Abstract
The ability of the immune system to eliminate and shape the immunogenicity of tumours defines the process of cancer immunoediting1. Immunotherapies such as those that target immune checkpoint molecules can be used to augment immune-mediated elimination of tumours and have resulted in durable responses in patients with cancer that did not respond to previous treatments. However, only a subset of patients benefit from immunotherapy and more knowledge about what is required for successful treatment is needed2,3,4. Although the role of tumour neoantigen-specific CD8+ T cells in tumour rejection is well established5,6,7,8,9, the roles of other subsets of T cells have received less attention. Here we show that spontaneous and immunotherapy-induced anti-tumour responses require the activity of both tumour-antigen-specific CD8+ and CD4+ T cells, even in tumours that do not express major histocompatibility complex (MHC) class II molecules. In addition, the expression of MHC class II-restricted antigens by tumour cells is required at the site of successful rejection, indicating that activation of CD4+ T cells must also occur in the tumour microenvironment. These findings suggest that MHC class II-restricted neoantigens have a key function in the anti-tumour response that is nonoverlapping with that of MHC class I-restricted neoantigens and therefore needs to be considered when identifying patients who will most benefit from immunotherapy.
Similar content being viewed by others
Main
Immune checkpoint therapy (ICT) demonstrates remarkable clinical efficacy in subsets of patients with cancer, but many patients do not show durable responses2,3,4. Although MHC class I (MHC-I)-restricted neoantigens are important targets of tumour-specific CD8+ cytotoxic T lymphocytes (CTLs) during successful ICT in both mice and humans5,6,7,8,9,10,11,12, current methods to predict patient response to ICT are imprecise and additional or better prognostic indicators are needed13,14,15,16,17. The influence of MHC class II (MHC-II)-restricted CD4+ T cell responses to tumour neoantigens during immunotherapy has only recently been addressed18,19. While some reports show that effective tumour immunity can occur in the absence of help from CD4+ T cells, most indicate that CD4+ T cells are important for generating tumour-specific CD8+ T cells20,21,22,23,24,25. However, as it has proven difficult to identify tumour-specific mutations that function as neoantigens for CD4+ T cells using existing MHC-II antigen prediction algorithms, considerable uncertainty remains as to whether strict tumour specificity in the CD4+ T cell compartment is required during spontaneous or ICT-induced anti-tumour responses24,26,27, especially for tumours that do not express MHC-II.
In this study, we used the well-characterized, MHC-II-negative T3 methylcholanthrene (MCA)-induced sarcoma line, which grows progressively in wild-type mice but is rejected following ICT in a CD4+ and CD8+ T cell-dependent manner9. Although we have identified point mutations in laminin-α subunit 4 (LAMA(G1254V); mLAMA4) and asparagine-linked glycosylation 8 glucosyltransferase (ALG8(A506T); mALG8) as major MHC-I neoantigens in T3 cells, the identities of T3-specific MHC-II antigens remain unknown9. Here we use new predictive algorithms to identify an N710Y somatic point mutation in integrin-β1 (mITGB1) as a major MHC-II neoantigen of T3 sarcoma cells. In nonimmunogenic oncogene-driven KP9025 sarcoma cells (KP), which lack mutational neoantigens, co-expression of single MHC-I and MHC-II T3 neoantigens rendered KP9025 cells susceptible to ICT. We find similar requirements for vaccines that drive rejection of T3 tumours. In mice bearing contralateral KP.mLAMA4.mITGB1 and KP.mLAMA4 tumours, ICT induced the rejection of tumours expressing both neoantigens but not tumours expressing mLAMA4 only, indicating that co-expression of both MHC-I and MHC-II neoantigens at the tumour site is necessary for successful ICT. These results show that the expression of MHC-II neoantigens in tumours is a critical determinant of responsiveness to ICT, personalized cancer vaccines and potentially other immunotherapies.
Predicting MHC-II neoantigens with hmMHC
The best currently available methods for predicting MHC-II-restricted neoantigens rely on tools (netMHCII-2.3 and netMHCIIpan-3.2) that are inaccurate, partially because the open structure of the MHC-II binding groove leads to substantial variation in epitope length18,26. Moreover, the existing tools cannot be re-trained on new data. We therefore developed a hidden Markov model (HMM)-based MHC binding predictor (hmMHC, Extended Data Fig. 1a) that inherently accommodates peptide sequences of variable length and is trained on recent Immune Epitope Database (IEDB) content (Extended Data Fig. 1b–d). Validation analyses showed that hmMHC displays substantially higher sensitivity for high-specificity values than other predictors (Extended Data Fig. 2a, b). Using hmMHC, we calculated the likelihood of each of the 700 missense mutations that are expressed in T3 (Supplementary Data 1) being presented by the MHC-II I-Ab molecule and refined our results by prioritizing candidates based on I-Ab binding affinity, mutant:wild-type I-Ab binding ratios, and transcript abundance18 (Fig. 1a, Extended Data Fig. 3a).
One candidate, mITGB1, met all our criteria (Fig. 1a, Extended Data Fig. 3a). Notably, mITGB1 was not selected using netMHCII-2.3 or netMHCIIpan-3.2 (Extended Data Fig. 3b, data not shown). Enzyme-linked immune absorbent spot (ELISPOT) analysis showed that the mITGB1 peptide induced high IFNγ production from CD4+ T3 tumour-infiltrating lymphocytes (TILs). Other mutant peptides that fulfilled some but not all of our criteria induced only weak or absent responses, thereby validating our hmMHC prediction method (Fig. 1b, Extended Data Fig. 3c, Supplementary Table 1). To confirm this result, we stained T3-derived CD4+ TILs with MHC-II tetramers carrying either the 707–721 mITGB1 peptide or an irrelevant peptide (CLIP). Whereas 5.9% of T3-infiltrating CD4+ T cells stained positively with the mITGB1–I-Ab tetramer, the CLIP–I-Ab tetramer stained only 0.7% of the cells (Fig. 1c, Extended Data Fig. 3d, e). Cytokine profiling of mITGB1-specific CD4+ TILs from T3 tumours revealed that they produced IFNγ, TNF, and IL-2 but not IL-4, IL-10, IL-17 or IL-22, indicating a phenotype resembling that of T helper type 1 (TH1) cells (Extended Data Fig. 3f). T3 tumour-bearing mice treated with ICT did not develop additional MHC-II neoantigen specificities (data not shown). To assess whether T3-specific CD4+ T cells selectively recognized the mutant, we compared mutant to wild-type ITGB1 peptides in ELISPOT analyses using freshly isolated T3 CD4+ TILs. Only the mITGB1 peptide induced positive responses (Fig. 1d). Similar data were obtained using CD4+ T cell hybridomas generated from T3 TILs (Extended Data Figs. 4, 5a).
