The hMet and S45Y ß-catenin Sleeping Beauty transposon system induces multi-focal immune suppressive HCC. Concomitant expression of activated hMet and ß-catenin leads to multi-focal liver tumors in murine models following hydrodynamic tail vein injection33, thereby mimicking an expression signature seen in a proportion of HCC patients33, 37, 38. We confirmed that the hMet + S45Y ß-catenin/Sleeping Beauty transposon plasmid system consistently induced multi-focal tumors in the livers of C57Bl/6 mice (Figs. 1A-B). Histologic signs of liver tumor development were observed as early as 4 weeks following hydrodynamic tail vein injection, initially with well-differentiated morphology but minimal immune infiltration. By 7 weeks post injection, livers were densely population by tumors, with large pseudo-cysts, fat deposits, and signs of extramedullary hematopoiesis and shortly thereafter inevitably reached a terminal endpoint (Figs. 1A-B).
Whilst CD8+ T cell infiltration into livers was 10-fold higher in tumor-bearing mice than in non-tumor bearing control animals at both days 22 and 29 post hydrodynamic injection, CD4+ T cell infiltration trended towards increased levels but did not reach significance (Figs. 1C-D). The majority of liver infiltrating CD8+ and CD4+ T cells in non-tumor bearing animals had minimal expression of either Tim-3 or PD-1, although there was a small but detectable population of Tim-3+/PD-1+ CD4+ T cells (Figs. 1E-F). In contrast, the proportion of Tim-3+/PD-1+ CD8+ T cells in tumor bearing livers at both days 22 and 29 was significantly increased (≥ 50% of the total CD8+ T cell population) compared to the control non-tumor bearing mice, suggesting that presence of tumor induced a profound state of CD8+ T cell dysfunction/exhaustion (Fig. 1E). This change was not observed in the liver infiltrating CD4+ T cell population, where trends towards increased proportions of Tim-3+/PD-1+ CD4+ T cells did not reach significance compared to non-tumor bearing mice (Fig. 1F). Consistent with development of an immune suppressive TME within hMet + S45Y ß-catenin liver (SB-HCC) tumors, expression of PD-L1 on liver infiltrating CD11c+/MHCII+ dendritic cells was significantly increased in tumor bearing compared to non-tumor bearing mice (Fig. 1G). Finally, macroscopic visualization of tumor bearing livers showed high levels of PD-L1 expression associated with the multi-focal tumor lesions which increased progressively with tumor development from 4 through 7 weeks post hydrodynamic injection (Fig. 1H). Taken together, these data indicate that SB-HCC tumors develop a highly immune suppressive TME within 20–30 days of hydrodynamic plasmid injection, thereby re-capitulating this aspect of human HCC in which a highly immunosuppressive TME39 is associated with exhausted CD8+ T cells with reduced anti-tumor cytotoxicity40, 41, 42.
SB-HCC is responsive to CD8+ T cell-mediated checkpoint blockade with anti-PD-L1. Consistent with SB-HCC liver tumors developing a highly immune suppressive, PD-L1 dependent TME analogous to the human disease, treatment with anti-PD-LI ICI led to cures of up to 50% of treated mice by day 150 post hydrodynamic injection (Fig. 2A). Therapy with anti-PD-L1 ICI was completely dependent upon CD8+ T cells (Fig. 2A) but not on CD4+ T cells (Fig. 2B). Over two separate experiments, depletion of NK cells showed a trend towards increased efficacy of anti-PD-L1 therapy compared to non-depleted mice, although this did not reach significance (Fig. 2C). While not definitive, this may suggest that NK-mediated immune suppression plays a role in HCC development in this model.
Combination VSV-IFNß oncolytic virotherapy with anti-PD-L1 ICI abolishes the therapy of ICI alone. Our initial hypothesis was that early IT treatment of HCC with VSV-IFNß would convert the immune-suppressive TME into a pro-inflammatory environment. In turn, this would liberate HCC tumor associated antigens (HCCTAA) leading to the priming of anti-HCCTAA T cells upon which immune checkpoint therapy could work. Contrary to this hypothesis, we observed that early IT treatment with VSV-IFNß (day 21–25 for three doses) prior to anti-PD-L1 ICI (day 28–39 for 6 doses) completely abolished the survival benefit generated by ICI alone and was no better than either virus alone or control treatment alone (Fig. 2D). When this sequencing was reversed and IT virus was administered later (day 40–44) than anti-PD-L1 ICI (day 21–33), the virus still inhibited the effects of anti-PD-L1 therapy alone (Fig. 2E), although there was a (non-significant) trend to an improvement of the combination (virus + ICI) therapy compared to the virus alone. Similar inhibition of the therapeutic effects of anti-PD-L1 therapy by IT virus were observed with other schedules of administration including when virus and ICI overlapped (Supplemental Fig. 1). Taken together, these data show that the addition of a highly immunogenic oncolytic virus to a weak, anti-PD-L1-sensitive anti-tumor immune response acted consistently to inhibit the therapeutic anti-HCCTAA immune response.
