Non-canonical antagonism of PI3K by the kinase Itpkb delays thymocyte β-selection and renders it Notch-dependent

β-selection is the most pivotal event determining αβ T cell fate. Here, surface-expression of a pre-T cell receptor (pre-TCR) induces thymocyte metabolic activation, proliferation, survival and differentiation. Besides the pre-TCR, β-selection also requires co-stimulatory signals from Notch receptors - key cell fate determinants in eukaryotes. Here, we show that this Notch-dependence is established through antagonistic signaling by the pre-TCR/Notch effector, phosphoinositide 3-kinase (PI3K), and by inositol-trisphosphate 3-kinase B (Itpkb). Canonically, PI3K is counteracted by the lipid-phosphatases Pten and Inpp5d/SHIP-1. In contrast, Itpkb dampens pre-TCR induced PI3K/Akt signaling by producing IP4, a soluble antagonist of the Akt-activating PI3K-product PIP3. Itpkb-/- thymocytes are pre-TCR hyperresponsive, hyperactivate Akt, downstream mTOR and metabolism, undergo an accelerated β-selection and can develop to CD4+CD8+ cells without Notch. This is reversed by inhibition of Akt, mTOR or glucose metabolism. Thus, non-canonical PI3K-antagonism by Itpkb restricts pre-TCR induced metabolic activation to enforce coincidence-detection of pre-TCR expression and Notch-engagement. DOI: http://dx.doi.org/10.7554/eLife.10786.001

However, they can also cause rheumatoid arthritis and other autoimmune diseases by attacking healthy tissue. T cells recognize target cells via receptor proteins on their surface. To maximize the variety of infections and cancers our immune system can recognize, we generate millions of T cells with different T cell receptors every day.
To ensure T cells work correctly, T cell receptors are tested at various checkpoints. The first checkpoint involves a process called beta (b) selection, during which T cells produce their first T cell receptor -the so-called pre-T cell receptor. This receptor causes T cells to divide and mature, and sets their future identity or "fate". To complete b-selection, T cells must also receive signals from another surface receptor -one that belongs to the Notch family, which determines cell fate in many different tissues.
The Notch receptor and the pre-T cell receptor both activate an enzyme called PI3K -a key mediator of b-selection. But the pre-T cell receptor also activates another enzyme called Itpkb that is required for T cell development. Westernberg, Conche et al. have now investigated how these different proteins and signaling processes work and interact during b-selection, using mice that lack several immune genes, including the gene that produces Itpkb.
The results of the experiments show that during b-selection, Itpkb limits the ability of PI3K to activate some of its key target proteins. This "dampened" PI3K signaling ensures that both the pre-T cell receptor and the Notch receptor must be activated to trigger T cell maturation. Without Itpkb, b-selection can occur in the absence of Notch signaling.
As Notch signaling is important for determining the fate of many different cell types, Westernberg, Conche et al.'s findings raise the possibility that Itpkb might also regulate cell fate determination in other tissues. Moreover, Itpkb may suppress tumor development, because excessive PI3K signaling drives many cancers.
Here, we present data which suggest that this non-canonical mechanism restricts pre-TCR induced pro-metabolic PI3K/Akt signaling to limit the kinetics and enforce the Notch-dependence of b-selection. Itpkb -/-DN3 cells were pre-TCR hyperresponsive with Akt/mTOR hyperactivation and evidence for metabolic hyperactivity. They showed an accelerated and Notch independent, but pre-TCR dependent differentiation to the DP stage. Pharmacologic inhibition of Akt, mTOR or glucose metabolism restored wildtype (WT) developmental kinetics and Notch-dependence of Itpkb -/-DN3 cells.
We next analyzed whether the reduced DN4 and ISP cell numbers in Itpkb -/mice might reflect reduced proliferation or viability. But similar Ki67-staining and in vivo BrdU incorporation suggest comparable steady-state proliferation of all thymocyte subsets between genotypes ( Figure 3C). Similar AnnexinV staining suggests comparable viability ( Figure 3D).

