Inactivation of nuclear GSK3β by Ser389 phosphorylation promotes lymphocyte fitness during DNA double-strand break response

Variable, diversity and joining (V(D)J) recombination and immunoglobulin class switch recombination (CSR) are key processes in adaptive immune responses that naturally generate DNA double-strand breaks (DSBs) and trigger a DNA repair response. It is unclear whether this response is associated with distinct survival signals that protect T and B cells. Glycogen synthase kinase 3β (GSK3β) is a constitutively active kinase known to promote cell death. Here we show that phosphorylation of GSK3β on Ser389 by p38 MAPK (mitogen-activated protein kinase) is induced selectively by DSBs through ATM (ataxia telangiectasia mutated) as a unique mechanism to attenuate the activity of nuclear GSK3β and promote survival of cells undergoing DSBs. Inability to inactivate GSK3β through Ser389 phosphorylation in Ser389Ala knockin mice causes a decrease in the fitness of cells undergoing V(D)J recombination and CSR. Preselection-Tcrβ repertoire is impaired and antigen-specific IgG antibody responses following immunization are blunted in Ser389GSK3β knockin mice. Thus, GSK3β emerges as an important modulator of the adaptive immune response.

G lycogen synthase kinase 3b (GSK3b) is a serine threonine protein kinase abundantly expressed in all cells and tissues 1 . GSK3b is present predominantly in the cytoplasm, but also within the nucleus in response to proapoptotic stimuli, although the function of nuclear GSK3b is unclear 2,3 . GSK3b plays a critical role in determining the balance between cell survival and death 4 . Deletion of GSK3b results in lethality during embryonic development 5 . Unlike most kinases, GSK3b is constitutively active and high levels of GSK3b activity are associated with its role in promoting cell death 4 . To maintain cell survival, active mechanisms are required to restrain GSK3b activity [6][7][8] . Although cell death also plays an important role during T-and B-cell development and the immune response, little is known about the contribution of GSK3b to adaptive immune responses. Pharmacological inhibitors that inhibit the activity of both GSK3b and its closely related kinase GSK3a, have been shown to interfere with thymocyte development at the double negative (DN)3 stage in vitro 9 . Pharmacological inhibition of GSK3b/ GSK3a activity can reduce activation-induced cell death of CD4 cells 10 . In contrast, other studies have proposed that GSK3a is primarily required for Th1 differentiation while GSK3b is primarily required for Th17 differentiation 11 . It remains unknown whether GSK3b plays a role in B-cell development or activation.
The best-characterized mechanism to repress GSK3b is the phosphorylation of Ser 9 primarily by Akt (refs 12,13). The flexible N terminus of GSK3b containing phospho-Ser 9 acts as an intrinsic competitive inhibitor of this kinase by folding into the active site and competing with substrate binding 14 . High levels of phospho-Ser 9 GSK3b can be detected without external stimulation in most cells of the immune system. However, unexpectedly, GSK3b Ser 9 Ala knockin (KI) mice do not have a defect in cell survival within the immune system or other tissues 15 . This could be due to the compensatory effect by GSK3a because increased activation-induced cell death of CD4 cells was found in double mutant mice for GSK3b Ser 9 Ala and GSK3a Ser 21 Ala (analogue to Ser 9 in GSK3b) 16 . Increased GSK3 activity in the Ser 9 Ala GSK3b/Ser 21 Ala GSK3a double mutant mice promoted polarization of CD4 T cells to Th17 (ref. 17). No studies, to the best of our knowledge, have reported the function of GSK3b Ser 9 inactivation in B cells.
We have identified that GSK3b can also be inactivated by phosphorylation at Thr 390 (human)/Ser 389 (mouse) by p38 mitogen-activated protein kinase (MAPK) 18 . Phosphorylation of GSK3b at Ser 389 /Thr 390 inactivates GSK3b to a similar degree as Ser 9 phosphorylation by Akt (ref. 18). While Akt, inhibits both GSK3a and GSK3b, p38 MAPK does not phosphorylate GSK3a (ref. 18). Interestingly, phosphorylation of Ser 9 is ubiquitously present under normal conditions, Ser 389 phosphorylation is restricted to the thymus, spleen and brain 18 . Thus, Ser 389 phosphorylation represents an alternative pathway to restrain GSK3b activity, but it remains unclear which stimuli require this pathway and its function in cell death/survival. Here we show that phosphorylation of GSK3b at Ser 389 /Thr 390 is specifically induced in response to DNA double-strand breaks (DSBs) and targets nuclear GSK3b. We show that DSBs naturally generated during V(D)J recombination in T-and B-cell receptors, and during class switch recombination (CSR) of the immunoglobulin genes in activated B cells 19 , inactivate nuclear GSK3b through Ser 389 phosphorylation. We demonstrate that this pathway plays an important role in the fitness of T and B cells during DSB repair.

Results
Phosphorylation of GSK3b on Ser 389 is induced by DSB. While Ser 9 phosphorylation of GSK3b is ubiquitous, phosphorylation of GSK3b Ser 389 occurs predominately in the thymus 18 . To determine whether phosphorylation of Ser 389 on GSK3b was regulated during T-cell development, we examined phospho-Ser 389 GSK3b in DN and double positive (DP) thymocytes as well as mature CD4 and CD8 cells. Interestingly, while total GSK3b and phospho-Ser 9 GSK3b did not change during development, high levels of phospho-Ser 389 GSK3b were detected in both DN and DP thymocytes, but not in mature CD4 and CD8 cells (Fig. 1a). Consistent with our previous results 18 , phosphorylation of GSK3b on Ser 389 correlates with the presence of active p38 MAPK (Fig. 1a).
