Thymic development beyond β-selection requires phosphatidylinositol 3-kinase activation by CXCR4

Although the PI3K pathway has been implicated in preTCR signaling, a direct link between the preTCR and individual PI3K catalytic subunits has yet to be shown. This is caused in part by the technical difficulties of studying signaling events initiated by the preTCR, as it is rapidly degraded and difficult to detect at the cell surface (Panigada et al., 2002). To determine whether p110γ and/or p110δ are required for preTCR-mediated Akt phosphorylation, we used the OP9-DL1 culture system. BM-derived hematopoietic stem cells (HSCs) from WT, p110δ^−/−, or p110γ^−/− were cultured in the continued presence of IL-7, which enabled differentiation to the DN3/DN4 stage of T cell development (Huang et al., 2005 and unpublished data). Under these culture conditions, we found that DN4 cells express detectable CD3ε at the cell surface (Fig. S1). DN4 cells from WT, p110δ^−/−, and p110γ^−/− cultures were similarly responsive to stimulation with anti-CD3ε antibody for the induction of Erk1/2 phosphorylation as measured by flow cytometry (Fig. 1). The staining profile for phospho-Erk appeared bi-modal and is thus similar to the Erk1/2 response in more mature TCR-stimulated thymocytes (Prasad et al., 2009). In contrast, the staining profile of Akt phosphorylation at residue serine 473 showed an overall shift in median fluorescence indicative of a graded analogue response. The increase in phospho-Akt in p110γ^−/− DN4 cells upon anti-CD3ε stimulation was similar to the WT, but p110δ^−/− cells failed to generate detectable p-Akt under identical conditions. These results suggested that the p110δ subunit is critical for PI3K signaling from the preTCR, whereas p110γ is dispensable. Moreover, preTCR activation of the Erk signaling pathway is independent of either p110γ or p110δ.

p110γ is recruited to the Gβγ subunits of GPCRs via its association with p101 or p84 regulatory subunits (Stephens et al., 1997; Suire et al., 2005). We found that p101 mRNA was expressed in DN3 and DN4 thymocytes, whereas p84 mRNA was undetectable (Fig. S2 A). Thymocytes from mice lacking p101 (Suire et al., 2006) had a CD4/CD8 FACS staining profile similar to the p110γ^−/− single mutants (Fig. 2 A), and they had a reduced number of DP thymocytes compared with WT (P = 0.01; using the unpaired Student’s t test; not depicted). However, mice deficient for both p101 and p110δ showed thymic atrophy to a similar degree to the p110γ^−/−δ^−/− mice (Fig. 2 A). These mice had a marked increase in the proportion of thymocytes that were CD4^–CD8^– DN and a decrease in the proportion of CD4⁺CD8⁺ DP cells. DN4 cells (CD25^–/CD44^–) from p110γ^−/−δ^−/− and p101^−/−/p110δ^−/− mice failed to show increased expression of the CD98 heavy chain of amino acid transporters and CD71 (transferrin receptor), nor did they show increased forward angle light scatter indicative of cell growth (Fig. 2, B–D). This defect was not secondary to a failure to express a productively rearranged tcrb gene as the frequency of small DN3 cells expressing the β polypeptide identified by intracellular flow cytometry was not different from the WT (Fig. 2 D). Additionally, the amounts of p110γ protein were unchanged in the absence of p101 (Fig. S2 B). Thus, the phenotype observed in the p101^−/−/p110δ^−/− mice is not a consequence of reduced expression of the p110γ. As p101 is required to regulate p110γ function we conclude that, in its absence, p110γ is uncoupled from its activating receptor.