Itpkb restricts pre-TCR signaling to delay b-selection and render it Notch-dependent
To determine if the Akt/mTOR and metabolic hyperactivation of Itpkb -/pre-TCR + DN3 cells causes their accelerated development to DP cells, we next studied if treatment with inhibitors of Akt (Aktinhibitor VIII), mTOR (rapamycin) or glucose-metabolism (2-deoxy-D-glucose, 2DG) could reverse the increased DP cell generation from equal numbers of sorted Itpkb -/versus Itpkb +/+ DN3 cells on OP9DL1 stroma. Strikingly, all three treatments yielded Itpkb -/-DP cell proportions similar to those of untreated or carrier-treated WT controls after 4-day co-culture ( Figure 9A, Figure 9-figure supplement 1A,C,E). As expected, the treatments also reduced WT DP cell generation below untreated controls. Their reduced efficacy towards Itpkb -/-DN cells is expected, as these had increased amounts of the respective active inhibitor-targets ( Figure 8A). Similarly complete rapamycin reversal of the accelerated DN-to-DP development in Itpkb -/versus Itpkb +/+ FTOC confirmed these important findings in a less reductionist system ( Figure 7C,D). These data suggest that the hyperactive Akt, mTOR and glucose metabolism of Itpkb -/-DN3 cells contribute to their accelerated DN-to-DP development.
Notch signaling depends on its cleavage by cellular g-secretases (Wong et al., 2004). To corroborate our findings in vivo, we thus analyzed DN3 cell maturation in Itpkb -/-Rag2 -/versus Rag2 -/mice treated p.o. with vehicle or the g-secretase inhibitor LY-411,575 for two days post a-CD3 injection. P.o. administered 10 mg/kg LY-411,575 potently inhibited g-secretase function in mice and impaired DN thymocyte maturation into ab T cells with a particularly profound loss of DP cells (Wong et al., 2004). We found that LY-411,575 strongly impaired the a-CD3 induced DN cell development into ISP and DP cells in Rag2 -/but not Itpkb -/-Rag2 -/mice ( Figure 10). Hence, Itpkb-loss reduces the Notch-dependence of DN thymocyte development to DP cells both in vitro and in vivo.

Discussion
Here, we identify Itpkb as a novel pre-TCR effector which restricts the kinetics of b-selection and establishes its Notch-dependence. Itpkb -/mice show a cell-autonomously accelerated and Notch independent, but pre-TCR dependent DN3-to-DP cell differentiation associated with DN3 cell pre-TCR hyperresponsiveness, Akt/mTOR hyperactivation and evidence for metabolic hyperactivity. Pharmacologic inhibition of Akt, mTOR or glucose metabolism restored WT kinetics and Notchdependence of Itpkb -/-DN3-to-DP cell development.
The increased DP cell production from Itpkb -/versus WT DN3 thymocytes in several different in vivo and in vitro models without concomitantly increased proliferation or viability suggests accelerated developmental kinetics, consistent with our mathematical simulations. We propose that this is caused by pre-TCR hyperresponsiveness based on the Akt/mTORC1 hyperactivation, increased Nr4a1/Nur77-GFP expression and hyperinduction of activation markers in Itpkb -/versus WT DN3 and later stage thymocytes, and on the phenotype reversal by Akt/mTORC1 and metabolic inhibitors. Importantly, the Nr4a1/Nur77-GFP and activation marker hyperinduction in Itpkb -/-DN3 cells indicate increased transcriptional responses. This might accelerate development by inducing required amounts of cell fate determinants earlier in Itpkb -/-DN3 cells than in WT cells. Alternatively or in addition, Itpkb-loss might increase the number of DN3 cells responding to pre-TCR signals or developmental cues present even in the reductionist OP9-DL1 cell co-culture system. Our present data do not allow us to discern the relative contributions of accelerated kinetics of cellular signaling Figure 11. Antagonistic signaling by PI3K and Itpkb controls the kinetics and Notch-dependence of b-selection. (A) We propose a model in which pre-TCR and Notch signaling both activate PI3K to produce PIP 3 in DN3/DN3-4 cells. PIP 3 then recruits and activates Akt to increase glucose metabolism via the Akt/mTOR pathway. This is required for DN3-to-DP cell differentiation. However, pre-TCR signaling also activates Itpkb to produce IP 4 , which competes with PIP 3 for Akt PH domain binding and limits Akt recruitment, Akt and mTOR activation in pre-TCR expressing DN3/DN3-4 cells. IP 4 may have additional effectors, indicated by the question mark. By limiting downstream glucose metabolism, this "IP 4 brake" delays the kinetics of bselection and renders this process dependent on Notch costimulation. (B) Without Itpkb, IP 4 no more dampens Akt activation and pre-TCR signaling alone sufficiently activates Akt/mTOR signaling to trigger DP cell development in the absence of Notch engagement. (C) In the presence of Notchsignals, Akt is now hyperactivated and causes an accelerated DN3-to-DP cell differentiation. DOI: 10.7554/eLife.10786.015 events versus increased proportions of responding cells due to lowered pre-TCR signaling thresholds or enhanced sensitivity to developmental cues upon Itpkb-loss. Distinguishing between these possibilities will require future detailed studies of the effects of Itpkb-loss on the sizes (amplitudes and proportions of cells responding), kinetics (rate constants) and shapes (analog, digital) of pre-TCR signaling events, transcriptional and functional responses in populations of individually analyzed DN3 cells, combined with mathematical simulations.