An immuno-dominant anti-viral rapid effector CD8+ T cell population replaces sub-dominant putative anti-HCCTAA T cells. We observed that, when IT virus was given early either before, or co-incidentally, with anti-PD-L1 ICI, the therapeutic effects of ICI therapy alone were lost (Fig. 2D, Supplemental Fig. 1), with the survival curves beginning to separate about 10 days after the last treatment with anti-PD-L1. Therefore, we used Cytometry by Time of Flight (CyTOF) analysis using t-distributed Stochastic Neighbor Embedding (tSNE) with Rphenograph, 22 populations to analyze tumor infiltrating lymphocyte populations (TIL) (Fig. 3A). Using differential immune marker expression, we defined a total of 9 distinct immune populations and reanalyzed through tSNE with FlowSOM, 9 populations (Fig. 3B). When comparing TIL populations between naïve and HCC tumor bearing mice, the presence of developing HCC tumor in the liver induced expansion of a CD8+ T cell population with markers of terminal effector cells, as defined by high Tim-3 (CD366) and PD-1 expression (Fig. 3A), at day 38 post hydro-dynamic injection (Fig. 3C-D). This population, which we identify as ‘Exhausted CD8 T Cells’, correlates to cluster 17 in our 22-population tSNE analysis (Fig. 3A). This tumor-expanded population also expressed high levels of Lymphocyte Activation Gene-3 (LAG3, CD223), T-cell immunoglobulin and mucin domain-3 (TIM3), T cell immunoreceptor with Ig and ITIM domains (TIGIT), programmed cell death protein 1 (PD-1) and CD39 which are also associated with terminal effector and exhausted CD8+ T cells43, 44, 45. These data validate the use of SB-HCC as a model for human HCC, as the TME of HCC is enriched with exhausted, PD-1 expressing CD8+ T cells that represent the main subset of TIL and display anti-tumor cytotoxic activity in HCC46, 47, 48.
Treatment of HCC-bearing mice with ICI alone (days 21–34) induced detectable changes in the landscape of CD8+ T cell populations, with a particularly significant further proportional increase in the ‘Exhausted CD8 T Cells’ (Cluster 17) population which is characterized by activation/inhibitory markers (TIM-3, TIGIT, LAG-3, CD39, and PD-1) (Fig. 3E). We hypothesize this population to be anti-HCCTAA CD8+ T cells upon which ICI therapy was operative.
In contrast, treatment of HCC-bearing mice with ICI (days 21–34) and VSV-IFNß early (days 21–25) - under which conditions the therapeutic effects of anti-PD-L1 ICI therapy were abolished by addition of virus (Supplemental Fig. 1) – induced a predominant population of putative viral specific CD8+ T cells (Clusters 18, 19, 20, 22, Fig. 3A; ‘Anti-viral CD8 T Cells,’ Fig. 3B&F) which proportionately downregulated most of all the other CD8+ T cell populations, including the putative anti-HCCTAA CD8+ T cells (cluster 17, Fig. 3A; ‘Exhausted CD8 T Cells,’ Fig. 3B&F). This population was characterized by relatively high levels of expression of Granzyme B, interleukin 7 receptor (IL-7R, CD127), leukocyte antigen B associated transcript 3 (Bat3), and PD-L1 as well as relatively low levels of CD69 compared to the terminally differentiated putative anti-tumor CD8+ T cell population – an overall profile consistent with the presence of anti-viral effector CD8+ T cells.
Taken together, these data strongly argue that the abolition of therapeutic benefit achieved with ICI alone by addition of virus to ICI therapy was due to a rapid induction/expansion of anti-viral effector CD8+ T cells at the expense of anti-tumor, PD-1 expressing terminal effector CD8+ T cells. Moreover, the repolarization of the effector response of CD8+ T cells away from anti-HCCTAA CD8+ T cells towards a population of anti-viral effector CD8+ T cells would allow the concomitantly administered anti-PD-L1 ICI to focus upon the anti-viral, rather than anti-HCCTAA CD8+ T cells – thereby re-invigorating the anti-viral, rather than anti-tumor, T cell response.
