Fcγ receptor–mediated cross-linking codefines the immunostimulatory activity of anti-human CD96 antibodies

New strategies that augment T cell responses are required to broaden the therapeutic arsenal against cancer. CD96, TIGIT, and CD226 are receptors that bind to a communal ligand, CD155, and transduce either inhibitory or activating signals. The function of TIGIT and CD226 is established, whereas the role of CD96 remains ambiguous. Using a panel of engineered antibodies, we discovered that the T cell stimulatory activity of anti-CD96 antibodies requires antibody cross-linking and is potentiated by Fcγ receptors. Thus, soluble “Fc silent” anti-CD96 antibodies failed to stimulate human T cells, whereas the same antibodies were stimulatory after coating onto plastic surfaces. Remarkably, the activity of soluble anti-CD96 antibodies was reinstated by engineering the Fc domain to a human IgG1 isotype, and it was dependent on antibody trans-cross-linking by FcγRI. In contrast, neither human IgG2 nor variants with increased Fcγ receptor IIB binding possessed stimulatory activity. Anti-CD96 antibodies acted directly on T cells and augmented gene expression networks associated with T cell activation, leading to proliferation, cytokine secretion, and resistance to Treg suppression. Furthermore, CD96 expression correlated with survival in HPV+ head and neck squamous cell carcinoma, and its cross-linking activated tumor-infiltrating T cells, thus highlighting the potential of anti-CD96 antibodies in cancer immunotherapy.


Introduction
The clinical success of agents targeting immune checkpoint receptors such as CTLA-4 and PD-1 has demonstrated that the immune system is a bona fide and key therapeutic target for the treatment of cancer. Despite the unprecedented durable anti-tumor responses seen in a subset of patients, the majority of patients fail to respond to these treatments or develop resistance after the initial response (1). This has galvanised the search of additional immune checkpoint receptors that could be targeted to extend the benefit of immunotherapy to the wider population (2). One such receptor that has recently received attention is CD96, also known as T-cell activated late expression (TACTILE). CD96 is a type I transmembrane protein comprising an extracellular region that consists of three immunoglobulin superfamily (IgSF) domains followed by an Oglycosylated stalk region (3,4). The cytoplasmic domain of CD96 contains a conserved short basic/proline-rich motif, which typically associates with SH3-domain containing proteins, followed by a single Immunoreceptor Tyrosine-based Inhibitory Motif (ITIM). In addition, a YXXM motif similar to that found in CD28 and ICOS is present in human but not mouse CD96.

Expression of CD96 is limited to immune cells, primarily T cells, NK cells and NKT cells and is
upregulated following T-cell activation (3,5). Two isoforms of CD96 that differ in the sequence of the second IgSF domain exist as a result of alternative splicing with the shorter isoform (CD96v2) being the predominant form expressed in human primary cells (6). CD96 shares an ability to bind proteins from the nectin and nectin-like family with two other IgSF receptors, namely T-cell immunoreceptor with Ig and ITIM domains (TIGIT) and CD226 (DNAX accessory molecule 1, DNAM-1). While TIGIT and CD226 bind to CD155 (necl-5) and CD112 (nectin-2), CD155 is the only known ligand for CD96 in humans (7). CD155 is weakly expressed on a variety of cells, including immune, epithelial and endothelial cells, and is upregulated on cancer cells (8,9). TIGIT and CD226 function as inhibitory and activating receptors, respectively, while both inhibitory and stimulatory functions have been ascribed to CD96. Initial studies demonstrated that engagement of CD96 stimulates human NK cell-mediated lysis of P815 cells in redirected killing assays, albeit less efficiently than CD226 (10,11).
Furthermore, unlike CD226, CD96 was dispensable for killing of CD155-expressing tumor cells, suggesting that the stimulatory effect of CD226 is dominant (12,13). In contrast, studies in mice showed that CD96 deficiency results in an exaggerated NK cell-mediated IFNγ production and resistance to carcinogenesis and experimental lung metastases (14), indicating that CD96 functions as an inhibitory receptor in murine NK cells. Additional studies employing anti-CD96 antibodies provided further support for targeting this pathway as a strategy to treat cancer (14,15), however the findings were confounded by the observation that anti-CD96 antibodies need not block the CD155-CD96 interaction to exert their anti-metastatic effect (16). More recently Chiang et al (17) showed that genetic ablation or antibody blockade of CD96 rendered murine CD8 T cells less responsive and conversely anti-CD96 antibody presented on microbeads promoted T-cell proliferation. Antibodies have the capacity to induce receptor clustering dependent on co-engagement of Fcγ receptors (FcγR) and this property has been exploited for the development of agonistic immunostimulatory antibodies that target costimulatory TNF receptor superfamily members (18)(19)(20).
Here we have addressed whether Fcγ receptor crosslinking potentiates the activity of antihuman CD96 antibodies. Through Fc domain engineering, we have identified the human IgG1 isotype as a key determinant that co-defines the activity of anti-CD96 antibodies. We show that anti-CD96 antibodies costimulate the proliferation of human peripheral CD4+ and CD8+ T cells and enhance cytokine production in an isotype and FcγRI-dependent manner. Costimulation by anti-CD96 antibodies was effective in countering suppression by regulatory T cells and in inducing the proliferation of tumor-infiltrating T cells. RNAseq analysis following CD96 costimulation revealed upregulation of multiple gene networks associated with T-cell proliferation and effector function. These results inform the design of immunostimulatory anti-CD96 antibodies for the reinvigoration of anti-cancer T cells.