Contrasting with Pten-loss, Notch or Akt hyperactivity, Itpkb-loss accelerates DN3 cell differentiation without significant effects on proliferation and viability, and overcomes the Notch dependence but not the pre-TCR dependence of b-selection. This is evidenced by the lack of accumulating intracellular TCRb -DP cells in Itpkb -/mice, and of post-DN3 cells in Itpkb -/-Rag2 -/mice ( Figures 2D, 6C, D). We speculate that this reflects the need for TCR signals to activate Itpkb and produce IP 4 (Chamberlain et al., 2005;Wen et al., 2004;Pouillon et al., 2003). By abrogating pre-TCR induced IP 4 -inhibition of pre-TCR and Notch induced Akt/mTOR signaling, Itpkb-loss mimics the effects of Pten-loss or dominant active Akt1 expression. This overcomes Notch-requirements and accelerates differentiation but not proliferation, because Notch-induction of c-Myc is PI3K-independent (Wong et al., 2012). The surprising lack of increased DN3/DN4 cell viability in Itpkb -/mice might reflect differing degrees of Akt/mTOR hyperactivation in Pten -/-, dominant active Akt1-expressing and Itpkb -/-DN3/DN4 cells, consistent with unaltered development of Inpp5d/SHIP1 -/thymocytes (Kashiwada et al., 2006). Altogether, the largely restored developmental kinetics and Notch-dependence of Itpkb -/-DN3 cells by treatment with Akt, mTORC1 or metabolic inhibitors support contributing roles for the Akt/mTOR-hyperactivity. Future studies with sub-optimal Akt/mTOR inhibitor concentrations not affecting WT thymocytes but still reversing the Itpkb -/phenotype, with complex genetic models and with inhibitors of b-selection effectors unaffected by Itpkb will be needed to more conclusively distinguish between specific causative roles for the Akt/mTORC1 and metabolic hyperactivity and mere remaining sensitivity of Itpkb -/-DN thymocytes to inhibition of this particular pathway. Such studies can also address whether additional mechanisms contribute to the b-selection phenotype of Itpkb -/mice.
Contrasting with dominant active Akt1 expression or loss of Pten, which has high constitutive PIP 3 -phosphatase activity (Leslie and Foti, 2011), Itpkb-loss cannot replace pre-TCR signals because Itpkb is inactive without them, so its loss has no further effect. Itpkb-loss might also reduce less essential positive Itpkb roles in pre-TCR signaling, such as augmenting PLCg1/Erk activation by Itk . Indeed, TCRb + DN3 cells from Itpkb -/vs. WT mice tended to have mildly reduced Erk activity ( Figure 8A). Erk signaling is required for DN cell proliferation and differentiation (Kortum et al., 2013). The mild defects in Itpkb -/mice are consistent with the only minor role of Itk in b-selection (Lucas et al., 2007) and the unaltered DN cell proliferation.
Hyper-upregulation of Glut1, CD71 and cell-size in Itpkb -/-TCRb + DN3 cells and reversal of their accelerated, Notch-independent differentiation by the glycolytic inhibitor 2DG suggest that Itpkb controls b-selection by ultimately restricting DN3 cell metabolic activation. Similar Akt-inhibitor and rapamycin effects indicate a causative role for Akt/mTOR hyperactivation. Akt promotes metabolism by increasing Glut1 expression and activity, regulating enzymes in glucose and lipid metabolism and promoting mTOR-dependent protein translation (Juntilla et al., 2007). In DN cells, Pdk1/Akt/ mTORC1 also upregulate surface CD71 and CD98 downstream of pre-TCR and Notch (Kelly et al., 2007;Fayard et al., 2010). Thus, upregulated iron uptake, glucose and amino acid metabolism and protein biosynthesis might all contribute to the accelerated, Notch-independent development of Itpkb -/-DN3 cells.
