocyte expansion and development occurs independently of the intra-thymic migration events seen in the thymus in vivo. Thus, we used it to evaluate if loss of WNK1 still affected thymic development even when migration was no longer relevant, by comparing the development of DN thymocytes from Wnk1 fl/+ /CD2Cre and Wnk1 fl/fl /CD2Cre mice. As expected, we found that WNK1-expressing DN thymocytes expanded 100-fold in number and the majority differentiated into DP, 4SP and 8SP cells ( Figure 5G). In contrast, WNK1-deficient cells did not increase in number and while most of them became DN4 cells (CD25 -CD44 -), almost none differentiated beyond the DN stage, resulting in a striking absence of DP thymocytes. Taken together, these results show that the block in T-cell development caused by WNK1 deficiency can be dissociated from its role in the regulation of thymocyte migration.

WNK1 controls entry of thymocytes into S-phase
RNAseq analysis revealed a particularly strong cell cycle signature at the 24h time point, though not at 0 or 3 h ( Figure 3E). While at 3 h post-stimulation, induction of genes associated with progression from the early to late G1 phase of the cell cycle To test whether S-phase entry is defective we measured DNA content of thymocytes from Wnk1 fl/+ /CD2Cre and Wnk1 fl/fl /CD2Cre mice. We found that while 11% of both control and WNK1-deficient DN2 and DN3 thymocytes were in the S/G2/M phases of the cell cycle, at the DN4 stage 30% of control cells but only 5% of WNK1-deficient cells had progressed beyond the G1 phase ( Figure 6D). Furthermore, loss of WNK1 abrogated the increase in cell size seen during the transition of DN3 cells to the DN4 stage, a hallmark of dividing cells ( Figure 6E).
Thus, WNK1 is required for DN4 thymocytes to enter the S/G2/M phases of the cell cycle.

WNK1 is required for pre-TCR induced regulation of MYC
To gain further insight into this cell cycle defect, we used a computational approach to identify transcription factors whose targets were most enriched among either all DEGs or only DEGs encoding cell cycle genes at 24 h ( Figure 7A) and extended the analysis to test the expression levels of the identified transcription factors themselves ( Figure   7B). The transcription factor, whose known target genes were both most numerous and most highly enriched within the list of all DEGs at 24 h was ETS1. Deletion of ETS1 impairs the development of DN3 into DP thymocytes because ETS1-deficient DN4 thymocytes are more likely to die (Eyquem et al., 2004). ETS1 regulates the expression of the anti-apoptotic Bcl2A1 gene in endothelial cells (Wei et al., 2009), suggesting that it may perform a similar function in thymocytes. However, analysis of the 5 anti-apoptotic members of the BCL2-family that are expressed in thymocytes, showed that WNK1-deficient thymocytes express substantially more Bcl2 and Bcl2l1 (Supplementary Figure 3A). In addition, WNK1-deficient DN2+DN3 and DN4 thymocytes did not show increased levels of active caspase 3 as would be expected if they were dying at a greater rate (Supplementary Figure 3B). Thus, it is unlikely that loss of WNK1 perturbs cell survival at the b-selection checkpoint. Furthermore, loss of ETS1 does not affect proliferation of DN4 cells (Eyquem et al., 2004), thus reduced levels of ETS1 are not the main cause of the defect in WNK1-deficient thymocytes.
Another high scoring transcription factor was MYC, whose targets were also strongly enriched within all DEGs and the more limited cell cycle gene set ( Figure 7A). MYC regulates the expression of cell cycle genes, including CDK-, CYCLIN-and E2F-family proteins, many of which are dysregulated in WNK1-deficient thymocytes ( Figure 6B) (Bretones et al., 2015). Indeed, 4 direct MYC targets, E2F2, E2F3, E2F4 and E2F7, were identified amongst the transcription factors most likely to cause dysregulation of cell cycle genes ( Figure 7A). Furthermore, thymic development is blocked in the absence of MYC, because of a failure of DN thymocytes to proliferate (Dose et al., 2006;Douglas et al., 2001), similar to the phenotype in WNK1-deficient thymocytes.
Thus, we hypothesized that WNK1 may be required for pre-TCR-induced proliferation in DN4 thymocytes because it regulates MYC activity.
Myc transcription in thymocytes is induced by Notch signaling (Weng et al., 2006), whereas pre-TCR signaling leads to increased MYC protein levels (Dose et al., 2006;Mingueneau et al., 2013). Myc mRNA levels were unaltered in WNK1-deficient DN3 thymocytes 24 h after anti-CD3e stimulation, implying that Notch regulation of Myc transcription is not affected in the absence of WNK1 ( Figure 7B). The expression of another Notch target, Hes1, was also unaltered by the loss of WNK1 ( Figure 4B), reinforcing the conclusion that WNK1 is not required for Notch function. In contrast, loss of WNK1 affected levels of MYC protein. Whereas anti-CD3e stimulation resulted in increased MYC in control DN3 cells, no such rise was seen in WNK1-deficient cells ( Figure 7C). Furthermore, analysis of Wnk1 fl/fl /CD2Cre mice also showed reduced levels of MYC in both DN2+DN3 and DN4 thymocytes compared to control mice ( Figure 7D). Thus, WNK1 is required for pre-TCR-induced post-transcriptional upregulation of MYC in DN thymocytes, and the failure to upregulate MYC in WNK1deficient DN cells is likely to account for their reduced proliferation and failure to differentiate into DP thymocytes.