Directing the anti-viral response to become an anti-tumor response. Our previous studies have shown that inclusion of a tumor antigen within VSV induces CD8+ T cell dependent anti-tumor therapy directed specifically against the virally encoded antigen as a result of the potent immune stimulating adjuvant properties of infection with VSV19, 34, 35, 36. Therefore, we hypothesized that it would be possible to amalgamate the potent anti-viral response (Fig. 3) with an anti-tumor T cell response by expressing immunologically relevant tumor associated antigens from within the virus. To test this in the hMet + S45Yß-catenin / Sleeping Beauty transposon system, we initially used the model tumor antigen OVALBUMIN (OVA) against which we could following anti-tumor T cell responses using SIINFEKL tetramer analysis. Following hydrodynamic injection of the hMet + S45Yß-catenin + pOVA / Sleeping Beauty transposon plasmid system, in which the model antigen OVA was co-expressed in the HCC tumors, anti-OVA T cells were detected in mice with hMet + S45Yß-catenin + OVA tumors from both tumor and spleen but not in naïve, non-tumor-bearing animals (Figs. 4A-B). Even though in this case OVA is a foreign non-tolerized antigen, these data were consistent with the generation of anti-HCCTAA T cells in these tumors as suggested from Figs. 3B-E. Furthermore, addition of anti-PD-L1 to tumor bearing mice significantly enhanced the anti-OVA T cell response approximately threefold in both tumor and spleen, confirming that anti-PD-L1 ICI was able to expand anti-HCCTAA CD8+ T cell responses in vivo (Figs. 4A-B).
Administration of ICI confirmed the anti-PD-L1 enhanced expansion of anti-OVA T cells in both liver and spleen (Figs. 4C-E). However, consistent with the effective replacement of anti-tumor T cells with anti-viral T cells seen in Fig. 3, the addition of either early (Fig. 3) or late (Figs. 4C-D) VSV (VSV-GFP or VSV-IFNß) to HCC-OVA tumor-bearing mice ablated almost entirely the anti-HCCTAA(OVA) T cell response in the liver, whether or not anti-PD-LI ICI was also administered (Figs. 4C-D). The diminution of the anti-HCCTAA(OVA) T cell response by addition of VSV-GFP or VSV-IFNß was less complete in the spleen (Figs. 4C&E) suggesting that the inflammatory anti-viral response is most dominant within the liver TME.
Therefore, we tested the hypothesis that it would be possible to amalgamate the anti-viral response with an anti-tumor response by expressing the HCCTAA(OVA) from within VSV. When ICI was given in the absence of virus to allow for maximal expansion of the anti-tumor T cell response, treatment with VSV-OVA alone at a late stage (Fig. 4F) generated significantly greater numbers of anti-OVA T cells in both liver and spleen compared to the levels spontaneously generated in response to growth of HCC-OVA tumors alone (Figs. 4G&H compared to Figs. 4A&B). In addition, treatment with early anti-PD-L1 ICI combined with late VSV-OVA significantly expanded the levels of anti-HCCTAA(OVA) CD8+ T cells to high levels as a percentage of total CD8+ T cells in both liver and spleen (Figs. 4G&H). Interestingly, the expansion of anti-HCCTAA(OVA) CD8+ T cells by the addition of anti-PD-L1 ICI was significantly greater than the relative expansion of anti-VSV T cells (measured by tetramer directed against the immune-dominant VSV-N52 − 59 peptide of the VSV N protein) (Figs. 4G&H).
Taken together these data show that by encoding a putative HCCTAA within the oncolytic VSV highly significant numbers of anti-HCCTAA T cells were generated in vivo which were substrates for expansion by anti-PD-L1 ICI therapy and which expanded preferentially over at least one component of the immune-dominant anti-viral response.
