Small Molecule Inhibition of GSK-3 Specifically Inhibits the Transcription of Inhibitory Co-receptor LAG-3 for Enhanced Anti-tumor Immunity.

Immune checkpoint blockade using antibodies against negative co-receptors such as cytolytic T cell antigen-4 (CTLA-4) and programmed cell death-1 (PD-1) has seen much success treating cancer. However, most patients are still not cured, underscoring the need for improved treatments and the possible development of small molecule inhibitors (SMIs) for improved immunotherapy. We previously showed that glycogen synthase kinase (GSK)-3α/β is a central regulator of PD-1 expression, where GSK-3 inhibition down-regulates PD-1 and enhances CD8+ cytolytic T cell (CTL) function, reducing viral infections and tumor growth. Here, we demonstrate that GSK-3 also negatively regulates Lymphocyte Activation Gene-3 (LAG-3) expression on CD4+ and CD8+ T cells. GSK-3 SMIs are more effective than LAG-3 blockade alone in suppressing B16 melanoma growth, while their combination resulted in enhanced tumor clearance. This was linked to increased expression of the transcription factor, Tbet, which bound the LAG-3 promoter, inhibiting its transcription, and to increased granzyme B and interferon-γ1 expression. Overall, we describe a small molecule approach to inhibit LAG-3, resulting in enhanced anti-tumor immunity.


In Brief
Lymphocyte Activation Gene-3 (LAG-3) is a checkpoint molecule that has been targeted alongside programmed cell death (PD-1) for immunotherapy. Here, Rudd et al. demonstrate inhibition of glycogen synthase kinase-3 to control LAG-3 expression and that its combination with anti-LAG-3 antibody enhances T cell responses against tumors, causing tumor regression.

INTRODUCTION
Immune checkpoint blockade (ICB) is a promising approach for the treatment of cancer. The approach involves the blockade of negative co-receptors on T cells such as cytolytic T cell antigen-4 (CTLA-4) and programmed cell death-1 (PD-1). PD-1 expression on tumor-infiltrating CD8 + T cells correlates with impaired function (Ahmadzadeh et al., 2009), while PD-L1 expression on tumor cells facilitates their escape from the immune response (Iwai et al., 2002). ICB reverses T cell exhaustion and restores T cell function-ality (Freeman et al., 2006;Wherry, 2011). The cancer treatment drug nivolumab (Opdivo) against PD-1 was approved as a firstline treatment for advanced melanoma and has been used alone, and in combination with ipilimumab (Yervoy), an anti-CTLA-4 antibody (Ab) resulting in high response rates (Hodi et al., 2010;Sharma et al., 2011;Wolchok et al., 2013). However, despite this success, many patients are still not cured and may suffer immune-related adverse events (irAEs). This poor prognosis continues to highlight a need to develop novel clinical interventions. In this context, we have identified the enzyme glycogen synthase kinase-3 (GSK-3) as a major nexus in the control of PD-1 expression on T cells (Taylor et al., 2016). We also showed that small molecule inhibitors (SMIs) of GSK-3 can suppress tumor growth in a manner comparable to PD-1 Ab blockade (Taylor et al., 2018).