Previous studies have shown that when WT BM is transplanted into p110γδ^−/− recipient mice, thymic development is impaired to the same extent as that seen in p110γ^−/− mice, suggesting a non–cell autonomous role for p110γ in T cell development (Swat et al., 2006). Therefore, to determine whether there was a thymocyte-intrinsic requirement for p110γ/p110δ or p101/p110δ, we created chimeras by reconstituting recipient mice doubly deficient in Rag2 and the common cytokine receptor γ chain (Rag2^−/− Il2rg^−/−) with BM from WT, p110γ^−/−δ^−/−, or p101^−/−/p110δ^−/− mutant mice. Chimeric mice reconstituted with p110γ^−/−δ^−/− or p101^−/−/p110δ^−/− BM had a reduction in the number of DP thymocytes compared with control chimeras reconstituted with WT BM (Fig. 3 A). Although the numbers of DN3/DN4 cells were equivalent to WT chimeras, these cells failed to become CD98^hi, CD71^hi or to increase in size (Fig. 3, B–F). In addition, pulse-labeling with the thymidine analogue 5-ethynyl-2′-deoxyuridine (EdU) revealed that both p110γ^−/−δ^−/− and p101^−/−/p110δ^−/− DN3/4 thymocytes incorporated EdU poorly compared with WT (Fig. 3 G-H). Defective expansion and differentiation was recapitulated in vitro using sorted DN3 cells from WT and PI3K single- or double mutant mice cultured on OP9-DL1 stromal cells (Fig. 4). After 3 d of culture, the majority of WT and PI3K single-mutant DN3 cells had proliferated and differentiated to become large CD4⁺CD8⁺CD98^hi cells. However, both p110γ^−/−δ^−/− and p101^−/−/p110δ^−/− DN3 cells remained undifferentiated, did not proliferate, and failed to increase in size. Thus, thymocytes have a cell autonomous requirement for p110γ/p110δ or p101/p110δ to progress beyond β-selection.

The in vivo and in vitro requirement for p101 at the β-selection checkpoint prompted us to seek evidence of a role for a GPCR in this process. We observed that CXCL12, the ligand for CXCR4, was secreted by the OP9-DL1 stromal line (Fig. S3 A). We examined whether CXCL12 was required for DN3 proliferation and differentiation by culturing sorted DN3 cells on OP9-DL1 in the presence of the CXCR4 antagonist AMD3100. This small molecule binds to CXCR4 with high specificity and inhibits chemotaxis of human PBMCs to CXCL12 with an IC₅₀ of 0.16 µM (Hatse et al., 2002). Culturing DN3 cells in the presence of AMD3100 inhibited their ability to become CD4⁺CD8⁺ (Fig. 5). The numbers of DP cells recovered after 3 d of culture were decreased in a dose-dependent manner, indicating that in the absence of CXCR4 signaling DN3 cells had reduced proliferation and/or survival. This effect on DN3 cells was also seen with a chemically distinct small molecule inhibitor of CXCR4 termed GSK812397A, which inhibits the CXCL12-mediated chemotaxis of the U937 cell line with an IC₅₀ of <1 nM (GlaxoSmithKline; personal communication). These results show that signaling through CXCR4 contributes to the development of DN3 cells to the DP developmental stage.

We sought evidence to determine whether CXCR4 activation of PI3K required p110γ and/or p110δ. Using intracellular flow cytometry, we measured p-Akt (S473) and p-Erk1/2 (T202/Y204) levels in freshly isolated thymocytes after in vitro stimulation with CXCL12. WT DN3, DN4, and DP, but not CD4⁺ SP, thymocytes showed increased p-Akt and p-Erk in response to CXCL12 stimulation (Fig. 6, A and B). DN3 cells lacking p110δ responded similarly to WT but the p-Akt response in CXCL12-stimulated p110γ^−/− and p101^−/− DN3 cells was reduced, although not eliminated. These results indicate that the PI3K signaling pathway is activated in DN3 cells in response to CXCL12 and that p110γ and p101 are required for this response. We also measured CXCL12-mediated Akt in p110γ^−/−δ^−/− DN3 cells and were unable to detect any increase in p-Akt (Fig. 6, A and D). Thus, in the absence of p110γ, p110δ also contributes to the CXCR4 activation of PI3K. Erk phosphorylation was found to be equivalent to WT in all PI3K mutant genotypes showing that p110δ, p110γ, and p101 were not required for CXCL12-mediated Erk phosphorylation and that some aspects of CXCR4 signaling are independent of these PI3Ks.

Our findings have identified the preTCR and CXCR4 as receptors on DN3 cells through which p110δ and p110γ are activated. However, it is unclear whether these receptors operate independently or cooperatively to induce activation of PI3K required for development beyond β-selection. To address this, we first compared PI3K activity before and after β-selection by measuring p-Akt (S473) in freshly isolated DN3 cells which were either CD98^lo or CD98^hi (Fig. 7 A) We have found that the expression of CD98 in DN3 cells correlates with expression of CD27 (data not shown), thus identifying DN3 cells with expression of TCRβ (Taghon et al., 2006). Our results show that CD98^hi DN3 cells have an ∼1.4-fold increase in detectable p-Akt compared with CD98^lo cells. This is consistent with increased PI3K activity upon β-selection.

