Regulation of thymocyte β-selection, development and positive selection by glycogen synthase kinase-3

Glycogen synthase kinase-3 (GSK-3) is a ubiquitously expressed serine/threonine kinase, that exists as two isoforms in mammals, GSK-3α and GSK-3β, that are key downstream mediators of the phosphatidylinositol 3’ kinase, Wnt, Notch and other pathways. Here, we report that simultaneous inactivation of both GSK-3α and GSK-3β during early thymocyte ontogeny has profound effects on both β-selection and positive selection, key checkpoints essential to producing functionally mature αβ T cells. Conditional GSK-3α/β knockout animals (LckCre+ GSK-3αβfl/fl) possessed pre-double positive (pre-DP) thymocytes (CD4−CD8−CD117−CD25−) with compromised TCRβ chain expression along with elevated levels of β-catenin and reduced Notch activity. β-selection was impaired allowing pre-DP thymocytes to differentiate to DP thymocytes (CD4+CD8+) while bypassing strict requirements for productive TCRβ chain rearrangements and functional expression. Also impaired was the requisite pre-TCR and Notch-mediated expansion that normally precedes differentiation to the DP stage. Consequently, LckCre+ GSK-3αβfl/fl mice initially generated fewer DP thymocytes that expressed significantly reduced levels of mature TCR. The aberrant DP thymocytes expressed high levels of the pro-survival Bcl-2 family member Mcl-1, failed positive selection and accumulated as CD4hiCD8lo positive selection intermediates resulting in loss of both mature CD4 and CD8 lineages. LckCre+ GSK-3αβfl/fl mice succumbed to oligoclonal peripheral lymphomas with high penetrance. These data reveal essential roles for GSK-3 in several checkpoints of early T cell development.


Introduction
Thymocyte development requires navigation of cells through a well-orchestrated series of differentiation and maturity events that are modulated by multiple signalling pathways and results in the selection of a functional, but not auto-reactive, T cell repertoire, ab initio. Early stages of αβ thymocyte development prior to CD4 and CD8 expression (CD4 -CD8 -(double-negative, DN) sub-population), have been separated into four stages (DN1-DN4) based on expression of CD117, CD44 and CD25. A critical checkpoint known as β-selection occurs at the DN3 stage, phenotypically defined as being CD4 -CD8 -CD117 -CD44 -CD25 + 1-3 . It is at this stage that rearrangements within the TCRβ chain locus produce a diverse repertoire of TCRβ chains with varied specificities including non-functional variants. Productively rearranged TCRβ chains go on to form complexes with invariant pre-Tα chains and CD3/ζ/η components to form what is referred to as the pre-TCR [4][5][6] . Signalling from the pre-TCR promotes survival of thymocytes that would otherwise die by neglect, along with down-regulation of CD25 to produce DN4 (CD4 -CD8 -CD117 -CD44 -CD25 -) thymocytes. Pre-TCR signalling also triggers a series of other events critical for transition to double-positive (DP) cells expressing both CD4 and CD8 that includes TCRβ chain allelic exclusion, differentiation and proliferation. Only thymocytes that have undergone successful rearrangement at the TCRβ locus receive the appropriate proliferative, survival and differentiation signals necessary to transition from CD117 -CD44 -CD25 -CD4 -CD8 -TCRβ + (pre-DP) to CD4 + CD8 + (DP) thymocytes 7,8 . Numerous studies have shown that various mouse models including: RAG1 null , RAG2 null , TCRβ -/and pTα -/mice are unable to form the pre-TCR and consequently arrest development at the DN3 stage [9][10][11][12] .
Several signalling pathways including Notch, CXCR4, p53, NFAT and Egr-3 have been shown to play important roles during β-selection [13][14][15][16][17][18][19][20][21][22] . Specifically, one model of Notch function posits that this molecule works in delicate balance with the pre-TCR by providing signals associated with the proliferative-self renewal of early post β-selected thymocytes prior to pre-TCR directed differentiation. Notch activity is transient and eventually gives way to signals from the pre-TCR that promote differentiation involving loss of CD25 and pTα expression along with cell cycle arrest, RAG activation in preparation of mature TCRα gene rearrangement and, finally, up-regulation of CD4 and CD8 2, 3, 23 .