Selection of putative Sleeping Beauty TAA through prediction of high MHC Class I binding epitopes. Although it was encouraging that encoding a TAA within VSV could generate ICI-expandable anti-tumor T cells, the experiments of Fig. 4 used the immunogenic model OVA antigen to which the C57Bl/6 mice were not tolerized. Therefore, to expand the concept of VSV-TAA to therapeutic efficacy against real putative HCCTAA, RNAseq analysis of SB-HCC tumors was used to identify the top ten genes whose expression was upregulated in tumors compared to normal liver (Fig. 5A). Predicted binding affinities of peptide epitopes from these genes to the relevant H2Kb and H2Db MHC Class I molecules of C57Bl/6 mice identified several strong binding epitopes from the Lcn2, Lect2, Smagp, Nsdh1 and Plrg1 genes (Fig. 5B-C), suggesting that over-expression of these genes in Sleeping Beauty HCC may expose potential neo-antigens for endogenous CD8+ T cell priming/recognition. To test the ability of the three highest over-expressed genes (Lcn2, Lect2, Smagp) encoded separately within VSV to break tolerance to these potential HCCTAA, VSV-IFNß co-expressing the relevant genes were tested as vaccinating agents in vivo. Splenocytes from mice vaccinated with separate VSV-IFNß-TAA did not generate any detectable Th17 responses following in vitro re-stimulation with a 1:1:1 mix of three Sleeping Beauty HCC explants recovered from untreated tumor bearing mice (SB-HCC 1,2,3) (Fig. 5D). However, splenocytes from mice vaccinated with VSV-IFNß-Lcn2 generated a Th1-like, IFN-γ response to these tumor targets which was significantly greater than that generated by VSV-IFNß alone (Fig. 5E), suggesting that this TAA may be a potential HCCTAA in this system.
VSV expressing a single putative HCCTAA may boost a pre-existing HCCTAA T cell response. From these data we reasoned that a treatment regimen in which early ICI was administered prior to virus treatment would optimize the expansion of the endogenous anti-tumor T cell response prior to a boosting effect with VSV-TAA. In this setting, and as before, anti-PD-L1 ICI alone cured ~ 50% of mice bearing Sleeping Beauty HCC tumors, a therapeutic effect which was completely and rapidly eradicated by the addition of VSV-IFN-ß (Fig. 6A). Interestingly, treatment with VSV-IFNß-Lcn2 in combination with anti-PD-L1 ICI was significantly more therapeutic than VSV-IFN-ß + anti-PD-L1 (Fig. 6A). However, although VSV-IFNß-Lcn2 in combination with anti-PD-L1 did not ablate the therapy seen with anti-PD-L1 ICI alone, it did not increase therapy either (Fig. 6A). Similarly, treatment with (VSV-IFNß-Lect2 + VSV-IFNß-Lcn2 + VSV-IFNß-Smagp) in a ratio of 1:1:1 in combination with anti-PD-L1 ICI was also significantly more effective than VSV-IFN-ß + anti-PD-L1, but also did not improve upon, the therapy of anti-PD-L1 alone (Fig. 6A). Consistent with the results of Figs. 5D-E, splenocytes from all groups showed no detectable Th17 anti-SB-HCC 1,2,3 responses (Fig. 6B). In contrast, splenocytes from mice treated with anti-PD-L1 alone showed significant Th-1-like IFN-γ recall responses against live SB-HCC 1,2,3 explants (Fig. 6C). Splenocytes from mice treated with VSV-Lcn2 + anti-PD-L1, showed a trend towards a recall Th-1 response against live SB-HCC 1,2,3 explants but this did not reach significance compared to splenocytes from mice treated with VSV-IFNß + anti-PD-L1 (Fig. 6C). Finally, splenocytes from mice treated with the combination of all three putative HCCTAA (VSV-IFNß-Lect2 + VSV-IFNß-Lcn2 + VSV-IFNß-Smagp) + anti-PD-L1 showed no detectable response to SB-HCC 1,2,3 explants (Fig. 6C), suggesting that addition of the Lect2 and/or Smagp antigens may even exert an inhibitory effect upon anti-tumor immunity as generated by Lcn2 alone.
VSV expressing multiple undefined HCCTAA boosts anti-tumor CD8+ T cell responses expanded by ICI. One possible interpretation of these data is that, when at least one potentially immunogenic HCCTAA, such as Lcn2, was added to VSV-IFNß + anti-PD-L1, early treatment with anti-PD-L1 ICI allowed for expansion of potentially tumor reactive CD8+ T cells (Cluster 17 in Fig. 3). Thereafter, the anti-Lcn2 component of the anti-tumor T cell response could then be boosted by late vaccination with VSV-IFNß-Lcn2. If this model were true, we predicted that by adding multiple further HCCTAA to the vaccinating VSV-IFNß virus, an increased number of anti-HCCTAA T cells could be boosted by late VSV-IFNß-HCCTAA vaccination. Therefore, cDNA from three separate Sleeping Beauty explants mixed at a 1:1:1 ratio was cloned into the VSV-IFNß virus to give a viral stock of VSV-IFNß-SB-HCC 1,2,3 cDNA (Fig. 7A). Presence of the three most highly expressed SB-HCC genes, Lcn2, Lect2, and Smagp (identified by RNAseq, Fig. 5) in the VSV-IFNß-SB-HCC 1,2,3 cDNA library, but not in the ASMEL VSV-cDNA library constructed previously from melanoma cells36, was confirmed by PCR. Conversely, the melanoma associated genes gp100 or TYRP1 were abundant in the ASMEL library but were essentially undetectable in the VSV-IFNß-SB-HCC 1,2,3 cDNA library (Fig. 7B).
