Non-redundant activity of GSK-3α and GSK-3β in T cell-mediated tumor rejection

Summary Glycogen synthase kinase-3 (GSK-3) is a positive regulator of PD-1 expression in CD8+ T cells and GSK-3 inhibition enhances T cell function and is effective in the control of tumor growth. GSK-3 has two co-expressed isoforms, GSK-3α and GSK-3β. Using conditional gene targeting, we demonstrate that both isoforms contribute to T cell function to different degrees. Gsk3b−/− mice suppressed tumor growth to the same degree as Gsk3a/b−/− mice, whereas Gsk3a−/− mice behaved similarly to wild-type, revealing an important role for GSK-3β in regulating T cell-mediated anti-tumor immunity. The individual GSK-3α and β isoforms have differential effects on PD-1, IFNγ, and granzyme B expression and operate in synergy to control PD-1 expression and the infiltration of tumors with CD4 and CD8 T cells. Our data reveal a complex interplay of the GSK-3 isoforms in the control of tumor immunity and highlight non-redundant activity of GSK-3 isoforms in T cells, with implications for immunotherapy.


INTRODUCTION
Glycogen synthase kinase-3 (GSK-3) was first discovered in 1980 as a regulatory kinase which phosphorylates and inhibits glycogen synthase (Embi et al., 1980) and has since been implicated in several disease states and key cellular processes, including Wnt, insulin, and Hedgehog signaling (Matthews and Cantrell, 2009). Two isoforms of GSK-3 have been reported in mammals; a 51 kDa GSK-3a and a 47 kDa GSK-3b isoform, encoded by the unlinked Gsk3a and Gsk3b genes, respectively. These two isoforms exhibit 98% homology in their kinase domains but only 36% identity in the last 76 C-terminal amino acid residues. Both isoforms are highly expressed in many tissues (Woodgett, 1990) and have been implicated in processes ranging from glycogen metabolism to gene transcription, apoptosis and microtubule stability (Eldar-Finkelman and Martinez, 2011;Frame and Cohen, 2001). GSK-3b is thought to be of prime importance in diabetes (McManus et al., 2005), Alzheimer disease (Phiel et al., 2003), and inflammation (Martin et al., 2005).
GSK-3 can phosphorylate more than one hundred substrates (Sutherland, 2011) and plays a key role in T cell activation (Ohteki et al., 2000;Rudd et al., 2020;Taylor et al., 2016;Taylor and Rudd, 2020). Active GSK-3 blocks T cell activation and cytokine production (Ohteki et al., 2000), and we previously showed that the inhibition of GSK-3 downregulate PD-1 and LAG-3 gene expression (Taylor et al., 2016). Other substrates include transcription factors such as cyclic AMP response element binding protein, the nuclear factor of activated T cells (NFATs), b-catenin, c-Jun, and NF-kB Eldar-Finkelman and Martinez, 2011;Grimes and Jope, 2001). In the case of NFAT, GSK-3 inactivates the pathway by phosphorylating NFAT and facilitating its exit from the nucleus in T cells (Beals et al., 1997;Neal and Clipstone, 2001). Active GSK-3 inhibits T cell proliferation (Ohteki et al., 2000), whereas T cell receptor (TCR) and

