Genetically stable poliovirus vectors activate dendritic cells and prime antitumor CD8 T cell immunity

Viruses naturally engage innate immunity, induce antigen presentation, and mediate CD8 T cell priming against foreign antigens. Polioviruses can provide a context optimal for generating antigen-specific CD8 T cells, as they have natural tropism for dendritic cells, preeminent inducers of CD8 T cell immunity; elicit Th1-promoting inflammation; and lack interference with innate or adaptive immunity. However, notorious genetic instability and underlying neuropathogenicity has hampered poliovirus-based vector applications. Here we devised a strategy based on the polio:rhinovirus chimera PVSRIPO, devoid of viral neuropathogenicity after intracerebral inoculation in human subjects, for stable expression of exogenous antigens. PVSRIPO vectors infect, activate, and induce epitope presentation in DCs in vitro; they recruit and activate DCs with Th1-dominant cytokine profiles at the injection site in vivo. They efficiently prime tumor antigen-specific CD8 T cells in vivo, induce CD8 T cell migration to the tumor site, delay tumor growth and enhance survival in murine tumor models.

In addition, the authors designed and implemented various in vitro and in vivo experiments to demonstrate that type I IFN was induced by the engineered virus, antigen presentation occurred, and that tumor antigen-specific CD8 T cells were induced and activated by treatment with the antigenexpressing virus. Finally, the authors demonstrate that CD8 T cells migrated into the tumor site, and that this infiltration was associated with a reduction in tumor burden and an increase in animal survival in one mouse model. Generalizability: Given the limited scope of tumor models explored, and limited foreign antigen transgene constructs evaluated, the likely generalizability of this approach to other systems and antigens cannot be assessed.

Technical questions & comments:
While the authors have shown that there was genetic stability preserved in the foreign insert after 20 passages of virus in cells in culture, they did not comment on the stability in the rest of the viral genome. Did they sequence the whole viral genome to rule out any potential change in other regions of the genome that were not around the locus of the foreign insert? This could have significant safety implications.
The authors should also consider evaluating this system in additional tumor models, and with additional antigen transgene constructs to demonstrate the likely generalizability of this approach to other systems and antigens.
In addition, it is likely that the oncolytic effect of this virus will be reduced by increasing its immunogenicity. This question should be addressed by evaluating the relative efficacy of a tumor antigen expressing PVSRIPO in a tumor model demonstrating potent oncolytic virus replication and efficacy.
Reviewer #3, expert in clinical glioma immunotherapy (Remarks to the Author): In this manuscript, Mubeen et al. designed a poliovirus/rhinovirus chimera-based vector, named mOVA2, which infects dendritic cells(DCs) such that antigen peptide (epitope) carried by the virus can be expressed by DCs, followed by antigen presentation to cytotoxic CD8+ T cells. mOVA2 overcomes the neuropathogenicity and genetic instability originated from the poliovirus, while maintaining its tropism to DCs. The authors demonstrated that mOVA2 infects, activates and induces epitope presentation in mouse bone marrow derived dendritic cells (BMDCs) and the following activation of CD8+ T cells in vitro. In addition, mOVA2 recruits and activates DCs, and triggers antigen-specific T cell activation in vivo. Using a murine model of melanoma, they showed that mOVA2 recruits CD8+ T cells to the tumor, reduces tumor size and increases survival. As a proof of concept experiment, Mubeen et al. engineered the virus [named RIPO(H3.3)] to express epitope containing H3.3-H27M, a signature mutation found in patients with diffuse intrinsic pontine glioma (DIPG). The authors showed that RIPO(H3.3) activates human BMDCs and triggers responses from H3.3-K27M-TCR+, CD8+ Jurkat cells. RIPO(H3.3) has a good translational value given that intratumoral infusion of PVSRIPO virus in patients with recurrent grade IV malignant glioma confirmed absence of neurovirulence and increased survival in past studies.
The study is of great potential clinical and scientific significance. However, the writing is so filled with jargon and lacking in crucial details that it is very difficult to follow the manuscript. The main concern is that the authors need to experimentally validate that this viro-immunotherapy shows preclinical efficacy in a valid experimental model of DIPG in vivo. I am unclear what the B16F10.9-OVA tumor model really is, but I do not believe it is a DIPG model.
Please see the following specific suggestions: 1. Fig. 2C: a lysate immunoblot showing the total Flag-tagged proteins should be provided for the IP experiment. In addition, what are the reasons for disappearance of the three flag-tagged bands at 6 and 8 hpi? 2. Fig. 3E: what are the results when the same experiments are done in FLT3L BMDCs? 3. In addition to eIF4G1 cleavage, are proliferation, cell death, and senescence affected in the infected BMDCs? 4. There is no reference or introduction for the B16F10.9-OVA melanoma tumor model. How is this relevant to H3.3K27M+ glioma? 5. Fig. 7F: what are the levels of Granzyme B and IFN-gamma in this co-culture system? 6. Is it possible to set up a co-culture system of human BMDC, CD8+ T cells and HLA-A2+ DIPG cells to demonstrate the efficacy of RIPO(H3.3)? 7. A T cell migration assay should be performed with conditioned medium collected from RIPO(H3.3)infected DCs to demonstrate effective T cell recruitment. 8. It would be nice if the virology terms (e.g. the processing of VP0, VP3 and VP1 and P2/2BC/2C) are explained for readers not in this field. 9. Sup Fig.1: please show experimental data demonstrating the viability of each vector.