OXSR1 and STK39 are required for pre-TCR induced proliferation and regulation of MYC
We have previously shown that in CD4 + T cells, WNK1 transduces signals via the related OXSR1 and STK39 kinases and the SLC12A2 ion co-transporter (Köchl et al., 2016). To test if these proteins are also required for T cell development, we evaluated thymocyte numbers in mice bearing a point mutation in Stk39 (Stk39 T243A ) and Oxsr1 (Oxsr1 T185A ). These residues are phosphorylated by WNK1 leading to the activation of STK39 and OXSR1, respectively and thus STK39-T243A and OXSR1-T185A can no longer be activated by WNK1. To analyze the effect of the Stk39 T243A mutation, we compared Stk39 T243A/T243A and control mice, whereas to assess the effect of the Oxsr1 T185A mutation we reconstituted irradiated RAG1-deficient mice with Oxsr1 T185A/T185A and control fetal liver cells, since this homozygous mutation is lethal in late gestation (Rafiqi et al., 2010). We found that thymocyte numbers in both mutant mouse strains were normal ( Figure 8A, B), suggesting that individually WNK1 activation of either STK39 or OXSR1 is not required for thymocyte development.
However, since OXSR1 and STK39 are related kinases, they could be functionally redundant with each other. We were unable to test this possibility by simply combining the Oxsr1 T185A and Stk39 T243A mutations, since embryos homozygous for both mutations die early in gestation before fetal liver cells can be harvested to make radiation chimeras (not shown). Instead, we intercrossed mice with a floxed allele of Oxsr1 (Oxsr1 fl ), the Stk39 T243A allele and RCE, the tamoxifen-inducible Cre to generate embryos whose liver cells we used to reconstitute the hematopoietic system of irradiated RAG1-deficient mice. Subsequent treatment of the mice with tamoxifen allowed us to compare development of control thymocytes expressing wild-type OXSR1 and STK39 (Oxsr1 +/+ /Stk39 +/+ /RCE), cells missing OXSR1 but expressing wildtype STK39 (Oxsr1 fl/fl /Stk39 +/+ /RCE) or double mutant cells missing OXSR1 and expressing a mutant STK39 (Oxsr1 fl/fl /Stk39 T243A/T243A /RCE). In the double mutant animals there was a large reduction in DN4, ISP and DP thymocytes compared to control or single mutant mice ( Figure 8C), demonstrating that OXSR1 and STK39 are required for development beyond the pre-TCR checkpoint, but that the kinases have redundant function with each other. Furthermore, DN4 thymocytes deficient in both OXSR1 and STK39 had reduced entry into S-phase and reduced levels of MYC, similar to the phenotype seen in WNK1-deficient thymocytes ( Figure 8D, E). Finally, we extended this to analysis of chimeras reconstituted with fetal livers from SLC12A2deficient embryos and found their thymocytes to also be substantially reduced at the DN4, ISP and 4SP stages, though the effect was not as large as that caused by loss of WNK1 ( Figure 8F, G).
Taken together, these results show that a WNK1-OXSR1-STK39-SLC12A2 pathway that regulates ion homeostasis plays a critical role at the pre-TCR checkpoint, by transducing signals that regulate levels of MYC and subsequent thymocyte proliferation and differentiation into DP thymocytes.

