A modular self-adjuvanting cancer vaccine combined with an oncolytic vaccine induces potent antitumor immunity

Functional tumor-specific cytotoxic T cells elicited by therapeutic cancer vaccination in combination with oncolytic viruses offer opportunities to address resistance to checkpoint blockade therapy. Two cancer vaccines, the self-adjuvanting protein vaccine KISIMA, and the recombinant oncolytic vesicular stomatitis virus pseudotyped with LCMV-GP expressing tumor-associated antigens, termed VSV-GP-TAA, both show promise as a single agent. Here we find that, when given in a heterologous prime-boost regimen with an optimized schedule and route of administration, combining KISIMA and VSV-GP-TAA vaccinations induces better cancer immunity than individually. Using several mouse tumor models with varying degrees of susceptibility for viral replication, we find that priming with KISIMA-TAA followed by VSV-GP-TAA boost causes profound changes in the tumor microenvironment, and induces a large pool of poly-functional and persistent antigen-specific cytotoxic T cells in the periphery. Combining this heterologous vaccination with checkpoint blockade further improves therapeutic efficacy with long-term survival in the spectrum. Overall, heterologous vaccination with KISIMA and VSV-GP-TAA could sensitize non-inflamed tumors to checkpoint blockade therapy.

further searches what antigen Adpgk is, or what type of cell line TC-1 is.
2. For some figure panels, the groups being compared are unclear (including but not limited to Figs 2f,2g,S3f,g) Reviewer #2 (Remarks to the Author): This well-written manuscript from Das et al reports on a series of preclinical studies that builds on their prior work with a novel chimeric protein cancer vaccine (KISIMA) combined with systemic administration of an oncolytic virus vaccine that encodes the same antigens, in a prime:boost approach. The KISIMA vaccine incorporates a multi-antigenic domain (Mad) with epitopes for CD8 and CD4 T cells + cell penetrating peptide (CPP) + peptide agonist for TLR2 and TLR4 (TLRag). This is abbreviated as K. The viral construct is built on a vesicular stomatitis virus (VSV) variant encoding tumor antigen, and abbreviated as V. The investigators report on extensive preclinical studies in 3 different tumor models, using model tumor antigens, viral tumor antigens, and mutated cancer neoantigens. The prime:boost strategy with KISIMA prior to oncolytic virus (KV), especially with followed with K boost (KVK) induces strong and durable T cell responses in circulation and infiltration of those cells in the tumors. T cell function status is assessed by flow cytometry and gene expression (Nanostring) and reveals both effector and memory T cells, and suggests less exhausted T cells infiltrating tumors when using KVK than other strategies. The tumor control with KVK vaccines is disappointing (1 of 7 tumors with durable control in some experiments), but addition of PD1 blockade (which is ineffective as monotherapy in the models tested) induces durable tumor control in about half of the mice. They also show that IV administration of V is more effective than IM, which is another novel finding. The manuscript includes extensive data in the main figures as well as in supplemental material. The methods are sound and are well-controlled and comprehensive. The work has strong implications for design of optimal cancer vaccines in a range of cancers. There is strong translational potential and high impact across the field of cancer immunology and immunotherapy. The statistical analyses are appropriate and rigorous.
The only criticism is in Figure 4, where the immunohistology images (4d) show higher CD8 densities with KV than with VV, but the data in 4b and 4c suggest that the CD8 infiltrates are likely greater on average with VV. It would be good to modify so that the images in 4d are more representative of the summary data in 4b and 4c, or to explain why there may be a discrepancy.
One additional question is whether the investigators have compared intratumoral injection of V instead of the IV approach that they have used. In humans, oncolytic viruses have mostly been administered intratumorally. Being able to induce benefit with IV administration is appealing, but it may be good to comment on this in the manuscript and to highlight the advantages of being able to administer IV (for tumors that are not readily accessible percutaneously.

Referee response letter
We would like to thank the reviewers for the very helpful and constructive suggestions. We have thoroughly revised the manuscript, added new data where required and discussed pertinent issues that lacked clarity. A point-by-point response to all of the referee's comments is listed below.
The response letter has been structured the following way:

Reviewers comments in grey Responses in indented text
Revised manuscript text passages in italics