Among the expanding numbers of inhibitory co-receptors on T cells are Lymphocyte Activation Gene-3 (LAG-3), T cell immunoreceptor with immunoglobulin (Ig) and ITIM domains (TIGIT), CD244 (2B4), T cell Ig and mucin-domain containing-3 (TIM-3), CD160, killer-cell lectin like receptor G1 (KLRG-1), and B and T lymphocyte attenuator (BTLA-4) (Baumeister et al., 2016). LAG-3 is an activation antigen that negatively regulates T cells (Workman et al., 2004;Workman and Vignali, 2003). It is a cell-surface molecule initially discovered on natural killer (NK) cells (Triebel et al., 1990) and then on activated T cells (Huard et al., 1994), B cells, and plasmacytoid dendritic cells (Workman et al., 2009). Furthermore, LAG-3 is a member of the Ig superfamily, containing four Ig loops, with structural homology to CD4. Unlike CD4, the membrane-distal IgG domain contains an extra loop region that appears important for binding to major histocompatibility complex class II (MHC class II) molecules (Triebel et al., 1990). Aside from MHC class II antigens, fibrinogen-like protein-1 is a major immune inhibitory ligand of LAG-3 . Studies in LAG-3 knockout mice have demonstrated an inhibitory role for LAG-3 in controlling both CD4 and CD8 T cell proliferation in vitro and in vivo (Workman et al., 2004). This negative effect on T cell function is exerted through a poorly understood but unique amino acid sequence (KIEELE) in the intracellular domain (Workman et al., 2002). Importantly, LAG-3 and PD-1 are co-expressed at high levels in tumor-infiltrating lymphocytes (TILs) from murine ovarian tumor and EG7-bearing mice, and blockade of both molecules enhances CD8 + effector T cell frequency and function, thus enhancing anti-tumor immunity (Andrews et al., 2017;Huang et al., 2015;Woo et al., 2012).
We originally demonstrated that the inhibition of serine/threonine kinase GSK-3 downregulates PD-1 expression, which enhances T cell responses against viral infections (Taylor et al., 2016(Taylor et al., , 2018. There are two ubiquitously expressed and highly conserved isoforms of GSK-3, GSK-3a and GSK-3b, which have shared and distinct substrates, as well as functional effects. Both forms have been implicated in processes ranging from glycogen metabolism to gene transcription, apoptosis, and microtubule stability. The notable aspect of GSK-3 is that it is constitutively active in resting T cells (Embi et al., 1980;Woodgett, 1990) and is inhibited by receptor-induced activation signals (Woodgett, 2001).
Our findings also showed that GSK-3 inhibition operated primarily via a reduction in PD-1 transcription on CD8 + T cells (Taylor et al., 2016(Taylor et al., , 2018. We further showed that SMIs against GSK-3 are as effective as anti-PD-1 Abs in restricting B16 and EL-4 tumor growth in mice (Taylor et al., 2018). In this article, we demonstrate that GSK-3SMIs and siRNAs can also downregulate LAG-3 expression on T cells and that combination therapy of SMIs with LAG-3 can significantly enhance the clearance of tumors in mice.

Small Molecule Inhibition of GSK-3 Downregulates PD-1 and LAG-3 in T Cells
We have previously shown that the inactivation of GSK-3a/b increases Tbet expression for the downregulation of PD-1 expression (Taylor et al., 2016(Taylor et al., , 2018Taylor and Rudd, 2017). By contrast, we found no effects on the expression of other T cell receptors, including CD3, CD4, CD8, CD28, CTLA-4, CD44, CD62L, TIM-3, CD3, BTLA, NKG2D, CD122, IL-2Ra, CD25, CD69, FasL, and CD8 (Taylor et al., 2016). However, given that Tbet can regulate the expression of multiple genes (Lazarevic and Glimcher, 2011), we widened our survey of potential GSK-3 targets. Of particular interest was the expression of other inhibitory receptors (IRs). To this end, CD8 + OT-1 T cells were stimulated with OVA 257-264 (SIINFEKL) peptide in the presence or absence of the GSK-3 SMI SB415286 over 7 days, during which samples were assessed for LAG-3 expression using flow cytometry and PCR analysis. As expected from previous work (Taylor et al., 2016(Taylor et al., , 2018Taylor and Rudd, 2017), the OVA-induced upregulation of PD-1 was markedly reduced by the presence of the GSK-3 SMI. Surprisingly, the expression of LAG-3 was also downregulated by the presence of SMI, while the expression of other IRs such as CTLA-4, BTLA-4, and TIM-3 was unaffected ( Figure 1A).