Discussion
T-cell development in the thymus is a highly regulated process, and critical checkpoints at specific thymocyte stages remain poorly understood. We show here that WNK1 and its kinase activity are absolutely required for the transition of DN to DP thymocytes at the b-selection checkpoint. Specifically, in the absence of WNK1 there is a failure of the proliferative expansion of TCRb + DN4 cells and subsequent differentiation into DP cells, hallmarks of b-selection. This developmental transition depends on signals from at least three receptors: pre-TCR, CXCR4 and NOTCH1 (Carpenter and Bosselut, 2010). NOTCH1 signaling appears to be unaffected by loss of WNK1, since WNK1deficient DN thymocytes have normal expression of the NOTCH1-regulated Hes1 and Myc genes. Instead, the developmental block in WNK1-deficient thymocytes is more likely caused by defective signaling from the pre-TCR, CXCR4 or both.
Our results demonstrate that WNK1 transduces signals from the pre-TCR and CXCR4 in DN thymocytes. Furthermore, we showed that WNK1-deficient cells have increased anti-CD3e-and CXCL12-induced LFA1-mediated adhesion and reduced CXCL12-induced migration. Thus, WNK1 is a negative regulator of adhesion and a positive regulator of migration in DN3 thymocytes. This mirrors our findings in CD4 + T cells, where we showed that WNK1 transduces signals from the TCR and CCR7 in CD4 + T cells to regulate both adhesion and migration (Köchl et al., 2016). WNK1 may be independently acting downstream of both the pre-TCR and CXCR4, but the two receptors may co-localize, with CXCR4 acting as a co-receptor for the pre-TCR (Trampont et al., 2010); other studies have shown that the TCR and CXCR4 physically associate to form a complex (Kumar et al., 2006). Thus, WNK1 may transduce signals from a pre-TCR-CXCR4 complex.
CXCR4-deficient mice show a strong developmental block at the b-selection checkpoint, similar to that seen in the absence of WNK1 (Janas et al., 2010;Trampont et al., 2010). This block was ascribed to mislocalization of DN3 thymocytes, with fewer cells reaching the sub-capsular zone (Trampont et al., 2010). Reduced CXCL12induced migration of WNK1-deficient DN3 thymocytes in vitro suggested that mislocalization of DN3 cells may also account for the developmental arrest in WNK1deficient thymi. However in vivo we found that WNK1-deficient DN3 thymocytes localized correctly to the sub-capsular zone and defective migration is unlikely to account for the developmental defect. This conclusion is further supported by the strong developmental arrest of WNK1-deficient thymocytes cultured on OP9-DL1 cells, a system where thymocyte development does not require migration. Since development of DP cells in this system requires CXCL12 produced by the OP9-DL1 cells, CXCR4 contributes to development through the b-selection checkpoint by more than just controlling thymocyte migration (Janas et al., 2010). Indeed, CXCR4, acting with the pre-TCR is required for proliferation of DN thymocytes as they undergo bselection (Trampont et al., 2010).
Analysis of WNK1-deficient DN3 thymocytes responding to pre-TCR signals showed that the initial responses of the cells are largely normal. They increase rates of translation, transiently induce expression of Nur77, downregulate expression of Il2ra, Rag2 and Ptcra, and lose surface expression of CD25, phenotypically becoming DN4 thymocytes. Furthermore, after 3 h of stimulation with anti-CD3e, they progress into the late G1 phase of the cell cycle as indicated by upregulation of Cdk6, Ccnd2 and Ccnd3. However, they have a profound block in entry into S-phase 24 h after stimulation, with reduced expression of Cdk2, Ccne1 and Ccne2, as well as genes encoding E2F-, ORC-and MCM-family proteins. In agreement with this, many fewer WNK1-deficient DN4 thymocytes entered S/G2/M phases of the cell cycle. Since proliferation of DN thymocytes is required for their differentiation into DP cells (Kreslavsky et al., 2012), this proliferative defect also accounts for the failure of WNK1deficient cells to develop into DP thymocytes.
MYC is upregulated at the b-selection checkpoint as thymocytes transit from the DN3 to DN4 stage and start to proliferate (Dose et al., 2006;Mingueneau et al., 2013). Importantly, we found that WNK1-deficient DN3 thymocytes failed to upregulate MYC in response to pre-TCR stimulation. Furthermore, many MYC target genes were dysregulated in the absence of WNK1, especially cell cycle genes. MYC-deficiency results in a similar thymic phenotype to that seen in the absence of WNK1 -reduced proliferation of DN thymocytes, failure to increase in cell size and a substantial reduction of DP thymocytes (Dose et al., 2006;Douglas et al., 2001).Taken together, we propose that the reduced proliferation of WNK1-deficient DN4 thymocytes is caused by failure to upregulate MYC, and hence reduced expression of genes required for S-phase entry.
The upregulation of MYC at the b-selection checkpoint is post-transcriptional -MYC protein increases despite constant Myc mRNA levels (Mingueneau et al., 2013). Loss of WNK1 affects this post-transcriptional regulation since mutant DN thymocytes had reduced MYC protein, but unchanged levels of Myc mRNA. MYC is regulated at the level of protein stability. It has a short half-life, estimated to be around 20 min, caused by rapid proteasomal degradation and its stability is regulated by phosphorylation; ERK phosphorylation stabilizes MYC, whereas GSK3b-dependent phosphorylation results in ubiquitination of MYC by E3-ligases such as FBXW7 and subsequent proteasomal degradation (Dang, 2012;Welcker et al., 2004;Yada et al., 2004). Notably, deletion of Fbxw7 in the thymus results in high levels of MYC, increased proliferation and hypercellularity (Onoyama et al., 2007). Interestingly, the expression of Fbxw7 is increased in WNK1-deficient DN3 thymocytes 24 h after anti-CD3e stimulation, which may contribute to destabilizing MYC.
Since loss of both OXSR1 and STK39 also leads to a block at the pre-TCR checkpoint, along with reduced MYC levels and decreased proliferation of DN4 thymocytes, we hypothesize that WNK1 transduces pre-TCR and CXCR4 signals via OXSR1 and STK39 to the upregulation of MYC, and hence to entry into S-phase and increased cell division. OXSR1 and STK39 phosphorylate multiple members of the SLC12A-family of ion co-transporters. Deficiency of SLC12A2, a Na + K + Cltransporter, showed a block at the pre-TCR checkpoint, which was less complete than that seen in the absence of WNK1. This suggests that OXSR1 and STK39 may be transducing signals that regulate MYC via several SLC12A-family proteins, including SLC12A2. The WNK1-OXSR1-STK39-SLC12A2 pathway regulates ion uptake in kidney epithelial cells. Our results suggest that regulation of ion homeostasis may be essential for pre-TCR induced proliferation of thymocytes. Further studies will be needed to establish if it is the ion movement per se or the subsequent osmotic movement of water into the cell which is essential for thymic development.
Interestingly, among >1400 cancer cell lines, the expression of WNK1 is highest in T- In summary we have shown that WNK1 is an essential regulator of early thymocyte development, where, during b-selection, it transduces signals from the pre-TCR and CXCR4 via OXSR1, STK39 and SLC12A2 that lead to upregulation of MYC, proliferation of DN4 thymocytes and differentiation into DP cells.

