Chimerization of the Anti-Viral CD8+ T Cell Response with A Broad Anti-Tumor T Cell Response Reverses Inhibition of Checkpoint Blockade Therapy by Oncolytic Virotherapy

Although immune checkpoint inhibition (ICI) has produced profound survival benefits in a broad variety of tumors, a proportion of patients do not respond. Treatment failure is in part due to immune suppressive tumor microenvironments (TME), which is particularly true of hepatocellular carcinoma (HCC). Since oncolytic viruses (OV) can generate a highly immune-infiltrated, inflammatory TME, we developed a vesicular stomatitis virus expressing interferon-ß (VSV-IFNß) as a viro-immunotherapy against HCC. Since HCC standard of care atezolizumab/bevacizumab incorporates ICI, we tested the hypothesis that pro-inflammatory VSV-IFNß would recruit, prime, and activate anti-tumor T cells, whose activity anti-PD-L1 ICI would potentiate. However, in a partially anti-PD-L1-responsive model of HCC, addition of VSV-IFNß abolished anti-PD-L1 therapy. Cytometry by Time of Flight showed that VSV-IFNß expanded dominant anti-viral effector CD8 T cells with concomitant, relative disappearance of anti-tumor T cell populations which are the target of anti-PD-L1. However, by expressing a range of HCC tumor antigens within VSV, the potent anti-viral response became amalgamated with an anti-tumor T cell response generating highly significant cures compared to anti-PD-L1 ICI alone. Our data provide a cautionary message for the use of highly immunogenic viruses as tumor-specific immune-therapeutics by showing that dominant anti-viral T cell responses can inhibit sub-dominant anti-tumor T cell responses. However, by chimerizing anti-viral and anti-tumor T cell responses through encoding tumor antigens within the virus, oncolytic virotherapy can be purposed for very effective immune driven tumor clearance and can generate anti-tumor T cell populations upon which immune checkpoint blockade can effectively work.


INTRODUCTION
Hepatocellular Carcinoma (HCC) represents the most common presentation of cancer within the liver and the 6th most common cancer worldwide 1,2,3 .Although immune checkpoint inhibition (ICI) therapy has produced signi cant survival bene ts in a broad range of tumor types, a proportion of patients with putatively ICI-sensitive tumors, such as hepatocellular carcinoma, either do not respond or experience diminishing returns from ongoing therapy.The anti-PD-L1 monoclonal antibody atezolizumab in combination with the anti-vascular endothelial growth factor (VEGF) antibody bevacizumab has recently been approved as front-line therapy for HCC 4 .In the phase III imbrave150 study, this combination reduced the risk of death by 56% and the risk of disease progression by 40% compared to the treatment control arm sorafenib.However, although atezolizumab/bevacizumab has clinical activity, response rates reach about 30% and the majority of patients ultimately succumb to the disease 5 .
A major factor contributing to treatment failure in HCC is believed to be the immune suppression mediated by the tumor microenvironment (TME) 6,7,8,9 .In this respect, oncolytic viruses (OV) infect and selectively replicate within tumor cells, leading to selective cancer cell lysis 10,11,12,13 .Pro-in ammatory viral oncolysis also modi es the TME and helps to reverse immune-sparse, or -suppressive, microenvironments into highly in ltrated in ammatory immune environments 14,15,16 .To exploit the in ammatory aspects of OV therapy, we have developed the Vesicular Stomatitis Virus (VSV) as an oncolytic and immunotherapeutic agent against different tumor types with the rationale that intratumoral (IT) delivery of VSV would disrupt the immune-suppressive TME and convert it into a virusin amed, immune in ltrated tumor site 15,17,18,19,20,21 .To increase both safety and immune-stimulating potency, we also expressed the IFN-ß gene from the virus 21,22,23 .Following IND-enabling toxicology studies 24,25 , we conducted a rst-in-human clinical study of VSV expressing human IFN-ß (VSV-hIFNβ) using IT injection under ultrasound guidance in patients with liver tumors [NCT01628640].Fifteen patients were treated with a dose range of 10 5 to 1.8x10 7 median tissue culture infectious dose (TCID 50 ) and the maximum tolerated dose was 3x10 6 TCID 50 .Preliminary evidence of e cacy included one partial response by Response Evaluation Criteria in Solid Tumors (RECIST) v1.1 criteria and two patients with stable disease (manuscript in preparation), making it clear that further improvements to the therapeutic regimen are required.
