HVEM is a novel immune checkpoint for prostate cancer immunotherapy in humanized mice

Prostate cancer is one of the deadliest cancer in which immunotherapy with current immune checkpoint inhibitors have shown limitations. Thus, it is crucial to investigate other checkpoints to determine whether immunotherapy could be feasible to prevent disease progression in prostate cancer patients. Here, we first show that a pair of molecules (HVEM/BTLA) are associated to disease progression in patients. We next show that immunotherapy aimed to target HVEM reduced tumor growth two-fold in vivo in a humanized mice model of the pathology. We determine the mode of action of the therapy to be dependent on CD8+ T cells and associated with improved T cell activation and reduced suppression. We also formally demonstrate that HVEM/BTLA are novel immune checkpoints for the anti-tumor response. Our results indicate that targeting HVEM might be an interesting therapeutic option for prostate cancer patients. Abstract The Herpes Virus Entry Mediator (HVEM) delivers a negative signal to T cells mainly through the B and T Lymphocyte Attenuator (BTLA) molecule and thus, could represent a novel immune checkpoint during an anti-tumor immune response. A formal demonstration that HVEM can be targeted for cancer immunotherapy is however still lacking. Here, we first show that HVEM and BTLA were associated to a worse prognosis in patients with prostate adenocarcinomas, indicating a detrimental role for this pair of molecule during prostate cancer progression. We then show that a monoclonal antibody to human HVEM significantly impacted the growth of a prostate cancer cell line in immuno-compromised NOD.SCID.gc-null mice reconstituted with human T cells. Using CRISPR/Cas9, we showed that HVEM expression by the tumor was mandatory to observe the therapeutic effect. Mechanistically, tumor control was dependent on CD8 + T cells and was associated to an increase in the proliferation and number of tumor-infiltrating leukocytes. Accordingly, the expression of genes belonging to various T cell activation pathways were enriched in tumor infiltrating leukocytes, whereas genes associated with immuno-suppressive pathways were decreased, possibly resulting in modifications of leukocyte adhesion and motility. Finally, we developed a simple in vivo assay in humanized mice to directly demonstrate that HVEM was an immune checkpoint for T-cell mediated tumor control. Our results show that targeting HVEM is a promising strategy for prostate cancer immunotherapy. Biosystems). Subsequently, up to 2.106 live cells were stained with viability dye (eF506, Fixable Viability Dye, ThermoFisher) for 12 min. at 4°C, Fc receptor were blocked with human FcR Blocking Reagent (120-000-442, Miltenyi Biotec) and anti-CD16/32 (clone 2.4G2) for 10 min. The followings antibodies were added for 35 min. at 4°C: hCD45-BUV805 (HI30, BD), hCD3-PECyn7 (SK7, BD), hCD4-PerCP (RPA-T4, Biolegend), hCD8-APC-H7 (SK1, BD), hKi67-AF700 (B56, BD), hCD270-BV421 (cw10, BD), mCD45-BUV395 (30-F11, BD), hGranzymeB-APC (GB11, eBioscience), and hPerforin-PE (B-D48, Biolegend). For intracellular staining, Foxp3/Transcription Factor Staining (eBioscience) or Cytofix/Cytoperm (BD) buffer sets were used. Cells were washed with 1X PBS before acquisition on an X20 cytometer (Becton Dickinson (BD), San Jose, CA). The absolute count of different

anti-tumor immune response. A formal demonstration that HVEM can be targeted for cancer immunotherapy is however still lacking. Here, we first show that HVEM and BTLA were associated to a worse prognosis in patients with prostate adenocarcinomas, indicating a detrimental role for this pair of molecule during prostate cancer progression. We then show that a monoclonal antibody to human HVEM significantly impacted the growth of a prostate cancer cell line in immuno-compromised NOD.SCID.gc-null mice reconstituted with human T cells. Using CRISPR/Cas9, we showed that HVEM expression by the tumor was mandatory to observe the therapeutic effect. Mechanistically, tumor control was dependent on CD8 + T cells and was associated to an increase in the proliferation and number of tumor-infiltrating leukocytes. Accordingly, the expression of genes belonging to various T cell activation pathways were enriched in tumor infiltrating leukocytes, whereas genes associated with immuno-suppressive pathways were decreased, possibly resulting in modifications of leukocyte adhesion and motility. Finally, we developed a simple in vivo assay in humanized mice to directly demonstrate that HVEM was an immune checkpoint for T-cell mediated tumor