Radiation chimeras
To generate radiation chimeras, either bone marrow cells were harvested from Slc12a2 +/and Slc12a2 -/-). RAG1-deficient animals (5-8 weeks of age) were irradiated with 5Gy using a 137 Cs-source, and then reconstituted intravenously with at least 1x10 6 bone marrow cells/recipient or 0.5x10 6 fetal liver cells/recipient. All chimeric animals received Baytril in their drinking water (0.02%, Bayer Healthcare) for at least 4 weeks post-transplantation. If required, 8-20 weeks after reconstitution chimeric mice were injected intraperitoneally for 3 d with 2 mg/day of tamoxifen (Sigma) re-suspended at 20 mg/ml in corn oil (Sigma) and analyzed 7 d after start of tamoxifen treatment.

Flow Cytometry, antibodies, cytokines and other materials
Flow cytometry was carried out using standard techniques with pre-titered antibodies.
Antibodies for stimulations and flow cytometry against the following proteins were release 86) and to obtain gene level counts. For RSEM, all parameters were run as default except "--forward-prob" that was set to "0". Differential expression analysis was carried out with DESeq2 package (Love et al., 2014). Genes were considered to be differentially expressed with adjusted p-value (padj) ≤ 0.05. For data processing in Metacore, only genes with an average expression of transcripts per million (TPM) > 3 over all conditions were considered. All RNAseq data has been deposited in Gene Expression Omnibus (GEO) with accession number GSE136210. Reviewers can access this data using the following token: gruleowavdgxlqr.
For the re-analysis of RNAseq data from thymocyte subsets (Hu et al., 2013), Fastq files were obtained from the Gene Expression Omnibus (GSE48138). Gene-level TPM abundance estimates were calculated with RSEM using STAR to align reads against the NCBIM37 (mm9) genome assembly with Ensembl release 67 transcript annotations.