T cells isolated from wild-type C57BL/6 mice activated with anti-CD3 showed an increase in LAG-3 expression that was reduced by the presence of SB415286 ( Figure 1B, left panel). As expected, the level of LAG-3 expression on anti-CD3-activated Pdcd À/À (PD-1-deficient) T cells was higher than seen on control T cells, consistent with the inhibitory effect of PD-1 on T cell activation. Nevertheless, SB415286 markedly reduced LAG-3 levels on anti-CD3-activated Pdcd À/À T cells ( Figure 1B, right panel). These data showed that GSK-3 inactivation downregulates LAG3 independent of PD-1 expression.
We next assessed whether the suppression of LAG-3 expression was mediated at the level of transcription of LAG-3 throughout the 7-day time course, as seen by use of real-time PCR ( Figure 1C). The level and time course of the inhibition of LAG-3 transcription was similar to that of PD-1, while CTLA-4 transcription was unaffected.
LAG-3 is also known for its role in regulating CD4 + T cell function (Workman et al., 2002); therefore, the effect of SMI in CD4 + T cells was investigated using OT-II transgenic mice. CD4 + OT-II T cells were activated with OVA 323-339 (ISQAVHAA-HAEINEAGR) peptide in the presence or absence of SMI for 7 days. Flow cytometry showed that GSK-3 inactivation reduced expression of LAG-3 on CD4 + T cells (Figures S1A and S1B). Similarly, inhibition of GSK-3 decreased LAG-3 expression on NK cells (defined as CD3 À NKp46 + cells) ( Figure S1C). Overall, our findings identified LAG-3 as a novel target regulated by the inhibition of GSK-3 in CD4 and CD8 + T cells.

Anti-LAG-3 Treatment Combined with GSK-3 SMIs
In Vitro Enhances CD8 + T Cell Cytolytic Responses Given that anti-LAG-3 and GSK-3 SMI treatment would not be expected to block LAG-3, we next assessed whether they cooperate to enhance antigen-specific CD8 + T cell cytolytic responses ( Figure 2). CD8 + OT-I T cells were stimulated with OVA 257-264 for 5 days in the presence of anti-LAG-3 Ab at different doses. After 5 days of activation, the killing of EL4 lymphoma target cells by cytolytic T cells (CTLs) using a lactate dehydrogenase (LDH) release detection assay. EL4 cells were pulsed with OVA peptide before the incubation to induce an antigen-specific response. Non-pulsed EL4 cells were used as a control for background killing. A trend was observed in which anti-LAG-3 increased killing; however, this did not achieve statistical significance (Figure 2A). By contrast, the addition of GSK-3 SMIs did increase in vitro OT-I killing, and the combination of GSK-3 SMIs and anti-LAG-3 increased the killing further at statistically significant levels ( Figure 2B). This was seen over a range of target ratios and was generally better than the cooperativity seen with anti-PD-1 and anti-LAG-3 combination therapy. Together, these data show that under conditions in which anti-LAG-3 blockade alone has limited effects in unleashing T cell responses against tumors, the inhibition of GSK-3, while having an effect on its own, sensitizes T cells to be more responsive to tumor antigens.
A similar cooperativity was observed when small interfering RNAs (siRNAs) against the GSK-3a and GSK-3b isoforms were used to knock down (KD) expression ( Figure 2C). Naive T cells were transfected before a 5-day activation period with OVA-pulsed EL4 cells and then assessed for cytolytic function using LDH release assays. siRNA treatment reduced GSK-3a and GSK-3b protein substantially, as seen by flow cytometry ( Figure 2C, top panel), in which with the use of fluorescent GFP-conjugated siRNAs, more than 80% of cells took up the siRNAs. siRNAs to GSK-3 increased OT-I-mediated cytolysis of EL4-OVA targets when compared with the scrambled (Scrm) siRNA control ( Figure 2C, bottom panel). This was significantly increased by the addition of anti-LAG-3 Ab. These data again showed that the combination of GSK-3 inactivation and anti-LAG-3 Ab therapy markedly increased the killing capacity of CD8 + OT-I CTLs.