Mapping experiments revealed that the MHC-II binding core of mITGB1 consists of nine amino acids (710YNEAIVHVV718), in which the mutant Y710 residue functions as an I-Ab anchor (Extended Data Fig. 5b). To verify that the mITGB1 epitope is physiologically presented by MHC-II, we transduced T3 cells with a vector encoding the mouse MHC-II transactivator CIITA (T3.CIITA cells), which induced high levels of I-Ab expression28 (Extended Data Fig. 5c). Elution of peptides bound to I-Ab on T3.CIITA cells and analysis by mass spectrometry identified two mITGB1 peptides encompassing the Y710 mutation (a 17-mer and a 14-mer; Fig. 1e, Extended Data Fig. 5d). Peptides with the corresponding wild-type sequence were not found. The mITGB1 epitope was also not detected in MHC-I eluates from IFNγ-stimulated T3 cells, and mITGB1-specific CD8+ T cells were not observed by cytokine production (data not shown). Together, these data demonstrate that mITGB1 is a major MHC-II-restricted neoantigen of T3 sarcoma cells.
ICT response requires CD4+ T cell help
Recent publications have highlighted the ability of CD4+ T cells to recognize tumour-specific antigens and promote tumour rejection in the absence of ICT18,29,30. To assess whether CD4+ T cells are required during ICT-induced rejection, we expressed MHC-I and/or MHC-II neoantigens from T3 sarcoma cells in an oncogene-driven sarcoma cell line generated from a KrasLSL-G12D/+ × Tp53fl/fl mouse injected intramuscularly with lentiviral Cre-recombinase (KP9025 cells)7. The unmodified KP9025 sarcoma line formed progressively growing tumours in either syngeneic wild-type mice treated with or without dual anti-PD-1 and anti-CTLA4 ICT or mice rechallenged with unmodified KP9025 after previously being cured of their KP9025 tumours via surgical resection (Fig. 2a, b). As this challenge–resection–rechallenge approach promotes immune control or rejection of even poorly immunogenic tumour cells used in the initial priming step31, these results supported the conclusion that KP9025 sarcoma cells were not immunogenic. Whole-exome sequencing revealed that KP9025 cells expressed only four nonsynonymous mutations (Supplementary Data 2) and none were predicted to be immunogenic (Extended Data Fig. 6a, b, Supplementary Table 2). Enforced expression of either mLAMA4 or mITGB1 alone did not render KP9025 cells immunogenic in wild-type mice in the presence or absence of ICT (Fig. 2c, Extended Data Fig. 6d, e). Progressively growing KP.mLAMA4 tumours maintained expression of their MHC-I tumour neoantigen, thereby ruling out antigen loss via immunoediting (Extended Data Fig. 7a). KP9025 cells expressing both mLAMA4 and mITGB1 formed tumours in immunodeficient Rag2−/− mice that grew with kinetics similar to those of KP.mLAMA4 or KP.mITGB1 cells (Extended Data Fig. 6c). However, growth of KP.mLAMA4.mITGB1 cells in wild-type mice treated with a control monoclonal antibody was noticeably slower than that of either single-antigen-expressing cell line, and KP.mLAMA4.mITGB1 tumours were rejected in wild-type mice following either dual or single agent ICT despite the absence of tumour cell MHC-II expression (Fig. 2c, Extended Data Fig. 6d, e, data not shown).
We considered the possibility that the enhanced immunogenicity of KP.mLAMA4.mITGB1 tumours was merely a function of antigen quantity. Therefore, we generated KP9025 cells that lacked MHC-II neoantigens but co-expressed two strong MHC-I neoantigens: the MHC-I epitope of ovalbumin (SIINFEKL) and the R913L mutant of spectrin-β2 (mSB2), which contributes to the spontaneous rejection of the MCA-induced d42m1 sarcoma line in wild-type mice6. KP.mSB2.SIINFEKL tumours grew progressively in mice treated either with a control monoclonal antibody or dual ICT, and the expression of both MHC-I antigens was maintained in growing tumours from ICT-treated animals (Fig. 2c, Extended Data Fig. 7b–d). Enforced expression of mITGB1 in KP.mSB2.SIINFEKL cells led to significantly (P = 1.5 × 10−5) increased survival of ICT-treated mice injected with the uncloned tumour line (Extended Data Fig. 7e). Thus, tumour rejection and ICT sensitivity are dependent on combinatorial effects of CD4+ and CD8+ T cells.
mITGB1 CD4+ T cells are TH1 polarized
We next investigated whether mITGB1-specific CD4+ TILs displayed a TH1 phenotype similar to that seen with T3 tumours. Seventy-four per cent of mITGB1 tetramer-positive CD4+ T cells in KP.mLAMA4.mITGB1 tumours from control-treated mice expressed the TH1-associated transcription factor T-BET, but not the regulatory T cell (Treg)-associated transcription factor FOXP3. An additional 17% expressed both T-BET and FOXP3. Conversely, tetramer-negative CD4+ T cells showed substantially diminished expression of T-BET (24%) and much higher expression of FOXP3 expression (61%). mITGB1-tetramer+ CD4+ T cells displayed a higher T-BET+:FOXP3+ ratio than tetramer-negative cells (4 versus 0.4, respectively) and this ratio was further increased in response to anti-CTLA4 treatment (33 versus 3.7, respectively; Extended Data Fig. 8a–c). On average, 83% of mITGB1-specific CD4+ T cells expressed high levels of PD-1 compared to only 19% of mITGB1-tetramer-negative cells (Extended Data Fig. 8d, e). CD4+ T cells specific for mITGB1 also expressed high levels of CD44, ICOS and CD150 (also known as SLAM), and low levels of KLRG1 (Extended Data Fig. 8f). The presence of an expanded population of TH1-like ICOS+ CD4+ T cells was recently reported in mice bearing B16 or MC38 tumours that were treated with anti-CTLA4, although the tumour antigen specificity of this population was not identified32. These data, together with the cytokine profiles described above, indicate that mITGB1-specific CD4+ T cells display an activated TH1 phenotype.
CTL generation requires CD4+ T cell help
To identify the mechanism by which tumour neoantigen-specific CD4+ T cells influence ICT-mediated anti-tumour responses, we assessed their effects on CD8+ T cell priming by comparing MHC-I tetramer staining of splenic mLAMA4-specific CD8+ T cells from mice bearing KP.mLAMA4 or KP.mLAMA4.mITGB1 tumours, treated with a control monoclonal antibody or ICT. In the absence of ICT, mLAMA4-H-2Kb tetramers stained only 1.2% of CD8+ T cells from mice bearing KP.LAMA4 tumours, but 5.3% of CD8+ T cells in mice bearing KP.mLAMA4.mITGB1 tumours (Fig. 3a, b). This staining percentage was unchanged in the presence of PD-1 blockade, but was increased by anti-CTLA4 treatment, either as monotherapy or in combination with anti-PD-1. This result is consistent with the observation that anti-CTLA4 treatment functions largely to enhance CD4+ T cell responses32,33.
To assess whether MHC-II neoantigens also enhanced CTL formation, we used an in vivo T cell cytotoxicity assay that monitored the capacity of naturally arising CTLs to kill peptide-pulsed splenocytes labelled with carboxyfluorescein succinimidyl ester (CFSE)34. Non-tumour-bearing control mice and mice bearing KP.mLAMA4 tumours were largely incapable of eliminating mLAMA4 peptide-pulsed splenocytes in either the presence or absence of ICT (Fig. 3c). By contrast, mice bearing KP.mLAMA4.mITGB1 tumours efficiently eliminated CFSEhi-labelled, mLAMA4 peptide-pulsed splenocytes but not CFSElo-labelled SIINFEKL-pulsed splenocytes, and the degree of elimination of the former was enhanced by ICT (Fig. 3c, d). The cytotoxic activity of control-treated mLAMA4-specific CD8+ T cells observed in the splenocyte killing assay was higher than would be expected from our in vivo tumour rejection experiments (Fig. 2e). This difference is likely to reflect differences in the susceptibility of splenocytes and tumour cells to T cell-mediated killing. Thus, CD4+ T cell help enhances both CD8+ T cell priming and maturation of CD8+ T cells into CTLs.