Reviewer #1 (Remarks to the Author):
This manuscript by Das et al. describes dual treatment of cancers with KISMA (a multidomain self-adjuvating vaccine) and VSV-GP (a modified oncolytic vesicular stomatitis virus expressing LCMV-GP) in increasing T cell specificity, increased cytokine production, tumor regression, and development of long-term immunity. While the experiments show impressive tumor rejections and increase in immunity, some major and minor concerns exist: We appreciate the constructive and insightful analysis of the reviewer. We have expanded the analysis of some included studies and added new results as well.
All comments and suggestions of this reviewer were addressed, as outlined below: Major Concerns: 1. Timelines, modes of treatment, and administration numbers vary throughout the studies. For example, some non-tumor studies use a Day 0 prime -Day 7 boost -Day 14 boost cycle with VSV given i.v. and KISMA s.c. with 3 total treatments. TC-1 tumor models used a Day 0 tumor graft -Day 7 prime -Day 14 boost model with 2 total treatments. OVA, MC-38, and subsequent TC-1 tumor models use a Day 0 tumor -Day 5 prime -aPD-1 twice -Day 7 boost -aPD-1 twice...repeat, --with variance between each cell line --and VSV given by i.t at times instead of i.v. What is the rationale for these changes between experiments and possible effect on outcomes?
We appreciate this comment and have added additional clarification to the manuscript. What may look like a very heterogenous and arbitrary interval setting between treatments actually follows a rather systematic approach. In the following paragraphs, we explain in detail the rationale behind the treatment regimen used. The administration schedule for the vaccination and checkpoint blockade was based on previous data, depending on the growth kinetics of the respective tumor model. The initial KISIMA vaccination schedule has been established in tumorfree animals, based on the frequency and the quality of the T cells response along with memory induction, as previously published (Belnoue E et al., JCI Insight. 2019 -cited in the manuscript). This schedule with 3 vaccinations 2 weeks apart followed by 3 monthly injections is currently investigated for ATP128 combined with anti-PD-1 blockade in the ongoing KISIMA-01 clinical trial (Phase 1b Study to Evaluate ATP128, With or Without BI 754091, in Patients With Stage IV Colorectal Cancer -ClinicalTrials.gov).
In various pre-clinical models, the vaccination schedule had been adapted to the fast growth of the mouse tumors to allow for enough time to mount an immune response. The schedule between 2 vaccinations was reduced to one week, and 4 administrations were performed as outlined in detail below. The anti-PD1 treatment was administrated twice a week as usually performed in many studies, and reduced to once a week for the TC-1 model. Regarding the administration of VSV, this was explored regarding scheduling and administration route. As shown in Figure 1, the i.v. route was found the best one after a KISIMA prime. Hence, one viral administration intravenously was performed for all the subsequent experiments.
For the TC-1 tumor model, prime vaccination was administered on day 7 after tumor implantation when palpable tumors were present. The following boost vaccinations were given on day 14, 28 and 49 after tumor implantation. This is consistent with the vaccination schedule applied for testing the immunogenicity of HPV vaccine in non-tumor bearing mice, as the interval between the prime and subsequent boost vaccinations was maintained (Fig 1 and suppl fig 2a- Regimen scheme for HPV vaccine / TC-1 tumor treatment In contrast, the MC-38 and the EG.7-OVA tumors have faster growth compared to the TC-1 tumor model. Hence, the interval between the prime and subsequent boosts was reduced to 7 days for both models. Importantly, the same interval was used for both tumor-bearing mice and immunogenicity studies as clarified in the figures below. The 4 th vaccination (or Boost 3) was not applied in tumor-free mice in Fig 1h and Fig 1a-g since these studies were performed to test the immunogenicity of the heterologous KVK vaccination against Adpgk (Fig 1h) and OVA (Fig 1a-g). In tumor-bearing mice, we applied a 3 rd boost in order to maintain the pool of antigen-specific CD8+ T cells induced by heterologous KVKK combination and prolong the therapeutic window.
Regimen scheme for Mad24 vaccine / MC-38 tumor treatment and OVA vaccine / EG.7 tumor treatment The following clarifications have been added to the immunization chapter of the Materials and Methods section:

We have added the following to Materials and Methods; page 25
Vaccination regimens were based on previously published homologous KISIMA vaccinations studies 4 . … … … Unless otherwise noted, vaccine treatment intervals in tumor bearing mice followed the application regimen from immunogenicity studies in non-tumor bearing mice for the respective vaccine combinations.
2. CD8+ T cell functionality is assessed by ex vivo restimulation of cells. Unfortunately, this afford cells (even functionally exhausted ones) an opportunity for additional synthetic restimulation after which they can produce cytokines ex vivo that would otherwise not be produced in vivo (particularly within a suppressive tumor microenvironment). Steps should be taken to assess actual capabilities of these cells (by staining for markers which do not require ex vivo restimulation, such as granzyme B and often IFN-gamma; or by direct analysis after tissue dissection/cell dissociation from animals that received protein export blockade [golgiplug/golgistop] in vivo a few hours prior to euthanasia).
Though ex vivo ELISPOT/ICS are a standard assay to monitor immune response/polyfunctionality in mice as well as in clinical trials, we appreciate this comment and the concern. We would like to emphasize why ex vivo ICS with 6h restimulation was applied in our studies. We also believe that heterologous KVK combination leading to enhanced functional and polyfunctional T cell pools is supported by corroborating data, as shown in details hereafter. First, regarding the experimental design. In this study we have demonstrated that TC-1 tumors are strongly infiltrated by CD8+ T cells after 1 st boost with VSV-GP-HPV in both homologous VV and heterologous vaccinated mice. This has been shown by transcriptome analysis (Fig 3E), flow cytometry analysis (4C) and immunohistochemistry (4D). Importantly, the tumors are infiltrated by both HPV-E7 (Fig 2D, E)

and VSV-N (Suppl fig 3 D, E) specific CD8+ T cells.
Performing intracellular cytokine staining for granzyme B or IFN-γ directly without ex vivo restimulation would not allow us to distinguish specificity of intra-tumoral CD8+ T cells, which is why we turned to ex vivo restimulation. While ex vivo re-stimulation can induce exhausted T cells to express IFNγ and TNFα, this effect is only partial and can't restore full T cell functionality compared to non-exhausted T cells (see Sandu et al., Nature Communication, 2020; doi: 10.1038/s41467-020-18256-4). In addition, analyzing the transcriptional levels of freshly harvested tumor tissue from mice treated with different regimen showed higher levels of Ifnγg, Tnfa and Gzmb in KV treated TC-1 tumors compared to VV treated tumors (Fig 3E, F and additional violin plots below). Another molecule associated with cytotoxicity, prf1 (perforin) was also upregulated after KV vaccination compared to VV vaccination ( Fig 1E and additional violin plots). Of note, the violin blots were derived from the same data set already presented as a heatmap in the manuscript (Fig 3); to avoid duplicating the data presentation this blot is not included in the revised manuscript but is rather used here to highlight the upregulation of functionality-associated gene signatures.
Gene expression levels of markers of T cell functionality using nanostring transcriptome analysis of freshly isolated tumor tissue Although the suggested in vivo golgiplug treatment experiment sounds indeed quite interesting, we were and still are limited by our animal experimentation license. As a license amendment would come with several months processing time, we opted for an ex vivo treatment of TILs with golgiplug in absence of other restimulation (see graphs below), and found that about 80% of tumor infiltrating CD8 T cells in KV treated mice expressed granzyme B (see below).
Finally, to corroborate the transcriptome analysis discussed above, we also added a new study focusing on the protein levels of T cell functionality markers using Luminex assays. Luminex results did not show differences between KV and VV and small difference vs mock, while we do see higher IFN-gamma and TNFalpha content in KV vs. Mock group (Figure below). These data underline the importance to assess functionality markers in an antigen-specific manner when comparing KV and VV.
We have modified and added the following to Discussion; page 20 Although homologous VSV-GP-TAA also showed efficacy in some models, tumor control was limited. The combination with KISIMA-TAA prime strongly enhanced the proportion of memory T cells, and tumor relapse was delayed. An additional key finding of our heterologous combination was the increase in polyfunctional CD8+ T cells, including IFN-γ, TNF-α, and CD107a as shown with ex vivo intracellular staining. Importantly, the transcriptome analysis of freshly isolated tumor tissue corroborates the notion that heterologous KV vaccination profoundly enhances T cell activity and functionality.