While it has been reported that constitutive activation of Akt in absence of pre-TCR and NOTCH signals can prevent programmed cell death of DN thymocytes independent of several Bcl-2 members there is evidence that Bcl-2 family members such as Mcl-1 play a crucial role in T cell survival, development and tumorigenesis [30][31][32][33][34][35] . GSK-3 regulates Mcl-1 stability through ubiquitin-mediated degradation 34,35 . Deletion of Mcl-1 impairs thymocyte survival causing reduced DP cell number and elevated apoptosis for DN and DP subsets. Only a small fraction of DP thymocytes receive the appropriate amount of TCR signal and are selected for further differentiation into mature CD4 and CD8 lineages in a process referred to as positive selection [36][37][38] .
In mammals, GSK-3 exists as two isoforms termed GSK-3α (51kDa) and GSK-3β (47kDa) encoded by distinct genes that are implicated in multiple pathways and diseases such as diabetes, Alzheimer's disease, bipolar disorder and cancer 39,40 . GSK-3β null embryos die late in development while GSK-3α null mice are viable though males are sterile [41][42][43] . Regulation of GSK-3 activity is atypical as it is negatively regulated in response to cellular signals leading to inactivation of GSK-3 and de-repression of certain substrates 44, 45 .
Wnt signalling regulates GSK-3 through a distinct mechanism and many studies employing gain-of-function and loss-of-function mutations have demonstrated that Wntsignalling and its central effector, β-catenin, figures prominently in developmental processes including thymopoesis 46-48 . In the thymus Wnt proteins, secreted mainly by the thymic epithelium, signal through the Frizzled/LRP5/6 co-receptor complex to divert GSK-3 activity to LRP5/6, allowing β-catenin to accumulate and associate with TCF/LEF transcription factors 49, 50 . TCF1/LEF1 null mice have impaired thymocyte development while ICAT, a natural inhibitor of β-catenin and TCF interaction, has been shown to block the DN to DP transition 51-53 . Signals from the pre-TCR and TCR lead to β-catenin stabilization and activation [54][55][56][57][58][59][60] .
Given the association of GSK-3 with signalling pathways shown to play key roles during thymocyte development, we crossed floxed alleles of GSK-3α and GSK-3β with transgenic animals expressing Cre-recombinase under control of the p56 lck proximal promoter 61 to selectively inactivate the GSK-3 isoforms early in T cell development.
While inactivation of one isoform or the other has little effect, loss of both GSK-3α and GSK-3β isoforms disrupted key regulators of thymocyte β-selection. We show that pre-DP's differentiate to form aberrant DP's that lack mature TCR without the normally requisite pre-TCR associated expansion and result in almost complete loss of both CD4 and CD8 mature T cell lineages.

Materials and Methods
Mice. Creation of GSK-3α flox/flox and GSK-3β flox/flox mice was previously described and were bred together to generate various allelic combinations of floxed GSK-3α and β    Thymocyte survival assay. To assess spontaneous death 1X10 6 thymocytes were plated in media for 24, 48 and 72 hrs. Thymocytes were then washed, blocked and stained for CD4 and CD8 to identify DP's. Apoptosis was assessed using the Annexin V-FITC apoptosis detection kit (BD cat # 556570). For anti-CD3-induced death thymocytes were plated at 0.5 X 10 6 cells/well of a 96 well plate (Nunc Maxisorb) coated with anti-CD3 antibodies (clone 145-2C11) and cultured for 24 and 48 hrs. For anti-Fas-induced death cells were plated similarly except in the presence of anti-Fas (2µg/ml) (BD cat # 554254) and protein G (2µg/ml) for 30 min. For dexamethasone-induced death, 2 X 10 6 thymocytes were cultured for 3 hrs with 4ml of media containing dexamethasone at various concentrations (1.0µM, 0.1µM, 0.01µM or 0µM). Following these treatments cells were washed, blocked and stained for CD4 and CD8 to identify DP and DN cells.