Vaccines require MHC-I and MHC-II antigens
As CD4+ T cell help was crucial for generating mLAMA4-specific CTLs during ICT, we tested whether mITGB1-specific CD4+ T cells were also important for vaccine-elicited anti-tumour responses (Fig. 4a). Vaccination of naive recipient mice with irradiated parental KP9025, KP.mLAMA4, or KP.mITGB1 cells was not sufficient to protect most mice from a subsequent challenge with T3 sarcoma cells. Vaccination with a mixture of irradiated KP.mLAMA4 and KP.mITGB1 cells provided protection against T3 challenge in 30% of mice. By contrast, vaccination with irradiated KP.mLAMA4.mITGB1 cells prevented T3 tumour outgrowth in 11 of 13 recipients (Fig. 4b, c). Furthermore, spleens from mice vaccinated with irradiated KP.mLAMA4.mITGB1 cells contained significantly (P = 0.0002) more mLAMA4-specific, IFNγ-producing CD8+ T cells than did the spleens of mice vaccinated with KP cells expressing only mLAMA4 (Fig. 4d). The differences in efficacy between mixed cellular vaccines and dual antigen-expressing KP.mLAMA4.mITGB1 vaccines support previous findings that effective vaccines are those in which the MHC-I and MHC-II epitopes reside on the same peptide strand, potentially leading to more efficient uptake and presentation of both antigens by the same antigen-presenting cell (APC)20,35. A similar situation would be expected to occur when both antigens were present in the same tumour cell used for vaccination.
MHC-II antigen expression at tumour site
To investigate whether CD4+ T cells are required beyond the priming and maturation of anti-tumour CTLs, we tested whether tumour cell expression of MHC-II neoantigens was necessary at the site of tumour rejection. We assessed the in vivo growth of contralaterally injected KP.mLAMA4.mITGB1 and KP.mLAMA4 tumours in either immunodeficient or immunocompetent mice treated with ICT. The contralateral tumours grew at equivalent rates in Rag2−/− mice (Extended Data Fig. 9a). However, ICT treatment of wild-type mice bearing contralateral tumours resulted in complete rejection of the KP.mLAMA4.mITGB1 tumour but only delayed outgrowth of the KP.mLAMA4 tumour on the opposite flank (Fig. 5a, b). This result shows that CTLs specific for mLAMA4 can control tumours expressing both the cognate MHC-I epitope and the helper MHC-II epitope locally, but function poorly against distant but related tumours that lack CD4 neoepitopes. In similar experiments, we investigated whether mITGB1-specific CD4+ T cells generated from KP.mLAMA4.mITGB1 tumours were sufficient to control the outgrowth of KP.mITGB1 tumours on the opposite flank. In this setting, contralateral KP.mITGB1 tumour growth was identical to that observed in mice bearing only a single KP.mITGB1 tumour (Extended Data Fig. 9b, c). Together, these results show that tumour cell expression of MHC-II-restricted neoantigens and the presence of tumour-specific CD4+ T cells in the tumour microenvironment are required to maintain tumour control during ICT but are not sufficient to mediate tumour rejection by themselves.
To expand this observation, we investigated whether CD4+ T cells and expression of MHC-II neoantigens in tumour cells are required to maintain functional CD8+ T cell memory. When mice that had been cured of T3 tumours by ICT treatment were rechallenged with T3 tumour cells, they rejected the cells. However, if mice were depleted of CD4+ T cells before being rechallenged, they did not control T3 tumour outgrowth (Extended Data Fig. 9d). In parallel experiments, mice previously cured of KP.mLAMA4.mITGB1 tumours by surgical resection were protected against subsequent rechallenge with KP.mLAMA4.mITGB1 but were unable to prevent outgrowth of KP.mLAMA4 or KP9025 tumours (Extended Data Fig. 9e). Thus, both expression of MHC-II neoantigens by tumour cells and CD4+ T cell help are required for the maintenance of tumour-specific immunologic memory.
Last, we investigated whether an MHC-II tumour neoantigen can significantly affect the local tumour microenvironment (gating strategy, Extended Data Fig. 10a). The expression of inducible nitric oxide synthase (iNOS) is higher in macrophages that populate tumours destined to be rejected after ICT than in macrophages from progressively growing tumours, and this expression is induced by ICT-dependent production of IFNγ33. iNOS+ macrophages were present at threefold higher levels in ICT-treated KP.mLAMA4.mITGB1 tumours than in contralateral KP.mLAMA4 tumours (Extended Data Fig. 9g, h). ELISPOT analysis of tumour-infiltrating CD4+ T cells showed 5.9-fold more IFNγ+ mITGB1-specific CD4+ T cells in KP.mLAMA4.mITGB1 tumours than in contralateral KP.mLAMA4 tumours (Fig. 5c, Extended Data Fig. 9f). Flow cytometry analysis of the lymphoid compartment (gating strategy, Extended Data Fig. 10b) identified 3.7-fold more CD8+ T cells and 9-fold more mLAMA4-specific CD8+ T cells in KP.mLAMA4.mITGB1 tumours than in KP.mLAMA4 tumours (Fig. 5d, e). We then investigated whether CD4+ T cells were sufficient to mediate these changes, by comparing iNOS+ macrophages in KP.mLAMA4.mITGB1 tumours with those in contralateral KP.mITGB1 tumours. KP.mLAMA4.mITGB1 tumours contained 83-fold more iNOS+ macrophages than did KP.mITGB1 tumours (Extended Data Fig. 9i, j). Together, these data show that MHC-II-restricted anti-tumour responses are necessary but not sufficient in ICT-sensitive tumour models to induce localized effects on the immune composition of tumours.
Discussion
The work described herein focuses on the functional role of MHC-II restricted tumour neoantigens in mediating ICT-dependent anti-tumour responses in a well-characterized mouse sarcoma model. Using an HMM-based tool (hmMHC), we have predicted and validated that an N710Y point mutation in the integrin ITGB1 forms a major MHC-II restricted neoepitope of the T3 MCA sarcoma. It is reasonable that mITGB1 represents a major MHC-II neoantigen of T3 tumour cells because ITGB1 is the second most highly expressed mutation in T3 tumour cells and the point mutation in mITGB1 generates a novel anchor residue that promotes high affinity binding to I-Ab. Moreover, others have proposed that secreted tumour proteins are favoured targets for CD4+ T cell responses because they are more easily taken up by professional APCs36. Localization of mITGB1 on the cell membrane would also be likely to facilitate efficient access by APCs, although we did not directly test this idea. Notably, we do not rule out the possibility that T3 cells express other MHC-II-restricted epitopes that might be elicited by vaccination18,19. Nevertheless, we have shown that mITGB1 functions as a major neoantigen of T3 cells during naturally occurring anti-tumour responses.