Conditional knockout of either or both isoform(s) of GSK-3 does not affect the total number of splenic T cells
Our previous studies have demonstrated a clear role for GSK-3 in the regulation of tumor growth, with the inhibition of GSK-3 through SMIs potentiating T cell reactivity leading to diminished growth (Rudd et al., 2020;Taylor et al., 2018). However, it is unclear if the different isoforms of GSK-3 work in a similar manner, if both isoforms are required for the function of GSK-3, or if the action of one is dominant over the other, particularly in the context of cancer. Several SMIs are available for GSK-3 and cited k i values suggest that it is possible for SMIs to preferentially target one isoform over the other; although in practice, particularly in vivo, evidence suggests that this is not the case and that inhibitors tend to act against both isoforms, albeit at varying levels (Lo Monte et al., 2012;Martinez and Perez, 2008). To unequivocally answer this question, we have used Gsk3a-flox/flox-bÀflox/flox mice crossed with an Lck-Cre line (Jackson labs) which uses the distal Lck promoter to delete both Gsk3a and Gsk3b in post-selection CD4+CD8+ DP cells without affecting the T cell repertoire (Chiang and Hodes, 2016) to generate Gsk3ab cKO mice. These mice were then used further to generate individual a and b isoform-specific cKO mice; Gsk3a-flox/flox-Lck-Cre+ (Gsk3a cKO) and Gsk3b-flox/flox-Lck-Cre+ (Gsk3b cKO), respectively, resulting in mice with T cells devoid of either GSK-3a or GSK-3b ( Figure 1). This was apparent in both CD4 and CD8 T cell populations as shown in Figure 1A. Importantly, the loss of GSK-3 expression in T cells had no effect on the total number of CD8+ and CD4+ splenic T cells when compared to control mice ( Figure 1B). iScience Article T cell depletion of Gsk3b, alone or in combination with Gsk3a suppresses growth of two distinct tumor types Previously, we showed that using GSK-3 SMIs in in vivo mouse models of cancer led to suppression of tumor growth (Taylor et al., 2018). These studies focused on the CD8+ T cell population as the effector cells. However, SMIs act across all tissues and although CD8+ T cells were investigated ex vivo there is still the possibility that other cells might play a role within these models. We have now utilized the newly generated cKO mice to test if GSK-3 deletion in T cells alone is sufficient to suppress tumor growth and whether deletion of both isoforms is required. We first determined the tumor growth rate and survival kinetics in the three different strains of cKO mice bearing EL4 tumors in the flank ( Figure 2A) The mean survival time of 13 days for wt littermate mice was similar to that seen in Gsk3a cKO mice, with the exception of 1 mouse surviving of 8. In contrast, depletion of Gsk3b alone or in combination with Gsk3a gave a clear increase in the overall survival rate, with 50% of Gsk3ab cKO mice and 60% of Gsk3b cKO still surviving at day 40. This survival was concurrent with diminished tumor growth in these animals and the eventual eradication of tumor which was apparent within 15-20 days post-tumor challenge (lower panel Figure 2A). In a repeat study, EL4 tumor bearing mice were sacrificed at day 10 (when tumors were present in all mice) to analyze splenic and tumor-infiltrating cells. In agreement with the results from tumor-free animals ( Figure 1B), no difference (E) Flow cytometric profiles for T cells isolated from tumor-infiltrating cells (TILs) (data taken from animals at day 10 and representative of 5 samples). Data represent three combined independent experiments of n R 5 mice per group. Data are represented as mean G SEM. Groups compared using unpaired t test. ***p < 0.0001; **p < 0.001; *p < 0.01; ns, no significant difference relative to controls. Log rank tests performed for Kaplan-Meier survival data.