Treatment of SB-HCC tumor bearing mice with a reduced dosage of anti-PD-L1 ICI (100µg/mouse/injection instead of 200µg/mouse/per injection) in order to accentuate any combinatorial therapeutic effects, generated some long-term survivors (Fig. 7C). As before, addition of VSV expressing at least one putative HCCTAA(Lcn2) no longer ablated, but neither did it improve, the therapy with ICI alone (Fig. 7C). However, with Early anti-PD-L1 ICI + Late VSV-SB-HCC1,2,3, for the first time, a combination of VSV and anti-PD-L1 ICI generated significantly improved therapy over anti-PD-L1 ICI treatment alone with 100% of mice surviving to day 150 (Fig. 7C). Interestingly, treatment with VSV-SB-HCC1,2,3 alone (no anti-PD-L1) was also very effective at generating long term cures (Fig. 7C). These data suggest that VSV-mediated display of multiple antigens was sufficient of itself to boost slow developing anti-HCCTAA CD8+ T cell responses even in the absence of expansion by anti-PD-L1 treatment. Consistent with this, splenocytes from mice treated with anti-PD-L1 ICI alone re-stimulated in vitro with live SB-HCC Explants 1,2,3 cells secreted IFN-γ but not IL-17 (Figs. 7D&E). In contrast, splenocytes from mice treated with either VSV-SB-HCC1,2,3 alone (no anti-PD-L1 ICI), or with VSV-SB-HCC1,2,3 + anti-PD-L1, generated a Th17 recall response when re-stimulated in vitro with SB-HCC cells (Fig. 7D). Splenocytes from mice treated with VSV-SB-HCC1,2,3 alone generated a Th1/IFN-γ recall response that was equivalent to that from splenocytes from mice treated with anti-PD-L1 alone (Fig. 7E), whereas splenocytes from mice treated with VSV-IFNß-SB-HCC1,2,3 + anti-PD-L1 generated a significantly increased Th1/IFN-γ recall response to SB-HCC cell explants (Fig. 7E). To confirm the mechanism associated with the highly significant therapy induced by treatment with VSV-SB-HCC1,2,3, SB-HCC tumor-bearing mice were depleted of different immune subsets prior to treatment with VSV-SB-HCC1,2,3 +/- ICI (Fig. 7F). As before, anti-PD-L1 ICI alone generated cures in just under 50% of the mice (Fig. 7F, light blue line) and mice treated with a control isotype IgG all succumbed to tumor (Fig. 7F, dark blue line). Also, as before, treatment with Early anti-PD-L1 + Late VSV-SB-HCC1,2,3 cured 100% of mice (Fig. 7F, brown line). Treatment with VSV-SB-HCC1,2,3 alone was not significantly different from Early anti-PD-L1 + Late VSV-SB-HCC1,2,3, although one mouse in this experiment succumbed to tumor (Fig. 7F, black line). Depletion of neither NK, nor CD4+, cells from mice treated with the Early anti-PD-L1 + Late VSV-SB-HCC1,2,3 regimen significantly decreased overall therapy, although both groups fell from 100% survival by day 150 (Fig. 7F, red and green lines respectively). However, depletion of CD8+ T cells almost completely abolished therapy (Fig. 7F, purple line). Finally, treatment of tumor bearing mice with Early anti-PD-L1 and Late VSV-cDNA library in which the cDNA was sourced from a melanoma cell line (VSV-ASMEL) was also ineffective (Fig. 7F, orange line), showing that CD8+ T cell boosting against cancer type specific antigens is critical for these effects.
Taken together, these data show that optimal therapy requires ICI-induced expansion of anti-HCCTAA-specific CD8+ T cells, which are subsequently further expanded by VSV-IFNß-HCCTAA boosting, through induction of both Th17 and Th1 component mechanisms.