Since current standard of care HCC therapy atezolizumab/bevacizumab incorporates ICI, in the current study we sought to develop our initial clinical approach by combining virotherapy with ICI.Therefore, here we tested the hypothesis that the heavily pro-in ammatory activity of vesicular stomatitis virus expressing interferon-ß (VSV-IFNß) virotherapy would convert the immune suppressive TME of HCC to an immune stimulatory environment.In turn, this would recruit/prime/activate potentially anti-tumor T cells, the activity of which anti-PD-L1 ICI therapy would be able to potentiate further, thereby generating superior anti-tumor therapy to either therapy alone 26,27,28,29,30,31,32 .This hypothesis is consistent with the model that in ammatory killing of tumor cells through viral oncolysis will recruit antigen presenting cells (APC) to the tumor site, release high loads of tumor associated antigens (TAA), and facilitate presentation of these TAA to CD8 and CD4 T cells in the draining lymph nodes, thereby generating systemically active anti-tumor T cell immunity 26,27,28,29,30,31,32 .
To test this hypothesis, we used a slow developing model of HCC, which resembles the etiology of human HCC 33 , in which a Sleeping Beauty transposon integrates ß-catenin and hMet oncogenes into the livers of neonatal mice resulting in multi-focal tumor formation (SB-HCC) over 100-150 days.Tumor bearing mice were partially susceptible to anti-PD-L1 therapy (~ 50% long term cures), thereby mimicking at least one aspect of the human disease and current therapy.To our surprise, rather than observing even additive bene t to therapy, and irrespective of the relative timings of both treatments, the addition of VSV-IFNß to anti-PD-L1 therapy largely abolished the therapeutic effects of anti-PD-L1 ICI alone.We show that VSV-IFNß treatment generated a highly signi cant expansion of one, or a few, dominant anti-viral effector CD8 + T cell populations with the concomitant relative disappearance of those putative anti-tumor T cell populations which are the target of anti-PD-L1 treatment.Based on our previous studies showing that inclusion of a tumor antigen within VSV induces CD8 + T cell dependent anti-tumor therapy directed speci cally against the virally encoded antigen 19,34,35,36 , we hypothesized that it would be possible to amalgamate the potent anti-viral response with an anti-tumor T cell response by expressing immunologically relevant tumor associated antigens from within the virus.Here we show that by expressing a multitude of putative, uncharacterized, tumor antigens within VSV using a cDNA library derived from three different SB-HCC explants (VSV-SB-HCC1,2,3) we observed highly signi cant improvements compared to anti-PD-L1 ICI alone -even in the absence of added anti-PD-L1 treatment.
Our data provide a cautionary message for the use of highly immunogenic viruses as tumor-speci c immune-therapeutics.We show that oncolytic virotherapy can induce anti-viral T cell responses which can actively inhibit, or obscure, weak anti-tumor T cell responses leading, potentially, to decreased antitumor therapy.However, by chimerizing anti-viral and anti-tumor T cell responses through encoding tumor antigens within the virus, at least a proportion of the anti-viral T cell response is co-opted into an antitumor response.In this way, oncolytic virotherapy can be purposed for very effective immune based tumor clearance and can generate anti-tumor T cell populations upon which immune checkpoint blockade can effectively work.

RESULTS
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 injection 33 , thereby mimicking an expression signature seen in a proportion of HCC patients 33,37,38 .We con rmed that the hMet + S45Y ßcatenin/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 in ltration.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 in ltration 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 in ltration trended towards increased levels but did not reach signi cance (Figs.1C-D).The majority of liver in ltrating 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 signi cantly increased (≥ 50% of the total CD8 + T cell population) compared to the control nontumor 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 in ltrating CD4 + T cell population, where trends towards increased proportions of Tim-3 + /PD-1 + CD4 + T cells did not reach signi cance 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 in ltrating CD11c + /MHCII + dendritic cells was signi cantly increased in tumor bearing compared to nontumor 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 TME 39 is associated with exhausted CD8 + T cells with reduced anti-tumor cytotoxicity 40,41,42 .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 e cacy of anti-PD-L1 therapy compared to non-depleted mice, although this did not reach signi cance (Fig. 2C).While not de nitive, 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 immunesuppressive TME into a pro-in ammatory environment.In turn, this would liberate HCC tumor associated antigens (HCC TAA ) leading to the priming of anti-HCC TAA 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 bene t 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-signi cant) 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-HCC TAA immune response.