Minor comments:
The statement "Unlike adult GBM, which is extremely heterogeneous, ~80% of Diffuse Intrinsic Pontine Gliomas (DIPG) and ~20% of pediatric GBMs carry a homogenously expressed driver mutation in histone 3.3 [H3.3K27M]12, 13" is neither complete nor entirely accurate. While it is true that 80% of DIPGs express H3K27M mutation, hemispheric pediatric GBMs express a different histone mutation in ~20% cases, H3G34R/V. Furthermore, other midline gliomas such as thalamic and spinal cord, also express the H3K27M mutation. This entity is now classified as H3K27M+ diffuse midline glioma (DMG) by the WHO and this formal reclassification would be referenced.

Reviewer #4, expert in preclinical development of oncolytic virus (Remarks to the Author):
This is a clearly written and presented manuscript that argues that a picornavirus vector created by the authors can be used as a T cell vaccine specifically for use in cancer indications. The authors argue that while in general picornavirus vectors are not useful for expressing transgenes due to inherent instability, their design of incorporating a peptide sequence into a critical regulatory region of the virus will favour retention of transgenic sequences. This is a strategy that this same group published on in 2003 (Dobrikova et al). They present very nice data that shows their clinically validated vector can stimulate respectable T cell responses against encoded peptides. Critique (1) the authors do not really provide any detailed discussion comparing and contrasting the results from this paper and their 2003 work. Is this an incremental improvement compared to what they have done previously or a significant improvement and why? (2) The observation that the SINNFEKL peptide encoding sequence evolved over time suggests that there may be significant restrictions on what can be encoded...this should be discussed.
(3) the virus cannot encode more than one peptide? Is this the case? These types of limitations of the technology should be discussed. (4) The authors present data that shows about 0.5% of T cells recognize the encoded peptide after PVRIPO immunization. Can this be boosted with more administrations. Do the authors feel this is a clinically relevant level of T cell stimulation and if so why? (5) Most of the population has already been vaccinated against polio, will this impact the ability of this vector to initiate immune responses?

21
The reviewer is right. We have added a complete polyprotein map, showing the proteolytic cleavages in the viral polyprotein and explaining the identity of the various viral products referenced in the manuscript for the readers' orientation (revised Figure 2a).