Although we have demonstrated that blocking the CXCR4 signaling pathway inhibited DN3 cell differentiation in vitro, we wanted to determine whether this also affected PI3K activity. To test this, CD98^lo DN3 cells were cultured on the OP9-DL1 stromal line in the presence of increasing concentrations of AMD3100. Cells were harvested at 24-h intervals, and their p-Akt levels were measured. We observed an increase in the p-Akt levels of DN3 cells with time, which was diminished in the presence of AMD3100 (Fig. 7, B and C). Thus, signaling through CXCR4 contributes to PI3K activity. We also measured p-Akt levels in DN3 cells isolated ex vivo from PI3K mutant mice (Fig. 7 D). Despite the requirement of p110δ for preTCR-mediated Akt phosphorylation, p-Akt levels in the DN3 cells of p110δ^−/− mice were similar to WT. In contrast, DN3 cells from p110γ^−/−, p101^−/−, and p110γ^−/−δ^−/− mice had 70–80% of WT p-Akt levels, demonstrating a role for the class IB subunits in the generation and/or maintenance of PI3K activity in DN3 thymocytes.

To examine the role of CXCR4 in thymocyte development, we analyzed thymi from CXCR4-deficient mice. CXCR4 is required for embryo viability (Ma et al., 1998; Tachibana et al., 1998; Zou et al., 1998), therefore we compared T cell development in E17.5 CXCR4^+/− and CXCR4^−/− embryos (Fig. 8). We found a decreased proportion of DP thymocytes and increased proportion of DN thymocytes in CXCR4^−/− mice (Fig. 8 A). We also observed significant thymic atrophy in CXCR4^−/− mice (Fig. 8 B), which corresponded to reduced numbers of all thymic subsets from DN2 to DP (Fig. 8 C). This phenotype may, in part, be attributed to the reduced number of T cell precursors colonizing the thymus. In addition, we noted a progressive decrease in the ratio of CXCR4^−/− to CXCR4^+/− cells at each successive developmental stage, indicating a continual requirement for CXCR4 as thymocytes progress through β-selection and become DP T cells. Finally, we compared the cell size of DN3 and DN4 thymocytes from CXCR4^+/− and CXCR4^−/− mice, as we had observed that DN4 cells from PI3K mutant mice were smaller compared with WT (Fig. 1 D). Unlike the adult thymus, we found that there was no increase in size between DN3 and DN4 thymocytes of the E17.5 embryo. However, both DN3 and DN4 cells were significantly smaller in CXCR4^−/− thymi (Fig. 8 D). Collectively, these results suggest CXCR4 is required during T cell development.

Our findings suggest that, in addition to the preTCR and Notch, DN3 cells also require signals through CXCR4 for their continued development. When sorted DN3 cells were cultured with the recombinant Notch ligand Delta-like 4 (rDL4), they failed to proliferate or differentiate and minimally increased expression of CD98, demonstrating the requirement for a factor(s) in addition to the preTCR and Notch (Fig. 9, A–C). In contrast, when DN3 cells were cultured in the presence of rDL4 with CXCL12, CD98 up-regulation occurred, the cells increased in size and DP cells were formed (Fig. 9, A–C). The effect of CXCL12 was found to be dose-dependent (Fig. S3, B-C). The absolute requirement for preTCR signaling in this culture system was also confirmed as Rag2^−/− DN3 cells failed to differentiate (unpublished data). The effect of CXCL12 could not be substituted by two other chemokines. Culturing DN3 cells in the presence of a CCR5 ligand CCL5 did not induce DP differentiation (Fig. 9 A), nor did it have any additive effects in combination with CXCL12 (not depicted). Similarly, the addition of CCL25 neither reproduced nor potentiated the effect of CXCL12 (Fig. 9 A). Together, these results demonstrate that signaling through the preTCR, Notch, and CXCR4 combine to promote in vitro thymocyte differentiation beyond the DN3 stage.