Unlike peripheral naive T cells, both DN and DP thymocytes are rapidly proliferating. To determine whether proliferation promoted Ser 389 phosphorylation of GSK3b, we activated CD4 cells with anti-CD3 and anti-CD28 antibodies (Abs). Although p38 MAPK was activated upon antigen stimulation, no phospho-Ser 389 GSK3b could be detected in activated CD4 cells indicating that proliferation does not induce phospho-Ser 389 GSK3b (Fig. 1b). We have shown that p38 MAPK is activated by V(D)J-mediated DSBs in thymocytes 20 . To investigate whether phospho-Ser 389 GSK3b was also triggered by DSBs, we examined Ser 389 phosphorylation of GSK3b in CD4 cells following treatment with doxorubicin, a chemotherapeutic agent known to generate DSBs 21 . Interestingly, doxorubicin elicited the phosphorylation of GSK3b on Ser 389 and this correlated with the activation of p38 MAPK (Fig. 1c) and the induction of DSBs as determined by the presence of gH2AX (Fig. 1d). The low dose of doxorubicin that we used to induce DSBs did not cause significant death in CD4 cells (Fig. 1e). Similar to mouse CD4 cells, phospho-Thr 390 was induced by doxorubicin in human CD4 cells isolated from healthy controls (Fig. 1f).
To examine induction of phospho-Ser 389 GSK3b in vivo by DSBs, mice were irradiated and CD4 cells were purified from spleen after exposure. X-irradiation induced Ser 389 phosphorylation of GSK3b in CD4 cells (Fig. 1g). Ser 389 phosphorylation was p38a MAPK dependent, since only marginal levels of phospho-Ser 389 GSK3b could be detected in CD4 cells from T-cell conditional p38a MAPK knockout (p38c KO) mice (Fig. 1g). To determine whether in vivo exposure to radiation could induce Thr 390 phosphorylation of GSK3b in humans, we performed a pilot study with breast cancer patients undergoing local radiotherapy as the first regimen of therapy. CD4 cells were isolated from peripheral blood collected before beginning the treatment (base line). Patients received a daily dose of radiotherapy for four consecutive days and CD4 cells were isolated from blood collected 24 h after the last dose. While total GSK3b levels remained unchanged by the treatment (Fig. 1h), following radiotherapy phospho-Thr 390 GSK3b was increased over baseline in all four patients (Fig. 1h). We also examined phospho-Thr 390 GSK3b levels at two different time points (4-6 days apart) in CD4 cells from healthy untreated volunteers, and no changes over time were detected (Fig. 1i). Thus, phosphorylation on Ser 389 /Thr 390 regulates GSK3b selectively in response to DSBs in both mouse and human. V(D)J induces phospho-Ser 389 GSK3b in the nucleus. DSBs are also naturally produced in lymphocytes during V(D)J recombination to generate the coding T-cell and B-cell receptor genes [22][23][24] . V(D)J-mediated DSBs also trigger DNA damage and repair responses 23 . At the DN3 stage of development, thymocytes undergo V(D)J recombination of the TCRb locus to generate a functional TCRb that provides a signal to terminate recombination and promote differentiation to the DN4 stage. Although the levels of total GSK3b were comparable between DN3 and DN4 thymocytes, high levels of phospho-Ser 389 were only detected in DN3 thymocytes (Fig. 2a). To show that phospho-Ser 389 GSK3b was dependent on V(D)J recombination, we examined DN3 thymocytes from wild-type (WT) and recombination activating gene (RAG)-deficient mice that cannot undergo V(D)J recombination due to the lack of RAG recombinase 25 . Phospho-Ser 389 GSK3b was much more abundant in WT DN3 thymocytes than in RAG KO thymocytes (Fig. 2b). To determine whether the increased level of phospho-Ser 389 GSK3b correlated with lower GSK3b activity, in vitro kinase assays were performed. Lower GSK3b activity was present in WT thymocytes than in RAG KO thymocytes (Fig. 2c). Ataxia telangiectasia mutated (ATM) is a kinase activated by DSBs including V(D)J-mediated DSBs and it is a major player in the DSB-repair response 26 . To address whether phosphorylation of GSK3b on Ser 389 in DN3 thymocytes was triggered by DSBs or the DSB-repair response we examined DN3 thymocytes in ATM KO mice. Phospho-Ser 389 GSK3b was diminished in DN3 thymocytes from ATM KO mice (Fig. 2d). Consistent with previous studies in vitro DSBs 27 , activated p38 MAPK was almost absent in ATM KO thymocytes (Fig. 2d). In addition, analysis of phospho-Ser 389 in DN3 thymocytes from T cell-conditional p38a MAPK KO mice showed minimal levels compared to WT DN3 thymocytes (Fig. 2e). In contrast, phospho-Ser 9 GSK3b levels were not affected (Fig. 2e). Thus, the restricted presence of GSK3b phosphorylated on Ser 389 in thymocytes in vivo is due to the presence of V(D)J-induced DSB-repair response as a signalling pathway to inactivate GSK3b through ATM and p38 MAPK.