Introduction
Immune escape by tumor is now considered a hallmark of cancer [1]. Many immune mechanisms are involved to explain the loss of tumor control, including defective MHC function and expression, recruitment of suppressive immune cells, and expression of co-inhibitory receptors such as PD-L1 [2]. In the last few years, targeting co-inhibitory molecules with antibodies has shown impressive results in tumor regression and overall survival, leading to the approval of anti-CTLA-4, anti-PD-1 and anti-PD-L1 in numerous cancers [3]. However, the success of immune checkpoint inhibitors (ICI) is still partial and many patients fail to respond. This is particularly true for prostate cancer (PCa), the second deadliest cancer in the industrialized world and in which numerous clinical trials using ICI monotherapy have been disappointing [4] . Limited tumor infiltrate (cold tumors) or low expression of the targeted molecule may explain the relative inefficiency of ICI [5,6]. To overcome these limitations, it is necessary to explore other pathways that might be involved in immune escape and that could complement actual therapies.
Recently, a new co-inhibitory pair has been highlighted in anti-tumor immune response: HVEM (Herpes Virus Entry Mediator, TNFRSF14) and BTLA (B and T lymphocyte attenuator) [7]. These two molecules can be expressed by many immune cells, including T-cells, in which signaling through BTLA is associated with inhibition of their activation [8,9]. Additionally, the HVEM network includes many additional partners, such as LIGHT, Herpes Simplex Virus-1 (HSV-1) glycoprotein D (gD), lymphotoxin α (LTα) or CD160 [7]. Like BTLA, binding of HVEM to CD160 on T-cells is associated with an inhibition of their activation [10]. In contrast, LIGHT is clearly a T-cell activator since transgenic expression of LIGHT in T cells leads to massive activation, especially in mucosal tissues [11]. On the other hand, stimulation of HVEM expressed by T-cells by any of its ligands is associated with proliferation, survival and production of inflammatory cytokines, such as IL-2 and IFN-γ [10,12]. Thus, the HVEM network is complex and T cell activation or inhibition may follow HVEM engagment depending on cis or trans activation by various ligands in various type of cells.
Although the role of HVEM on T cell activation is well established, much less is known on the role of HVEM expressed by tumor cells on the immune system. Several clinical studies have shown that HVEM expression is up regulated in many types of cancers including colorectal cancers [13], melanomas [14], esophageal carcinomas [15], gastric cancers [16], hepatocarcinomas [17], breast cancers [18], lymphomas [19] or PCa [20]. In these studies, high levels of HVEM expression by tumors or soluble HVEM in the sera were associated with a worse prognosis and lower survival. Moreover, HVEM expression by tumors was also associated with a reduction in the numbers of tumor-infiltrating leukocytes (TILs) [13,15,17], indicating a detrimental role for HVEM in various cancers. Few studies have considered targeting the HVEM network to affect tumor growth. In fact, various strategies to inhibit HVEM expression or function lead to increased T cell proliferation and function in syngeneic tumor mouse models [15,21,22]. However, to our knowledge, no study to date has assessed the possibility to use a monoclonal antibody (mAb) to HVEM to favor the antitumor immune response in a humanized context in vivo.
However, PDX mice models are difficult to generate on a regular and consistent basis. Humanized mice grafted with human cancer cell lines have been used in hundreds of studies to evaluate various treatments efficacy for cancer immunotherapy [29] . Whatever the mode of reconstitution of immune cells (from progenitors or peripheral blood, as herein), they all suffer from the same drawback: human T cells will be allogenic to the tumor. This would prevent the tumor-antigen specific response to be faithfully recapitulated in humanized mice. Nevertheless, accumulating evidences indicate that, besides their different proportions in the native T cell repertoire, a high degree of similarity exists between the antigen-specific and allogenicspecific immune responses [30,31] . Furthermore, T cell response to the tumor is not restricted to antigenspecific T cells but can also imply bystander T cells, not specific but actively engaged in tumor control [32,33] . Thus, despite some limitations, humanized mice may still provide valuable clues on therapeutic strategies aimed at enhancing the allogenic immune response mediated by human T cells to the tumor, with relevance to T cell-mediated tumor-antigen specific and bystander immune responses in cancer patients.
Herein, we investigated the therapeutic potential of a monoclonal antibody targeting human HVEM in humanized mice and the underlying cellular and molecular mechanisms associated to therapeutic efficacy.