By defining authentic MHC-I and MHC-II neoantigens of T3 sarcoma cells, we have shown that, in a minimal antigen system, a single clonally expressed MHC-I neoantigen (mLAMA4) and a single clonally expressed MHC-II neoantigen (mITGB1) are necessary and sufficient to render nonimmunogenic, oncogene-driven KP9025 sarcoma cells sensitive to ICT. Using KP9025 sarcoma cells that express different combinations of mLAMA4 and/or mITGB1, we have shown that CD4+ T cell responses are required for optimal priming of MHC-I restricted CD8+ T cells and their maturation into CTLs, in either the presence or absence of ICT. We have also shown that optimal anti-tumour responses occur when tumour cells express both MHC-I and MHC-II neoantigens. In part, this requirement reflects the potential need for CD4+ T cell responses in the tumour microenvironment and, from previous work, appears to be at least partially due to production of IFNγ by tumour-specific CD4+ T cells33. We find it of particular interest that the generation of effective tumour immunity requires MHC-II neoantigens following either vaccination with tumour-specific neoantigen vaccines or ICT. These results provide new insights into the role of MHC-II neoantigens in natural and therapeutic immune responses to tumours. They also suggest that patients with tumours that are predicted to contain immunogenic MHC-I neoantigens or have favourable tumour mutational burdens could still be unresponsive to immunotherapies, owing to the absence of immunogenic MHC-II-restricted CD4+ T cell antigens. This possibility has not been critically evaluated yet, owing to the past absence of reliable MHC-II prediction algorithms. Future work is needed to test this hypothesis in patients with cancer undergoing immunotherapy.
Note added in proof: As this Article was being prepared for publication, an independent paper was published online describing an MHC-II prediction algorithm for human tumours37.
Methods
Mice
Male wild-type 129S6 mice (for experiments involving T3 cells) were purchased from Taconic Farms. Male wild-type 129S4 mice (for experiments involving KP9025 cells) and 129S6 Rag2−/− mice were bred in our specific-pathogen free facility. All in vivo experiments were performed in our specific-pathogen free facility and used mice between the ages of 8 and 12 weeks. All experiments were performed in accordance with procedures approved by the AAALAC-accredited Animal Studies Committee of Washington University in St Louis and were in compliance with all relevant ethical regulations.
Tumour transplantation
T3 MCA-induced sarcoma cells were previously generated in our laboratory in 129S6 wild-type mice. KP sarcoma cell lines were provided by T.J., and were generated following intramuscular injection of lentiviral Cre-recombinase into 129S4 KrasLSL-G12D/+ × Tp53fl/fl mice. Tumour cells were cultured in Roswell Park Memorial Institute (RPMI) medium (Hyclone) supplemented with 10% fetal calf serum (FCS) (Hyclone). Cell lines were authenticated using whole-exome sequencing and verification of specific antigen expression. All cell lines used tested negative for mycoplasma contamination. For transplantation, cells were washed extensively in PBS and resuspended at a density of 13.34 × 106 cells per ml (T3) or 6.67 × 106 cells per ml (KP sarcomas) in PBS. Then, 150 μl was injected subcutaneously into the rear flanks of syngeneic recipient mice. For irradiated tumour cell vaccines, KP.mLAMA4, KP.mITGB1 or KP.mLAMA4.mITGB1 sarcoma cells were lethally irradiated with 10 Gy and 500,000 cells were injected subcutaneously into 129S6 mice. T3 challenge following vaccination occurred on the opposite flank. Following tumour transplantation, animals were randomly assigned to treatment groups. No statistical methods were used to determine group size. Tumour growth was measured by calipers and individual growth curves are represented as the average of two perpendicular diameters. Tumour measurements were performed blinded to treatment group. In accordance with our IACUC-approved protocol, maximal tumour diameter was 20 mm in one direction, and in no experiments was this limit exceeded.
Tumour rechallenge
For tumour rechallenge following surgical resection, primary tumours were allowed to grow until they reached 10 mm in size or to the time point indicated. Following surgical removal of the established tumour, animals were rested for 30 days. Animals were then rechallenged on the opposite flank with either the same tumour line as was used in the primary tumour challenge or the tumour line indicated. For tumour rechallenge following ICT-mediated rejection, primary tumours were rejected following treatment with combination anti-PD-1 and anti-CLTA4 ICT. After tumours were no longer apparent, animals were rested for 30 days followed by rechallenge on the opposite flank with the same tumour line as was used in the primary challenge or the tumour line indicated.
Epitope prediction
The identification of point mutations in T3 and KP sarcomas and the prediction of MHC-I epitopes in KP and F244 sarcomas were performed as previously described9. To predict neoepitopes, we applied hmMHC, our newly developed HMM-based binding predictor, trained on the most recent IEDB data. HMMs inherently accommodate inputs of variable length and have already demonstrated reasonable performance for prediction of MHC binding affinity38. Our predictor uses a fully connected HMM with emissions representing amino acids (see a pedagogical example in Extended Data Fig. 1a). We trained the model on a set of known binders using the Baum–Welch algorithm39, as implemented by the GHMM library (http://ghmm.sourceforge.net/). A trained HMM returns the likelihood of a peptide to be a binder, which we represent as the −log10 odds ratio, where a smaller value indicates that a peptide has a higher likelihood of being a binder. The model that we apply in this study was trained on murine H2-I-Ab binders taken from the IEDB full MHC ligand export (downloaded on 25 November 2018, containing 1,072,460 entries). Non-binders were not used in model training. The categorization of the data into binders and non-binders was based on the qualitative and quantitative fields of IEDB entries: binders are peptides with IC50 ≤ 500 nM or with positive, positive-high or positive-intermediate binding quality. These data came largely from mass spectrometry assays. We validated the model using the Monte Carlo (shuffle-split) cross-validation approach, with ten random partitions of H2-I-Ab binders from IEDB into training and validation sets, with a relative validation set size of 0.2. As the number of non-binders in the IEDB dataset was insufficient for validation, we used decoy sets composed of random natural peptides as non-binders. Protein-coding transcript translation sequences for Mus musculus were obtained from GENCODE release M19 (GENCODE project, 2018); there are 65,257 translations. For every cross-validation partition, the translations were randomly cut into fragments uniformly distributed in the interval [12, 24], which generated about 1.5 × 106 fragments. Of this set of random natural peptides, a random sample 100 times the number of binders in the validation set was taken. The 100-fold bias in the number of generated non-binders and uniform distribution of their lengths are in line with recent work on MHC binding prediction, in particular netMHCpan-4.040. We have also performed experiments in which the distribution of random natural peptide lengths followed the distribution of lengths in the IEDB dataset (Extended Data Fig. 1d) and found no significant difference in results in our setting compared to uniform distribution. The rationale for the 100-fold bias is that for a sample of peptide fragments from an organism, it is commonly considered that about 1–2% will bind to MHC receptors. On average, there were 4,412 binders in a training set, and 771 binders and 77,086 random natural peptides in a validation set. Classification performance of our predictor was significantly higher than the performance of the two best-known class II binding predictors41 (netMHCII-2.3 and netMHCIIpan-3.2), compared on our ten validation datasets. This is due, in part, to the large amount of new mass spectrometry data compared to the data on which the recent netMHCII(pan) predictors were trained (netMHCIIpan-3.2 public dataset available at http://www.cbs.dtu.dk/suppl/immunology/NetMHCIIpan-3.2/ contains 1,794 measurements for H-2-I-Ab, all qualitative, of which 431 are binders and 1,363 are weak or non-binders). We do not exclude the possibility that netMHCII(pan), as a method, performs better than the HMM method. As the published netMHCII(pan) tools lack re-training capability, we cannot compare the methods and draw conclusions on netMHCII(pan) performance on new qualitative data. We determined the threshold for strong binders by calibrating the predictor to return a percentile rank against a large decoy set of random natural peptides. We used the approach taken by the existing neural network-based predictors, in which strong binders are predictions in the second percentile of the empirical distribution of predictions on random natural peptides40. The decoy set was generated from the mouse proteome in the same way as for validation and consists of about 1.5 × 106 fragments with lengths in the interval [12, 24]. Predicted neoantigens were further prioritized using the NER: the ratio between the binding predictions for the mutant and wild-type peptides. Expression of each mutation is represented as FPKM generated from cDNA capture sequencing.