We have previously shown that DSBs result in the selective accumulation of p38 MAPK in the nucleus 28 . Phospho-Ser 389 GSK3b localized selectively to the nucleus as shown by immunostaining and confocal analysis in WT DN3 thymocytes (Fig. 2f) and this was confirmed by western blot analysis (Fig. 2g). A similar nuclear distribution was found for human phospho-Thr 390 GSK3b in a human cancer cell line following treatment with doxorubicin ( Supplementary Fig. 1a), and this was prevented by inhibiting ATM (Supplementary Fig. 1b). The focal distribution of phospho-Ser 389 in the nucleus of DN3 thymocytes was reminiscent of gH2AX foci at the DSBs generated by V(D)J recombination at the TCRb locus. To determine whether phospho-Ser 389 GSK3b was recruited to the DSBs, we performed co-immunostaining for both gH2AX and phospho-Ser 389 in DN3 thymocytes. Interestingly, the most prominent phospho-Ser 389 foci were adjacent to the gH2AX foci (Fig. 2h). Thus, phosphorylation of GSK3b on Ser 389 is induced by V(D)J-mediated DSBs in thymocytes to preferentially inactivate nuclear GSK3b. WT and T-cell-specific p38a conditional knockout (cKO) mice were left unexposed or exposed to 4 Gy of X-rays. After 1.5 h, CD4 splenocytes were isolated and P-S 389 GSK3b and total GSK3b assessed by western blotting. Actin is shown as a loading control. (h) Relative levels of total GSK3b, P-T 390 GSK3b and the relative ratio of P-T 390 GSK3b to total GSK3b at baseline (BL) and after radiotherapy (Rad) in CD4 cells from breast cancer patients, determined by the Odyssey system (n ¼ 4). (i) Relative ratio of P-T 390 GSK3b to total GSK3b in CD4 cells from healthy donors isolated in two different days (baseline 1 and 2) for each subject (n ¼ 3). *P valueo0.05 as determined by paired t-test.  The levels of P-p38 MAPK, total p38 MAPK, P-S 389 GSK3b and total GSK3b in DN3 thymocytes from WT and ATM KO mice were determined by western blot analysis. (e) DN3 thymocytes from WT and p38a conditional KO (p38cKO) mice were examined for p38a, P-S 389 GSK3b, P-S 9 GSK3b and total GSK3b by western blot analysis. Actin is shown as a loading control. (f) DN3 thymocytes were examined by immunostaining and confocal microscopy for the presence of P-S 389 GSK3b (red) and TOPRO nuclear stain (blue). Scale bar, 5 mm. (g) Western blot analysis for P-S 389 GSK3b and total GSK3b using nuclear and cytosolic extracts from DN thymocytes. Histone is shown as a marker for the nuclear fraction. (h) DN3 thymocytes were examined by immunostaining and confocal microscopy for the presence of gH2AX (green), P-S 389 GSK3b (red) and TOPRO nuclear stain (blue). Scale bar, 2 mm. Data are representative of three or more independent experiments. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10553 machinery 22,24,29 . B cells were isolated from the spleen and activated with lipopolysaccharide (LPS) and interleukin-4 (IL-4) to induce CSR for the production of IgG1. Phospho-Ser 389 GSK3b was induced following activation and correlated with the presence of DSBs as determined by gH2AX (Fig. 3a). Analysis of GSK3b activity in whole-cell lysates from activated B cells showed lower activity at 48 h of activation correlating with the induction of GSK3b Ser 389 phosphorylation (Fig. 3b).
To demonstrate that AID-mediated DSBs were required to induce phosphorylation of Ser 389 , we examined B cells from WT and AID KO mice following stimulation. Correlating with a significant reduction in DSBs as determined by gH2AX (Fig. 3c), phospho-Ser 389 GSK3b was reduced in the AID KO B cells compared with WT B cells (Fig. 3c). In contrast, levels of phospho-Ser 9 GSK3b were not affected in AID KO B cells (Fig. 3c). We investigated the subcellular distribution of phospho-Ser 389 GSK3b in B cells undergoing CSR. Phospho-Ser 389 GSK3b was predominately in the nucleus of activated B cells (Fig. 3d) while total GSK3b was present in both the nucleus and cytosol (Fig. 3d). We also examined whether phospho-Ser 389 GSK3b was associated with CSR-generated DSBs by co-immunostaining of gH2AX and phospho-Ser 389 GSK3b. Phospho-Ser 389 foci were adjacent to the gH2AX foci in activated B cells (Fig. 3e). Thus, DSBs generated in B cells undergoing CSR selectively cause the inactivation of nuclear GSK3b through phosphorylation on Ser 389 .
TCRb rearrangement is impaired in GSK3b Ser 389 KI mice.