Immobilized and Fcγ γ γ γR-crosslinked anti-CD96 antibodies promote human T-cell proliferation
We evaluated three different anti-CD96 mAbs that either fully (19-134 and 4-31) or partially  inhibit the CD155-CD96 interaction (Table I Figure 2, C and D). As shown in Figure 1A and 1B, clone 19-134 did not significantly alter the proportion of dividing T cells. Similar results were obtained using two additional anti-CD96 mAbs (clones 19-14 and 4-31; Figure 1, A, C and D). Taken together, these data demonstrate that CD96 blockade does not confer a proliferative advantage to anti-CD3 stimulated T cells. Next we tested whether the activity of anti-CD96 mAbs could be potentiated through antibody immobilization on tissue culture plates, an experimental strategy used for inducing antibody-mediated receptor crosslinking. In contrast with the findings using soluble antibodies, plate-bound anti-CD96 mAbs were able to costimulate the proliferation of CD4+ and CD8+ T cells (Figure 2, A and B). We also tested if blocking the CD155-CD96 interaction with an anti-CD155 mAb can affect T-cell proliferation. As shown in Figure 2C, the addition of a blocking anti-CD155 mAb failed to enhance T-cell proliferation and did not affect the increase in cell proliferation afforded by plate-bound anti-CD96 mAb. As CD96 is expressed by T cells and NK cells in resting human PBMCs (3, 10), we examined whether anti-CD96 mAbs could costimulate purified CD3+ T cells. As shown in Figure 2D, immobilized anti-CD96 mAb significantly boosted the proliferation of isolated CD4+ and CD8+ T cells, demonstrating that  (27). We evaluated two anti-CD96 clones (19-134 and 19-14) in the IgG1 V12 format, but neither mAb was active ( FcγRs coated onto plastic, together with highly purified CFSE-labelled CD3+ T cells, and showed that FcγRI was uniquely able to restore the activity of soluble anti-CD96 human IgG1 ( Figure 4B).
Collectively, our data demonstrate that soluble anti-CD96 mAbs of the IgG1 subclass enhance the proliferation of CD4+ and CD8+ T cells, dependent on mAb crosslinking through Fc domain trans-interaction with FcγRI.

Agonistic anti-CD96 mAb counters suppression by Tregs
Tregs exert a dominant role in maintaining self-tolerance and suppressing anti-tumor Tcell responses (29), but the role of CD96 on Treg is currently unknown. Flow cytometric analysis revealed that peripheral blood Tregs expressed CD96 similarly to conventional CD4+ and CD8+ T cells ( Figure 5A). To assess if the presence of increasing numbers of Tregs would negate the costimulatory effect of anti-CD96 mAbs, highly purified, CFSE-labelled CD3+ CD25-CD127+ (98.1±0.5%) conventional/effector T cells (Tconv) were stimulated with anti-CD3 and either anti-CD96 or an isotype-matched control antibody. In some cultures, purified unlabelled CD4+ CD25+ CD127-Treg (93.6±1.8% purity) were added to obtain a Tconv:Treg of 2:1 or 3:1. Tconv proliferation and activation were determined by measurement of CFSE dilution and upregulation of CD25, respectively, after four days. As expected, the addition of purified Tregs suppressed the proliferation of CD4+ and CD8+ Tconv and reduced expression of CD25 ( Figure 5, B-E).
However, when anti-CD96 mAb was present, both Tconv proliferation and CD25 expression were restored to levels seen in the absence of Tregs ( Figure 5, B-E). These data support the notion that costimulation of Tconv by anti-CD96 mAb overcomes to a large extent the suppression exerted by Tregs.