Apoptosis was also assessed using the Annexin V-FITC apoptosis detection kit (BD cat # 556570).

Real Time PCR.
Mouse CD4 Dynabeads (Invitrogen cat# 114.45D) were used to enrich the CD4 -CD8 -(DN) fraction of thymocytes from 3-4 wk old mice which were then sorted for the CD117 -CD25lin neg (DN4) subpopulation. DP thymocytes were sorted according to CD4 and CD8 co-expression. 7AAD was included in sorts to ensure viability. RNA was extracted from CD117 -CD25lin neg (DN4) and CD4 + CD8 + (DP) thymocytes using the RNeasy Plus Mini Kit (Qiagen) and SuperScript lll reverse transcriptase was used to make cDNA using the protocol supplied by Invitrogen (cat # 18080-093). Power SYBR Green PCR master mix (Applied Biosystems cat # 4367659) was used to prepare samples before running on an Applied Biosystems 7500 Real-Time PCR System. Analysis was performed using the comparative CT method (DDCT Method) described by Applied Biosystems with GAPDH used as the endogenous reference gene. PCR product values were expressed relative to LckCre -GSK-3 control samples. An average of three wells was used to calculate PCR product values for each experiment and each experiment was repeated at least four times with standard error used to measure variability.

GSK-3β
Floxed GSK-3α and GSK-3β mice were crossed to transgenic animals expressing Cre-recombinase under control of the p56 lck proximal promoter. Inter-crosses between floxed LckCreGSK-3α and floxed LckCreGSK-3β mice produced T-cell conditional mice with disrupted GSK-3 expression according to various allelic combinations of GSK-3α and GSK-3β to allow investigation of potential isoform-specific differences and/or gene dosage effects on thymocyte development (Suppl. Fig. 1).
Significant excision of exon 2 of GSK-3α and GSK-3β was evident by the DN3 stage of thymocyte development for LckCre + GSK-3αβ fl/fl mice. This is in keeping with the well-

Inactivation of both GSK-3α and GSK-3β causes marked loss of mature T cells, reduced thymocyte cell number and accumulation of positive selection intermediates
Comparison between LckCre + GSK-3αβ fl/fl and littermate LckCre -GSK-3αβ fl/fl control mice at 3 wks of age showed significant loss of mature T cells in mice having lost both isoforms of GSK-3 when first analyzed in terms of CD4 and CD8 surface expression in the thymus. We also noted an increase in the proportion of DN cells (13.2% vs. 3.5%) indicating a potential block in development prior to DP differentiation. Subsequent analysis of the CD3 + CD24subpopulation, which identifies mature thymocytes revealed the extent to which mature CD4 and CD8 (SP) thymocytes are missing from the LckCre + GSK-3αβ fl/fl thymus (CD4: 3.7% vs. 61.8% and CD8: 5.4% vs. 34.7%). Furthermore, the majority of CD3 + CD24thymocytes from LckCre + GSK-3αβ fl/fl mice continue to express both CD4 and CD8, which is another indication of aberrant development. Analysis of mice at 6-8 wks of age showed similar findings (Fig. 1a, 1b, 1d and Suppl. Fig. 2c).
Thymocyte-specific loss of either GSK-3α or GSK-3β alone had no discernible effect on thymocyte development as defined through CD4 vs. CD8 expression. Notably even the presence of a single allele of either GSK-3α (GSK-3α fl/fl β fl/wt (α 3/4 )) or GSK-3β (GSK-3α fl/wt β fl/fl (β 3/4 )) in the presence of the p56 lck Cre transgene produced mice with entirely normal ratios and cell numbers of DN, DP, CD4 (SP), CD8 (SP) sub-populations along with normal levels of CD3 and TCR expression in the thymus, compared to littermate controls. These data show functional redundancy between GSK-3α and GSK-3β with a single allele of either GSK-3α or GSK-3β sufficient to mediate the complex signalling outcomes governing thymocyte development. Given the severity of the phenotype observed for LckCre + GSK-3αβ fl/fl mice versus the lack of obvious phenotype for either of the singly deleted animals, we focused further characterization on the LckCre + GSK-3αβ fl/fl mice (Fig. 1c, Suppl. Fig. 3a and data not shown).