The combination of rDL4 and CXCL12 defined minimal requirements for a stromal cell–free in vitro system that promoted thymocyte differentiation. However, these conditions did not support DN3 cell proliferation or differentiation to the same extent as the OP9-DL1 stromal cell line, suggesting that additional factors were contributed by the OP9-DL1 cells. Glycogen synthase kinase 3β (GSK3β) has been implicated as a regulator of multiple signaling pathways, and we tested whether inhibition of GSK3β in DN3 cells could further augment the stromal-free in vitro culture. DN3 cells from WT and p110γ^−/−/δ^−/− mice were cultured on rDL4 with the GSK3β inhibitor CHIR99021 in the presence or absence of CXCL12 (Fig. 9, D and E). CHIR99021 enabled the differentiation of WT DN3 cells in the absence of CXCL12, but together CHIR99021 and CXCL12 increased the proportion of cells that became CD4⁺CD8⁺ and increased cell numbers to that achieved when DN3 cells were cultured on OP9-DL1. This result suggests a role for the GSK3β signaling pathway in DN3 cells and further demonstrates that CXCL12 is required for optimal differentiation and proliferation. In contrast to WT DN3 cells, p110γ^−/−/δ^−/− DN3 cells failed to differentiate when cultured with rDL4 and CHIR99021. Further, the addition of CXCL12 did not enhance the numbers of p110γ^−/−/δ^−/− cells recovered or support their differentiation. Thus, for DN3 cells cultured under in vitro conditions, p110γ and p110δ activation is absolutely required for CXCR4-dependent expansion and differentiation.

Until recently, T cell development in vitro was studied in the context of fetal thymic organ cultures. The creation of the OP9-DL1 stromal cell line allowed the differentiation of HSCs into functional T cells in a two-dimensional culture system (Schmitt and Zuñiga-Pflucker, 2002). Exploitation of this system has allowed T cell precursors to be genetically manipulated and subsequently used in the treatment of tumor models in mice (Zakrzewski et al., 2008). Accessory cell–free culture conditions for T cell precursors have been described previously, but these did not permit cells to differentiate beyond the DN2 stage (Dallas et al., 2007). This suggests that additional factors elaborated by the stromal cell line contribute to the proliferation and differentiation of thymocytes. However, the identity of these additional factors has remained unknown. Here, we have provided evidence that CXCR4 and its ligand CXCL12, which is produced by OP9-DL1, are important for the proliferation and differentiation of DN3 cells. CXCR4 was identified as being essential in the OP9-DL1 culture system through the use of highly selective receptor antagonists. CXCL12 promoted DN3 differentiation that was dependent on Notch and the preTCR. We suggest these three receptors are functionally nonredundant, with each receptor–system interacting to promote continued thymic development beyond β-selection. Our findings offer the possibility of studying thymocyte development in a stromal cell–free system in which the effects of extrinsic factors or genetic manipulation can be controlled.

The combination of rDL4 and CXCL12 alone did not support DN3 cell proliferation or differentiation to the same extent as the OP9-DL1 stromal cell line. This might have been caused by the suboptimal presentation of the Notch ligand, which varies in effectiveness depending on density (Dallas et al., 2005). Similarly, presentation of CXCL12, which binds to extracellular matrix (Amara et al., 1999), may not be optimal in our system. It is likely that an additional factor besides Notch, preTCR, and CXCL12 are required for the proliferation and differentiation of DN3 cells. Utilization of the GSK3β inhibitor CHIR99021, in conjunction with CXCL12, appeared to compensate for this and allowed the proliferation and differentiation of DN3 cells to a similar degree to that observed when cultured with OP9-DL1. CHIR99021 did not potentiate the differentiation of p110γ^−/−δ^−/− DN3 cells; however, indicating that it was not able to compensate for the lack of PI3K. It is possible that the GSK3β inhibitor may promote Wnt/β-catenin signaling that has been identified as important for T cell development at the β-selection checkpoint (Gounari et al., 2001; Xu et al., 2003). Alternatively, CHIR99021 may be potentiating the effects of Notch (McKenzie et al., 2006) or regulating Hedgehog signaling, which has also been shown to regulate DN thymocyte differentiation (Outram et al., 2000). These issues can be explored further using an accessory cell–free culture system.

Although we have shown here that T cell development is impaired in CXCR4 knockout mice, controversy exists regarding the role of CXCR4 in T cell development (Ma et al., 1998; Zou et al., 1998; Kawabata et al., 1999; Ara et al., 2003; Plotkin et al., 2003). These discrepancies may be caused by the differing requirements for T lymphopoiesis in the fetal and adult thymus. For example, fetal T cell development has been shown to be independent of the IL-7 receptor (Crompton et al., 1998) and α4 integrin (Arroyo et al., 1996), but these molecules are essential for postnatal T cell development. However, a complete block in adult T cell development upon the conditional deletion of CXCR4 in DN1 cells in chimeric mice suggests that CXCR4 is important for directing T cell progenitors beyond the corticomedullary junction in the adult thymus (Plotkin et al., 2003). To fully explore the role for CXCR4 at the later developmental stage of β-selection in the adult, a conditional model whereby CXCR4 is deleted upon the expression of TCRβ is required.