Peptides
All 27-mer peptides used for neoantigen screening (Supplementary Table 1) were purchased from Peptide 2.0 and purified by high-performance liquid chromatography (HPLC) to >95% purity. The T3-specific mutant amino acid was placed in the centre of the peptide and was flanked on both sides with 13 amino acids of wild-type peptide sequence.
ELISPOT
Cells from tumours or lymph nodes were enriched for CD4+ or CD8+ T cells using the Miltenyi mouse CD4+ or CD8+ enrichment kits according to the manufacturer’s protocols. Ten thousand TIL-derived T cells or 50,000 tumour-draining lymph node (TDLN)-derived T cells were stimulated with 500,000 splenocytes isolated from naive mice pulsed with 2 μg ml−1 29-mer peptide (class II) or 1 μM 15-mer peptide (class I). For analysis from spleens, 500,000 cells from whole-spleen preparations were used. Cells were stimulated overnight in anti-mouse IFNγ-coated ELISPOT plates (Immunospot). Plates were developed according to the manufacturer’s protocol and spots were quantified using a CTL ImmunoSpot S6 Universal machine and Professional 6.0.0 software.
Mass spectrometry
For isolation of I-Ab-bound peptides, 5 × 108 T3.CIITA cells were washed twice with PBS and snap-frozen. MHC-II molecules were isolated by immunoaffinity purification using the I-Ab-specific antibody Y-3P (BioXCell) coupled to cyanogen bromide-activated sepharose 4B (GE Healthcare) as described42. Peptides were eluted with 0.2% trifluoroacetic acid, cleaned by detergent removal (Pierce Detergent Removal Spin Columns, Thermo Scientific) and desalting (Pierce C-18 Spin Columns, Thermo Scientific), dried, and resuspended in 2% acetonitrile (ACN) and 0.1% formic acid (20 µl). For mass spectrometry, a Dionex UltiMate 1000 system (Thermo Scientific) was coupled to an Orbitrap Fusion Lumos (Thermo Scientific) through an Easy-Spray ion source (Thermo Scientific). Peptide samples were loaded (15 µl/min, 3 min) onto a trap column (100 µm × 2 cm, 5 µm Acclaim PepMap 100 C18, 50 °C), eluted (200 nl/min) onto an Easy-Spray PepMap RSLC C18 column (2 µm, 50 cm × 75 µm ID, 50 °C, Thermo Scientific) and separated with the following gradient (all percentages indicate buffer B: 0.1% formic acid in ACN): 0–110 min, 2–22%; 110–120 min, 22–35%; 120–130 min, 35–95%; 130–150 min, isocratic at 95%; 150–151 min, 95–2%, 151–171 min, isocratic at 2%. Spray voltage was 1,900 V, ion transfer tube temperature was 275 °C, and RF lens was 30%. Mass spectrometry scans were acquired in profile mode (375–1,500 Da at 120,000 resolution (at m/z 200)); centroided HCD MS/MS spectra were acquired using a Top Speed method (charge states 2–7, 3 s cycle time, threshold 2 × 104, quadrupole isolation (0.7 Da), 30,000 resolution, collision energy 30%) with dynamic exclusion enabled (5 ppm, 60 s). Raw data files were uploaded to PEAKS X (Bioinformatics Solutions) for processing, de novo sequencing and database searching against the UniProtKB/Swiss-Prot Mouse Proteome database (downloaded 1 December 2019; 22,286 entries), appended with a truncated sequence of mITGB1 (±20 amino acids from the site of mutation), with mass error tolerances of 10 ppm and 0.01 Da for parent and fragment, respectively, no enzyme specificity, and methionine oxidation as a variable modification. False discovery rate (FDR) estimation was enabled, and proteins were filtered for −log10P ≥ 0 and one unique peptide to give 1% FDR at the peptide-spectrum match level. Peptides matching to mITGB1 were manually verified by visual inspection.
Antibodies
For immune checkpoint therapy, rat IgG2a anti-PD1 (RMP1-14, Leinco) and mouse IgG2b anti-CTLA4 (9D9, Leinco Technologies) antibodies were used. Mice were injected intraperitoneally with 200 μg of each antibody on days 3, 6 and 9 after tumour transplantation. For multi-colour flow cytometry, we used antibodies against CD45 (30-F11), CD11B (M1/70), THY1.2 (30H12), CD4 (RM4-5), CD8β (YTS156.7.7), I-E/I-A (M5/114.15.2), CD64 (X54-5/7.1), LY6G (1A8), T-BET (4B10), CD150/SLAM (TC15-12F12.2), KLRG1 (2F1), ICOS (15F9), CD44 (IM7), PD-1 (29F.1A12), SIINFEKL-H-2-Kb (25-D1.16) (BioLegend), CD24 (M1/69), F4/80 (T45-2342) (BD Biosciences), FOXP3 (FJK-16 s, eBiosciences) and iNOS (CXNFT, Invitrogen). Zombie NIR (BioLegend) was used to stain for cellular viability. The BD Cytofix/Cytoperm Plus kit (BD Biosciences) was used according to the manufacturer’s protocol for intracellular staining of iNOS, T-BET and FOXP3.
Tetramer staining
Tetramer staining for mLAMA4-specific CD8+ T cells was performed as previously described9. I-Ab monomers bound to CLIP or mITGB1 were a gift from K. Wucherpfennig. For staining, biotinylated pI-Ab monomers were labelled at a 4:1 molar ratio with streptavidin–APC or streptavidin–PE (Prozyme). One million cells from whole-tumour digests were stained with equal amounts of APC and PE tetramer at 20 μg ml−1 for 2 h at room temperature. Tetramer staining was stabilized through the use of anti-PE and anti-APC cells beads (Miltenyi), similar to previously published methods for MHC-I tetramers43, followed by surface staining for CD11B, THY1.2 and CD4.
Multi-cytokine assay
CD4+ T cells were enriched from tumours 12 days after transplantation using the Miltenyi mouse CD4+ enrichment kit. Ten thousand enriched CD4+ T cells were stimulated in serum-free medium with 500,000 splenocytes isolated from naive mice pulsed with 2 μg ml−1 peptide. Following a 24-h incubation, secretion of IL-10, IL-1B, IL-2, IL-4, IL-5, IL-6, IL-22, IL-9, IL-13, IL-27, IL-23, IFNγ, IL-12 p70, GM-CSF, TNF, IL-17A and IL-18 was measured using a flow-based ProcartaPlex TH1/TH2/TH9/TH17/TH22/Treg cytokine panel (Luminex Technologies) following the manufacturer’s protocol.