Examination of total thymic cell number revealed a ~3-fold loss of cellularity in 3 wk old LckCre + GSK-3αβ fl/fl mice compared to littermate LckCre -GSK-3αβ fl/fl mice. This was due to an absolute loss of DP and SP thymocytes since the number of thymocytes in the 3 wk old DN compartment did not change relative to LckCre -GSK-3αβ fl/fl mice.
Further comparison of DP and DN cell numbers between 3, 6 and 14 wk old animals revealed a striking increase in DP cellularity in LckCre + GSK-3αβ fl/fl mice at six weeks of age, in addition to an increase in DN cells. However, these animals remained unable to produce mature T cells as judged by comparison of CD4 + and CD8 + thymocyte numbers within the CD3 + CD24sub-population. LckCre + GSK-3αβ fl/fl DP cell numbers did not continue to rise unchecked. There was a return to lower LckCre + GSK-3αβ fl/fl DP cell numbers, comparable to LckCre -GSK-3αβ fl/fl animals by 14 wks of age, suggesting that processes governing age-associated thymic atrophy remained intact (Fig. 1d) 67 . We also observed an increase in the fraction of CD4 hi CD8 lo cells in LckCre + GSK-3αβ fl/fl mice at 3 wks of age (8.7% vs. 2.7%) suggesting that thymocytes of LckCre + GSK-3αβ fl/fl mice are unable to complete positive selection, resulting in accumulation of positive selection intermediates (Fig. 1e).

Reduction of LckCre + GSK-3αβ fl/fl Pre-DP thymocytes expressing TCRβ chain and escape of checkpoints for productive TCRβ-chain rearrangement during β-selection
The proportion of DN thymocytes in 3 wk old LckCre + GSK-3αβ fl/fl mice was significantly higher when compared with LckCre -GSK-3αβ fl/fl littermates suggesting a partial block or delay in transition from the DN to DP stages (13.2% vs. 3.5%) (Fig. 1a).
Further characterization of DN thymocytes into DN1-DN4 subpopulations was established according to CD117 vs. CD25 expression within the lin neg population as described in Methods. LckCre + GSK-3αβ fl/fl mice were found to have an accumulation of DN3 thymocytes indicating impairment of DN3 to DN4 transition (49% vs. 29.1%).
TCRβ chain production did not appear to be compromised at the DN3 stage although there was a small increase in pre-Tα expression (DN3: 10.1% vs. 4.9%) (Fig. 2a, 2c).
Subdivision of DN3 thymocytes into pre-(DN3a) and post-(DN3b) β-selected cells according to CD71 surface expression and forward scatter also failed to reveal any notable impairments during early β-selection as judged by the proportion of DN3b vs.
Only αβ thymocyte development appeared impaired since expression of γδ TCR was undiminished. In fact, we found the proportion of TCRγδ thymocytes within the DN4 sub-population to be higher for LckCre + GSK-3αβ fl/fl mice than for LckCre -GSK-3αβ fl/fl mice (38% vs. 18%) which is reflected in their increased cell number. This may reflect impaired αβ thymocyte production skewing the normal ratio of αβ to γδ thymocytes, rather than enhancement of γδ T cell production per se (Fig. 2e).