OPEN ACCESS
iScience 24, 102555, June 25, 2021 3 iScience Article was seen in the frequency of splenic CD4+ or CD8+ T cells between the different cKO mice bearing EL4 tumors ( Figure 2B). Previous studies using GSK-3 SMIs have shown that control of tumor growth is associated with a decrease in PD-1 expression, correlated with an increase in the expression of the transcription factor T-bet (Taylor et al., 2016). Here, using quantitative real-time polymerase chain reaction (qRT-PCR) of splenic CD8+ T cells, we found Pdcd1 gene expression to be significantly lower in the Gsk3ab cKO compared to wt littermates, and this was associated with increased Tbx21 expression ( Figure 2C). A decrease in Pdcd1 expression was also seen in the Gsk3b cKO, whilst statistically significant, the reduction in expression of Pdcd1 in Gsk3b cKO T cells was not as great as that seen in the double cKO. There was no effect on Pdcd1 expression in Gsk3a cKO cells but all three cKO strains showed increased Tbx21 gene expression ( Figure 2C).
Analysis of the tumors revealed no difference in the infiltration of CD4+ or CD8+ T cells in the double cKO mice when compared to wt littermates, in agreement with results obtained using GSK-3 SMIs (Taylor et al., 2018). However, deletion of Gsk3b alone resulted in significantly increased infiltration of both CD4+ and CD8+ T cells ( Figure 2D). Tumor-infiltrating CD8+ T cells from the double cKO mice had significantly higher levels of granzyme B expression and increased numbers of IFNg expressing cells. Although the latter was true for all three strains of GSK-3 cKO, the levels of granzyme B were only increased when both GSK-3 isoforms were depleted ( Figure 2D). These findings indicate that the regulatory effects of GSK-3 on these events are more nuanced that previously appreciated. It appears the required events that determine tumor infiltration are different from those needed for the differentiation of T cells leading to Granzyme B expression. Unlike these differences, cell surface PD-1 expression was measured on tumor-infiltrating lymphocytes (TILs) and showed a clear reduction in the double and Gsk3b cKO mouse cells ( Figure 2E) and a lesser reduction in TILs from the Gsk3a cKO mice, consistent with the Pdcd1 gene expression data from the spleen ( Figure 2C).
We next sought to confirm the results obtained from cKO mice bearing EL4 tumors with those bearing a second tumor type, B16 melanoma, and to determine whether deletion of one GSK-3 isoform could lead to suppression of both flank tumors and pulmonary metastasis. B16 tumor cells tagged with luciferase were injected intradermally into the three different strains of cKO mice and control animals and the tumor growth rate and survival kinetics assessed. The wt littermate mice and Gsk3a cKO mice bearing B16 flank tumors demonstrated a mean survival of 15 and 18 days, respectively, with the exception of 1 surviving mouse of 8 in the Gsk3a cKO ( Figure 3A). In contrast, the double GSK-3 cKO demonstrated a 50% survival rate at day 50. This was slightly lower in the Gsk3b cKO, with 40% of mice surviving at day 50; here increased survival was associated with diminished tumor growth and the eventual eradication of tumor was apparent within 15-20 days post-tumor challenge (lower panel Figure 3A). It should be noted that there was an overall prolonged survival of the Gsk3b cKO mice of 5-10 days compared to all other strains, with tumor growth being slower in these mice ( Figure 3A).
Splenic and tumor-infiltrating cells were analyzed from B16 flank tumor bearing mice which were sacrificed at day 12 (a time point at which tumors were present in the majority of animals to enable TIL isolation). As expected, no difference was seen in the number of splenic CD4 + or CD8 + T cells from the different cKO mice compared to wt littermates ( Figure 3B). qRT-PCR of splenic CD8 + T cells demonstrated Pdcd1 transcription to be significantly lower in the double GSK-3 cKO compared to wt littermates, and this was associated with increased Tbx21 gene expression ( Figure 3C). Furthermore, as seen in the EL4 lymphoma model, a significant decrease in Pdcd1 gene expression was also seen in the Gsk3b cKO but not in Gsk3a cKO cells ( Figure 3C). All three cKO strains showed increased Tbx21 gene expression.
The Gsk3b cKO had a significantly higher number of infiltrating CD4+ and CD8+ T cells compared to the other cKOs and wt littermates ( Figure 3D). Further analysis demonstrated tumor-infiltrating cells from GSK-3 double cKO mice to have higher levels of granzyme B expression compared to the single cKO mice. All three strains of GSK-3 cKO had increased levels of IFNg expressing TILs compared to wt littermates. iScience Article tumor types in the flank. Human cutaneous melanoma metastasizes to multiple sites, including the lungs and the B16 model can seed tumors in the lung following tail vein injection. We therefore analyzed the requirement for T cell GSK-3 isoforms in B16 pulmonary metastasis in our cKO mice. Luciferase tagged B16 tumor cells were injected intravenously into the different cKO and control mice. At day 24, mice were injected intraperitoneally with luciferin and scanned by IVIS Lumina imaging; data from five mice from each group are shown along with the total flux measurements from these groups ( Figure 4A). As found in the EL4 and B16 flank models, rejection of B16 pulmonary tumors was enhanced by deletion of GSK-3b, revealed by the Gsk3b cKO and double cKO mice but not the Gsk3a cKO mice ( Figure 4A). This was shown further by tracking tumor spread within the lung at various timepoints ( Figure 4B). This method also aided in the tissue collection from mice, ensuring that spleen and tumor samples were taken when tumors were apparent in all the different strains (Day 18). As found in the flank models, no difference was seen in the number of splenic CD4 + or CD8 + T cells between the different cKO mice ( Figure 4C). Splenic CD8 + T cells from the double cKO and Gsk3b cKO mice showed the expected reduction in Pdcd1 gene expression and all three strains of cKO mice showed increased Tbx21 expression ( Figure 4D). Analysis of TILs revealed increased infiltration of CD4+ and CD8+ T cells in the Gsk3b cKO mice and all three strains of cKO mice exhibited an increased frequency of CD8+ cells expressing granzyme B or IFNg compared to controls ( Figure 4E). Collectively, these data demonstrate a key role for the GSK-3b isoform in controlling T cell mediated anti-tumor immunity. Data represent three combined independent experiments of n R 5 mice per group. Data are represented as mean G SEM. Groups compared using unpaired t test. ***p < 0.0001; **p < 0.001; *p < 0.01; ns, no significant difference relative to controls. Log rank tests performed for Kaplan-Meier survival data.

