Reversion analysis reveals the in vivo immunogenicity of a poorly MHC I-binding cancer neoepitope

High-affinity MHC I-peptide interactions are considered essential for immunogenicity. However, some neo-epitopes with low affinity for MHC I have been reported to elicit CD8 T cell dependent tumor rejection in immunization-challenge studies. Here we show in a mouse model that a neo-epitope that poorly binds to MHC I is able to enhance the immunogenicity of a tumor in the absence of immunization. Fibrosarcoma cells with a naturally occurring mutation are edited to their wild type counterpart; the mutation is then re-introduced in order to obtain a cell line that is genetically identical to the wild type except for the neo-epitope-encoding mutation. Upon transplantation into syngeneic mice, all three cell lines form tumors that are infiltrated with activated T cells. However, lymphocytes from the two tumors that harbor the mutation show significantly stronger transcriptional signatures of cytotoxicity and TCR engagement, and induce greater breadth of TCR reactivity than those of the wild type tumors. Structural modeling of the neo-epitope peptide/MHC I pairs suggests increased hydrophobicity of the neo-epitope surface, consistent with higher TCR reactivity. These results confirm the in vivo immunogenicity of low affinity or ‘non-binding’ epitopes that do not follow the canonical concept of MHC I-peptide recognition.

General Summary: The authors assess the impact of low binding affinity neoepitopes on the natural immunogenicity of tumor. Current MHC-I binding affinity algorithms are being used to predict CD8+ immunogenicity with the assumption that a 'high-binder' is more likely to bind to MHC-1 and elicit CD8+ T cell response and thus result in greater anti-tumor immunity. However, studies have now suggested that low affinity binding neoantigens also elicit immune responses and an anti-cancer response. Similarly, the human trials that have based vaccine design on MHC-I binding affinity have not resulted in robust CD8+ T cell responses that are essential for anti-tumor activity. The authors perform both vaccination experiments and CRISPR to show that both vaccination against and expression (and lack thereof) of low binding neoantigens significantly impacts the overall immunogenicity of the tumor microenvironment. The authors use CRISPR to reverse expression of a pre-specified low-affinity neoantigen to WT sequence which reverses the immunogenic phenotype. Conversely, when the neoantigen is re-introduced, the immunogenic transcriptomic signature is restored. They also perform structural modeling to show that the MHC-neoepitope complex of these low-binding epitopes have increased hydrophobicity consistent leading to TCR reactivity.
Major Points: Overall, this study is timely and is of interest to the field. These data underscore the idea that low affinity binding neoantigens should still be evaluated for immunogenicity and do impact the overall composition of the TME. These data are also important to address optimal neoantigen-targeted vaccine studies. Although the studies support the hypothesis that low-binding neoantigens are a source of immunogenicity and impact the TME, further studies to evaluate mechanism should performed. Additionally, evaluation and characterization of neoantigen-specific T cells should be performed and ideally functional experiments should be carried out to directly demonstrate cytotoxic function of the low-binding neoantigen T cells.
Specific Points: 1. In the set of experiments 1a-1b, DCs pulsed with WT sequence should also be used as an additional control. Cross-reactivity to WT sequence should be tested by pulsing DCs with WT sequence as well, although we assume that WT is non-immunogenic. This is later done in fig 2a but using a peptide vaccine approach (details should be provided as to the consistency of vaccineis an adjuvant used? 2. It does look like there is some minimal anti-tumor activity of WT sequence vaccine in Fig 2a. ELISPOT assays could be used to measure immunogenicity of both the mutated and WT in conjunction with the tumor response curves. 3. There is no reference to a figure for the depletion study mentioned on lines 157-168. Additionally, CD4 depletion may also be indicated since vaccination is with a long peptide which could contain both Class I and II epitopes. 4. Although the authors conduct single-cell analyses, they do not characterize (mutation specific) neoantigen-specific T cells. Do these specific neoantigen-specific T cells mediate tumor rejection? Direct proof of this is necessary to fully support their hypotheses.
Reviewer #2 (Remarks to the Author): Summary: The authors examined the effects of a low affinity class I neopeptide on CD8 T cell responses in mice challenged with Meth A sarcoma cells. For a particular neoantigen and corresponding peptides confirmed by MS in BMDCs, they used CRISPR to study homozygous and WT lines derived from the original heterozygous tumor. Although the neoantigens had low affinity, they found it could induce a CD8 T cell response characterized by a greater diversity of TCR clonotype and higher overall expression of early response genes involved in TCR engagement.
Predicting which mutations in a tumor are likely to elicit an immune response remains an important question for immunotherapy design and clinical application. High affinity and high expression of mutations remain the most predictive criteria for identifying mutations that have potential to generate responses, but are suboptimal, and field continues to search for understanding of what determines whether a mutation will be immunogenic. This study illustrates that low affinity may not be sufficient to rule out a mutation's potential to be immunogenic. Although the study focuses only a single mutation (with a second high affinity mutation provided for contrast), the cases study is instructive and will be of interest to the broader community investigating immunotherapies and antigen presentation in general.