Given that LckCre + GSK-3αβ fl/fl pre-DP thymocytes had low levels of TCRβ chain production with accumulation of aberrant CD3 + CD24 -CD4 + CD8 + (DP's), we hypothesized that loss of GSK-3 would allow thymocytes to bypass the strict requirements for productive TCRβ chain rearrangements during β-selection and allow differentiation to DP thymocytes (Fig. 1b, 2a). To test this LckCre + RAG1 -/-GSK-3αβ fl/fl mice were generated by inter-crossing the LckCre + GSK-3αβ fl/fl mice with RAG1 -/mice. RAG1 (recombination activation gene 1) is required for V(D)J recombination at the TCRβ chain locus 68 . Without RAG1 -/thymocytes cannot assemble the required pre-TCR complex and experience an acute block at the DN3 stage 9 . Here, we show that loss of GSK-3 in the absence of RAG1 was permissive for differentiation from DN3's to DP thymocytes (16.6% vs. 0.1%) (Fig. 2f). We also found an age-dependent increase in thymocyte numbers for LckCre + GSK-3αβ fl/fl RAG1 -/mice (Fig. 2g). This finding underscores the extent to which loss of GSK-3 can impact the regulation of β-selection by overriding requirements for competent TCRβ chain rearrangement and pre-TCR dependent differentiation. Of note, a significant proportion of CD4 and CD8 SP's were also generated which were CD3 + /CD24 -, markers normally associated with maturity.
However, these cells lacked TCRβ chain surface expression and therefore remained defective ( Fig. 2f and data not shown).  (Fig. 3a). We also assessed the TCR status of DP's in LckCre + GSK-3αβ fl/fl mice. Thymocytes from 3 wk old mice were stained for surface expression of TCRβ along with CD4 and CD8. Loss of TCRβ expression persisted throughout DP developmental stages, which we defined as CD4 hi CD8 hi (25% vs 66% and median 0.32 vs. 1.43), CD4 lo CD8 lo (45% vs. 95% and median 0.51 vs. 5.58) and CD4 hi CD8 lo (34% vs. 90% and median 0.44 vs. 11.42) respectively (Fig. 3b). Together, these data support the view that pre-DP cells in LckCre + GSK-3αβ fl/fl mice bypass normal β-selection checkpoints for proper TCRβ chain rearrangement and expression prior to becoming DP thymocytes.

Rescue of defective positive selection through enforced transgenic TCR expression
Loss of TCR surface expression on thymocyte positive selection intermediates (CD4 hi CD8 lo ) of the LckCre + GSK-3 double knockout mice suggests impaired TCRmediated signalling during the positive selection process itself (Fig. 4a). Normally, CD5 and CD69 expression is directly correlated with signalling through the mature TCR 69 . We observed a significant reduction in CD69 and CD5 expression in the positive selection intermediates (CD4 hi CD8 lo ) in six wk old LckCre + GSK-3αβ fl/fl mice, a finding consistent with impaired TCR-dependent signalling associated with positive selection 70 (Fig. 4a). Lack of TCR-mediated signalling may help to explain accumulation of CD4 hi CD8 lo positive selection intermediates as these cells failed positive selection toward mature CD4 and CD8 SP cells (Fig. 1e). We also noted a small population of CD24 -/CD4 + and CD24 -/CD8 + SP thymocytes that appear to have relatively high levels of TCRβ and CD3 expression (data not shown). This result suggests that it could be the loss of mature TCR expression that contributes to the inability of most LckCre + GSK-3αβ fl/fl DP thymocytes to pass through positive selection to become mature SP cells.
To test this idea we used a P14 transgenic TCR model that enforces P14-specific TCR expression and induces positive selection to the CD8 lineage 63 . When supplied with a functional, mature TCR in the form of the P14 transgene, LckCre + P14 + GSK-3αβ fl/fl mice produced a greater proportion of CD8 SP's compared to LckCre + P14 -GSK-3αβ fl/fl mice (8.7% vs. 2.2%) (Fig. 4b). Further characterization of mature thymocytes (CD3 + /CD24 -) in these animals revealed that substantial numbers of mature CD8 SP cells were being generated with expression of the P14 TG (42.8% vs. 8.6%). Co-expression of the Vβ 8.1 and Vα2 T-cell receptors expressed by the P14 transgene confirmed that these CD8 cells were of P14 origin (96.9%). An increase in absolute numbers of CD8 + /CD3 + /CD24cells was also noted for LckCre + P14 + GSK-3αβ fl/fl thymocytes compared to LckCre + P14 -GSK-3αβ fl/fl thymocytes that lacked the P14 transgene (Figs.   4b, 4d and suppl. Fig. 3c). Together these results indicate that the presence of a functional TCR can at least partially rescue the inability of thymocytes lacking GSK-3 to undergo positive selection.