The identification of a critical role for CXCR4 signaling arose from an investigation into the functional redundancy between PI3K isoforms during T cell development. This study has identified receptors through which each of these catalytic isoforms is activated. PreTCR-mediated PI3K activation is p110δ dependent. In contrast, activation of PI3K by CXCR4 is predominantly mediated by p110γ in a p101-dependent manner. Activation of PI3K signaling through CXCR4 at the point of β-selection is critical, as in the presence of CXCR4 antagonists the phosphorylation of the PI3K effector Akt in DN3 cells is impaired and cells fail to differentiate. The relatively normal phenotypes of the p110δ^−/−, p110γ^−/−, and p101^−/− DN thymocytes suggest that the loss of one catalytic isoform is compensated for by the presence of the other and that major defects in thymocyte development are only observed when class IA and IB PI3K subunits are inactivated simultaneously. Thus, p110δ^−/− DN3 cells are unable to phosphorylate Akt downstream of the preTCR, but possess a robust p-Akt response to CXCR4 stimulation caused by the presence of p110γ. Conversely, the ability of p110γ^−/− and p101^−/− thymocytes to phosphorylate Akt upon CXCR4 stimulation is poor, but their PI3K response to preTCR signaling, mediated by p110δ, is intact. In each of these scenarios, PI3K activity in DN3 cells is sufficient for continued thymocyte development. The mechanistic rationale for having multiple cell surface receptors activate the same growth promoting pathway may be to promote proliferation in defined regions of the thymus. Thus, CXCL12 is concentrated at the subcapsular zone (Plotkin et al., 2003), which is where the most intensive proliferation of thymocytes is known to occur.

Although p110γ appears to be relatively more important for Akt phosphorylation after CXCR4 activation in DN3 cells, p110δ clearly plays an ancillary role as the p-Akt response of the p110γ^−/−δ^−/− or p101^−/−/p110δ^−/− double mutants was further reduced when compared with the p110γ^−/− or p101^−/− single mutants. The potential of class IA PI3K to contribute to CXCR4 signaling has been noted previously in studies using Jurkat cells (Curnock et al., 2003). In peripheral T cells, CXCR4 signaling leads to the phosphorylation of Lck and Zap-70 (Patrussi et al., 2007), possibly via direct interactions with the TCR (Kumar et al., 2006; Contento et al., 2008). It is not yet known whether these mechanisms are relevant to p110δ-mediated CXCR4 signaling in DN thymocytes; however, we have noted that pAkt responses to CXCL12 are intact in DN3 cells from Rag-2 mutant mice (unpublished data). Thus, TCRβ is dispensable for PI3K-mediated CXCR4 signaling in DN3 cells.

CXCL12 is predominantly considered as a chemokine but it has been recently shown to enhance the proliferation and activation of CD3-stimulated peripheral T cells (Molon et al., 2005). Indeed CXCL12 was originally identified through its preB cell growth factor properties (Nagasawa et al., 1994). Our results with stromal cell–free in vitro cultures suggest that CXCL12 also promotes the growth and differentiation of developing thymocytes. Presumably, in stromal cell–free culture, this function of CXCL12 is independent of any chemotactic activity. Notch has been documented as a regulator of DN3 metabolism (Ciofani and Zuñiga-Pflucker, 2005) but our results demonstrate that the provision of a Notch signal alone to DN3 thymocytes is not sufficient to support the metabolic changes associated with β-selection and that CXCL12 was additionally required. The contribution of CXCL12 to DN3 cell survival, proliferation, and differentiation have yet to be fully addressed, but further investigation using specific CXCR4 inhibitors in combination with mathematical modeling of time course data may provide further insight (Callard and Hodgkin, 2007).