Q-PCR for Wnk1 mRNA
Up to 5x10 4 cells of each thymic population from thymi of control or WNK1-deficient animals were sorted directly into RLT lysis buffer, total RNA was extracted with an RNAEasy Plus Micro Kit (Qiagen) and cDNA was synthesized with a Superscript III kit (Life Technologies). Samples were analyzed on an ABI 7900 using a TaqMan gene expression assay (Life Technologies) spanning exon1 and exon 2 of Wnk1. Data was normalized to Hprt1 and analyzed using the comparative threshold cycle method.

Development of double positive thymocytes was induced by a single intraperitoneal
injection of 30 µg anti-CD3e antibody into RAG1-deficient mice, followed by harvesting the thymus for flow cytometric analysis 4 d later. For all other experiments where pre-TCR signaling responses were being measured, RAG1-deficient mice were injected with 150 µg anti-CD3e.

MetaCore analysis
The lists of filtered differentially expressed genes were uploaded to MetaCore and analyzed with the Process Networks analysis and Transcription Factor analysis modules (MetaCore version 19.1 build 69600).

Heatmaps
Selected genes from a number of enriched processes were used to generate a heatmap using the R package pheatmap. Genes were clustered using an Euclidean distance matrix and complete linkage clustering. Red indicates higher expression and blue indicates low expression relative to the mean expression of the gene across all samples.

Protein synthesis assay
Mice were injected intraperitoneally with 150 µg anti-CD3 or left untreated. 24h later thymi were harvested. Single cell suspensions were prepared and either pre-treated with 10 µM cycloheximide for 30 min or not, before being incubated with 5 µM O-

Adhesion assays
Binding of ICAM1 complexes to primary mouse thymocytes was analyzed as described (Konstandin et al., 2006). Soluble ICAM1-Fc-F(ab')2 complexes were generated by diluting APC-labeled goat anti-human IgG F(ab')2 fragments (109-135-098, Jackson  analyzed at least 10 µm apart, and the frequency of CD45.2 + CD25 + CD11b -CD11ccells was recorded in the subcapsular zone, cortex and cortico-medullary junction (CMJ) in a set area as described (Cowan et al., 2014). The CMJ was defined by ERTR5 staining for medullary epithelial cells. Images were obtained using a Zeiss LSM 780 microscope and analyzed using Zeiss LSM software.

Statistical Analysis
All statistical comparisons were carried out using the nonparametric two-tailed Mann-Whitney test, Hypergeometric test, Wald test, or two-way ANOVA. The specific analysis used is described in each Figure Legend,

Supplementary Table 1. RNAseq analysis of control and WNK1-deficient DN3 thymocytes following injection of anti-CD3e
Analysis of RNAseq experiment described in Figure 3A, showing expression of all genes measured in TPM at all time points.

Supplementary Table 2. Differential gene expression in control and WNK1deficient DN3 thymocytes following injection of anti-CD3e
Analysis of RNAseq experiment described in Figure 3A, showing differential gene expression analysis with DESeq2, taking into account all genes.

WNK1-deficient DN3 thymocytes following injection of anti-CD3e
Analysis of RNAseq experiment described in Figure 3A, showing differential gene expression analysis with DESeq2 as in Supplementary Table 2 but showing only genes that were statistically significantly differentially expressed (padj ≤ 0.05) and had an average expression value of TPM>3 over all conditions.

Supplementary Figure 1. Expression of genes encoding WNK pathway proteins
Mean±SEM expression of genes in the WNK1 pathway in the indicated thymocyte subsets, re-analyzed from previously published data (Hu et al., 2013).

Supplementary Figure 2. WNK1 is not required for progression of DN3 thymocytes into late G1
Analysis of RNAseq experiment described in Figure 3A, showing mean±SEM expression of selected genes associated with progression from the early to late G1 phase of the cell cycle at 0 h and 3 h following injection of anti-CD3e. * 0.01 < P < 0.05.
Significance displayed is from the analysis of the RNAseq data and was calculated by the Wald test (Supplementary Table 3     P-value (-log 10 ) Minimal P-value (-log 10 ) 0 hours P-value (-log 10 ) 3 hours P-value (-log 10 ) 24 hours P-value (-log 10 ) 48 hours