HVEM and BTLA are associated to lower progression-free intervals in prostate cancer patients
Analysis of the TCGA (The Cancer Genome Atlas) revealed that HVEM mRNA was expressed at high levels in the PRAD (PRostate ADenocarcinomas) dataset (normalized log2 counts >10) and at higher levels in primary tumors than in patient-matched normal tissues ( Figure 1A). In contrast, BTLA was expressed at low (normalized log2 counts <3) and similar levels in the tumor and in the patient-matched normal tissues ( Figure 1B). For patients with Gleason scores of 6 or 7 (low grade adenocarcinomas), above-median expression of HVEM was associated to a lower progression-free interval (PFI) over a 5-year period whereas this was not observed for BTLA ( Figure 1C). It was the opposite in patients with more advanced disease (Gleason scores of 8 or 9), in which HVEM mRNA expression levels were not correlated with good or bad prognosis, but where above-median expression of BTLA was associated to a lower PFI ( Figure 1D). Overall, above-median expression of HVEM and BTLA together was a bad PFI prognosis factor irrespective of the clinical score ( Figure 1E). Interestingly, this was not observed for the PD1/PD-L1 co-inhibitory signaling pathway ( Figure 1F). These results indicate a detrimental role of HVEM/BTLA in PCa at all stages of the disease and suggest that blocking the HVEM/BTLA pathway might be an interesting option for PCa patients.
Accordingly, we focused the rest of the study on PCa cell lines in vivo.

Targeting HVEM with a mAb improves tumor control in humanized mice
We first determined whether targeting HVEM with a mAb could impact tumor growth in vivo. For that, we implanted prostate cancer cell lines in NSG mice and grafted human PBMCs few days after. No differences in tumor growth were observed in mice grafted with the prostate cancer cell line DU145, which did not express HVEM (Figure 2A). In contrast, a two fold reduction of tumor growth was observed in mice grafted with the HVEM-positive PC3 prostate cancer cell line ( Figure 2B). The presence of human cells was mandatory for the efficacy of the mAb since tumor growth was unaffected by the treatment in nonhumanized NSG mice ( Figure S1A). The lack of an effect in non-humanized mice also indicates that the mAb did not kill directly the PC3 cell line in vivo. To formally demonstrate that HVEM on the tumor was required for therapeutic efficacy of the mAb, we generated an HVEM-deficient PC3 cell line (clone 1B11)

Tumor control is dependent on CD8 + T cells
To dig into the mode of action of the mAb, we determined the composition of human CD45 + cells in the PC3 tumor by flow cytometry. Human CD3 + T cells represented more than 95% of hCD45 + cells, in which CD4 + and CD8 + cells represented 40 to 60%, irrespective of the treatment ( Figure S2). In contrast, we observed an increase in CD8 T cell numbers (normalized across experiments) in the anti-HVEM-treated group relative to the control group ( Figure 3A). Additionally, frequencies of cells expressing the proliferation marker Ki-67 were significantly elevated in both CD4 + and CD8 + T cells (Figures 3B). However, frequencies of CD8 + T cells expressing both Granzyme B and Perforine-1 were not elevated ( Figure 3C). To directly determine the contribution of CD8 + T cells to tumor control in anti-HVEM-treated mice, we compared tumor growth in humanized mice depleted or not of human CD8 + T cells. Depletion of CD8 + T cells before the initiation of the treatment reverted the effect of the anti-HVEM mAb ( Figure 3D), showing that tumor control was dependent on CD8 + T cells.