Thymocytes lacking GSK-3 express high levels of β-catenin with reduced EGR3 and the Notch targets DTX1 and HES1
Given the established role of GSK-3 in suppressing Wnt signalling DP thymocytes were probed for cytoplasmic β-catenin (non-cadherin-associated fraction).
β-catenin stabilization has effects on Notch signalling as well as EGR expression 57,59 . Since both Notch and EGR activity are associated with the pre-DP (DN4) to DP expansion phase following β-selection we used real-time PCR analysis to assess their gene targets in sorted DN4 subpopulations 2, 21, 22, 30 and observed a 2.5-fold reduction of DTX1, a 2-fold reduction in HES1 and a 3-fold reduction of Egr3 transcripts in LckCre + GSK-3αβ fl/fl DN4 thymocytes compared to control littermates. These findings are consistent with the loss of pre-DP proliferation and DP hypocellularity seen in 3 wk old LckCre + GSK-3αβ fl/fl mice ( Fig. 1d and Fig. 5b). Notch activity, as indicated by DTX1 and HES1 transcript levels, was also significantly reduced in LckCre + GSK-3αβ fl/fl DP thymocytes (10-and 2-fold, respectively). Since we observed enhanced DP survival and an increase in the Bcl-2 family survival protein Mcl-1, we also examined Bcl-XL, another Bcl-2 family member that has been linked to DP survival, but found no significant change (Fig. 5c and data not shown).

LckCre + GSK-3αβ fl/fl mice produce anergic peripheral T cells and succumb to lymphomas
Peripheral lymphoid splenic CD4 and CD8 SP's were significantly reduced in both percentage of population and cell number in GSK-3 knock-out mice but expressed high levels of CD3 and TCR (Fig. 6a, 6b and suppl. Fig. 4b). However, these lymphocytes failed to proliferate in response to either anti-CD3 +/-IL-2 or anti CD3/CD28 +/-IL-2 mediated stimulation, as measured by the CFSE dye incorporation, indicating TCR mediated signaling is impeded ( Fig. 6c and data not shown). LckCre + GSK-3αβ fl/fl mice developed peripheral lymphomas with a penetrance of 77% and reached endpoint at a median of 157 days in comparison to LckCre -GSK-3αβ fl/fl and LckCre + control mice, which showed no significant decrease in survival within a comparable time period. LckCre + RAG -/-GSK-3αβ fl/fl mice had a median endpoint survival of 131 days with a penetrance of 43% and therefore do not show the enhanced survival observed for stabilized β-catenin mice within a RAG null background 60 (Fig. 6d).
Furthermore, unlike stabilized β-catenin, thymic lymphomas occurred with much lower frequency, presented heterogeneous CD4 and CD8 expression and exhibited high levels of Mcl-1 ( Fig. 6e and data not shown). Peripheral lymphomas also had heterogeneous CD4 and CD8 expression and were determined to be oligoclonal through PCR analysis of the 22 TCRVβ clone families within the CDR3 region, in addition to screening for overrepresentation of TCRVβ gene families by flow cytometry (Fig. 6f, suppl. Fig. 4c and data not shown).

Discussion
Given the central role GSK-3 in mediating PI3K and Wnt signals along with implication in a variety of signaling pathways which impact b-selection, the goal of this study was to identify steps in early T cell development and beyond affected by loss of GSK-3 13-20; 25, 75 . Indeed, pharmacological inhibition of GSK-3 within an in vitro, stromal-free cell culture system have suggested that DP thymocytes could be produced in the absence of preTCR and Notch1-mediated signalling 16 .