Gene expression profiling reveals augmentation of multiple T-cell activation pathways by CD96
To gain further insights into the downstream events triggered by anti-CD96 mAbs, we were also enriched in the anti-CD96 treatment group. Consistently, the hallmark of the unfolded protein response, which is known to contribute to the regulation of T-cell proliferation and effector function (31), was significantly upregulated following anti-CD96 treatment ( Figure 6C).

Quantification of cytokine production in the supernatant of T cells stimulated for 6 or 22 hours
showed that anti-CD96 significantly upregulated IL-2 production by CD3+ T cells at both time points, while IFN-γ production was augmented at 6 hours ( Figure 6, D and E). Hence, increased gene transcription correlated with elevated protein levels for IL-2 and IFN-γ. Moreover, we showed that agonist anti-CD96 mAb provided direct costimulation to CD4+ and CD8+ isolated T cells, resulting in enhanced IL-2 production from each of these cell types in addition to promoting independent signals for CD4+ and CD8+ T-cell proliferation (supplemental Figure 4).
Furthermore, IPA identified a broad range of upstream regulators predicted to be activated and a smaller number of regulators predicted to be inhibited by CD96 stimulation (supplemental Figure 5, A and B). TCR, CD3 and CD28 were highlighted as potential positive upstream regulators of the gene signature induced by anti-CD96 mAb, suggesting that CD96 engagement elicits signaling pathways that overlap and strengthen those emanating from the engagement of the TCR and CD28 (supplemental Figure 5A). In agreement with this, transcription factors and signaling kinases triggered by the integrated response to TCR and CD28 engagement, such as Myc, Jun, NFκB, Mek/MAP2K1/2, PI3K/Akt and p38 MAPK, were additionally identified as upstream activating regulators (supplemental Figure 5A).
Collectively our transcriptomic data indicated that CD96 engagement triggers multiple signaling pathways associated with increased T-cell proliferation and effector function and identified several candidate molecules that could mediate signaling downstream of CD96.

Agonist anti-CD96 mAb augments the proliferation of tumor-infiltrating T cells
Given that anti-CD96 mAbs were able to costimulate peripheral blood T cells, we asked if this approach could also promote the proliferation of tumor-infiltrating T cells (TIL), which are known to exist in various dysfunctional states (32). Using publicly available data from the Cancer Genome Atlas (TGCA) database through the Tumor Immune Estimation Resource (33) Next, we used flow cytometry to examine CD96 expression on T-cell subsets isolated from fresh HNSCC tumor biopsies (patient characteristics are included in Table III). CD96 was expressed on CD8+ T cells, CD4+ Foxp3-Tconv and CD4+ Foxp3+ Tregs ( Figure 7B). Although ranging widely between patients, on average expression of CD96 on CD8+ T cells was higher than that seen on the other T-cell subsets analyzed ( Figure 7B). Furthermore, we evaluated if CD96 is co-expressed with the inhibitory receptor PD-1, typically found on chronically stimulated and/or exhausted tumor-infiltrating CD8+ T cells (36). Figure 7C shows that PD-1 expression on CD8+ T cells from HNSCC tumors varied amongst patients and expression of CD96 could be detected on a significant proportion of the PD-1 bright and PD-1 dim T cells ( Figure 7C).
To test whether anti-CD96 mAbs are capable of costimulating tumor-infiltrating T cells, we isolated lymphocytes from HPV+ HNSCC tumors and measured T-cell proliferation in response to plate-bound anti-CD3 and anti-CD96. On average, the percentages of tumoral CD8+ T cells, CD4+ Tconv and CD4+ Tregs out of the CD3+ T cells were 35.6 ± 5.2, 42.7 ± 5.9 and 15.9 ± 2.2, respectively. The data presented in Figure 7D show that tumor-infiltrating T cells proliferated more extensively when cultured with anti-CD3 and anti-CD96 mAb compared to incubation with anti-CD3 and a control mAb, highlighting CD96 as a potential target to reinvigorate anti-cancer T cells.