Combined Treatment of Anti-LAG-3 Ab with GSK-3 SMIs In Vivo Enhances CD8 + T Cell Cytolytic Responses To assess whether the combination of anti-LAG-3 with GSK-3 inhibition was effective in limiting tumor growth, B16 tumor cells tagged with luciferase were injected intravenously into C57BL/6 mice and treated with the GSK-3 SMI SB415286, anti-LAG-3, and/or anti-PD-1 (Figure 3). The optimal dose of SB415286, LAG-3, and PD-1 was established at 200 mg/mouse/ treatment. SB415286 was administered every 2 days on six occasions, beginning 7 days post-injection of tumor cells, with the Ab given on four occasions (days 7, 10, 13, and 16). At days 10, 12, and 19, mice were injected intraperitoneally with luciferin and scanned by IVIS Lumina imaging ( Figure 3A; Figure S2). Importantly, anti-LAG-3 treatment alone did not have a statistically significant effect on B16 tumor growth. Anti-PD-1 or SB415286 treatment alone reduced the luciferase signal relative to control at all time points. However, the combination of anti-LAG-3 with SB415286 treatment was seen to be the most effective in decreasing tumor growth. As reported (Huang et al., 2015;Zhang and Vignali, 2016), anti-LAG-3 plus anti-PD-1 also showed a trend in cooperating to reduce tumor size ( Figure 3A); however, the combination of anti-LAG-3 and SB415286 was even more effective than anti-LAG-3 plus anti-PD-1. The efficacy of anti-LAG-3 with SB415286 treatment was observed by total flux measurements and the percentage of mice showing complete tumor clearance ( Figure 3A, left panels). In both instances, the combination of anti-LAG-3 and GSK-3 inactivation was more effective than anti-LAG-3 plus anti-PD-1.
The cooperativity between anti-LAG-3 and SB415286 was particularly evident when we measured the complete loss of tumor ( Figure 3A, right histogram). In this representative experiment, the combination was significantly more potent in clearing the tumor (3/5 mice) when compared with the effects of anti-LAG-3 alone (0/5 mice) and SB415286 alone (1/5 mice). While the synergistic effects of PD-1 and LAG-3 on T cell function have been reported previously (Huang et al., 2015;Nguyen and Ohashi, 2015;Okazaki et al., 2011), our findings show that the combination of anti-LAG-3 plus GSK-3 inhibition is more effective than this previously documented combination against B16 melanoma. No direct effect of SB415286 on tumor cells was observed as previously demonstrated by treating B16 tumor-bearing Rag2À/À mice (devoid of B and T cells) with SB415286. Further, SB415286 had no obvious effect on tumor growth in vitro (Taylor et al., 2018).
The presence of SB415286 alone increased the presence of GZMB + CD8 + CD3 + T cells and TNF + IFNg + CD3 + CD8 + T cells ( Figure 3B), while anti-LAG-3 alone also increased GZMB + T cells. However, the combination of anti-LAG-3 plus SB415286 or anti-LAG-3 plus PD-1 further increased the presence of TNF + IFNg + CD3 + CD8 + T cells and GZMB-expressing CD8 + CD3 + TILs in accordance with the increased efficacy of tumor killing. The combination of anti-LAG-3 with SB415286 was generally more effective than anti-LAG-3 combined with anti-PD-1. Lastly, an examination of TILs showed that both LAG-3 and PD-1 on TILs were downregulated in this in vivo model (Figure 3C). Further analysis by CytoBank showed that anti-LAG-3 and PD-1 defined several overlapping populations of CD8 + TILs ( Figure 3D). Interestingly, the three CD8 + islands (i-iii) that were the most brightly stained islands with anti-LAG-3 (upper panel) were also stained with anti-PD-1 (lower panel). TILs from mice treated with SB415286 showed a reduced expression of LAG-3 on each of the three islands. These data showed that three subpopulations of cells expressing LAG3 and PD-1 were affected by SB415286.