Treatment with the anti-HVEM mAb does not increase Graft-Vs-Host-Disease nor number or proliferation of human T cells
Our observations so far are consistent with the hypothesis that targeting HVEM on the tumor released an inhibitory signal on CD8 + T cells, allowing their proliferation in situ. We wanted to rule out the possibility that the mAb behave as an agonist, directly activating HVEM + CD8 + T cells in vivo, leading to better tumor control. One prediction of that hypothesis would be that GVHD occurring in NSG mice grafted with the tumor and reconstituted with human T cells would be exacerbated following anti-HVEM administration.
However, despite the impact of the mAb on tumor growth (Figure 2), we observed similar weight loss and mortality in anti-HVEM or isotype control treated mice ( Figure 4A-B). Furthermore, and in contrast to TILs, the activation of human T cells in the spleens were the same in both groups, as judged by similar numbers of CD4 + and CD8 + T cells and similar frequencies of Ki-67 + cells ( Figure 4C-D). Thus, anti-HVEM therapy in humanized mice reduced the growth of HVEM + tumors by a mechanism that was independent of a systemic agonist effect of the mAb on human T cells.  Figure 5B). In addition, up-regulated genes of the anti-HVEM group were enriched in members of several ontologies related to lymphocyte activation, including the tumor necrosis factormediated signaling pathway, cytokine and chemokine binding and activity, and T cell receptor complex ( Figure S3). On the other hand, some genes belonging to immuno-suppressive pathways were clearly downregulated in HVEM-treated TILs such as ENTPD1 (CD39), IL10 and the co-inhibitory receptors BTLA, TIGIT, LAG3 and HAVCR2 (TIM3), as well as the "don't eat me" receptor CD47 ( Figure 5A). In addition, GSEA showed that the "immunoregulatory interactions between a lymphoid and a non lymphoid cell" signature was significantly repressed in the DEG signature ( Figure 5C), signing altered adhesion and motility of TILs. Overall, anti-HVEM treatment was associated with profound modifications of TILs, with an increased expression of genes belonging to activation and proliferation signaling pathways and a decreased expression of genes signing an exhausted phenotype.

HVEM is an immune checkpoint during anti-tumor T cell immune response in humanized mice
Finally, and to directly demonstrate that HVEM expression by the tumor was indeed an immune checkpoint, we devised a simple in vivo assay. We implanted the HVEM-positive or the HVEM-negative PC3 cells in NSG mice, and compared tumor growth with or without human PBMCs ( Figure 6). Both cell lines grew equally well in non-humanized NSG mice ( Figure 6A), showing that HVEM-deficiency did not impact in vivo tumor development per se. In contrast, tumor growth of the 1B11 clone was reduced two-fold compared to the parental PC3 cell line in humanized mice ( Figure 6B). Thus, removing HVEM from the tumor released a brake on the allogenic T cell response to the tumor, demonstrating that HVEM was an immune checkpoint in this experimental condition.