An immuno-dominant anti-viral rapid effector CD8 + T cell population replaces sub-dominant putative anti-HCC TAA 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 in ltrating lymphocyte populations (TIL) (Fig. 3A).Using differential immune marker expression, we de ned 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 de ned 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 cells 43,44,45 .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 HCC 46,47,48 .
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 signi cant 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-HCC TAA 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 speci c 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-HCC TAA 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 pro le consistent with the presence of anti-viral effector CD8 + T cells.
Taken together, these data strongly argue that the abolition of therapeutic bene t 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-HCC TAA 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-HCC TAA 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 speci cally against the virally encoded antigen as a result of the potent immune stimulating adjuvant properties of infection with VSV 19,34,35,36 .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-HCC TAA T cells in these tumors as suggested from Figs. 3B-E.Furthermore, addition of anti-PD-L1 to tumor bearing mice signi cantly enhanced the anti-OVA T cell response approximately threefold in both tumor and spleen, con rming that anti-PD-L1 ICI was able to expand anti-HCC TAA CD8 + T cell responses in vivo (Figs.4A-B).
Administration of ICI con rmed 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-HCC TAA(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-HCC TAA(OVA) T cell response by addition of VSV-GFP or VSV-IFNß was less complete in the spleen (Figs.4C&E) suggesting that the in ammatory 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 HCC TAA(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 signi cantly 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 signi cantly expanded the levels of anti-HCC TAA(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-HCC TAA(OVA) CD8 + T cells by the addition of anti-PD-L1 ICI was signi cantly greater than the relative expansion of anti-VSV T cells (measured by tetramer directed against the immunedominant VSV-N 52 − 59 peptide of the VSV N protein) (Figs.4G&H).
Taken together these data show that by encoding a putative HCC TAA within the oncolytic VSV highly signi cant numbers of anti-HCC TAA 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 immunedominant 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 e cacy against real putative HCC TAA , 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 a nities of peptide epitopes from these genes to the relevant H2K b and H2D b MHC Class I molecules of C57Bl/6 mice identi ed 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 overexpressed genes (Lcn2, Lect2, Smagp) encoded separately within VSV to break tolerance to these potential HCC TAA , 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 signi cantly greater than that generated by VSV-IFNß alone (Fig. 5E), suggesting that this TAA may be a potential HCC TAA in this system.
VSV expressing a single putative HCC TAA may boost a pre-existing HCC TAA 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 signi cantly 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 signi cantly 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 signi cant 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 signi cance 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 HCC TAA (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 unde ned HCC TAA boosts anti-tumor CD8 + T cell responses expanded by ICI.One possible interpretation of these data is that, when at least one potentially immunogenic HCC TAA , 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 HCC TAA to the vaccinating VSV-IFNß virus, an increased number of anti-HCC TAA T cells could be boosted by late VSV-IFNß-HCC TAA 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 (identi ed 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 cells 36 , was con rmed 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).
Taken together, these data show that optimal therapy requires ICI-induced expansion of anti-HCC TAAspeci c CD8 + T cells, which are subsequently further expanded by VSV-IFNß-HCC TAA boosting, through induction of both Th17 and Th1 component mechanisms.

DISCUSSION
Here, we used a Sleeping Beauty Transposon-based model of HCC which re-capitulates many features of the human disease genetically, phenotypically, and in its partial sensitivity to anti-PD-L1 ICI therapy.We show that, in a tumor model where there is a pre-existing, weak/immuno-subdominant anti-HCC TAA CD8 + T cell response, addition of a highly immuno-dominant oncolytic virus effectively occluded the anti-tumor response with a rapid effector anti-viral CD8 + T cell response.However, by encoding a range of HCC TAA within the virus in the form of an HCC-derived cDNA library, the potent immune adjuvant properties of the virus allowed for boosting of the pre-existing anti-HCC TAA CD8 + T cells leading to high rates of tumor cures.