Here, we found that the normally strict requirements for successful β-selection are We observed a block during the DN3-DN4 transition in LckCre + GSK-3ab fl/fl mice with significant loss of TCRb chain from both the surface and inside the cell, but without loss of TCRgd expression or numbers (Fig 2a, b and e). Furthermore, Notch targets DTX and HES1 were reduced suggesting lower Notch1 activity, followed by a four-fold reduction in DP cellularity in 3 wk old LckCre + GSK-3ab fl/fl mice (Fig 5b and   Fig 1d). Resulting DP thymocytes expressed significantly less TCRb chain along with a reduction in generation of productive TCRb V(D)J fragments (Fig 3a and b). It has been reported that b-selected DN3b and DN4 cells can forgo proliferation and differentiate into DP cells in absence of Notch signals in vitro 2,78 . Taken together, one possibility is that the strict requirements for productive TCRb chain expression during b-selection are bypassed promoting aberrant differentiation without prerequisite clonal expansion of bselected precursors. Consistent with this loss of proliferative burst was an observed decrease in EGR3 expression in pre-DP DN4 cells, since EGR3 is known to play a proliferative role during b-selection (Fig 5b) 21,22 . The fact that DN thymocytes show elevated survival argues against the idea that loss of DP cellularity is due to enhanced cell death during transition (Fig 3d). Along with pre-DP self-renewal and expansion, survival signals are necessary before thymocytes can complete transition to DP's. Several studies have shown components of PI3K signaling to be important for pre-TCR mediated survival, relevant since GSK-3 is a direct target of PI3K-PKB/AKT signaling 24,25,29 . Specifically, loss of various combinations of PKB/AKT isoforms has been shown to affect pre-TCR mediated survival and result in DP hypocellularity due to elevated apoptosis, while ablation of PTEN, a negative regulator of PKB/AKT, has the opposite effect. Furthermore, studies employing a knock-in mutation of PDK1 (PDK1 L155E ), which is restricted to signaling through PKB/AKT, suggest that PDK1-mediated differentiation, but not proliferation, signals through PKB/AKT during b-selection [26][27][28][29]79 . Supporting this finding, DN3 cells expressing an activated mutant of PKB/AKT were shown to have elevated survival and to not require Notch signals for progression to the DP stage using the OP-9 cell culture system. Constitutively active Akt1 (Myr-Akt1) was also able to rescue DP production in RAG null mice, which have curtailed TCRb-chain rearrangements and therefore do not possess a functional pre-TCR 26,27 . To examine whether LckCre + GSK-3ab fl/fl pre-DP's could escape requirements for pre-TCR expression during b-selection we tested the effect of a RAG1 null mutation which displays a severe block in pre-DP-DP transition 10,80 .
LckCre + GSK-3αβ fl/fl RAG1 null mice exhibited robust differentiation to DP's despite impaired TCRβ-chain rearrangement suggesting that LckCre + GSK-3αβ fl/fl pre-DP's can bypass the normal requirements of pre-TCR expression during b-selection (Fig 2f). We observed a similar age-dependent increase in LckCre + GSK-3αβ fl/fl RAG1 null thymocytes to that found in activated PKB/Akt RAG1 null mice (Fig 2g).
Sen and colleagues have shown that β-catenin is transiently expressed during βselection and that it facilitates pre-TCR signalling. They concluded that under normal circumstances β-catenin expression is tightly regulated and must be down-regulated following β-selection, otherwise sustained expression impairs development through oncogene-induced senescence followed by p53-dependent apoptosis, which, they suggest, contributes to reduced DP cell numbers observed in β-catenin-Tg mice 57,81 . Successful double-strand break repair is required during TCRb V(D)J recombination and is dependent upon p53 to mediate cell cycle arrest required for productive V(D)J rearrangement. Loss of p53 function in this context has been associated with genomic instability and lymphomagenesis in both mice and humans, along with DN-DP transition 17,82 . Interestingly GSK-3 is known to regulate p53 activity through phosphorylation of the Tip60 histone acetyltransferase 18 . It is therefore tempting to speculate that loss of GSK-3 may contribute to escape of LckCre + GSK-3αβ fl/fl preDP's from b-selection through lack of p53-mediated cell cycle arrest and apoptosis, thereby contributing to lymphomagenesis.