OPEN ACCESS
iScience 24, 102555, June 25, 2021 5 iScience Article Tumor rejection in Gsk3b cKO is associated with higher numbers of tumor-infiltrating T cells Mice injected intravenously with B16 melanoma cells were culled 18 days later and lungs removed for immunohistochemistry ( Figure 5). This timepoint was chosen to ensure that all mice were tumor bearing as confirmed prior to harvest by in vivo imaging. Staining showed CD4 + and CD8 + T cells to be scattered throughout the tumor region ( Figures 5A and 5B, respectively) however, significantly increased numbers of CD4 + T cells were apparent in the Gsk3b cKO sections as determined using Imagescope software (Figure 5A, right panel). Further staining demonstrated all strains to contain PD-1 expressing cells throughout the tumor ( Figure 5). A slight decrease in PD1 + cells was observed in the Gsk3b cKO, but this reduction was only significant when both isoforms were depleted in the double cKO. Interestingly, Foxp3 expression was significantly increased in all three cKOs compared to wt littermates.
T cell distributions within the peribronchial regions (areas surrounding the tumors rather than within the tumor itself) were also examined revealing accumulations of CD4 + T cells in the double cKO and Gsk3a cKO but not the Gsk3b cKO mice ( Figure 6A). Furthermore, there was a significant reduction in CD8 + T cells in the Gsk3b cKO mice ( Figure 6B). This reduction in peribronchial CD4+ and CD8+ T cells in the Gsk3b cKO mice was also reflected in lower numbers of Foxp3 and PD1 expressing cells ( Figures  6C and 6D). Taken together, these data demonstrate a clear difference in T cell distribution in response to the presence of tumor growth with a higher level of tumor-infiltrating T cells being seen in the Gsk3b cKO mice. and IFNg expressing CD8+ tumor-infiltrating cells (TILs) (taken from animals at day 18) as determined by flow cytometry. ***p < 0.0001; **p < 0.001; *p < 0.01; ns, no significant difference relative to controls. Data represent three combined independent experiments of n R 5 mice per group. Data are represented as mean G SEM. Groups compared using unpaired t test. ***p < 0.0001; **p < 0.001; *p < 0.01; ns, no significant difference relative to controls. iScience Article Further to this, lung tissue was taken from a single Gsk3b cKO mouse which had completely rejected the B16 tumor by day 18 and IHC performed ( Figure S1). There was a significant increase in the number of CD4 + T cells and Foxp3 expressing cells in the lungs of this mouse that rejected the tumor. This suggests that there is an increase in the cells migrating to the tumor site within Gsk3b cKO mice and that they efficiently infiltrate any tumors and contribute to controlling disease progression.