With the rationale of disrupting the immune-suppressive TME of HCC and converting it into a virusin amed immune in ltrated TME, we tested VSV-IFNß gene as an oncolytic and immunotherapeutic agent 15,17,18,19,20,21,24,34 in a clinical trial [NCT01628640] in which VSV-IFNß was administered to patients with sorafenib-refractory HCC.In that study, we observed modest activity associated with VSV-IFNß virus alone.Therefore, we sought to improve clinical utility of the virus in the context of the current standard of care regimen atezolizumab/bevacizumab which includes ICI.We used a slow developing model of HCC, which resembles the etiology of human HCC, in which a Sleeping Beauty transposon integrates the ß-catenin and hMet oncogenes into the livers of neonatal mice resulting in multi-focal tumor formation over 100-150 days (Fig. 1).Tumors grew progressively in the liver, showed characteristics of immune suppression, expressed PD-L1, and were partially susceptible to anti-PD-L1 therapy (~ 50% long term cures), thereby mimicking several aspects of the human disease and its current therapy (Figs.1&2A-C).
Contrary to our initial hypothesis, the addition of a highly immunogenic oncolytic virus to a weak, preexisting anti-PD-L1-sensitive anti-tumor immune response inhibited the therapeutic anti-HCC TAA immune response (Fig. 2D-E) irrespective of treatment sequencing and the combination of virus and ICI therapy was unable to improve on the e cacy of ICI alone.One possible interpretation of these data is that a highly immuno-dominant anti-viral CD8 + T cell response induced by VSV-IFNß out competes, or interferes with, the weak, immuno-subdominant, slow developing, anti-PD-L1-sensitive anti-HCC TAA T cell response.
Using CyTOF analysis, Fig. 3 showed that VSV-IFNß treatment did indeed generate a highly signi cant expansion of one, or a few, dominant anti-viral effector CD8 + T cell populations with the concomitant relative disappearance of those putative anti-tumor T cell populations which are the likely target of anti-PD-L1 treatment.The immune pro le of the virally induced CD8 + T cell populations matches that of antiviral effector CD8 + T cells, as granzyme B characterizes antiviral CD8 + T cell responses 49,50 .Other markers also are consistent with this population being an anti-viral effector CD8 population.IL-7R has been shown to be consistently overexpressed on antigen-speci c effector CD8 + T cells during viral infections 51 , while Bat3 promotes immune cell proliferation and cytokine production in persistent viral infections by promoting a T helper type 1 response 52,53 .Taken together, these data strongly argue that the abolition of therapeutic bene t achieved with ICI alone by addition of virus to ICI therapy was due to a rapid induction/expansion of anti-viral effector CD8 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-HCC TAA CD8 + T cells towards a population of anti-viral effector CD8 + T cells suggest that inappropriately timed anti-PD-L1 ICI could actually lead to in vivo enhancement of the immunodominant anti-VSV T cell response over the immuno-subdominant anti-HCC TAA T cell response -expanding antiviral T cells and not anti-HCC TAA T cells.An implication of these data is that early, prolonged ICI may allow for strengthening/expansion of the immuno-subdominant anti-HCC TAA T cell response with time such that a later immuno-dominant anti-viral T cell response is less able to out-compete it.However, the data of Fig. 2E suggest that even late virus treatment has a dominant negative effect on the anti-tumor T cell response.