Reviewers' comments:
Reviewer #1 (Remarks to the Author): Mosaheb et al. have provided a revised version of their manuscript. The revisions, experiments and explanations have been extensive and they should be commended for the tremendous amount of work and thought that went into this. The most relevant revision is that they know have generated the tools and reagents needed to address their hypothesis. Although title and abstract focus on the technical aspects of having engineered a strategy to express foreign antigens in polio, the proof-of-principle experiments are based on the Ova model and then on physiologically relevant H3.3 K27M antigen expressed in DIPG. The technical aspects related to stable expression of peptide antigens (figures 1 and 2) are better explained and more robust. Figure 3, Figure 4a-c, Figure 6 show the immune response in the Ova model. The more physiologic relevant experiments are in Figure 4d, 5e, f, figure 7, 8 which show the transgenic mouse model and the immune experiments related to this model. These experiments clearly make this paper more robust than the previous version. My major comments are: 1-The back and forth between OVA and H3.3 is very confusing. Since the most relevant part here is H3.3 (the physiologically relevant target), please consider placing the less interesting OVA experiments in the supplementary data 2- Figure 2e: the asterisks are confusing and the band that is meant to represent H3.3. are also not clear. I understand that there is time-related processing but it takes quite some time for a reader to look at these blots and try to figure which bands are H3. The authors have improved the quality of the manuscript. In addition to the in vitro data from the previous version, they developed an in vivo model to address the questions that were asked by the different reviewers. In terms of novelty they argue that their virus can be differentiated from others, raising points in terms of stimulating antigen presentation and T-cell co-stimulation instead of repressing as might occur with other DNA viruses. They do present mechanistic data to support their claims that PVSRIPO-based vectors achieve mainly 3 things: 1) antigen expression, 2) type I/III-dominant pro-inflammatory signaling stimulation, and 3) maturation marker induction in infected dendritic cells.
They have presented new data in this submission, including specifically for this vector to be used as a platform to express any foreign transgene antigen sequence. This may allow the use of this vector for the vaccination against other tumor types in humans, and hence this may be more generalizable than previously noted. The authors have not generated data from new animal models for each antigen that they propose to test.
Reviewer #3 (Remarks to the Author): The manuscript is improved and the authors have answered many of my concerns. However, the writing still needs work. The authors should endeavor to make this easier to read, trying wherever possible to avoid "alphabet soup".
Reviewer #4 (Remarks to the Author): The authors have gone to great lengths to address all of the reviewers concerns including a considerable amount of new data. I think they provide excellent and well controlled experiments. While I agree that the immune responses they are seeing in their mouse experiments are comparable to immune responses seen in clinical studies (e.g. they cite the work from Ugur Sahin and colleagues), they are assessing responses in mice (as they acknowledge) which in general notoriously overestimate responses that are actually found in humans. None-the-less an interesting application of their technology and I believe it merits testing in the clinical setting to understand its full potential. John Bell

Reviewer #1 (Remarks to the Author): Mosaheb et al. have provided a revised version of their manuscript. The revisions, experiments and explanations have been extensive and they should be commended for the tremendous amount of work and thought that went into this. The most relevant revision is that they know have generated the tools and reagents needed to address their hypothesis.
Although title and abstract focus on the technical aspects of having engineered a strategy to express foreign antigens in polio, the proof-of-principle experiments are based on the Ova model and then on physiologically relevant H3.3 K27M antigen expressed in DIPG. The technical aspects related to stable expression of peptide antigens (figures 1 and 2) are better explained and more robust. Figure 3, Figure 4a-c, Figure 6 show the immune response in the Ova model. The more physiologic relevant experiments are in Figure 4d, 5e, f, figure 7, 8 which show the transgenic mouse model and the immune experiments related to this model. These experiments clearly make this paper more robust than the previous version.

My major comments are: 1-The back and forth between OVA and H3.3 is very confusing. Since the most relevant part here is H3.3 (the physiologically relevant target), please consider placing the less interesting OVA experiments in the supplementary data
We have followed this suggestion and moved non-essential information relating to the mOVA2 vector to the Supplement (see comment #2, pg. 5-6).
However, we feel that principal OVA data, reported in Figs. 3-6, must remain in the main body of the manuscript. We describe a new technology, slated for first-in-man investigation, with claims of mechanistic superiority over conventional approaches. Empirical rigor demands that our claims are supported by investigations in standardized models with validated epitopes, without interference from host-, or target epitope-specific variables. We used rigorous assays based on standardized OVA tools (e.g. OT-1 CD8 T cells, H2Kb-SIINFEKL pentamer, etc.), which will enable readers to better gauge the merits of our technology.
H3.3 K27M is not a mouse epitope and the AAD/CD155 double-transgenic mouse model we must use to enable presentation of an HLA-A2 epitope in mice is somewhat contrived (it has an artificial, chimeric mouse-human MHC). Because of its spontaneous origin and relatively low intrinsic immunogenicity, B16F10 arguably is the most respected immunocompetent rodent tumor model. For example, it was instrumental in establishing the paradigm-shift of immune checkpoint blockade. 16 mOVA2 immunization experiments against B16F10-OVA involve the endogenous mouse MHC I (H2Kb) and demonstrate the mechanisms of our vector approach without interference from host-specific confounders. Figure 2e: the asterisks are confusing and the band that is meant to represent H3.3. are also not clear. I understand that there is time-related processing but it takes quite some time for a reader to look at these blots and try to figure which bands are H3.3

2-
The reviewer is right, the figure was difficult to read. The complicated depiction was a result of including data of divergent assays (immunoprecipitation followed by immunoblot, vs. 'direct' immunoblot), involving two different vector constructs, in the same figure.
We now moved the description of foreign epitope expression with mOVA2 to the new Supplementary Figure 5. This yielded a much simpler, easier to read revised Figure 2 in the main manuscript, which exclusively focuses on RIPO(H3.3) and the expression of the H3.3 K27M epitope. We now clearly labeled the viral fusion-polypeptides including the H3.3 K27M signature (see revised Figure 2b).