3.
There appear to be some inconsistencies amongst the data that need further investigation and/or consideration. It is mentioned that reduction in viral-specific T cells occurs (in a non-tumor model when compared to OVA) when either KVV or KVK are used (SFig 2i-k). However, in a tumor model (SFig 3b) KV priming appears to increase virus-specific T cells over VV. Also, in Figure 4c it appears that VV dramatically increases T cell infiltrates over KV in flow data, which is shown to be the opposite in 4d histology.
We appreciate these comments and hope to provide helpful clarification here as well as in the manuscript. We have reported the kinetics of the anti-viral T cell response in tumor-bearing mice for 3 different tumor models in supplementary fig  7 B,C, F, G, J, K. It is true that we consistently observed a strong VSV-N specific CD8+ T cell response in the heterologous KVKK group after the boost with the respective VSV-GP-TAA. This is consistent with the results shown in Suppl fig 3b as this analysis was performed after the boost with VSV-GP-HPV. However, longterm (after the 2 nd and 3 rd boost with the respective KISIMA-TAA) we observed that the proportion of anti-viral CD8+ T cell reduced for all 3 different tumor models analyzed. This is consistent with the data shown in suppl fig 2i-k since this analysis was performed after 2 nd boost. It is important to note that the overall reduction of viral-specific T cells in heterologous combination is also evident in enhanced ratio of anti-tumor and anti-viral CD8+ T cells in heterologous KVKK (+/-checkpoint blockade) compared to homologous VVVV vaccination. We generated three additional analysis graphs for each of the applied vaccination models, displaying the ratio of anti-tumor vs antiviral T-cell responses (see figure below). These graphs have been added to supplemental Figure 7. We have rephrased the respective passages in the mansucript and highlighted the KVK effect on the ratio rather than on antiviral immune responses per se.

Ratio of anti-tumor vs antiviral T-cell responses. Dotted lines indicate time of immunization.
We have added the following to Results; page 17 Surprisingly, circulating anti-viral CD8 + T cells were unaffected by checkpoint inhibition in all three tumor models (Supplementary Fig. 7b,c,g,h,l,m)

but the ratio of anti-tumor to anti-viral CD8 + T cells in circulation was not greatly enhanced by combining checkpoint blockade antibodies with KVK vaccination (Supplementary fig 7d, i, n).
In response to the discrepancy on display between TILs composition and TC-1 tumor histology in Figure 4c vs 4d, we have to admit that an error occurred in the final figure composition. In the flow cytometry TILs data set, KV and VV sets were mixed up. The KV set depicts the data from the VV group and vice versa. This mistake happened while rebuilding the final submission figure to comply with the required RGB style. The figure had originally been composed as a CMYK file.
To substantiate the nature of this mistake, we provide the original previous figure draft (see below) as a file with a date from Nov 23 rd (manuscript submission was on Dec 9 th ) as well as the original Graphpad Prism file from Sept. 9 th . These files have been uploaded to the submission system and also provided as a zip file (with original date stamp) to the editor. Figure 4 c has been corrected accordingly. We would like to thank the reviewer for the recent references and comments, we have re-formulated the following sections in the discussion. Whereas it is established that "turning a cold tumor hot" supports checkpoint inhibition therapy and that oncolytic virus are promoting these changes in the tumor microenvironment as TLR agonists do, we would like to strengthen the importance of balancing antigen specific T cells and bystander T cells. Indeed, removing the break with anti-PD-1 would require the presence of killer T cells able to specifically recognize the tumor cells for clinical efficacy. Hence, in the heterologous prime-boost setting, antigen specific T cells are significantly amplified while innate immunity and anti-viral T cells are contributing to the immune-promoting changes in the tumor microenvironment. By limiting the injection of the VSV to one administration, the balance is tilting toward antigen specific T cells as shown above in figure in response to comment #3. In addition to these new graphs included in supplemental Figure 7, we modified the respective section in the Introduction, including incorporation of some of the suggested references. Although the discussion already addresses the potentially beneficial role associated with OV-triggered antiviral immune responses (such as discussion page 20 below), a statement was added in the introduction page (as shown below).

Addition to the Introduction; page 4
One potential limitation is inherently linked to the strong immune activation that comes with two dominant antiviral forces, an initial innate and a subsequent adaptive response 13, 14 , although these very same mechanisms may also counter tumor associated immune suppression 15,16 . Arming OVs with antigens associated with the tumor can additionally enhance the tumor-specific T cell portion and therefore positively affect the balance of antitumor versus antiviral immune responses 17 .
Statistical analysis has been completed. The respective sections with additional analyses have been marked in the manuscript: We added information on the number of experimental repetitions to each figure. In brief: for our in vivo experiments, heterologous KISIMA prime and VSV-GP-TAA vaccination was performed in tumor-bearing mice more than once. In vivo study repetitions would also include additional experimental groups, such as immune checkpoint combinations. These additional groups were not always subject to repetition. A detailed description of studies being performed once or multiple times has been added to each figure legend.