Our previous studies have shown that inclusion of a tumor antigen within VSV induces CD8 + T cell dependent anti-tumor therapy directed speci cally against the virally encoded antigen as a result of the potent immune stimulating adjuvant properties of infection with VSV 19,34,35,36 .Therefore, we hypothesized that it would be possible to amalgamate the potent anti-viral response with an anti-tumor T cell response by expressing immunologically relevant tumor associated antigens from within the virus.As a model system, we incorporated the OVALBUMIN antigen into the Sleeping Beauty ß-catenin/hMet system such that tumors would express the OVA protein.ß-catenin/hMet/OVA tumors generated spontaneous anti-HCC TAA(OVA) endogenous T cells in both spleen and tumor and addition of anti-PD-L1 ICI expanded these CD8 + T cell responses (Fig. 4).By encoding the putative HCC TAA(OVA) within oncolytic VSV-ova highly signi cant numbers of anti HCC TAA(OVA) T cells were generated in vivo even in the absence of ICI.In addition, those anti-HCC TAA(OVA) CD8 + T cell responses were substrates for expansion by anti-PD-L1 ICI therapy.Moreover, in contrast to the ablation of the anti-HCC TAA CD8 + T cell responses by the addition of VSV virotherapy observed in Figs.2&3, combined treatment with virus (TAA) and ICI was able to expand the anti-HCC TAA CD8 + T cell response preferentially over at least one component of the immuno-dominant anti-viral response (Fig. 4).Nevertheless, in this system, OVA represents a highly immunogenic, non-self, non-tolerized HCC TAA and does not re ect the situation in which HCC TAA are likely to be highly immuno-subdominant with low numbers of precursor T cells available to recognize and reject them (Fig. 3).Therefore, using RNAseq analysis of Sleeping Beauty tumors recovered from untreated mice, we identi ed three separate antigens predicted to be strong MHC Class I binders and, therefore, potential T cell targets when over-expressed in tumors.Of these three candidates, only one, Lcn2, generated detectable Th1-like T cell responses following expression from VSV (Fig. 5).Treatment of SB-HCC tumors with VSV-IFNß-Lcn2 VSV was still not able to enhance therapy compared to anti-PD-L1 alone, although the presence of Lcn2 did prevent the inhibition of the effects of anti-PD-L1 alone (Fig. 6).
Cumulatively, these data are consistent with a model in which the immuno-dominant anti-VSV CD8 + T cell response ablates the weaker, immuno-subdominant anti-HCC TAA CD8 + T cell response causing loss of the ICI therapy alone when ICI was combined with VSV-IFNß (Figs. 2&3).However, when at least one potentially immunogenic HCC TAA , 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), the anti-Lcn2 component of which could then be boosted by late vaccination with VSV-IFNß-Lcn2.In contrast, late VSV-IFNß with no additional HCC TAA simply activates a rapid new anti-viral CD8 + T cell population, which replaces/outcompetes the developing, anti-PD-L1-expanded HCC TAA -speci c CD8 + T cells.
In this model, an endogenous anti-HCC TAA CD8 + T cell response against multiple (weak) HCC TAA , expanded by early ICI treatment, would be further boosted by VSV-HCC TAA vaccination.If this model were true, we predicted that by adding multiple further HCC TAA to the vaccinating VSV-IFNß virus, an increased number of anti-HCC TAA T cells could be boosted by late VSV-IFNß-HCC TAA vaccination.In the absence of effective models to predict the nature of multiple possible HCC TAA (Fig. 5), we hypothesized that cloning a cDNA library from Sleeping Beauty HCC tumor lines into VSV would allow for the highly immunogenic display of multiple HCC TAA with the advantage that the HCC TAA would not need to be molecularly identi ed beforehand to be effective 35,36 .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-HCC1,2,3cDNA (Fig. 7A).By expressing a multitude of putative, uncharacterized, tumor antigens within VSV, for the rst time, a combination of VSV and anti-PD-L1 ICI generated signi cantly improved therapy over anti-PD-L1 ICI treatment alone with 100% of mice surviving to day 150 (Fig. 7C).Anti-tumor therapy against HCC was not observed if the VSV-cDNA was derived from melanoma instead of HCC tumors, was dependent upon CD8 + but less so on CD4 + T cells and was associated with a Th17 phenotype of T cell activity (Fig. 7).Interestingly, treatment with VSV-IFNß-SB-HCC1,2,3cDNA alone (no anti-PD-L1) was also very effective at generating long term cures.These data suggest that VSV-mediated display of multiple antigens was su cient of itself to boost slow developing anti HCC TAA CD8 + T cell responses even in the absence of expansion by prior anti-PD-L1 treatment.
A major advantage of using a tumor-derived cDNA library as the source of multiple HCC TAA is that there is no need to de ne the identity of multiple TAA from each patient.On the other hand, a concern about the use of a cDNA library is that tolerance may be broken to multiple normal liver associated antigensalthough we did not observe any toxicity associated with autoimmune liver disease in the mice cured of their tumors when treated with VSV-IFNß-SB-HCC1,2,3cDNA either with, or without, anti-PD-L1.We believe that by constructing the cDNA library from tumor tissue the representation of normal self-antigens against which any non-tolerized T cells exist in vivo may be reduced relative to tumor associated markers which can form immunological targets.Further studies are underway to assess the full range of toxicities associated with VSV-IFNß-SB-HCC1,2,3cDNA-mediated liver damage.