Discussion
Despite the success of targeting the PD-1/PD-L1 inhibitory axes, there remains a strong incentive to discover additional immunomodulatory targets driven primarily by the need to extend the response rate and durability offered by current treatments. Herein we provide data to suggest that mAbs targeting human CD96, a member of IgSF, expressed at low levels on naive T cells, but strongly upregulated during T-cell activation, are potent stimulators of T-cell activation and proliferation. Although earlier studies, which primarily focused on murine NK cell responses, suggested that CD96 could function as an inhibitory receptor (14,15), our data using human T cells do not support this notion. Instead, we provide evidence that CD96 is a bona fide costimulatory receptor for human T cells. First, we showed that soluble 'Fc silent' mAbs that block the interaction of CD96 with its ligand CD155 did not exert functional effects (Figure 1), whereas the same mAbs were stimulatory when coated on tissue culture plastic ( Figure 2).
Second, the conversion of 'Fc silent' anti-CD96 mAbs to Fc competent mAbs of the IgG1 subclass endowed them with the capacity to costimulate T cells without the need for coating ( Figure 3). Third, we demonstrated that the T cell costimulatory effects of soluble anti-CD96 IgG1 are critically dependent on crosslinking mediated through trans-binding to FcγRI (Figures 3   and 4). We interpret these results as evidence that immobilization of anti-CD96 mAbs either by coating on synthetic surfaces, or more physiologically through co-engagement of FcγRI, results in CD96 clustering on the T-cell surface, which subsequently leads to stimulation of intracellular signaling. Our findings are consistent with a recent study demonstrating that coupling of anti-CD96 mAbs to beads provided a costimulatory signal to T cells (17). Our data extend previous findings by demonstrating the importance of the antibody Fc domain in driving the functional activity of anti-CD96 mAbs. These findings should therefore guide future development of agonist anti-CD96 mAb aimed towards enhancing sub-optimal anti-tumor responses. In this context it is well known that anti-tumor T-cell responses are hindered by Tregs and therefore our data showing that anti-CD96 mAb was highly effective in overcoming suppression by Tregs is noteworthy ( Figure 5). Therefore, we anticipate that anti-CD96 mAbs remain capable of augmenting Tconv responses in spite of the presence of increasing numbers of Tregs within the tumor microenvironment.
Mechanistically CD96 costimulation could lessen Treg-mediated suppression in a number of ways. First, by augmenting IL-2 secretion ( Figure 6 and supplemental Figure 4) and the expression of CD25 on CD4+ and CD8+ Tconv (Figure 5), the ability of Tregs to deprive responder T cells of IL-2 (29) is likely to be reduced, thus increasing the bioavailability of IL-2 to Tconv. Second, our transcriptomic data and pathway analysis suggested convergence of CD96 signaling pathways with those downstream of CD3 and CD28 ( Figure 6 and supplemental Figure   5). This is predicted to reduce the dependency of Tconv on costimulation via CD80/86-CD28 and therefore could circumvent Treg-mediated suppression exerted by CTLA-4 expressing Tregs (29). Third, our transcriptomic analysis also showed that CD96 costimulation upregulated several costimulatory receptors and ligands, including OX40, GITR, 4-1BB, CD40 ligand and CD226, which could further lower the activation threshold of Tconv and impede Treg suppression.
Although our data offer plausible mechanisms of how Tconv resist suppression, an alternative hypothesis might be that anti-CD96 antibodies modulate Tregs directly as these cells also express CD96, a possibility that will be examined in future studies.
From the perspective of developing new anti-cancer immunotherapies, the finding that CD96 costimulation is able to augment the proliferation of intratumoral T cells from HPV+ HNSCC is particularly encouraging. A recent study showed that intratumoral HPV-specific PD-1+ CD8+ T cells can be distinguished by expression of TCF-1 and TIM-3, markers that are used to identify stem cell-like and terminally differentiated T cells, respectively (36). Interestingly, the authors of that study demonstrated that it is the stem cell-like CD8+ T cells that proliferate extensively upon in vitro stimulation with the cognate HPV peptide (36). Herein we showed that fraction of PD-1 bright as well as on PD-1 dim T cells. Therefore, it would be interesting to dissect the role of CD96 further by examining how CD96 costimulation impacts on different HPV-specific CD8+ T-cell subsets. Such studies will inform of more effective strategies to reinvigorate anti-cancer T cells in patients.             Tables   Table I.