To determine the function of the phosphorylation of GSK3b at Ser 389 /Thr 390 relative to classical Ser 9 phosphorylation, we generated GSK3b-KI mice where Ser 389 of GSK3b was replaced with Ala to prevent the C-terminal phosphorylation of GSK3b (Fig. 4a). Western blot analysis confirmed the absence of phospho-Ser 389 GSK3b in thymocytes from homozygous GSK3b-KI mice, while normal levels were present in heterozygous GSK3b-KI mice (Fig. 4b). Ser 9 phosphorylation of GSK3b was not affected by the Ser 389 Ala mutation (Fig. 4b), showing that these two residues represent independent pathways to regulate GSK3b activity. To address the relative contribution of Ser 389 phosphorylation to the overall GSK3b activity, we assayed GSK3b activity in DN thymocytes and mature CD4 cells from WT and   GSK3b-KI mice. An increase in GSK3b activity was detected in DN thymocytes from GSK3b-KI mice but no difference was detected in mature CD4 cells correlating with the absence of phospho-Ser 389 GSK3b in mature CD4 cells (Fig. 4c). Thus, phosphorylation of Ser 389 contributes to restraining the activity of GSK3b, independent of Ser 9 phosphorylation. S 389 A-GSK3b-KI mice were fertile and no developmental abnormalities or gross alterations were detected under physiological conditions. The percentage and number of mature CD4 and CD8 cells in the spleen and lymph nodes was not altered in GSK3b-KI mice ( Supplementary Fig. 2a-c). The fact that no phospho-Ser 389 GSK3b could be found in activated CD4 cells (Fig. 1b) suggested that this pathway is not essential for activation of CD4 cells. Analysis of proliferation of activated CD4 cells from WT and GSK3b-KI mice showed no obvious difference ( Supplementary Fig. 3a). The number of cells recovered upon activation was also comparable ( Supplementary Fig. 3b) and there was no difference in IL-2 production ( Supplementary Fig. 3c). Thus, correlating with the selective inactivation of GSK3b through Ser 389 phosphorylation in response to DSBs, this pathway does not seem to be required for normal activation of CD4 cells.
We next assessed the contribution of this pathway to thymocyte development. No difference in the total number of thymocytes (Fig. 4d), or in the numbers and percentage of DP and single positive (SP) thymocytes ( Supplementary Fig. 4a,b), was detected between WT and S 389 A-GSK3b-KI mice. We examined Va to Ja rearrangements in DP thymocytes to determine if inability to inactivate GSK3b during Va-Ja recombination could impair the TCRa repertoire. Analysis of Va to Ja rearrangements using genomic DNA from isolated total DP thymocytes suggested that GSK3b Ser 389 Ala mutation had no effect in survival of DP thymocytes during TCRa recombination ( Supplementary Fig. 5). The number of total and percentage of DN thymocytes in GSK3b-KI mice was not significantly altered ( Supplementary Fig. 4a,b). However, examination of early thymocyte development within the DN population showed a reduction in the percentage and total number of DN4 thymocytes in GSK3b-KI mice ( Fig. 4e; Supplementary Fig. 4a). The reduction of DN4 thymocyte number was associated with an increase in the DN3/DN4 ratio (Fig. 4e). We examined whether preventing inactivation of GSK3b by Ser 389 phosphorylation during TCRb-V(D)J recombination affected the number of DN3 thymocytes that successfully survive b-rearrangement. Intracellular staining for TCRb in DN3 thymocytes showed a lower frequency of DN3 thymocytes with successfully rearranged TCRb in GSK3b-KI mice (Fig. 4f). Thus, inactivation of GSK3b through Ser 389 phosphorylation contributes to the fitness of the DN3 thymocyte pool undergoing V(D)J recombination.
To determine whether this could result in an altered distribution of TCRb specificities in mature T cells, CD4 cells freshly isolated from the spleen of WT and GSK3b-KI mice were used for TCRb spectrotyping analysis ( Supplementary Fig. 6). We detected a lower frequency of TRBV19 and TRBV13-2 in CD4 cells from GSK3b-KI mice relative to WT CD4 cells ( Fig. 4f; Supplementary Fig. 6). Other TCRb rearrangements (for example, TRBV15) were not affected in any of the analysed GSK3b-KI mice ( Fig. 4g; Supplementary Fig. 6). In addition, we performed quantitative analysis of a panel of Vb with Jb1 and Jb2 rearrangements using genomic DNA from WT and GSK3b-KI DN thymocytes 30 . The frequency of TRBV19 with Jb2 rearrangement was lower in GSK3b-KI DN thymocytes (Fig. 4h), but there was no statistically significant difference for the frequency of TRBV19 with Jb1 rearrangement. The TRBV16-Jb2 rearrangement frequency was also reduced in GSK3b-KI thymocytes, while the frequency of TRBV16-Jb1 was not significantly different, although a trend was observed (Fig. 4h). The frequency of the other examined rearrangements of Vb segment with Jb1 or Jb2 were not significantly affected ( Supplementary Fig. 7). Thus, the inactivation of GSK3b by phosphorylation on Ser 389 promotes the survival of DN3 thymocytes during V(D)J recombination of the preselection TCRb locus. Failure to inactivate GSK3b through this pathway diminishes DN3 thymocyte fitness and success in recombining the TCRb locus leading to an altered TCRb repertoire.
In vivo antibody response in GSK3b Ser 389 KI mice is impaired. Since GSK3b phosphorylation on Ser 389 was induced in B cells by CSR-generated DSBs, we investigated the effect of inactivation of GSK3b through this pathway on the in vivo humoral response. The percentage and the number of B cells in GSK3b-KI and WT mice was comparable ( Supplementary Fig. 8a,b). No significant difference was detected in the baseline levels of total IgM or IgG1 ( Supplementary Fig. 8c,d). To examine the effectiveness of CSR in an antigen-specific immune response, WT and GSK3b-KI mice were immunized with ovalbumin (OVA). After 14 days, immunized mice were euthanized and serum collected to assay antibody responses. The level of serum OVA-specific IgM was comparable in both WT and GSK3b-KI mice (Fig. 5a), as expected considering that production of IgM isotype-Abs does not require CSR. In contrast, the production of OVA-specific IgG, which requires CSR, was largely abrogated in GSK3b-KI mice (Fig. 5b). Moreover, IgG isotype analysis revealed reduced OVA-specific IgG1 (Fig. 5c) and OVA-specific IgG2c (Fig. 5d) in GSK3b-KI mice. Together, these results show for the first time that inactivation of GSK3b by phosphorylation of Ser 389 is essential for efficient CSR and antibody production in vivo.