Plasmids
Full-length mLAMA4 and mITGB1 were cloned from T3 cDNA and full-length CIITA was cloned from 129S6 splenocytes. Gene blocks encoding SIINFEKL and the minimal epitope of mSB2 were purchased from Integrated DNA Technologies. All constructs were cloned into the BglII site of pMSCV-IRES GFP (mLAMA4, CIITA, and mSB2) or pMSCV (mITGB1 and SIINFEKL) using the Gibson Assembly method (New England Biolabs). To generate neoantigen-expressing KP sarcoma cell lines and T3.CIITA cells, constructs were transiently transfected into Phoenix Eco cells using Fugene (Promega). After 48 h, viral supernatants were subsequently used for transfection of KP sarcoma line 9025 or T3 cells. KP.mLAMA4, KP.mITGB1, KP.mLAMA4.mITGB1, KP.mSB2.SIINFEKL and T3.CIITA clones were obtained by limiting dilution.
CD4+ T cell hybridomas and CTLL assay
Bulk CD4+ T cells from T3 tumours were isolated 12 days after transplantation and stimulated with lethally irradiated T3.CIITA cells to establish a rapidly dividing cell line. CD4+ T cells were fused with BW5147 cells and cloned via limiting dilution. To assess antigen specificity and to map the mITGB1 MHC-II binding core, splenocytes were collected from naive mice and pulsed with 10 μg ml−1 peptide unless otherwise stated. Fifty thousand hybridoma cells were incubated with 100,000 peptide-pulsed splenocytes overnight and culture medium was collected. Production of IL-2 was assayed by proliferation-dependent thymidine incorporation using the IL-2 dependent CTLL-2 cell line. Data are represented as counts per million (cpm).
Measuring IFNγ production by CD8+ T cell clones
Tumour cells were treated with 100 U ml−1 IFNγ for 48 h before use. One hundred thousand CTL cells specific against mLAMA4 (74.17) or mSB2 (C3) were co-cultured with 50,000 tumour cells for 48 h. IFNγ in supernatants was quantified using an IFNγ ELISA kit (eBioscience) according to the manufacturer’s protocol.
In vivo cytotoxicity assay
For targets, splenocytes were collected from naive mice, stained with either 5 μM or 0.5 μM CFSE (CFSEhi and CFSElo) (Thermo Fisher Scientific) and pulsed with either mLAMA4 (CFSEhi) or SIINFEKL (CFSElo) peptide, respectively, at 1 μM overnight. Cells were washed extensively and combined at a 50:50 ratio in PBS, and 20 × 106 cells were injected retro-orbitally into tumour-bearing mice 11 days after tumour transplantation. Naive, non-tumour bearing mice were used as a control. Spleens from tumour-bearing or control naive mice were removed 24 h after cell transfer, stained with Zombie NIR viability dye (Biolegend) and quantified for the presence of CFSE-labelled target cells. On histograms, equivalent heights of CFSEhi and CFSElo peaks indicate that equivalent numbers of each cell population are present, and that no cytotoxicity was observed. Peaks that differ in height, where the CFSElo population is more abundant than the CFSEhi population, indicate that cytotoxicity was observed specifically against the mLAMA4 peptide-pulsed, CFSEhi population of cells. The equation used for calculating per cent specific lysis was [1 − (naive control ratio/experimental ratio)] × 100 with ratio = irrelevant percentage/specific epitope percentage.
Statistics
Statistical analysis was performed using GraphPad Prism software version 7. Unless otherwise noted, significance was determined with an unpaired, two-tailed Student’s t-test.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
Code availability
Code for the hmMHC algorithm used to predict presentation of neoantigens by I-Ab can be accessed at https://github.com/artyomovlab/hmmhc.
References
Schreiber, R. D., Old, L. J. & Smyth, M. J. Cancer immunoediting: integrating immunity’s roles in cancer suppression and promotion. Science 331, 1565–1570 (2011).
Larkin, J. et al. Combined nivolumab and ipilimumab or monotherapy in untreated melanoma. N. Engl. J. Med. 373, 23–34 (2015).
Motzer, R. J. et al. Nivolumab versus everolimus in advanced renal-cell carcinoma. N. Engl. J. Med. 373, 1803–1813 (2015).
Borghaei, H. et al. Nivolumab versus docetaxel in advanced nonsquamous non-small-cell lung cancer. N. Engl. J. Med. 373, 1627–1639 (2015).
Lennerz, V. et al. The response of autologous T cells to a human melanoma is dominated by mutated neoantigens. Proc. Natl Acad. Sci. USA 102, 16013–16018 (2005).
Matsushita, H. et al. Cancer exome analysis reveals a T-cell-dependent mechanism of cancer immunoediting. Nature 482, 400–404 (2012).
DuPage, M., Mazumdar, C., Schmidt, L. M., Cheung, A. F. & Jacks, T. Expression of tumour-specific antigens underlies cancer immunoediting. Nature 482, 405–409 (2012).
Robbins, P. F. et al. Mining exomic sequencing data to identify mutated antigens recognized by adoptively transferred tumor-reactive T cells. Nat. Med. 19, 747–752 (2013).
Gubin, M. M. et al. Checkpoint blockade cancer immunotherapy targets tumour-specific mutant antigens. Nature 515, 577–581 (2014).
Wölfel, T. et al. A p16INK4a-insensitive CDK4 mutant targeted by cytolytic T lymphocytes in a human melanoma. Science 269, 1281–1284 (1995).
Snyder, A. et al. Genetic basis for clinical response to CTLA-4 blockade in melanoma. N. Engl. J. Med. 371, 2189–2199 (2014).
Strønen, E. et al. Targeting of cancer neoantigens with donor-derived T cell receptor repertoires. Science 352, 1337–1341 (2016).
Rizvi, N. A. et al. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348, 124–128 (2015).
Van Allen, E. M. et al. Genomic correlates of response to CTLA-4 blockade in metastatic melanoma. Science 350, 207–211 (2015).
Spranger, S. et al. Density of immunogenic antigens does not explain the presence or absence of the T-cell-inflamed tumor microenvironment in melanoma. Proc. Natl Acad. Sci. USA 113, E7759–E7768 (2016).
Hugo, W. et al. Genomic and transcriptomic features of response to anti-PD-1 therapy in metastatic melanoma. Cell 165, 35–44 (2016).
Hellmann, M. D. et al. Genomic features of response to combination immunotherapy in patients with advanced non-small-cell lung cancer. Cancer Cell 33, 843–852 (2018).
Kreiter, S. et al. Mutant MHC class II epitopes drive therapeutic immune responses to cancer. Nature 520, 692–696 (2015).
Ott, P. A. et al. An immunogenic personal neoantigen vaccine for patients with melanoma. Nature 547, 217–221 (2017).
Ossendorp, F., Mengedé, E., Camps, M., Filius, R. & Melief, C. J. Specific T helper cell requirement for optimal induction of cytotoxic T lymphocytes against major histocompatibility complex class II negative tumors. J. Exp. Med. 187, 693–702 (1998).