Excessive Notch signaling causes thymocyte transformation and T-ALL. This is augmented by pre-TCR signals (Campese et al., 2006;Fayard et al., 2010). Excessive Akt activity in thymocytes due to PI3K hyperactivity, Pten inactivation or dominant-active Akt1 expression causes leukemia/lymphoma (Aifantis et al., 2006;Fayard et al., 2010). The intermediate b-selection phenotype of Itpkb-loss between those of Pten-loss (Hagenbeek et al., 2004;Kelly et al., 2007;Shiroki et al., 2007;Wong et al., 2012) and Inpp5d/SHIP1-loss (Kashiwada et al., 2006) raises the possibility that IP 3 3kinases could have tumor suppressor functions by limiting Akt signaling. But we have not seen thymocyte neoplasia or accumulation of intracellular TCRb -DP cells in Itpkb -/mice. One possible explanation consistent with low residual IP 3 3-kinase activity and IP 4 -production in Itpkb -/thymocytes (Wen et al., 2004;Pouillon et al., 2003) is partial Itpkb redundancy with other IP 3 3-kinases. Moreover, their premature lethality due to infections (Pouillon et al., 2003) and anemia (Siegemund et al., 2015) limits aging studies with Itpkb -/mice. It will be important to re-assess in a germ-free vivarium whether conditional Itpkb-disruption in thymocytes causes T-ALL as the mice age, or on a sensitized Trp53 -/background as seen for p65 PI3K transgenics (Borlado et al., 2000). Then again, IP 4 -loss might simply not augment PIP 3 cellular activity sufficiently to transform thymocytes, reminiscent of Inpp5d/SHIP1 -/--loss (Kashiwada et al., 2006). Also, the potential reduction of required Akt-unrelated IP 4 functions such as promoting Itk/Erk signaling  might prevent thymocyte transformation in Itpkb -/mice. Clearly, more studies are needed to assess the tumor suppressor potential of Itpks.
By unveiling Itpkb antagonism with PI3K/Akt/mTOR signaling as a key determinant of the kinetics and Notch-dependence of thymocyte b-selection, our findings expand our limited knowledge about physiological IP 3 3-kinase functions (Sauer and Cooke, 2010). They unveil a novel molecular mechanism that integrates pre-TCR signaling with costimulatory Notch signaling to specifically restrict DN3 cell differentiation uncoupled from proliferation and survival. Broad expression of IP 3 3-kinases and PI3Ks, IP 4 detection in multiple tissues (Sauer and Cooke, 2010) and common PI3K implication in costimulation raise the possibility that 'metabokinetic' control and costimulation-enforcement through non-canonical PI3K antagonism by IP 3 3-kinases are broadly relevant.

Mice
Our C57BL/6 Itpkb -/mice were described in . All animal studies were approved by the Scripps Research Institute animal care and use committee and conform to all relevant regulatory standards. Mixed bone marrow chimeras were generated as in . For in vivo induction of Rag2 -/-DN3 cell differentiation, 10 mg anti-CD3 antibodies (BD Biosciences, San Jose, CA, clone 145-2C11) were injected i.p. 1-3 days later, the mice were euthanized and analyzed. Where indicated, the mice were treated orally once daily with 10 mg/kg LY-411,575 (Wong et al., 2004) or vehicle (5% polyethylene glycol, 3% propylene glycol, 1% ethanol, 0.4% methylcellulose). The first dose was administrated 3-4 hr prior to a-CD3 injection.
For BrdU incorporation assays, we injected mice i.p. with 100 ml BrdU [10 mg/ml] and analyzed thymi 4 hr later. For preparation of thymocyte suspensions, thymi were placed in M199 medium/2% FCS/1x penicillin/streptomycin/glutamate at room temperature and single-cell suspensions prepared by passage through a 40 mm mesh (BD Biosciences).