Our findings may be relevant in the context of some T cell acute lymphoblastic leukemias (T-ALLs) as altered Notch and PI3K signaling frequently underpins the transformed phenotype of these cells, yet primary explants of these tumors are prone to apoptosis and difficult to propagate in vitro (Armstrong et al., 2009). Recently, some primary T-ALLs were propagated on stromal cell lines expressing recombinant Notch ligand (Armstrong et al., 2009). Our finding that CXCL12 promotes growth in concert with Notch and the preTCR may be relevant to the in vitro culture of T-ALL and prompts the hypothesis that CXCL12 may have an in vivo role in T-ALL. Examining the responses of T-ALL to AMD3100 alone or in combination with other drugs with efficacy against T-ALL may suggest a new therapeutic modality.

Mice with deletions of p110γ, 110δ, p101, and CXCR4 have been described previously (Tachibana et al., 1998; Webb et al., 2005; Suire et al., 2006). Mice were on 129/B6 mixed backgrounds and 5–13 wk of age, except CXCR4-deficient mice which were on a B6 background and analyzed at embryonic day 17.5. For chimera studies, irradiated Rag2^−/−Il2rγ^−/− mice were injected intravenously with 3 × 10⁶ BM cells from WT, p110γ^−/−δ^−/−, or p101^−/−/p110δ^−/− mice and analyzed 8–12 wk later. All mice husbandry and experimentation were in accordance with UK Home Office regulations and local ethical review (except CXCR4 embryo studies, which were approved by the animal ethics committee of the Institute for Frontier Medical Sciences, Kyoto University).

BM-derived HSCs were isolated from single-cell suspensions of BM by a two-step process using lineage depletion followed by ScaI⁺ enrichment by MACS (Miltenyi Biotec). Typically, the ScaI⁺-enriched population represented 0.1% of the starting BM preparation. DN3 and DN4 thymocytes were isolated by FACS. Single-cell thymocyte suspensions were first enriched for DN cells by incubation with biotinylated anti-CD8α antibody followed by streptavidin-coated MACS beads and magnetic depletion before staining for DN subsets. DN3 and DN4 cells were isolated as Thy1⁺, [CD4/CD8/CD44/B220/CD11b/NK1.1/Gr1/Ter119/γδTCR]^neg and either CD25^hiCD98^lo (DN3) or CD25^loCD98^hi (DN4).

OP9 and OP9-DL1 cells were maintained as previously described (Ciofani and Zuñiga-Pflucker, 2005). BM-derived HSCs were cultured in α-MEM supplemented with 20% FCS, 100 U/ml penicillin, 100 µg/ml streptomycin, 5 × 10⁻⁵M 2ME, 10 ng/ml recombinant IL-7, and 5 ng/ml recombinant Flt3L (PeproTech) at a density of 4 × 10³ cells/ml in 24-well tissue culture plates that had been seeded with 2 × 10⁴ OP9-DL1 cells. After 7 d, and at weekly intervals thereafter, cells were harvested, passed through a 40-µM filter, and recultured at 2 × 10⁵ cells/ml in 24-well tissue culture plates that had been seeded with 2 × 10⁴ OP9-DL1 cells in supplemented α-MEM complete with IL-7 and Flt3L. For assays using sorted DN3 thymocytes on OP9-DL1 cell lines, 4 × 10⁴ cells were cultured in supplemented α-MEM, without additional cytokines, in 96-well plates that had been previously seeded with 4,000 OP9-DL1 cells. AMD3100 (Sigma-Aldrich) and GSK812397A (GlaxoSmithKline) were added with the DN3 cells at the start of culture. For stromal cell–free cultures, 10⁵ DN3 cells were cultures in 96-well plates that had been coated overnight with 10 µg/ml of recombinant mouse Delta-like ligand 4 (R&D Systems) supplemented with α-MEM in the presence or absence of recombinant CXCL12 (10 nM), CCL5 (10 nM), CCL25 (10 ng/ml; PeproTech), and/or CHIR99021 (10 µM; Axon MedChem). For controls, equivalent cell numbers were cultured on OP9 or OP9-DL1 stromal cells. Assays using sorted DN3 cells were harvested after 3 d of culture and analyzed by FACS. Cell numbers were determined by reference to the inclusion of a fixed number of microbeads (Spherotech).

Single cell suspensions from the thymus or cultured cells were stained using various combinations of antibodies conjugated to fluorochromes for analysis by FACS (LSRII; BD). DN3/DN4 thymocytes were identified as Thy1⁺ and [CD4/CD8/CD44/B220/CD11b/NK1.1/Gr1/Ter119/γδTCR]^neg. For intracellular staining, cells were treated with BD Cytofix/Cytoperm solutions before staining for cell-surface markers in conjunction with PE-conjugated anti-mouse TCRβ (clone H57) or intracellular Notch1 (clone mN1A; BD).