Corthay, A. et al. Primary antitumor immune response mediated by CD4+ T cells. Immunity 22, 371–383 (2005).
Wong, S. B. J., Bos, R. & Sherman, L. A. Tumor-specific CD4+ T cells render the tumor environment permissive for infiltration by low-avidity CD8+ T cells. J. Immunol. 180, 3122–3131 (2008).
Bos, R. & Sherman, L. A. CD4+ T-cell help in the tumor milieu is required for recruitment and cytolytic function of CD8+ T lymphocytes. Cancer Res. 70, 8368–8377 (2010).
Zhu, Z. et al. CD4+ T cell help selectively enhances high-avidity tumor antigen-specific CD8+ T cells. J. Immunol. 195, 3482–3489 (2015).
Borst, J., Ahrends, T., Bąbała, N., Melief, C. J. M. & Kastenmüller, W. CD4+ T cell help in cancer immunology and immunotherapy. Nat. Rev. Immunol. 18, 635–647 (2018).
Andreatta, M. et al. An automated benchmarking platform for MHC class II binding prediction methods. Bioinformatics 34, 1522–1528 (2018).
Mittal, P. et al. Tumor-unrelated CD4 T cell help augments CD134 plus CD137 dual costimulation tumor therapy. J. Immunol. 195, 5816–5826 (2015).
Suri, A., Walters, J. J., Rohrs, H. W., Gross, M. L. & Unanue, E. R. First signature of islet β-cell-derived naturally processed peptides selected by diabetogenic class II MHC molecules. J. Immunol. 180, 3849–3856 (2008).
Tran, E. et al. Cancer immunotherapy based on mutation-specific CD4+ T cells in a patient with epithelial cancer. Science 344, 641–645 (2014).
Linnemann, C. et al. High-throughput epitope discovery reveals frequent recognition of neo-antigens by CD4+ T cells in human melanoma. Nat. Med. 21, 81–85 (2015).
Old, L. J. & Boyse, E. A. Immunology of experimental tumors. Annu. Rev. Med. 15, 167–186 (1964).
Wei, S. C. et al. Distinct cellular mechanisms underlie anti-CTLA-4 and anti-PD-1 checkpoint blockade. Cell 170, 1120–1133 (2017).
Gubin, M. M. et al. High-dimensional analysis delineates myeloid and lymphoid compartment remodeling during successful immune-checkpoint cancer therapy. Cell 175, 1014–1030 (2018).
Marzo, A. L. et al. Tumor-specific CD4+ T cells have a major “post-licensing” role in CTL mediated anti-tumor immunity. J. Immunol. 165, 6047–6055 (2000).
Bennett, S. R. M., Carbone, F. R., Karamalis, F., Miller, J. F. A. P. & Heath, W. R. Induction of a CD8+ cytotoxic T lymphocyte response by cross-priming requires cognate CD4+ T cell help. J. Exp. Med. 186, 65–70 (1997).
Corthay, A., Lundin, K. U., Lorvik, K. B., Hofgaard, P. O. & Bogen, B. Secretion of tumor-specific antigen by myeloma cells is required for cancer immunosurveillance by CD4+ T cells. Cancer Res. 69, 5901–5907 (2009).
Abelin, J. G. et al. Defining HLA-II ligand processing and binding rules with mass spectrometry enhances cancer epitope prediction. Immunity https://doi.org/10.1016/j.immuni.2019.08.012 (2019).
Mamitsuka, H. Predicting peptides that bind to MHC molecules using supervised learning of hidden Markov models. Proteins 33, 460–474 (1998).
Welch, L. R. Hidden Markov models and the Baum–Welch algorithm. IEEE Inf. Theory Soc. Newsl. 53, 10–13 (2003).
Jurtz, V. et al. NetMHCpan-4.0: improved peptide–MHC class I interaction predictions integrating eluted ligand and peptide binding affinity data. J. Immunol. 199, 3360–3368 (2017).
Jensen, K. K. et al. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology 154, 394–406 (2018).
Kowalewski, D. J. & Stevanović, S. Biochemical large-scale identification of MHC class I ligands. Methods Mol. Biol. 960, 145–157 (2013).
Tungatt, K. et al. Antibody stabilization of peptide-MHC multimers reveals functional T cells bearing extremely low-affinity TCRs. J. Immunol. 194, 463–474 (2015).
Acknowledgements
We thank all members of the Schreiber laboratory for discussions and technical support. This work was supported by grants to R.D.S. from the National Cancer Institute of the National Institutes of Health (RO1CA190700), the Parker Institute for Cancer Immunotherapy, the Cancer Research Institute, Janssen Pharmaceutical Company of Johnson and Johnson and the Prostate Cancer Foundation, and by a Stand Up to Cancer-Lustgarten Foundation Pancreatic Cancer Foundation Convergence Dream Team Translational Research Grant. Stand Up to Cancer is a program of the Entertainment Industry Foundation administered by the American Association for Cancer Research. E.A. and D.M.L were supported by a postdoctoral training grant (T32 CA00954729) from the National Cancer Institute. D.M.L. and M.M.G. were supported by the Irvington Postdoctoral Fellowship from the Cancer Research Institute. M.D. is a St Baldrick’s Scholar with support from Hope with Hazel and a Pew-Stewart Scholar for Cancer Research supported by the Pew Charitable Trusts. J.P.W. is supported by the National Cancer Institute of the National Institutes of Health Paul Calabresi Career Development Award in Clinical Oncology (K12CA167540). M.M.G. is supported by a Parker Bridge Scholar Award from the Parker Institute for Cancer Immunotherapy. K.W.W. receives support from the National Institutes of Health (R01CA238039). T.J. receives support from a National Institutes of Health Cancer Center Support Grant (P30CA14051) and the Howard Hughes Medical Institute. E.R.U. receives support from the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health (AI114551 and DK058177). Aspects of the studies, including ELISPOT, were performed by D. Bender at the Immunomonitoring Laboratory (IML), which is supported by the Andrew M. and Jane N. Bursky Center for Human Immunology and Immunotherapy Programs and the Alvin J. Siteman Comprehensive Cancer Center which, in turn, is supported by a National Cancer Institute of the National Institutes of Health Cancer Center Support Grant (P30CA91842).
Author information
Authors and Affiliations
Contributions
E.A. conceived and designed the experiments, collected the data, performed and interpreted the analyses, and wrote the manuscript. D.M.L. and A.P.M. planned experiments, and collected and analysed data. I.K. conceived of and designed the hmMHC algorithm and performed analyses using it, and wrote the methodological description. M.D. generated the KP9025 sarcoma cell line. A.M.L provided technical assistance and helped to plan experiments using MHC-II tetramers. W.M. and C.F.L. planned, performed and analysed mass spectrometry experiments. E.E. assisted with bioinformatics analyses. A.N.V. assisted with the generation of the CD4+ T cell hybridomas, and helped to design and perform experiments using them. D.R. designed experiments involving multi-colour flow cytometry and collected and analysed the data. J.P.W. provided technical support for MHC-I tetramer staining. M.M.G. assisted in experiment planning. R.F.V.M. collected and analysed data for experiments involving multi-colour flow cytometry. C.D.A., K.C.F.S. and J.M.W. provided technical assistance throughout the study. A.C. collected data. K.W.W. provided mITGB1–MHC-II monomers and provided assistance in experimental design. T.J. provided support in experimental design and data analysis regarding the KP9025 sarcoma line. M.N.A. conceived and designed the hmMHC algorithm and provided bioinformatics support. E.R.U. provided assistance with experimental design. R.D.S. conceived experiments, interpreted data, and wrote the manuscript. All authors contributed to manuscript revision.