3-All in vitro experiments are conducted at a MOI of 10. IS this in vitro dose physiologically relevant to what is administered in vivo?
This is a valid point. Poliovirus infections in vitro are naturally inefficient, as 90% of input virus is 'sloughed off' from cells upon CD155 receptor interaction as 135S particles that do not enter host cells. 17 Thus, in vitro, only ~10% of input poliovirus successfully completes the entry process and initiates infection. Therefore, the 'actual' MOI with poliovirus is far lower than the input virus.

It is unknown if such sloughing off occurs in vivo.
To account for the sloughing phenomenon, in vitro studies with poliovirus classically are conducted with MOIs of 10, as to reach a 'true' MOI of 1 with the objective to target every cell in the culture. Synchronized infection at an MOI of 10 (corresponding to a 'true' MOI of 1) has been the mainstay of poliovirus research for >50 years, as it provides a rigorous empirical framework for deciphering virus:host relations during the infection process.
Our studies are geared to achieve fundamental mechanistic insight into virus/vector:host interactions in human and mouse dendritic cells. Infections at MOIs <10 would yield samples where only a subset of cells are infected. This would obfuscate the many immunoblot and flow cytometry assays we have conducted and greatly complicate the interpretation of our results.
Based on the experience with PVSRIPO in human subjects, 18 we anticipate a clinical dose in the range of ~10^7-10^8 TCID for intramuscular administration of RIPO(H3. 3). An empirical MOI of 10 in our mechanistic in vitro studies is consistent with this anticipated clinical dose range.
We have previously characterized the DC phenotype of the vector parent (PVSRIPO) at an MOI of 1 and observed similar sub-lethal infection with type I interferon-dominant stimulation as the data reported in the present study. 19 We also point the reviewer to the in vivo assays of PVSRIPO vector DC infection in our report (see Figure 4). I.M. immunization leads to loco-regional induction of CCL2, CCL5, CXCL1 and CXCL10 in vivo (Figure 4a). CCL2 (MCP-1) and CCL5 (RANTES) are potent DC chemokines, 20,21 inducing their migration to the immunization site. Accordingly, flow cytometry analyses of the immunization site demonstrated an influx of CD11c+ and CD11b+ cells, suggesting migration of antigen presenting cells (DCs) to the site of immunization (Figure 4b). Once they migrate to the immunization site, such DCs would become targets for PVSRIPO vector infection. This is indeed reflected in our studies, as locoregional DCs induced the CD40/CD86 maturation markers (Figure 4c). Therefore, hallmarks of PVSRIPO vector infection of DCs at an MOI of 10 in vitro, also were observed in vivo. Figure 4d-This is important data related to immunization. A lymph node with H3.3 positive IHC is shown. The authors say in the legend that these are APCs but there is no evidence of this. Staining for CD11c is also shown but it is not clear if these are the same cells as the H3.3 positive cells. There is also no quantitative analyses of number of cells that are H3.3 positive and number of draining Lymph Nodes assayed when compared to controls of SUppl. Figure 11, 12. The reviewer is correct. Mouse popliteal lymph nodes are minuscule -we were not able to collect consecutive sections (7-micron thickness) showing the same germinal center (containing the H3.3 K27M + cells) for CD11c co-staining.