Given the potential of GSK-3 to mediate survival and our observation that DP cellularity in LckCre + GSK-3αβ fl/fl mice rises substantially by seven weeks, we noted substantial resistance of these DP cells to apoptotic stimuli, likely contributing to their increased number over time (Fig 3c). This contrasts with other models that exhibit high levels of b-catenin, notably Ctnnb1 loxpex3 , and Apc lox/lox468 , which each show continued DP hypocellularity attributed to increased apoptosis 59,60 . In keeping with our findings of enhanced LckCre + GSK-3αβ fl/fl DP cell survival, we found high levels of the Bcl-2 family survival protein Mcl-1 (Fig 3e). Mcl-1 is normally expressed throughout thymocyte development and conditional deletion of Mcl-1 impairs DN/DP survival. It has also been suggested that MCL-1 contributes to pre-TCR-mediated survival 31,32 . Mcl-1 stability can be regulated via GSK-3, which supports the idea that survival of LckCre + GSK-3αβ fl/fl DP cells is GSK-3-dependent and can override pro-apoptotic signals derived from β-catenin stabilization 35 .
We noted that although DP cells were generated in LckCre + GSK-3αβ fl/fl mice, few mature CD4 or CD8 SP's resulted and their numbers did not increase as the DP thymocyte population rose over time (Fig 1a, b, c, and Suppl Fig 2c). Loss of mature CD4 or CD8 SP's was also evident in peripheral lymphoid organs and these cells failed to respond to TCR stimulation, even though they expressed relatively high levels of TCR and CD3 (Fig 6a, b and Suppl. Fig 4b). Given that positive selection requires appropriate levels of mature αβTCR it is not surprising that DP thymocytes of LckCre + GSK-3αβ fl/fl mice fail to produce significant SP numbers and accumulate as positive selection intermediates (Fig 1e). Previous studies have indicated that deregulation of both β-catenin and Notch can impact DP-SP differentiation and we found significantly less Notch activity, along with high levels of β-catenin, throughout DP development including the positive selection intermediate stage. Furthermore, b-catenin has been shown to be post-transcriptionally regulated by TCR signaling through PI3K involving GSK-3 (Fig 5c   and Suppl. 4a) 46,83,84 . Accumulated TCR lo CD4 hi CD8 lo positive selection intermediates lacked CD5 or CD69 expression, which is a further indication of compromised TCR signalling. This suggests that while positive selection may be initiated, TCR functions are impaired resulting in failure of DP's to undergo positive selection. However, we did find that, while significantly reduced in number, some CD4 + CD24and CD8 + CD24 -SP thymocytes were generated in LckCre + GSK-3αβ fl/fl mice with relatively high levels of mature TCR as measured through TCR b chain and CD3 expression (data not shown).
We therefore tested the possibility that enforced TCR expression might restore aspects of positive selection blocked by loss of GSK-3 by using a transgenic P14 cassette used in previous studies of positive selection to supply a functional TCR 73 . This TCR cassette directs positive selection of CD8 SP's and generated a significant proportion of P14specified mature CD8 SP thymocytes in LckCre + P14 + GSK-3αβ fl/fl mice. We observed only modest restoration of SP CD8 numbers in LckCre + P14 + GSK-3αβ fl/fl mice but this may be attributed to a documented lack of engineered TCR's ability to fully restore the proliferative burst associated with pre-DP expansion 7,25,85,86 .      Suppl. Fig. 4. A. Intracellular β-catenin expression expressed as median fluorescent intensity within gated thymocyte sub-populations. B. Splenic CD4 + and CD8 + SP populations from LckCre + GSK-3αβ fl/fl mice had comparable levels of TCR and CD3 expression to LckCre -GSK-3αβ fl/fl mice even though numbers were significantly reduced. C. FACS of 5 individual LckCre + GSK-3αβ fl/fl lymphomas reveal heterogeneity as judged by CD3, CD4, CD8, and CD19 markers.