Since cytokines derived from CD4 cells play an important role in promoting isotype switching of B cells, the impairment in IgG production could be due to an impaired cytokine production rather than to an intrinsic defect in B cells. To address this possibility splenocytes from immunized WT and GSK3b-KI mice were restimulated ex vivo with OVA, and the levels in the supernatant of IL-4 and IFNg, two major cytokines that promote IgG switching, were determined. The levels of both cytokines in WT and GSK3b-KI cells were comparable (Fig. 5e,f), indicating that T-cell helper responses are not impaired in GSK3b-KI mice. Thus, the failure to undergo in vivo class switching and to mount an antibody immune response in GSK3b-KI mice was most likely due to an intrinsic defect in B cells due to inability to inactivate nuclear GSK3b.
Phospho-Ser 389 GSK3b promotes fitness in response to CSR. To demonstrate that the GSK3b Ser 389 Ala mutation in KI mice interferes with efficient CSR, we examined Ab production in isolated B cells activated with LPS in the presence of IL-4 in vitro. After 4 days, we examined the presence of IgG1 on the cell surface of B cells, as a parameter for CSR. The frequency of IgG1-expressing B cells from GSK3b KI mice was markedly reduced (Fig. 6a). Thus, phosphorylation of GSK3b on Ser 389 is important for B cells undergoing CSR to successfully express cell surface IgG1. To demonstrate that the low frequency of IgG1-expressing B cells from GSK3b KI mice was due to the inability of these cells to inactivate GSK3b during CSR, B cells from WT and GSK3b-KI mice were activated in the presence or absence of the pharmacological inhibitor of GSK3 (GSK3 Inhibitor X). The GSK3 inhibitor had no effect on the generation of IgG1-expressing B cells from WT mice (Fig. 6b), supporting the existence of endogenous mechanisms that suppress GSK3b during CSR. Interestingly, the GSK3 inhibitor fully restored the frequency of IgG1-expressing B cells from GSK3b-KI mice (Fig. 6b). In addition, inhibition of p38 MAPK by SB203580 significantly reduced the frequency of IgG1-expressing WT B cells but it had no effect on the frequency of IgG1-expressing B cells from the GSK3b-KI mice (Fig. 6c) indicating that the effect of p38 MAPK on CSR in B cells is mediated through Ser 389 phosphorylation. Thus, inactivation of GSK3b through the Ser 389 phosphorylation by p38 MAPK is important for successful Ab isotype switching.
To determine whether the reduced frequency of IgG1producing B cells was due to impaired expansion, we performed carboxyfluorescein succinimidyl ester (CFSE) labelling. No difference in proliferation was observed between WT and GSK3b-KI B cells (Fig. 6d). Similarly, there was no difference in cell cycle profiles between WT and GSK3b-KI B cells following activation ( Supplementary Fig. 9). However, there was a significant reduction in survival of GSK3b-KI B cells relative to WT B cells based on cell recovery (Fig. 6e). Annexin V staining confirmed increased cell death in activated GSK3b-KI B cells ( Fig. 6f; Supplementary Fig. 9b). Increased death in activated GSK3b-KI B cells was also supported by analysis of cell size (Supplementary Fig. 9c) and live/dead blue staining ( Supplementary Fig. 9d,e). Together, these results show that GSK3b Ser 389 Ala mutation decreases the fitness of B cells while undergoing CSR, compromising their ability to produce antibodies.
Mcl-1 degradation causes necroptosis in KI B cells. One of the mechanisms by which GSK3b can mediate cell death is by phosphorylation of b-catenin in the cytosol, leading to the rapid degradation of b-catenin by APC complex 31,32 . However, bcatenin levels were comparable in activated GSK3b-KI B cells and WT B cells (Fig. 7a). GSK3b also promotes the degradation of Mcl-1 through phosphorylation on Ser 159 (mouse Ser 140 ) 33 . In response to DSBs, Mcl-1 accumulates in the nucleus, where it associates with gH2AX and DSB 34,35 , and it is required for DSB response through different mechanisms 35 . Mcl-1 deficiency results in increased genomic instability and it is essential for B-cell survival following activation [34][35][36] . Interestingly, western blot analysis showed that the levels of Mcl-1 were almost undetectable in activated GSK3b-KI B cells (Fig. 7b). To rule out that the reduced levels of Mcl-1 were not the result of increased mitochondrial dysfunction in GSK3b-KI B cells we examined the levels of other Bcl2 family members. The levels of Bcl2, BclX (anti-apoptotic) as well as Bid, Bax and Bad (pro-apoptotic) were not different between WT and GSK3b-KI B cells (Fig. 7c). Mcl-1 has been shown to regulate PUMA 37 that also mediates apoptosis in response to DNA damage 38 . PUMA was highly expressed in thymocytes exposed to irradiation (Fig. 7d), but only marginal levels could be detected in either WT or GSK3b-KI B cells (Fig. 7d). We also examined mitochondria integrity by staining with Mitotracker (Fig. 7e), as well as mitochondrial membrane potential by staining with TMRE (Fig. 7f), but no differences between WT and GSK3b-KI B cells were detected. The reduction of Mcl-1 in GSK3b-KI B cells therefore was not associated with dysfunctional mitochondria. These results also suggested that increased death found in GSK3b-KI B cells might not be through apoptosis mediated by mitochondria and caspase-3. Analysis of caspase-3 activation showed comparable levels of active caspase-3 (cleaved form) in activated WT and GSK3b-KI B cells (Fig. 7g). In addition, most of the Annexin V-positive cells were propidium iodide positive ( Supplementary Fig. 10), suggesting the presence of other death pathways. DNA damage has also been shown to trigger necroptosis, a mitochondria-independent mechanism of death 39,40 . RIPK1, the most upstream component of the ripoptosome pathway, is normally cleaved to prevent the formation of the necrosome and death by necroptosis 41 . Cleaved RIPK1 was present in activated WT B cells, but only low levels could be detected in GSK3b-KI B cells (Fig. 7g).