Our results here provide a cautionary message for the use of highly immunogenic viruses as tumorspeci c immune-therapeutics.We show here that oncolytic virotherapy can induce anti-viral T cell responses which can actively inhibit, or obscure, pre-existing weak anti-tumor T cell responses leading, potentially, to decreased anti-tumor therapy with ICI which targets that immuno-subdominant anti-tumor T cell population.However, by ensuring that at least a proportion of the anti-viral T cell response can also act to boost an anti-tumor response by encoding multiple relevant tumor antigens within the virus the CD8 + T cell response associated with oncolytic virotherapy can be purposed for very effective tumor clearance.In the model system used here, oncolytic virotherapy was used in the context of a poorly immunogenic developing tumor against which an endogenous T cell response existed and which could be enhanced by ICI therapy.We are currently investigating the alternative situation in which a developing tumor is completely non-immunogenic -that is it does not raise any anti-tumor T cell response.In such a case it may be that oncolysis by the virus helps to generate a previously non-existent anti-tumor T cell response upon which subsequent ICI treatment may be able to work.Our ndings can, therefore, inform the rational sequencing of a combinatorial use of oncolytic virotherapy and ICI therapy such that viral oncolysis is used productively to generate anti-tumor T cell populations upon which immune checkpoint blockade can effectively work.In summary, our data are consistent with a model in which a poorly immunogenic tumor is partially sensitive to ICI therapy through the selective re-invigoration/expansion of anti-tumor CD8 + T cell populations presumably recognizing weak, sub-dominant antigens.In these circumstances, treatment with an OV encoding a set of highly immunogenic, immuno-dominant viral antigens induces populations of anti-viral CD8 + T cells which overwhelm the pre-existing anti-tumor CD8 + T cell populations and which, therefore, signi cantly inhibit therapy associated with ICI alone.However, by expressing multiple TAA within the OV, the immune adjuvant properties of the virus become advantageous rather than detrimental by promoting boosting of the pre-existing anti-TAA CD8 + T cell populations in vivo through induction of both Th17 and Th1 component mechanisms.In this way, the anti-viral T cell response is chimerized to become, at least in part, an anti-tumor T cell response leading to a positive interaction between ICI therapy and oncolytic virotherapy.

METHODS
Pathology.Livers were xed in 10% formalin and stained with hematoxylin and eosin by the Mayo Clinic Histology Core Facility.Immuno uorescence staining was as described previously 54 .PD-L1 antibody was purchased from BioX-cell, clone 10F.9G2.
Cell lines and viruses.VSV expressing murine IFN-ß (VSV-IFN-ß), ovalbumin (VSV-OVA), Lcn2 (VSV-IFNß-Lcn2), Lect2 (VSV-IFNß-Lect2), SMAGP (VSV-IFNß-SMAGP), or green uorescent protein (VSV-GFP) were rescued from the pXN2 cDNA plasmid using the established reverse genetics system in BHK cells as described previously 21,22,34 .In brief, BHK cells were infected with MVA-T7 at an MOI of 1.Cells were incubated at 37°C and 5% CO 2 .After 1 h, cells were transfected with pVSV-XN2 genomic VSV plasmid (10 µg), pBluescript (pBS)-encoding VSV-N (3 µg), pBS-encoding VSV-P (5 µg), and pBS-encoding VSV L proteins (1 µg) using Fugene6 according to the manufacturer's recommendations.Cells were incubated at 37°C and 5% CO2 for 48 h.After 48 h, supernatant was collected and clari ed by passing through a 0.2µm lter.All transgenes were inserted between viral G and L genes using the XhoI and NheI restriction sites with the murine IFN-ß gene cloned between the viral M and G genes.Virus titers were determined by plaque assay on BHK cells.VSV-cDNA libraries were generated as described previously 35,36 .Brie y, cDNA from two human melanoma cell lines, Mel624 and Mel888, (ASMEL library) 36 or from three murine Sleeping Beauty HCC tumor-bearing livers, (SB HCC 1,2&3 Library), was pooled, cloned (Clone Miner cDNA Construction kit) (Invitrogen) and ampli ed by PCR.The PCR-ampli ed cDNA molecules were size fractionated to below 4 kbp for ligation into the parental VSV genomic plasmid pVSV-XN2 55 between the G and L genes since lower sized cDNA inserts were associated with both higher viral titers and lower proportions of Defective Interfering particles.The complexity of the ASMEL and SB HCC 1,2&3 cDNA libraries cloned into the VSV backbone plasmid between the Xho1 and Nhe1 sites was 7.0 × 10 6 36 and 4.1 × 10 6 colony forming units respectively (at dilutions of 10 − 6 and 10 − 5 there were 8 and 74 colonies with the SB HCC 1,2&3 Library).For the SB-HCC1,2,3 Library, of 30 colonies picked at random, 1 had no insert, 12 had an insert of less than 0.5 kbp and 17 had inserts between 0.5kbp and 4kbp.Virus was generated from BHK cells by co-transfection of pVSV-XN2-cDNA library DNA along with plasmids encoding viral genes as described 55 .Virus was expanded by a single round of infection of BHK cells and puri ed by sucrose gradient centrifugation.