Discussion
Here, we report for the first time that HVEM can be targeted by a mAb to improve tumor control by human T cells in vivo. Moreover, we deciphered the mode of action of the mAb in vivo using complementary technologies. Furthermore, we developed a simple in vivo assay for immune checkpoint discovery and validation. To our knowledge, the present report is the first that combine CRISPR/Cas9-mediated deletion of putative checkpoints with assessment of tumor growth in humanized mice. One limitation of the assay is that PBMC-humanized mice are mostly reconstituted with T cells, as shown herein, limiting the usefulness of the assay to T cell-specific immune checkpoints. Another limitation is the allogenic nature of the immune response to the tumor in humanized mice, that would be only circumvented in PDX models with autologous human immune cells, an ongoing effort but still difficult to set up [34] . Despite these limitatons, we believe that our simple in vivo assay will be of great help to investigate other candidates in more advanced models of humanized mice, i.e mice reconstituted with human hematopoietic progenitors.
We show that the HVEM/BTLA checkpoint could be exploited for therapy in humanized mice using a mAb to human HVEM. We found that HVEM expression by the tumor was necessary and sufficient to elicit tumor control by the mAb, since it had no effect on HVEM-negative cell lines and no agonist activity on human T cells. Park et al. showed in a syngeneic mouse model that transfecting an agonist scFv anti-HVEM in tumor cells resulted in increased T-cell proliferation, as well as improved IFN-γ and IL-2 production and better tumor control [21]. Aside the species differences, the discrepancy with our results could be explained by the fact that T-cells are strongly activated in huPBMC mice [35]. The down regulation of HVEM expression upon activation [36] may have limited the binding of the anti-HVEM antibody on T-cells in our model. Thus, it remains possible that the mAb would behave differently in humans. On the other hand, BTLA is up regulated upon T-cell activation [37], increasing the susceptibility of T-cells to inhibition by HVEM + tumor cells [13,15,17,38]. We observed quite the opposite in the tumor micro environment following treatment, with an increase in HVEM and a reduction of BTLA gene expression, with a concomitant increase in LTA and LIGHT, two other ligands for HVEM. It is important to note that the binding sites of LIGHT and BTLA differ on HVEM [39]. So, the anti-HVEM mAb might have limited inhibition of activated T-cells through blockade of HVEM binding with BTLA but not with the other ligands that are T-cell activators. An alternative possibility would be that LIGHT and LTA in their soluble forms inhibit the interaction of HVEM with BTLA [40]. As of today, reciprocal regulation of HVEM and BTLA has not been reported but our observation is reminiscent of earlier findings showing reciprocal regulation of HVEM by LIGHT [36].
Previous studies in mice also showed that inhibiting HVEM expression on the tumor or its interaction with its ligands has a positive effect on T cells. Injection of a plasmid encoding a soluble form of BTLA (to compete with endogenous BTLA for HVEM) was associated with an increase in inflammatory cytokines production by TILs and a decrease in anti-inflammatory cytokines at the RNA level [22]. In the same line, vaccination to a tumor-associated antigen was more efficient if HVEM interactions with its ligands were blocked by HSV-1 gD, allowing regression of large tumor mass [41]. Moreover, silencing HVEM expression in the tumor with siRNA was also associated with an increase in CD8 T cells and inflammatory cytokine production in a murine colon carcinoma model [15]. In addition, use of siRNA to HVEM on ovarian cancer in vitro promoted T-cells proliferation and TNF-α and IFN-γ production [42].
Numerous results from our study also support increased T cell activation in the absence of HVEM/BTLA signaling: TILs from mice treated with anti-HVEM were enriched in ontologies signing activation by cytokines, chemokines and signaling pathways that are well known inducers of proliferation, differentiation, migration and apoptosis. Comparison between TILs from mice treated with the anti-HVEM or isotype control mAb also highlighted decreased expression of many co-inhibitory receptors genes (BTLA, TIGIT, LAG3, and HAVCR2 [43,44]) or with immunosuppressive functions (ENTPD1 and IL10), suggesting a lower exhaustion status. Overall, we propose a model in which treatment with the anti-HVEM mAb would block the BTLA-associated inhibitory signaling on CD8 + TILs that would increased their proliferation and numbers, reduced their exhausted phenotype and improved their migration and/or adhesion to the tumor, ultimately leading to better tumor control.

Conclusions
The recent success of ICI for cancer immunotherapy (anti-CTLA-4, anti-PD-1/PD-L1) has confirmed the hypothesis that the immune system can control many cancers but disappointing results were obtained for PCa [4] , in line with our observation that PD1/PD-L1 are not associated to lower PFI in PCa patients. In light of the promising results reported herein, anti-HVEM therapy might be combined with ICI and/or chemotherapy to further enhance anti-tumor immunity in PCa.

Preparation of human peripheral mononuclear cells
Human peripheral blood was obtained from Etablissement Francais du Sang (EFS) after informed consent of the donor. Human peripheral blood mononuclear cells (PBMC) were isolated on a density gradient (Biocoll).
Cells were washed in PBS 3% FCS and diluted at the appropriate concentration in 1X PBS before injection into mice.