In vitro, blocking of caspase-8 promotes death by necroptosis 42 . Negligible levels of cleaved caspase-8 were detected in both activated WT and GSK3b-KI B cells (Supplementary Fig. 11a), suggesting that there is minimal activity of caspase-8 in activated B cells overall. As a positive control for cleaved caspase-8 we used activated B cells treated with Fas ligand known to trigger caspase-8 cleavage (Supplementary Fig. 11a). Thus, the increased death found in activated B cells that fail to inhibit GSK3b could be mediated by necroptosis. We therefore examined the effect of necrostatin-1s, a selective inhibitor of necroptosis through RIPK1 (ref. 43). WT and GSK3b-KI B cells were activated. Necrostatin-1s was added after 2 days of activation, when DSBs and phospho-Ser 389 GSK3b were detected (Fig. 3) and before differences in survival were observed ( Supplementary  Fig. 11b). Cell survival was determined 24 h after treatment with necrostatin-1s (day 3 of activation). Interestingly, treatment with necrostatin-1 restored survival of activated GSK3b-KI B cells (Fig. 1h) Similar results were obtained adding necrostatin-1s at 24 and 36 h following activation (Supplementary Fig. 11c). During necroptosis active RIPK1 activates the downstream kinase RIPK3, and RIPK3 then mediates the phosphorylation of MLKL, one of the mediators of necroptosis that translocates to the membrane, causes membrane rupture and death 40,44,45 . Increased levels in phospho-MLKL were detected in GSK3b-KI B cells (Fig. 7i). Thus, inactivation of nuclear GSK3b by phosphorylation on Ser 389 is a mechanism to minimize necroptosis-mediated cell death of B cells during CSR.
Since the subcellular localization of Mcl-1 in activated B cells has not been addressed, we performed immunostaining for Mcl-1 and confocal microscopy analysis. Abundant levels of Mcl-1 were present both in the cytosol and the nucleus of WT B cells (Fig. 7j). In contrast, the remaining Mcl-1 levels in GSK3b-KI B cells were almost exclusively outside the nucleus (Fig. 7j). Treatment of GSK3b-KI B cells with the GSK3 inhibitor restored Mcl-1 levels in the nucleus to those in WT B cells (Fig. 7j). In addition, western blot analysis for Mcl-1 showed comparable levels in the mitochondrial fraction, however, levels of Mcl-1 in the nuclear fraction were drastically decreased in GSK3b-KI B cells (Fig. 7k). To address whether the decreased nuclear levels of Mcl-1 contribute to the impaired survival of GSK3b-KI B cells we ectopically expressed WT Mcl-1, or a previously described mutant form of Mcl-1 (mMcl-1) where Ser 140 is replaced by Ala (Ser 140 Ala) 46 and therefore it cannot be phosphorylated by GSK3b and degraded 33,46 . Thus, WT Mcl-1 will still be degraded by GSK3b Ser 389 Ala in KI B cells, but the mMcl-1 mutant should be protected from degradation. Accordingly, WT Mcl-1 expression did not increase survival of GSK3b-KI B cells relative to empty vector (Fig. 7l). In contrast, expression of Ser 140 Ala mMcl-1 restored survival of GSK3b-KI B cells (Fig. 7l). Efficiency of transduction with both retroviruses in GSK3b-KI B cells was comparable ( Supplementary Fig. 12a), and the total levels of Mcl-1 expression as determined by qRT-PCR (quantitative PCR with reverse transcription) was also comparable ( Supplementary  Fig. 12b). Therefore, nuclear Mcl-1 present in activated B cells is selectively targeted by GSK3b for degradation, and inactivation of GSK3b through phosphorylation on Ser 389 is necessary to sustain Mcl-1 levels and fitness of B cells undergoing CSR.

Discussion
DSBs are dangerous types of DNA damage because they create genomic instability and impair cell survival. However, V(D)J and CSR-mediated DSBs are essential for generating the diversity of our adaptive immune system. Here we identify phosphorylation of GSK3b at Ser 389 /Thr 390 as a survival pathway that is triggered specifically by DSBs, in contrast to Ser 9 phosphorylation that is used as a general mechanism to restrict GSK3b activity. In addition, we show that Ser 389 /Thr 390 phosphorylation regulates GSK3b in the nucleus where GSK3b associates with gH2AX in response to DSBs. The selective induction of Ser 389 phosphorylation by DSBs could explain the selective tissue distribution of this phosphorylation, mainly in the thymus and spleen. Interestingly, phospho-Ser 389 is also abundant in brain 18 where recent studies have reported the presence of DSBs caused by normal brain activity and stress 47 . We show here that the inability of cells to inactivate GSK3b through Ser 389 phosphorylation (GSK3b Ser 389 Ala mutant cells) causes a decline in the fitness of cells undergoing DSB induced by V(D)J recombination or CSR. We have also shown that failure to inactivate GSK3b through Ser 389 phosphorylation in CD4 cells from GSK3b KI mice makes these cells more sensitive to death caused by low dose of irradiation ( Supplementary Fig. 13a) or the chemotherapeutic drug doxorubicin (Supplementary Fig. 13b,c). Thus, phosphorylation of GSK3b on Ser 389 in response to chemotherapy or radiotherapy may have implications in the treatment of cancer.