Mice.All experiments utilized 6 to 8-week-old male and female C57Bl/6 mice (stock 000664) obtained from The Jackson Laboratory with the exception of Fig. 1H, which utilized FVB mice (stock 001800) obtained from The Jackson Laboratory.All animals were maintained in a speci c pathogen-free BSL2 biohazard facility.Experimental mice were co-housed and exposed to a 12:12-h light-dark cycle with unrestricted access to water and food.The ambient temperature was restricted to 68°F to 79°F and the room humidity ranged from 30-70%.All animal studies were conducted in accordance with and approved by the Institutional Animal Care and Use Committee at Mayo Clinic.
RNA-Sequencing Analysis of Gene Expression.hMet + S45Y ß-Catenin Sleeping Beauty transposon system liver tumor-induced mice sacri ced on day 18 with livers harvested for RNA extraction (Rneasy, Qiagen).RNA samples were then subjected to RNA-seq analysis at the Genome Analysis Core, Mayo Clinic.The 10 most-expressed genes identi ed and full-length sequences were ltered through NET MHC 2.0 binding a nity algorithm to identify octamer or nonamer peptides whose binding a nity for H2K b or H2K d was below a threshold of 500 nM, and whose corresponding wild-type peptides had a binding a nity to the same molecules above 500 nM.A list of 10 candidates was further re ned using the EMBL-EBI Expression Atlas for expression in liver tissue to identify 4 high a nity peptides.This methodology is discussed in detail in our previous work 57 .
PCR Validation of ASMEL SB-HCC1,2&3 Libraries.RNA was prepared from the equivalent of 10 8 pfu of the VSV-based ASMEL 36 or SB-HCC1,2&3 Library with the QIAGEN RNA extraction kit.1µg total RNA was reverse transcribed in a 20µl volume using oligo-(dT) as a primer.A cDNA equivalent of 1ng RNA was ampli ed by PCR with gene speci c primers to the HCC-speci c Lcn2, Lect2, and Smagp, or melanoma speci c TYRP1 and GP100 genes.Statistical analysis.analysis was performed within GraphPad Prism software (GraphPad).Multiple comparisons were analyzed using one-or two-way analysis of variances with a Tukey's post hoc multicomparisons test.Survival data were assessed using a log rank Mantel-Cox test.Data are expressed as group mean ± SD unless otherwise stated.