DISCUSSION
Despite our establishment of the importance of GSK-3 in the activation of T cell function needed for tumor rejection, the relative roles of the different GSK-3 isoforms in tumor rejection have been unclear. Our results demonstrate that GSK-3beta activity in T cells provides a more permissive environment for the tumor models tested here and that deletion of this isoform alone can enhance T cell mediated tumor rejection. However, the mechanism behind this remains to be fully elucidated. We have previously shown that inhibition of GSK-3 using SMIs leads to down regulation of checkpoint molecules PD-1 and LAG-3, whilst upregulating Tbet (Rudd et al., 2020;Taylor et al., 2016). These findings are supported here when genes encoding both GSK-3 isoforms are targeted in a cKO model. Furthermore, we show that the two isoforms contribute in different ways to the various T cell functions linked to tumor rejection. Overall, the GSK-3b isoform appeared to be dominant over the GSK-3a isoform, although both isoforms were needed for optimal tumor rejections.
The targeting of the GSK-3b isoform alone also provided a tumor suppressive effect, but the mechanism behind this may not be the same. There is some reduction in tumor growth, but this is not as striking as that seen when both isoforms are depleted. This suggests that deletion of the GSK-3a isoform is also needed for optimum suppression of tumor growth, although depletion of the GSK-3a isoform alone failed to show a phenotype. Whether its contribution was additive in promoting T cell rejection or it operated cooperatively is not clear. It is possible that the function of the GSK-3b is in some manner cross-regulated by the GSK-3a isoform but how both isoforms operate remains unclear. Similar to tumor rejection, depletion of Gsk3a alone, unlike Gsk3b, has no direct effect Data representative of two independent experiments. Groups compared using unpaired t test. ***p < 0.0001; **p < 0.001; *p < 0.01; ns, no significant difference relative to controls.

OPEN ACCESS
iScience 24, 102555, June 25, 2021 7 iScience Article on PD-1 expression, whereas depletion of both together caused the effect to be magnified. One possibility is that the GSK-3b isoform is directly involved in the PD-1 pathway, whereas GSK-3a may act indirectly, affecting another protein which then acts on PD-1 expression. As, GSK-3a alone has no obvious effect on PD-1 expression, this would suggest that the protein it does act upon is also regulated by GSK-3b and therefore, the effects are only seen when both isoforms are switched on or off.
Another surprising finding was the suggestion that the two isoforms might differentially affect the CD4 and CD8 TILs in mice. CD4 + T cells were found to be at higher levels in tumors in Gsk3b cKO mice compared to the levels seen in Gsk3a cKO mice. More surprisingly, these levels were also higher than those seen in the double cKO mice. Although the increase in the numbers of infiltrating CD8+ T cells did not reach significance, it was apparent that some tumor regions had very high levels of CD8+ T cells, whilst others had very low levels. It may be speculated that had the samples been taken over a time course, an increase in T cell infiltration would have been more apparent and that perhaps this time point reflects a point at which CD8+ T cells were moving into the tumor site to attack tumor cells, or indeed they were already migrating away from these sites after fulfillment of their role. As seen in 40% of these cKO mice, a later time point may have resulted in no tumor being present and therefore it is possible that this is the recession following the peak of the immune response. Further analysis into the dynamics of T cell migration and infiltration are warranted to fully understand this aspect of GSK-3 regulation, as are a detailed analysis of the activation and maturation status of T cells in the different GSK-3 backgrounds; however, this is beyond the scope of the current manuscript. Regardless of these dynamics, the comparison of the infiltration patterns between these cKO mice at a specific time point suggests that the beta isoform plays a dominant role in controlling cellular movement and T cell migration. This finding is consistent with our previous work demonstrating the inhibition of GSK-3 (by SMIs) reduces motility and increase cell-to-cell contact, this coupled with enhanced CTL potential, led to tumor suppression (Taylor and Rudd, 2020).
The results showing higher tumor infiltration of T cells due to the beta isoform may be due to chemotactic effects enhancing migration of these T cells to the tumor site, or T cell retention at the tumor site, resulting iScience Article in prolonged contact and efficient killing of the tumor cells. Oddly, this effect is not seen in the Gsk3a cKO mice, nor in the double cKO which suggests a possible opposing effect of the alpha isoform. The data outlines a possible complex interplay between the two isoforms, each playing their own part in different pathways, but yet together synergizing to reduce PD-1 expression, an effect which was optimally seen when both isoforms are depleted.
Previously our work has focused on CD8 T cells, more specifically cytotoxic T cells and their enhanced ability to kill tumor cells when GSK-3 is inactivated. However, these results clearly demonstrate that the GSK-3b isoform also impacts upon CD4+ T cells in the context of tumor immunity. Previous work has shown that GSK-3b is important for the differentiation of Th17 cells (Beurel et al., 2011), whilst GSK-3a is needed for Th1 cells (Beurel et al., 2013). Further investigation into the role of both isoforms in the different subsets is warranted to fully understand the mechanisms behind GSK-3 regulation in controlling the immune response. Such studies could potentially lead to new targeted therapeutic approaches.