7.
Many of the treatment regimens appear to start before tumors are well established, and a prophylactic rationale for such studies is unclear.
As mentioned before (comment 1), the tumor rapid growth of some models allows for a limited therapeutic window in order to mount a potent adaptive immunity. Although the administration of the first vaccination was initiated early in some models (e.g. D3 for the MC38), the vaccination was always performed post-tumor challenge and can not be strictly considered as prophylactic. Prophylactic vaccinations are usually performed D-21 and D-7 before the tumor challenge. It shall be noted that for TC-1 (Vac1 performed at d7) and B16-Ova (Vac1 performed at d5), the tumors were already palpable. 2. For some figure panels, the groups being compared are unclear (including but not limited to Figs 2f, 2g, S3f,g) We have revised the figures accordingly to clearly indicate the groups compared. See detailed listing of modified graphs in comment to point # 5

Reviewer #2 (Remarks to the Author):
This well-written manuscript from Das et al reports on a series of preclinical studies that builds on their prior work with a novel chimeric protein cancer vaccine (KISIMA) combined with systemic administration of an oncolytic virus vaccine that encodes the same antigens, in a prime:boost approach. The KISIMA vaccine incorporates a multiantigenic domain (Mad) with epitopes for CD8 and CD4 T cells + cell penetrating peptide (CPP) + peptide agonist for TLR2 and TLR4 (TLRag). This is abbreviated as K. The viral construct is built on a vesicular stomatitis virus (VSV) variant encoding tumor antigen, and abbreviated as V. The investigators report on extensive preclinical studies in 3 different tumor models, using model tumor antigens, viral tumor antigens, and mutated cancer neoantigens. The prime:boost strategy with KISIMA prior to oncolytic virus (KV), especially with followed with K boost (KVK) induces strong and durable T cell responses in circulation and infiltration of those cells in the tumors. T cell function status is assessed by flow cytometry and gene expression (Nanostring) and reveals both effector and memory T cells, and suggests less exhausted T cells infiltrating tumors when using KVK than other strategies. The tumor control with KVK vaccines is disappointing (1 of 7 tumors with durable control in some experiments), but addition of PD1 blockade (which is ineffective as monotherapy in the models tested) induces durable tumor control in about half of the mice. They also show that IV administration of V is more effective than IM, which is another novel finding. The manuscript includes extensive data in the main figures as well as in supplemental material. The methods are sound and are wellcontrolled and comprehensive. The work has strong implications for design of optimal cancer vaccines in a range of cancers. There is strong translational potential and high impact across the field of cancer immunology and immunotherapy. The statistical analyses are appropriate and rigorous. We thank the referee for the positive notion and supportive comments. We have addressed the points raised below and in the manuscript.
The only criticism is in Figure 4, where the immunohistology images (4d) show higher CD8 densities with KV than with VV, but the data in 4b and 4c suggest that the CD8 infiltrates are likely greater on average with VV. It would be good to modify so that the images in 4d are more representative of the summary data in 4b and 4c, or to explain why there may be a discrepancy.
As addressed above for reviewer 1, point 3: In response to the discrepancy on display between TILs composition and TC-1 tumor histology in Figure 4c vs 4d, we have to admit that an error occurred in the final figure composition. In the flow cytometry TILs data set, KV and VV sets were mixed up. The KV set depicts the data from the VV group and vice versa. This mistake happened while rebuilding the final submission figure to comply with the required RGB style. The figure had originally been composed as a CMYK file.
To substantiate the nature of this mistake, we provide the original previous figure draft (see below) as a file with a date from Nov 23 rd (manuscript submission was on Dec 9 th ) as well as the original Graphpad Prism file from Sept. 9 th . These files have been uploaded to the submission system and also provided as a zip file (with original date stamp) to the editor. Figure 4 c has been corrected accordingly.
One additional question is whether the investigators have compared intratumoral injection of V instead of the IV approach that they have used. In humans, oncolytic viruses have mostly been administered intratumorally. Being able to induce benefit with IV administration is appealing, but it may be good to comment on this in the manuscript and to highlight the advantages of being able to administer IV (for tumors that are not readily accessible percutaneously. Although VSV-GP initial studies were performed in part with intra-tumoral injection in oncolytic settings without tumor-antigen carrying VSV-GP (Urbiola et al., 2018;Schreiber et al., 2016;Koske et al., 2019), other routes of administration (i.m, s.c. and i.v as shown in Figure 1) were privileged for the development of the heterologous prime-boost toward clinical trial. Even if the approach supporting intra-tumoral injections of oncolytic virus was developed toward accessible tumors like melanoma and head neck as well as administration to liver metastasis, it also reduces the number of possible targeted indications along with narrowing down the number of patients that could be included in the trial. Indeed, the process of intra-tumoral injections for non-superficial tumors is heavy and can be painful for the patients. Hence, the intravenous route was preferred for the virus, which is not only an oncolytic vector in this setting but also vaccine boost.