4-
To account for the Reviewer's concern, we removed the CD11c staining image because we cannot demonstrate co-staining for H3.3 K27M and CD11c on the same cells (see revised Figure 4). We also tempered our description and simply state that 'RIPO(H3.3)-mediated H3.3 K27M expression occurs in the draining lymph node' (see revised Results, pg. 11). We do not claim that H3.3 K27M positive cells are DCs in our paper.
Our descriptive immunohistochemical (IHC) analyses detecting H3.3 K27M -positive cells in popliteal lymph nodes after i.m. immunization, do not provide definitive evidence that such cells (presumably antigen presenting cells) are driving H3.3 K27M -directed CD8 T cell responses. However, our IHC results illustrate that RIPO(H3.3)-infected cells drain to local lymph nodes and, thus, corroborate important mechanistic in vitro analyses. For example, RIPO(H3.3) infection of human DCs in vitro strongly induced CCR7 (Fig. 8c), a mediator of lymph node migration in DCs.
Staining of mock lymph nodes ( Supplementary Fig. 12) and isotype control staining ( Supplementary Fig. 13) show that the H3.3 K27M IHC staining in lymph nodes of RIPO(H3.3)-immunized mice is specific. Regarding the number of cells in lymph nodes expressing H3.3 K27M : lymph nodes consist primarily of T and B cells that poliovirus cannot infect and, thus, should not express the H3.3 K27M epitope. DCs are extremely rare in vivo; peripheral DCs migrate to the draining lymph node after activation [e.g. after infection with mRIPO(H3.3)] and CCR7 upregulation (Fig. 8c) to present antigen to T cells (please see comprehensive reviews for detail 22,23 ). Because DCs are exceedingly rare in vivo, the distribution and extent of H3.3 K27M staining in a 7-micron section of a mouse popliteal lymph node is expected. While DCs are rare, they are extremely potent inducers of T cell responses; a single DC can stimulate thousands of T cells 24,25 . It has been estimated that only 85 antigen-presenting DCs are required to elicit a T cell response in humans 24 .
DC-mediated CD8 T cell responses are not necessarily dependent on the frequency of activated, epitope-presenting DCs in lymph nodes, but are contingent on the quality of proinflammatory stimulation and activation provided by the context of antigen uptake and presentation. We have compelling evidence buttressed by multiple state-of-the-art complimentary assays, that PVSRIPO vectored epitope delivery induces antigen-specific CD8 T cell responses. Therefore, we have not endeavored to quantify the presence of H3.3 K27M -positive cells in lymph nodes. While comparison of our PVSRIPO vector approach with conventional peptide + adjuvant regimens in rodent tumor models may appear sensible, such assays are not feasible or advisable for two main reasons:

5-The
First, such efforts are rendered almost impossible due to the dizzying variety of adjuvant regimens and administration methods/schedules that are being employed in conjunction with peptide immunization in the clinic. To illustrate this point, we provide details on the adjuvants used for recent peptide immunization trials in malignant glioma patients: Since there is no agreement on prioritizing adjuvant strategies and there are no benchmark immune monitoring standards guiding their use, it is impossible to define the correct regimen to which PVSRIPO should be compared. We often use high molecular weight poly(I:C) in our studies as a mechanistic comparator to PVSRIPO, because it resembles viral dsRNA as an innate immune trigger.
Second, the record of rodent tumor models in predicting immunotherapy efficacy in the clinic is dismal (see our detailed response to comment #10 of Reviewer 1 in the prior rebuttal letter). We kindly point out to the reviewer that our stance on the utility of rodent tumor models in predicting clinical outcomes is shared by Reviewer 4 (see pg. 7-8).
Ample precedent teaches that comparing PVSRIPO to conventional approaches in such models will almost certainly prevent reaching empirical conclusions that hold up in the clinic. A much better approach towards rigorous, robust preclinical analysis are mechanistic studies in relevant (human) model systems. This is the purpose of our study.

Reviewer #2 (Remarks to the Author):
The authors have improved the quality of the manuscript. In addition to the in vitro data from the previous version, they developed an in vivo model to address the questions that were asked by the different reviewers. In terms of novelty they argue that their virus can be differentiated from others, raising points in terms of stimulating antigen presentation and T-cell co-stimulation instead of repressing as might occur with other DNA viruses. They do present mechanistic data to support their claims that PVSRIPO-based vectors achieve mainly 3 things: 1) antigen expression, 2) type I/III-dominant proinflammatory signaling stimulation, and 3) maturation marker induction in infected dendritic cells. They have presented new data in this submission, including specifically for this vector to be used as a platform to express any foreign transgene antigen sequence. This may allow the use of this vector for the vaccination against other tumor types in humans, and hence this may be more generalizable than previously noted. The authors have not generated data from new animal models for each antigen that they propose to test.
The SIV and IDH2-epitope expressing vectors were included to address a prior review comment that documenting the specific mOVA2 and RIPO(H3.3) constructs did not properly support the claim of generalizability of our vector design (see our response to Reviewer 2, pg. 17-18 in the prior rebuttal letter). We added the additional vector prototypes in Supplementary Figure 3 to demonstrate that our vector design is agnostic to specific IRES insert sequences. These vector prototypes represent proof-of-principle constructs that were evaluated for genetic stability in vitro only (we did not propose to advance these designs to animal testing).
Testing the vector prototypes shown in Supplementary Figure 3 in in vivo immunization studies is outside the scope of this report and would be a massive endeavor, since none of the target epitopes are predicted to bind to murine MHC Class I.