This study reveals a role of GSK3b in B cells. We demonstrate that inactivation of GSK3b by Ser 389 phosphorylation upon activation of B cells is required optimum survival during CSR and for in vivo antigen-specific IgG antibody production. Suppression of p53 by BCL6 has been reported to be a prosurvival signal in germinal centre (GC) B cells 48,49 . In addition, both Mcl-1 (ref. 36) and c-Myc 50 are also essential for GC B-cell survival although it is unclear how they are regulated. Our studies show that inactivation of nuclear GSK3b by phosphorylation at Ser 389 is essential for the stabilization of the nuclear pool of Mcl-1. While Mcl-1 localizes predominantly in the mitochondria it is also found in the nucleus specifically in response to DSBs 34,35 . GSK3b has been shown to increase Mcl-1 instability through phosphorylation on Ser 140 (ref. 46). Here we show that only Ser 140 Ala mutant Mcl-1, but not WT Mcl-1, restores viability of activated GSK3b-KI B cells. Thus, preventing phosphorylation of Mcl-1 on Ser 140 by nuclear GSK3b could be a key mechanism to prolong survival during the repair of DSBs. Interestingly, our studies also reveal necroptosis as an important mechanism of death in B cells late during activation. DSBs are one of the reported stimuli that can promote necroptosis 51,52 . Inactivation of GSK3b through phosphorylation of Ser 389 restrains necroptosis, promoting successful CSR and antibody production. Future studies will be needed to address if the crosstalk between GSK3b, Mcl-1 and necroptosis is also observed in other DSB responses.
Our studies show that inactivation of GSK3b by Ser 389 phosphorylation contributes to the survival of DN3 thymocytes undergoing TCRb rearrangement, leading to a reduction in the number of DN4 thymocytes. Similar phenotypes have been also described for mice deficient in ATM, Nbs1 gH2AX, that have impaired DSB repair responses [53][54][55] . A number of parameters such as chromatin environment influence the frequency of the TCRb V-DJ rearrangements before TCRb selection 30 . Impaired survival of thymocytes undergoing specific TCRb rearrangements could also affect the TCRb repertoire. Here we show decreased frequencies of some TCRb rearrangements in GSK3b-KI mice. TCRb repertoire is also impaired in the absence of ATM through an antigen selection-independent mechanism 54 . Our data suggest that inactivation of GSK3b by Ser 389 is not essential for survival of DP thymocytes during TCRa recombination. Sequential recombination of D to J and DJ to V of the TCRb locus may need additional survival signals. Likewise, the secondary recombination event in the TCRb locus between V to DJ2, after an out of frame V to DJ1 rearrangement, may be more dependent on survival signals, as suggested by our results.
In summary, our study reveals for the first time a specific pathway for regulation of nuclear GSK3b in response specifically to DSBs, including V(D)J recombination and CSR (Fig. 8). We also demonstrate that this inactivation pathway of GSK3b is essential for fitness of cells undergoing DSBs and therefore has an important role in the adaptive immune response.

Methods
Mice. For the generation of GSK3b KI mice, a targeting vector constructed (InGenious Targeting Laboratory, Inc.) from a 9.27 kb region subcloned from the C57BL/B6 BAC clone (RPCI-23: 454D5) into the vector backbone (2.4 kb pSP7 (Promega)). The long homology arm extended 6.36 kb 5 0 to the T-G point mutation in exon 11 and the LoxP/FRT flanked Neo cassette was inserted 511 bp 3 0 to the T-G point mutation. The short homology arm extended 2.40 kb 3 0 to the loxP flanked Neo cassette. The targeting vector was constructed using Red/ET recombineering technology and the T-G point mutation was created by overlap PCR. The linearized targeting vector electroporated into BA1(C57BL/6 Â 129/ SvEv) hybrid embryonic stem cells. Following G418 selection, surviving clones were screened by PCR to identify homologous recombinant clones for injection into blastocysts. The neomycin cassette was excised by breeding heterozygous GSK3b KI mice with EIIa-cre mice 56 60 . For studies with DN3 thymocytes, 3-4-week old male and female mice were used. Male and female mice 8-16 weeks of age were used in all other experiments. Mice in each experiment were age and sex matched. All procedures that involved mice were approved by the University of Vermont's institutional guidelines for animal care.
Human samples. Healthy volunteers and patients who agreed to donate peripheral blood signed a consent form that was approved by the Institutional Review Board of The University of Vermont. Breast cancer patients who were undergoing local radiation therapy as a first regimen of therapy and had not received any chemotherapy or hormonal based-therapy were enroled in the study. Blood was drawn before beginning radiotherapy treatment (baseline), and after the last of the four consecutive doses of the first cycle with radiotherapy (X-rad). CD4 T cells were purified from peripheral blood mononuclear cells by positive selection using the CD4 MACS kit (Miltenyi Biotec) as recommended by the manufacturer.