DECLARATIONS Figures
The    3 IT treatment with VSV-mIFNß generates a dominant CD8 T cell response which replaces the anti-tumor T cell response.A. Treatment regimen.Following hydrodynamic injection of hMet + S45Y ß-Catenin on day 0, animals were treated with anti-PD-L1 on day 21 for 6 doses concurrently with VSV-mIFNß for 3 doses.All animals were sacri ced on day 38.Livers were processed into single-cell suspensions and analyzed by the Mayo CyTOF core facility.N=16, 4 animals per group, groups were pooled for Rphenograph tSNE analysis, 22 groups.Scatterplot and mean heat map shown.B. Following initial analysis (A), 9 distinct immune populations were identi ed and pooled data was re-analyzed using FlowSOM tSNE analysis.'Exhausted CD8 T cells' were identi ed by expression of CD8, Tim-3, LAG-3, CD39, and PD-1 expression, 'B cells' by CD38 and CD19; 'CD4 T cells' by CD4; 'Memory CD4 T Cells' by CD4, CCR7, and CD62L; 'Memory CD8 T Cells' by CD8, CCR7, and CD62L; 'NK Cells' by NK1.1; 'NKT Cells' by NK1.1 and CD8; and 'Anti-viral CD8 T Cells' by CD8, GranzB, and PD-L1.C. Pooled analysis of lymphocytes within livers of tumor-free mice, rst 10,000 events.D. Pooled analysis of lymphocytes within livers of SB-HCC-bearing mice, rst 10,000 events.E. Pooled analysis of lymphocytes within livers of SB-HCC bearing mice treated with anti-PD-L1, rst 10,000 events.F. Pooled analysis of lymphocytes within livers of SB-HCC-bearing mice treated with anti-PD-L1 and VSV-mIFNß, rst 10,000 events.Putative HCC TAA selected for high level expression and MHC binding a nity. A. mice sacri ced on day 18 with livers harvested for RNA extraction.RNA samples were then subjected to RNA-seq analysis, identifying 10 highest relative gene expression compared to non-tumor bearing livers.B-C.The 10 most-expressed genes identi ed and full-length sequences were ltered through NET MHC 2.0 binding a nity algorithm to identify octamer or nonamer peptides whose binding a nity for H2Kb (B) or H2Kd (C) was below a threshold of 500 nM, and whose corresponding wild-type peptides had a binding a nity to the same molecules above 500 nM.corresponded Lcn2, Lect2, and SMAGP.D-E.C57Bl/6 mice (n=12, 3 per group) were injected intravenously with 10 7 plaque-forming units (pfu) of VSV-mIFNß, VSV-Lcn2, VSV-Lect2, or VSV-SMAGP.10 days later 10 6 splenocytes were cocultured with a 1:1:1 mixture of SB-HCC explants 1, 2 and 3 at an effector:target (E:T) ratio of 10:1.
hMet + S45Y ß-Catenin Sleeping Beauty transposon system liver tumors which are in ltrated by PD-1 + CD8 T cells.Following hydrodynamic injection of hMet + S45Y ß-Catenin plasmids, animals were sacri ced at 4 and 7 weeks for pathologic studies.A. Livers (animals A1 and A4) demonstrated appreciable neoplasia with a well-differentiated morphology and minimal immune in ltration by week 4.This progressed by week 7 (animals C1 and C4) with neoplastic cells overtaking the majority of the tumor, with large pseudo-cysts, fat deposits, and signs of extramedullary hematopoiesis (more prominent in C4).B. Livers from a control mouse and from a mouse 7 weeks after hMet + S45Y ß-Catenin hydrodynamic tail-vein injection.C-D.Following hydrodynamic injection of hMet + S45Y ß-Catenin, animals were sacri ced at day 22 and day 29 and analyzed for CD8+ (C) and CD4+ (D) T cells by ow cytometry.D. Similar to C, tumor-in ltrating CD4 T cells were evaluated.E-F.Proportions of CD8+ (E) and CD4+ (F) T cells expressing Tim3 and/or PD1 + in livers of tumor bearing or tumor naïve mice at days 22 and day 29.G. Proportions of PD-L1 + CD86 + dendritic cells (DCs) in mice 22 and 29 days after hMet + S45Y ß-Catenin injection compared to tumor naïve mice.H. Hematoxylin and eosin staining combined with PD-L1 immunohistochemistry of the livers of FVB mice at weeks 4, 5, 6, and 7 after hMet + S45Y ß-Catenin hydrodynamic tail-vein injection.10x magni cation.Signi cance for C-D determined by ordinary one-way ANOVA.Signi cance for E-G was determined by 2-way ANOVA with multiple comparisons using Tukey's multiple comparisons test.Statistical signi cance was set with * indicating a p value less than 0.05, ** < 0.01, *** < 0.001, and **** < 0.0001.

Figure
Figure 3

Figure 4 Expression
Figure 4