The manuscript is well and clearly written. There are some concerns that should be addressed to further aid interpretation of the reported findings. In particular, more attention should be given to the role of MHC II presentation in driving the observed immune responses.
Major Comments: The title and some of the language in the manuscript are misleading. The authors are likely trying to express that low affinity binders that would otherwise be ignored (as if they were 'non-binding') can drive immune responses. They should revise their title and manuscript to avoid the term 'nonbinding' as that seems to suggest a non-MHC related mechanism which is clearly not the case.
The authors should further investigate the role of CD4 responses. That the cell surface density of class II in complex with WT and mutant peptides does not differ suggests that the binding affinity is likely similar, however it would be helpful to also show the binding affinity in a similar way to what was done for MHC I. A quick analysis using NetMHCII v3.2 shows that the 3 peptides mentioned (TYIRPFETKVK, YIRPFETKVK and IRPFETKVK) appear to have some potential to bind to H-2-IAb and H-2-IEd. Interestingly, this does not seem to be the case for the high affinity Alm1mut peptide (See pasted image). Is it possible that class II binding is driving the anti-tumor immune response to ccdc85cMut? Perhaps rather than contrasting to Alm1mut LYLDSKSDTTV, the authors could try to identify a peptide that has high class I affinity AND class II affinity characteristics? Could the role of CD4 be assessed by including antibodies to CD4 T cells during the priming phase? If class II affinity is equally or more important than class I affinity, this would be important information for vaccine designs. Notably, in Ott et al (Nature 2017), they found predominantly CD4 responses to a peptide vaccine designed around peptides designed to elicit CD8 responses. The manuscript from Alspach et al in 2019 (also in Nature) showed the requirement for both CD8 and CD4 responses for effective immunity, though it is not clear if the same peptide must elicit both.
Note: See attached version for figure.
Minor comments: Class II MHC typically presents longer peptides than class I. Did the authors consider using longer (e.g. 15mer) peptides for class II cell surface density assays?
For completeness, it would be helpful to provide a list of mutations in the tumor, their DNA and RNA variant allelic fraction, and perhaps predicted class I and class II affinities for peptides overlapping those mutations, perhaps limiting to only peptides were at least one of the 2 has weak affinity. This would allow the ccdc85c mutation to be placed into context of the overall mutation landscape.
The authors state "At the same time, two retrospective human studies analyzing the genomic and clinical outcome data from nearly 7,000 patients with 27 cancer types, have shown that better clinical outcomes and T cell infiltration of tumors are associated with the presence of cancer neoepitopes with low affinities for HLA I molecules, and not with the presence of high affinity HLA I-binding neoepitopes" [12,13] This sentence does not fully represent the two papers. The first seems to focus only on part of the TCGA and fewer than 1000 patients treated with ICB. Perhaps the authors could be more explicit about the clinical outcomes and how many patients were associated with each type. The second paper focuses on agretopicity (i.e. the relative mutant vs wildtype affinity) which they term ADNs or alternatively defined neopeptides, and finds these are predictive of survival/outcomes. The authors should more clearly point to the evidence in these papers that demonstrates that clinical outcomes are not associated with high affinity binders as they sate "not with the presence of high affinity HLA-I binding neoepitopes".
The authors may wish to clarify in the introduction that there are many reports where higher affinity neoepitopes are immunogenic. The problem is that the majority of mutations are unlikely to generate immunogenic neoepitopes and the use of affinity as a criteria for prioritizing neopeptides can help reduce false positives. However, as the authors demonstrate here, throwing out all mutations with low affinity may in some cases be "throwing the baby out with the bath water". This manuscript illustrates this with a compelling case study, though further analysis Tumor rejection score (TRS) is mentioned in the first results paragraph to be described in the Methods but no section was found.
There is a typo in CRISPR methods section: "CRIPR" in last line Reviewer #3 (Remarks to the Author): The demonstration that MHC-I-restricted presentation of low affinity peptides represents a relevant source of rejection antigens is indeed an important contribution of these authors to the cancer immunology and immunotherapy fields. The leading investigators of the present manuscript recently published an article demonstrating this using the MethA mouse tumor model (Ebrahimi-Nik et al, JCI Insight 2019). The present study aims to deepen these findings by studying changes in the tumor microenvironment of MethA tumors carrying a point mutation in the Ccdc85c gene. Similar to what was shown in their previous work, immunization with an 18-aminoacid-length peptide containing the mutant peptide (L-to-F; Ccdc85cMUT) near the center induces strong tumor rejection. Authors aimed to map the MHC-class I-restricted epitope peptide by immunizing mice with overlapping truncated peptides and evaluating loss of anti-tumor protection. The authors concluded that the 10-mer YIRPFETKVK is the MHC-class I-restricted epitope responsible of antitumor protection. Based on MS analyses of MHC-I eluted peptides, the authors concluded that the 11-mer TYIRPFETKVK and 10-mer YIRPFETKVK neoepitopes are actually presented by BMDCs pulsed with the 18-mer Ccdc85cMUT peptide. Then, the authors analyze the tumor microenvironment of mice immunized with Ccdc85cMUT or Alms1MUT, as a control, by scRNAseq observing that the former leads to higher lymphoid infiltration, including activated CD4 and CD8 T cells displaying higher levels of transcripts associated with cytotoxicity, effector function and TCR engagement. Then, authors analyzed the tumor microenvironment employing an experimental setting that does not involve immunization by comparing MethA tumors expressing the mutant (Lto-F; Ccdc85cMUT) gene in either one (MUT1), neither (REV) or both (MUT2) alleles. Similar to the immunization setting, they observed transcript expression skewed towards cytotoxic activity, effector function and TCR engagement in tumors carrying one (MUT1) or two (MUT2) mutant alleles as compared to REV. This differential transcriptomic state is also observed in clonally expanded T cells.