X-ray exposure. For in vivo X-radiation (RadSource 2000), mice were exposed to the indicated dose and sacrificed following a 1.5 h recovery period. For X-irradiation of cells, cells were exposed to the indicated dose and harvested 18 h post exposure. Freshly isolated thymocytes were exposed to 5 Gy and allowed to recover for 1 h before harvest.
Western blot analysis. Whole-cell extracts were prepared in Triton lysis buffer (20 mM Tris (pH 7.4), 137 mM NaCl, 2 mM EDTA, 1% Triton X-100, 10% glycerol, 25 mM b-glycerophosphate and 1 mM sodium vanadate) supplemented with mammalian protease inhibitor cocktail (Sigma). Nuclear and cytosolic extracts were prepared as follows: fresh cell pellet was resuspended in 400 ml cold buffer A (10 mM HEPES pH 7.9, 10 mM KCI, 0.1 mM, EDTA, 0.1 mM EGTA, 1 mM DTT and 0.5 mM PMSF) supplemented with mammalian protease inhibitor cocktail (Sigma). After 15 min on ice, 25 ml of a 10% solution of IGEPAL (Sigma) was added while vigorously vortexing for 10 s. The homogenate was centrifuged for 30 s at 13,000 r.p.m. The supernatant containing cytoplasm was transferred to a fresh tube. The nuclear pellet was resuspended in 50 ml ice-cold nuclear lysis buffer (20 mM HEPES pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA,1 mM DTT and 1 mM PMSF) supplemented with mammalian protease inhibitor cocktail (Sigma) and the tube is vortexed at 4°C for 15 min. The nuclear extract was centrifuged for 5 min at 13,000 r.p.m. Mitochondrial extracts were prepared using the Mitochondrial Fractionation Kit (Active Motif).
Cell culture. MCF-7 cells were maintained in RPMI (Lonza) supplemented with 5% FBS, L-glutamine, penicillin and streptomycin (Invitrogen). Cells were stimulated with doxorubicin (0.1 mg ml À 1 ) (Sigma) for 18 h. For all inhibitor treatments, SB203580 (5 mM), ATM inhibitor (5 mM) and GSK3 inhibitor X (2.5 mM; Calbiochem) pretreatment was performed at 37°C for 30 min, before cell stimulation. CD4 T cells were isolated from spleens by positive selection using anti-CD4 magnetic beads (Miltenyi). The purity obtained was about 90-95% ( Supplementary Fig. 14A). CD4 cells were treated with doxorubicin (0.1 mg ml À 1 ) or stimulated with plate bound anti-CD3 (5 mg ml À 1 ) and soluble anti-CD28 (1 mg ml À 1 ) mAbs (BD Biosciences) for 18 h before analysis. B cells were isolated from spleens by positive selection with anti-CD19 magnetic beads (Miltenyi). Purity for B cells obtained using this approach was 95% ( Supplementary Fig. 14B). B cells were stimulated with LPS (25 mg ml À 1 ; Sigma) and recombinant mouse Il-4 (10 ng ml Annexin V (Biolegend) staining with and without propidium iodine (Sigma) was performed as recommended by the manufacturer. CFSE (Molecular Probes) labelling was performed according to the manufacturer's instructions. Cells (1 Â 10 7 cells per ml) were incubated with 2.5 mM CFSE in 0.05% BSA in PBS at 37°C for 10 min and activated to stimulate proliferation. For cell cycle analysis, cells (10 6 cells) were resuspended in low salt staining solution (3% (w/v) polyethylene glycol PEG 8000, 50 mg ml À 1 propidium iodide, 180 U ml À 1 RNase A, 0.1% Triton X-100, 4 mM sodium citrate) and incubated on ice for 30 min. An equal volume of high salt solution (3% (w/v) polyethylene glycol PEG 8000, 50 mg ml À 1 propidium iodide, 180 U ml À 1 RNase A, 0.1% Triton X-100 and 400 mM sodium chloride) was added just before analysis by flow cytometry. Mitochondrial functional analysis was performed by staining with MitoTracker (Life Technologies) or TMRE (Molecular Probes) for 20 min at 37°C as recommended by the manufacturer.
Tcrb sprectrotyping. CD4 T cells were isolated from spleens by positive selection using anti-CD4 magnetic beads (Miltenyi). Spectrotyping was carried out using the SuperTCRExpress CDR3 Size Diversity Determination Kit as per manufactures' instructions (BioMed Immunotech).
Tcra rearrangement. Tcra rearrangement was analysed by real-time PCR using a QuantiFast SYBR Green PCR kit (Qiagen) and a Roche LightCycler 480. Primers are listed in (Supplementary Table 1). PCR reactions for individual experiments were run in duplicate using the following amplification program: 95°C for 5 min, followed by 45 cycles of 95°C for 10 s and 62°C for 30 s. Results were normalized to signals for the Tcra enhancer (E a ).
Tcrb rearrangement. Tcrb rearrangement was analysed by real-time PCR using PerfeCta SYBRGreen SuperMix (Quanta Biosciences) and AB 7500 Fast Real-time PCR System (Applied Biosystems). Primers are listed in (Supplementary Table 2). PCR reactions for individual experiments were run in duplicate using the following amplification program: 95°C for 2 min, followed by 40 cycles of 95°C for 15 s, 60°C for 1 min and 72°C for 1 min. Results were normalized to signals for the MAPK14 gene.
Statistical analysis. Statistical significance between two groups was determined using Prism (Graphpad), by standard Student t-test. Statistical significance among more than two groups was determined by one-way ANOVA. For all analyses, Pr0.05 was considered significant.