Humanized mice tumor model
All animals used were NSG mice (stock ≠005557) purchased from the Jackson Laboratory (USA). To assess therapeutic activity, 8-20-week-old NSG mice (males and females) were injected subcutaneously with 2.10 6 tumor cells. One week later, mice were irradiated (2 Gy) and grafted the same day with 2.10 6 huPBMC by retro orbital injection. Four to 5 days after transplantation, the anti-huHVEM antibody or isotype control was injected intra-peritoneally at 2 mg/kg. General state, body weight and survival of mice were monitored every 3-4 days to evaluate Graft-vs-Host-Disease (GVHD) progression. Mice were euthanized when exhibiting signs of GVHD, such as hunched back, ruffled fur, and reduced mobility. For CD8 depletion, mice were injected intra-peritoneally with 10mg/kg of the anti-CD8 MT807R1( Rhesus recombinant IgG1 provided by the Nonhuman Primate Reagent Resource [45]) or the isotype control (clone DSPR1) the day following humanization.

Antibodies
The clone 18.10 has been described previously [46]. Briefly, 18.10 is a murine IgG1 anti-human HVEM mAb and was produced as ascites and purified by protein A binding and elution with the Affi-gel Protein A MAPS II Kit (Bio-rad). A mouse IgG1 isotype control (clone MOPC-21 clone) was purchased from Bio X Cell (West Lebanon, NH, USA).

Cell lines
PC3 (non-hormonal-dependent human prostate cancer cells) and DU145 (prostate cancer cells) were grown in high glucose DMEM media supplemented with 10% FCS, L-glutamine and antibiotics (Penicillin/Streptomycin). The PC3 cell line was genetically authenticated before the initiation of the experiments (Eurofins). All cells were confirmed to be free of mycoplasmas before injection into mice by the MycoAlert detection kit (Lonza). Tumor growth was monitored using an electronic caliper and volumes were determined using the following formula: [(length*width²)/2].

NanoString nCounter expression assay
For Nanostring® experiment, 14 to 15 weeks-old NSG mice were humanized and treated with anti-HVEM or isotype. Day 28 post humanization, tumors were harvested and TILs were isolated as described above. To maximize mRNA recovery, TILs were pooled by treatment groups (4 mice in the anti-HVEM group and 5 in the isotype control group). Then, cells were stained with viability dye (eF506) and anti hCD45-APC (HI30, Biolegend). Live hCD45 + cells were sorted using Aria II cell sorter. After centrifugation, cells were suspended in RLT buffer (Qiagen) before freezing at -80°C until analysis. Data were normalized through the use of NanoString's intrinsic negative and positive controls according to the normalization approach of the nSolver analysis software (Nanostring).

Bioinformatics analysis
For ontologies enrichment analysis, only genes up regulated by the treatment were analyzed using the enrichment analysis visualization Appyter to visualize Enrichr results [47] . DEG (up-an down-regulated) were ranked by fold-change for pre-ranked GSEA. Enrichment was performed with the C2 Canonical

Statistical analysis
All statistical tests were performed with Prism v8 (Graph Pad Inc, La Jolla, CA, USA) or JASP v0.14.3 (available at https://jasp-stats.org). The nature of the statistical test used to compare results is indicated in each legend of the figures. When necessary, the p-values of these tests are indicated on the figure panels.
Statistical power of the analyses (alpha) was arbitrarily set at 0.05. No test was performed a priori to adequate the number of samples with statistical power. Figure S1: Anti-HVEM therapy in humanized mice.

Ethics approval and consent to participate
Human peripheral blood mononuclear cells were collected by Etablissement Français du Sang from healthy adult volunteers after informed consent in accordance with the Declaration of Helsinki. Mice were bred in our animal facility under specific pathogen-free conditions in accordance with current European legislation.
All protocols were approved by the Ethics Committee for Animal Experimentation Charles Darwin (Ce5/2012/025).

Consent for publication: All authors concur with the submission of the article in its present form
Competing interests: DO declares competing interests as being the co-founder and shareholder of Imcheck Therapeutics, Alderaan Biotechnology and Emergence Therapeutics and has research funds from Imcheck Therapeutics, Alderaan Biotechnology, Cellectis and Emergence Therapeutics.

Authors' contributions
SB and NA performed the experiments, analyzed the data and contributed to the writing of the manuscript, DO provided essential reagents and edited the manuscript, GM designed the study, analyzed the data and wrote the manuscript.