Even if some of the results presented in this manuscript are potentially interesting, a series of concerns have been raised by this reviewer. Major comments: The present study falls short to provide a mechanistic insight on why low MHC-I affinity neoepitopes can function as rejection antigens. The 11-mer epitope described here was already identified in their previous study (Ebrahimi-Nik et al, JCI Insight 2019), rendering some of the reported experiments overlapping. Actually, the contribution of CD8 T cells to anti-tumor protection in this immunization-challenge model, refers to anti-CD8 depletion experiments reported in their previous publication.
It is not clear to this reviewer the experimental setting of the immunization/challenge experiments (one or more immunizations how many days apart?, the interval between immunizations and tumor challenge, etc). A full description of the experimental setting is necessary. Also, a scheme of the experimental setting would be helpful. The way to show tumor protection data, as Tumor rejection score (TRS), is not defined in methods' section or elsewhere, and represents a rather unconventional way to do so. More importantly, grouping tumor growth curves from mice immunized with different truncated peptides does not allow to compare the protection elicited among the different truncated epitopes. It would be much more informative to display the individual tumor growth curves (and average perhaps) for the different truncated peptides to be able to understand how protective are the different peptides. In this regard, it is not clear why tumor protection diminishes from 5.0 to 3.0-3.3 (significant?) when mice are immunized with peptides STYIRPFETKVK, PSSTYIRPFETKVKLL and SSTYIRPFETKVKL, all which contain the predicted MHC-class-I-restricted epitopes.
To address the contribution of T cell populations to anti-tumor activity in the immunization setting, these experiments should include CD4 and CD8 depletion experiments. In their previous paper, the authors performed CD8 depletion experiments during the priming phase? It would be better to do the depletion right before the tumor challenge, in particular for CD4 T cells which may be important during the priming phase. It seems quite odd that tumor growth data of mice immunizated with the wild-type non-mutated Ccdc85cWT peptide (Fig. 2a) is displayed here. It seems more obvious to display these data at the beginning in Fig. 1b. Instead, this figure could start with the evidence showing that control immunization with Alms1MUT, does not lead to tumor rejection because this is the control used in the experiments described in Figure 2b-d. Figure 2c helps to define the different tumor-infiltrating cells, importantly clusters 1 and 5 which are further analyzed. How can immunization with another neoepitope such as Alms1MUT LYLDSKSDTTV, which has a higher MHC-I affinity, not induce comparable responses In the experimental setting that does not involve immunization, enhanced immunogenicity of the low MHC-I affinity neoepitope should be functionally demonstrated in terms of tumor growth. If the presence of the mutant neoepitope is relevant, tumor growth should be affected either spontaneously or after administration of checkpoint blockade immunotherapies such as anti PD1 and/or CTLA-4. Depletion experiments should also be performed to address the contribution of CD8 and CD4 T cell populations. This manuscript would greatly improve by a robust analysis of Ccdc85cMUT-specific CD8 T cell responses. Authors should quantify and phenotypically characterize CD8 T cell responses specific for both wild-type (T)YIRPLETKVK-and mutated (T)YIRPFETKVK epitopes. To this end, they could perform flow cytometry-based analyses, such as MHC-multimer staining and/or intracellular cytokine staining after ex vivo peptide stimulation. Along the different figures, statistical analysis is not informed for all relevant comparisons. The number of individual experiments performed for each analysis should be described.

Minor comments
The title does not quite reflect the main findings of the manuscript. It is not needed to remark a technical aspect of the study. In addition, to name neoepitopes with low affinity for MHC-class-I molecules as "non-MHC binders" seems misleading. It is not possible to assure that the 18-mer is presented through cross-presentation by BMDCs. Alternatively, this peptide may potentially be degraded extracellularly generating shorter peptides that can be loaded onto MHC-I molecules, bypassing antigen cross-presentation. This concern is particularly latent, when using relatively high concentrations (100 uM) of the peptide, as inferred by the authors' previous publication.