GSK-3 Downregulation or Inhibition Blocks LAG-3 Transcription
We have previously shown that inhibition of GSK-3 led to an increase in the transcription of Tbx21 (encodes T-box transcription factor Tbet), which was inversely correlated with Pdcd1 transcription (Taylor et al., 2016). We therefore examined whether the upregulation of Tbet due to GSK-3 inhibition was also connected to the downregulation of LAG-3. Mice were subjected to B16 melanoma and treated with, or without, SB415286. CD8 + T cells were isolated from spleens harvested on day 19, and RT-PCR was performed ( Figure 4A). Concurrent with the preceding findings, LAG-3 expression was reduced by the presence of SB415286. Furthermore, this was associated with an increase in Tbx21 transcription. Chromatin immunoprecipitation (ChIP) with anti-Tbet Ab followed by RT-PCR confirmed that GSK-3 inhibition increased Tbet binding to the LAG-3 promoter in anti-CD3-activated primary T cells ( Figure 4B).
Tbet siRNA expression decreased Tbet transcripts while increasing LAG-3 transcription, a result consistent with the negative regulation of LAG-3 by Tbet (Durham et al., 2014;Graydon et al., 2019;Kao et al., 2011). Importantly, SB415286 failed to reduce LAG-3 expression in cells expressing Tbet siRNA, as assessed by flow cytometry ( Figure 4D). Tbet siRNA impaired CTL killing over different E:T ratios ( Figure 4E). This inhibition was partially restored by LAG-3 blockade; however, GSK-3 inhibition had no effect. This result indicated that the modulatory effects of GSK-3 on LAG-3 expression operated in a pathway that required the expression of Tbet.
Overall, our findings demonstrating GSK-3 regulation of LAG-3 expression introduce a new pathway in the regulation of T cell responses in cancer immunotherapy. We have shown that GSK-3 now targets the two central IRs that are linked to T cell non-responsiveness or exhaustion (McLane et al., 2019;Wherry, 2011). Anti-PD-1 therapy, although effective, generates residual CD8 + T cells expressing PD-1 + LAG-3 + T cells with an exhaustion phenotype (Wei et al., 2019). GSK-3 SMIs, by reducing the expression of both PD-1 and LAG3, should promote the functionality of T cells in tumors, either in the context of anti-PD-1 or LAG-3 immunotherapy. Future studies will assess whether the GSK-3-PD1/LAG3 axis also affects other key players like the transcription factor TOX in determining the response of intra-tumoral T-cells (Alfei et al., 2019). Importantly, our findings showed that the combination of anti-LAG-3 with GSK-3 inactivation was as effective as, or more effective than, anti-LAG-3 and anti-PD-1 in B16 melanoma cancer therapy. The synergic effect of LAG-3 with PD-1 blockade has been studied extensively (Woo et al., 2012). One advantage of our approach in targeting GSK-3 involves the potential oral use of SMIs, which could be used in combination with anti-LAG-3. Intriguingly, in this context, we found that the combination of anti-LAG-3 plus SB415286 was more effective in the complete clearance of tumor mass than the combination of anti-LAG-3 and anti-PD-1. Its ability to compliment anti-LAG-3 presumably results from the limited effectiveness of Ab blockade in maintaining the complete blockade of receptors in vivo, as well as its ability to increase the expression of Tbet (Taylor et al., 2016), a key transcription factor for the development of CTLs and the transcription of granzymes and interferon-g1 (Glimcher, 2007). Overall, our findings show that GSK-3 inactivation can potentiate the anti-tumor effects of anti-LAG-3 even more effectively than the combination anti-LAG-3 and PD-1 to eliminate tumor cells.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: Taylor, A., and Rudd, C.E. (2017). Glycogen Synthase Kinase 3 inactivation compensates for the lack of CD28 in the priming of CD8+ cytotoxic T-Cells: Implications for anti-PD-1 Immunotherapy. Front. Immunol. 8, 1653.

STAR+METHODS Isolation of tumor infiltrating lymphocytes (TILs)
Lungs were harvested from mice at the time indicated. Tissue was disrupted using a blade and then incubated in HBSS solution containing 200units/ml of collagenase at 37 C for 2 hr. 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.

DATA AND CODE AVAILABILITY
This study did not generate/analyze datasets/code