Mice were injected intraperitoneally with 100 µg of EdU in PBS. After 4 h, thymi were harvested and single-cell suspensions were prepared and stained for cell surface markers. Cells were then prepared for EdU detection as per the manufacturer’s instructions (Invitrogen; Salic and Mitchison, 2008).

For preTCR stimulation studies, cultured cells were harvested, washed, and passed through a 40-µM filter before resuspension at 10⁷ cells/ml in supplemented α-MEM. Cells were incubated at 37°C for 20 min before the addition of 10 µg/ml of the α-CD3ε monoclonal antibody 2C11. At various time intervals, 10⁶ cells were fixed in 2% paraformaldehyde, washed in PBS, and resuspended in 90% ice-cold methanol at −20°C for at least 30 min. Cells were then washed three times in PBS with 2% FCS before incubation with antibodies to surface markers as well as to phosphorylated Akt (clone #4058; Cell Signaling Technology) or phosphorylated-Erk (clone #4370; Cell Signaling Technology) for 30 min at room temperature. Phosphospecific antibodies were detected by subsequent incubation with Cy5-conjugated donkey anti–rabbit antibody. Cells were then analyzed by FACS. CXCL12 stimulation studies were conducted as above, except that unfractionated thymocyte suspensions were resuspended at 10⁷ cells/ml in DMEM with 0.1% BSA. Cells were incubated at 37°C for 20 min before the addition of 10 nM CXCL12. For data analysis, fold induction was calculated as the median fluorescence of stimulated sample/median fluorescence of unstimulated sample. Unstimulated samples showed no increase in phospho-Akt or phospho-Erk throughout the timecourse.

RNA was extracted from 10⁶ isolated cells using TRIzol (Invitrogen) and converted to cDNA. Controls without reverse transcription were included for all samples. p101 and p84 mRNA was quantified using a Taqman probe and normalized to the expression of β₂ microglobulin (Applied Biosystems).

Lysates from 10⁶ thymocytes were separated by SDS-PAGE and transferred to nitrocellulose. Blots were probed with rabbit anti-p110γ (Onxy Pharmaceuticals) and antibody binding was revealed using HRP-conjugated anti–rabbit antibody (Dako) followed by ECL (GE Healthcare). Blots were subsequently probed using mouse anti-β actin.

Data presented throughout represent the mean and either the SD or SEM (if n ≥ 5). For statistical analyses containing two groups, data were analyzed using unpaired Student’s t test and P values are shown in the figures. For statistical analyses containing more than two groups, data were analyzed by one-way analysis of variance and statistical significance represented as: *, P < 0.05; **, P < 0.01; and ***, P < 0.001 compared with WT, except for the data in Fig. 4 where statistical significance is shown compared with cells cultured without agonist.

Fig. S1 presents the FACS profile of DN3 and DN4 cells derived from long-term BM HSC cultures. Fig. S2 presents relative RNA expression levels of the PI3K isoforms p101 and p84 in WT DN3 and DN4 cells, as well as p110γ protein levels in WT, p101^−/− and p110γ^−/− thymus lysates. Fig. S3 presents the secretion of CXCL12 by OP9-DL1 stromal cells as well as the cellular responses of WT DN3 cells cultured in the presence of increasing concentrations of CXCL12.

We thank A.M. Condliffe, L.R. Stephens, and P.T. Hawkins for the provision of the p101^−/− mice and the anti-p110γ antibody, S. Kulkarni for providing isolated mouse neutrophils, J.C. Zúñiga-Pflücker for provision of the OP9-DL1 stromal cell line, G. Butcher, A. Corcoran, P. Hawkins, K. Okkenhaug, M. Sims, E. Vigorito, and L. Webb for critical discussion, K. Bates and SABU staff for technical support and G. Morgan for FACS expertise.

This work was supported by the Biotechnology and Biological Sciences Research Council and the Ministry of Education, Culture, Sports, Science and Technology of Japan. G.V. was supported by a Da Vinci scholarship (Italy) and M.T. by a fellowship from the Medical Research Council (U.K.).

For the course of this work K. Gudmundsson was an employee of GlaxoSmithKline. The authors have no further conflicting financial interests.