Biochemistry
Thymocytes were lysed as previously described  in 1% Triton X-100/60 mM octylglucoside/150 mM NaCl/25 mM Tris-HCl, pH 7.5/1 mM EDTA containing Roche Complete Mini Protease Inhibitor and PhosSTOP Phosphatase Inhibitor Cocktails. Lysates were incubated for 20 min at 4˚C, then cleared by centrifugation at 14,000 g for 10 min at 4˚C. For immunoprecipitations, pre-cleared lysates were incubated for 1.5 hr with anti-Itpkb antibodies (G-20, Santa-Cruz Biotechnology) followed by incubation with Protein G-conjugated beads for 1.5 hr. Beads were washed 3 times with 1x lysis buffer, denaturated in 1x sample buffer at 99˚C for 10 min and analyzed via SDS-PAGE/immunoblot. For immunoblot analysis, nitrocellulose membranes were incubated overnight at 4˚C with anti-Itpkb (#AP8167b, Abgent) or anti-PLCg1 (#2822, Cell Signaling Technology) antibodies and then for 45 min with anti-rabbit-HRP secondary antibodies (Bio-Rad Laboratories) in TBS. Bound antibodies were detected by enhanced chemiluminescence (ECL kit, GE Healthcare).

Mathematical modeling
The kinetics of DN thymocyte differentiation were modeled by a set of linear ordinary differential equations (ODE). In these, the rate constants for DN3 cell generation and for thymocyte subset turnover were similar between genotypes. The rate constants for DN3-to-DP cell differentiation were increased over two2-fold for Itpkb -/cells. The ODE were solved by pen and paper calculations and results verified using BIONETGEN.
Modeling steady-state thymocyte developmental kinetics in WT vs. Itpkb -/mice We have built a linear ordinary differential equation (ODE) based model for the kinetics of b-selection in the presence or absence of Itpkb. We approximated the arrival of hematopoietic progenitors followed by their successive maturation via DN1 to DN3 cells as a constant influx rate K ( Figure 5A). DN3 cells transit to the DN4 stage with a differentiation rate K 1 or turn over at a turnover rate K d1 . Similarly, cells at the DN4 stage can further differentiate into ISP at a rate K 2 or turn over at a rate K d2 . ISP cells further mature into DP cells at a rate K 3 , or turn over a rate K d3 . Most DP cells turn over through death by neglect at a rate K d4 . Only very few DP cells mature to SP thymocytes (Starr et al., 2003); these are ignored. In the ODE, we denote the numbers of DN3, DN4, ISP and DP cells as C 1 , C 2 , C 3 and C 4 , respectively.
It is reasonable to assume that the thymocyte subset population sizes represent steady state solutions of the above ODEs. The ODEs reach the steady state at time scales much longer than the times associated with the kinetic rate constants. At the steady state, dC 1 /dt=dC 2 /dt=dC 3 /dt=dC 4 /dt=0. This implies that the influx of cells to a particular stage due to differentiation is balanced by outflux due to both differentiation into the next stage and cell death. The steady state solutions of the ODEs are then given by This implies that regardless of the initial values of the population sizes, the system will reach the above concentrations at long times. DN4 cells and ISP are highly proliferative compared to DN3 and DP cells but show similar overall viability as those, resulting in lower turnover rates ( Figure 3C,D). The DN4 cell and ISP cell turnover rates (K d2 , K d3 ) are thus small compared to the respective differentiation rates (K 2 , K 3 ), such that K 2 /(K d2 +K 2 ) » 1 and K 3 /(K d3 +K 3 ) » 1. In this case, equation (S2) can be further simplified as From equation (S3), we observe that when differentiation rates K 1 (DN3 to DN4 cells), K 2 (DN4 cells to ISP) and K 3 (ISP to DP cells) increase while progenitor influx rate K is constant, then the numbers of DN3 cells (C 1 ) and DP cells (C 4 ) remain relatively unaffected, while the numbers of DN4 cells (C 2 ) and ISP (C 3 ) decrease. Indeed, increasing K 1 , K 2 and K 3 to the values in Table 1 modeled the steady state DN cell subset number distribution in WT vs. Itpkb -/-(increased K 1/2/3 ) mice ( Figure 2C, Figure 5B).

Statistical analyses
Aggregated results are shown as mean ± SEM. p values for the indicated comparisons were calculated by two-tailed unpaired Student's t-test. Group sizes are described in the figure legends. Significant p values are denoted by asterisks: *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. All statistical analyses were performed in Prism.