Limitations of the study
This study focused on determining the relative contributions of the two isoforms of GSK-3 on T cell-mediated tumor immunity. Our investigation extended to CD8+ T cells and the expression of PD-1 and Tbet by these cells as indicated from our previous work using GSK-3 inhibitors. Our results reveal both overlapping and non-redundant functions of the isoforms, but the detailed mechanisms by which the two isoforms regulate the T cell phenotype remain unclear at this time. A detailed understanding of the mechanisms by which GSK-3 regulates T cell activity requires analysis of T cell activation and maturation status in the different GSK-3 backgrounds. In addition, determination of the pathways downstream of the two isoforms and further mechanistic studies of T cell migration and tumor infiltration will reveal the pathways by which GSK-3 regulates T cell activity. These mechanistic studies were beyond the scope of the current study.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:  Picton, C., Woodgett, J., Hemmings, B., and Cohen, P. (1982). Multisite phosphorylation of glycogen synthase from rabbit skeletal muscle. Phosphorylation of site 5 by glycogen synthase kinase-5 (casein kinase-II) is a prerequisite for phosphorylation of sites 3 by glycogen synthase kinase-3. iScience Article Primary T cell cultures Spleen cells were treated with a hypotonic buffer with 0.15M NH4CL, 10mM KHCO3 and 0.1mM EDTA, pH 7.2 to eliminate red blood cells before suspension in RPMI 1640 medium supplemented with 10% FCS, 50uM beta-mercaptoethanol, sodium pyruvate, 2 mM L-glutamine, 100 U/ml penicillin and streptomycin (GIBCO). T cells were isolated from tumor-infiltrating cells, spleen and lymph node samples using negative section of magnetically labeled cells (Miltenyi). In some cases, whole lymphocyte samples were used for flow cytometry to determine effect on other cell sub-types.

Cell lines
Tumor cells included B16 F10 melanoma and EL4 lymphoma cells (obtained from the ATCC). Cell lines were cultured in DMEM medium supplemented with 10% FCS, 50uM beta-mercaptoethanol, sodium pyruvate, 2 mM L-glutamine, 100 U/ml penicillin and streptomycin (GIBCO). Each cell line was grown to achieve adequate numbers for freezing, followed by repeated thawing for use in the described experiments. The length and time between thawing and use in experiments was on average 2-3 weeks. The cell lines were authenticated by means cell surface staining and flow cytometry for characteristic markers and by their growth properties as described in the literature. Cell cultures were occasionally tested for mycoplasma (last tested in 2011).

Flow cytometry
Flow cytometry of antibody staining of surface receptors was conducted by suspending 10 6 cells in 100ml PBS and adding antibody (1:100) for 2hr at 4 C. Cells were then washed twice in PBS and in some cases suspended in 100ml PBS with secondary antibody for a further 1h at 4 C. Cell staining was analyzed on a Beckman Coulter CytoFLEX S flow cytometer and by CytExpert software. For intracellular staining, cells were fixed in 4% paraformaldehyde (PFA), permeabilized with 0.3% saponin (Sigma-Aldrich) and stained with the desired antibody in saponin containing PBS for 2hr at 4 C, followed by a secondary Ab incubation where primary antibodies were not conjugated.

OPEN ACCESS
iScience 24, 102555, June 25, 2021 iScience 24, 102555, June 25, 2021 iScience Article Tissue was then passed through a strainer and cells collected and layered onto ficoll before centrifugation. Tumor-infiltrating cells were then collected from the lymphocyte layer.

QUANTIFICATION AND STATISTICAL ANALYSIS
The mean and SE of each treatment group were calculated for all experiments. The number of samples is indicated in the figure legends. Unpaired Student's t tests or ANOVA tests were performed using the InStat 3.0 software (GraphPad). In certain instances, statistics were done using 2-way ANOVA, or by non-parametric Mann Whitney at each time point. * P < 0.05, ** P < 0.01, *** P < 0.001. Log rank tests for Kaplan-Meier survival data were performed to determine significant differences.