Corresponding author
Ethics declarations
Competing interests
R.D.S. is a cofounder, scientific advisory board member, stockholder, and royalty recipient of Jounce Therapeutics and Neon Therapeutics and is a scientific advisory board member for A2 Biotherapeutics, BioLegend, Codiak Biosciences, Constellation Pharmaceuticals, NGM Biopharmaceuticals and Sensei Biotherapeutics. K.W.W. serves on the scientific advisory board of Tscan Therapeutics and Nextechinvest, and receives sponsored research funding from Bristol-Myers Squibb and Novartis; these activities are not related to the findings described in this publication. T.J. is a member of the Board of Directors of Amgen and Thermo Fisher Scientific. He is also a co-founder of Dragonfly Therapeutics and T2 Biosystems, and serves on the Scientific Advisory Board of Dragonfly Therapeutics, SQZ Biotech, and Skyhawk Therapeutics. None of these affiliations represent a conflict of interest with respect to the design or execution of this study or interpretation of data presented in this manuscript. The laboratory of T.J. currently also receives funding from the Johnson & Johnson Lung Cancer Initiative and Calico, but this funding did not support the research described in this manuscript.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Peer review information Nature thanks Lelia Delamarre, Cornelis Melief and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
Extended Data Fig. 1 The hmMHC predictive algorithm and IEDB′18 H2-I-Ab training dataset composition.
a, An example of a fully connected HMM with three hidden states, and emissions corresponding to amino acids. b–d, Composition of IEDB dataset (MHC full ligand export downloaded on 25 November 2018) represented as number of peptides per binding category and measurement type (b, c) and binding category and peptide length (d). Strong binders: IC50 ≤ 50 nM; binders: 50 nM < IC50 ≤ 500 nM; weak binders: 500 nM < IC50 ≤ 5,000 nM; non-binders: all remaining peptides.
Extended Data Fig. 2 Performance of hmMHC compared to netMHCII-2.3 and netMHCIIpan-3.2.
a, hmMHC (orange shapes) underwent 10× cross-validation. In each of the ten cross-validation partitions, on average there were 4,412 binders in the training set, and 771 binders and 77,086 random natural peptides in the validation set. Performance was compared in terms of AUROC to the performance of netMHCII-2.3 (blue triangles) and netMHCIIpan-3.2 (purple triangles) applied on the same validation sets. For hmMHC, performance for different numbers of hidden states is shown. For netMHCII-2.3 and netMHCIIpan-3.2, performance is shown for both predicted affinity and percentile rank (PR). b, Receiver operating characteristic (ROC) curves showing the performance of hmMHC on the H2-I-Ab dataset compared to existing predictors. ROC curves of all peptides and per specific peptide length for every cross-validation partition are shown. c, Illustration of percentile rank for strong binder classification calibrated on random natural peptides. Red lines indicate the percentile ranks of peptides screened for CD4+ T cell reactivity.
Extended Data Fig. 3 mITGB1 is a major MHC-II-restricted neoantigen in T3 sarcomas.
a, b, T3 MHC-II neoantigen predictions for all expressed mutations were made using hmMHC (a) and netMHCII-2.3 (b) (netMHCIIpan-3.2 predictions yielded very similar results, data not shown). The predictions are shown as −log10odds predictor value or logIC50 (smaller values indicate higher likelihood of being presented by I-Ab) and expression level (FPKM). Strong binders are defined as mutations residing in the second percentile of I-Ab binding predictions for random natural peptides for each algorithm (−log10odds ≤ 26.21 or logIC50 ≤ 343.8 nM). The N710Y mutation in ITGB1 met the strong binder threshold in the hmMHC predictions but not in the netMHCII-2.3 predictions. Red dots indicate all mutations that were screened for CD4+ T cell reactivity. Green line denotes high-expression cut-off (FPKM = 89.1). Blue line indicates strong binder cut off for each algorithm. c, Two million T3 sarcoma cells were injected subcutaneously into syngeneic mice and CD4+ TILs were isolated on day 12. IFNγ ELISPOT was performed using naive splenocytes pulsed with 2 μg ml−1 of the indicated peptides. Data are shown as mean of three independent experiments ± s.e.m. d, Gating strategy for pI-Ab tetramer staining of whole TILs. e, Quantification of mITGB1–tetramer and CLIP–tetramer staining of CD4+ T cells from whole T3 TILs 12 days after transplantation. Data are shown as mean ± s.e.m. per cent tetramer-positive cells of CD4+ cells from three independent experiments. f, Syngeneic 129S6 mice were injected subcutaneously with 2 × 106 T3 sarcoma cells and TIL-derived CD4+ T cells were collected 12 days after transplantation. CD4+ T cells were stimulated with naive splenocytes pulsed with 2 μg ml−1 OVA323–339 control or mITGB1697–724 peptide for a flow-based multi-cytokine array. Representative data from one of two independent experiments using pools of five tumours each are shown as average of technical triplicate wells from three pooled tumours.
Extended Data Fig. 4 T3 TIL-derived CD4+ T cell hybridomas are reactive against mITGB1.
CTLL assay of T3 TIL-derived CD4+ T cell hybridoma lines stimulated with naive splenocytes pulsed with 2 μg ml−1 of the individual indicated peptides. Representative data from one of three independent experiments are shown as average cpm from technical duplicate wells.
Extended Data Fig. 5 The mITGB1 epitope is presented on I-Ab.
a, T3 CD4+ T cell hybridomas were stimulated with 2 μg ml−1 mITGB1(710Y) or wild-type ITGB1(710N) peptide-pulsed splenocytes. Activation was measured by CTLL assay. Representative data from three independent hybridoma lines are shown as average of technical replicate wells. b, Mapping of the mITGB1 MHC-II binding core was performed using the CD4+ T cell hybridoma line 41 stimulated with naive splenocytes pulsed with 2 μg ml−1 of overlapping peptides covering mITGB1697–724. Red denotes the T3-specific mutant amino acid at position p1 of the minimal epitope; underlining denotes the validated binding core. Green amino acids represent random residue substitutions used to specifically define valines at residues 715 and 718 as the p6 and p9 MHC-II binding positions and the complete MHC-II binding core. Representative data from two independent experiments are shown as the average of technical triplicate wells. c, MHC-II I-Ab staining of parental T3 cells, IFNγ-stimulated T3 cells and T3 cells transduced with a vector encoding CIITA (T3.CIITA). Representative data from one of three independent experiments are shown. d, Mirror plot showing match between MS/MS spectra of the 14-mer peptide sequence encompassing the N710Y site of mITGB1 eluted from T3.CIITA cells (positive axis) and a corresponding synthetic peptide (negative axis). Labelled m/z values reflect those experimentally observed for the endogenous peptide, with peaks representing b ions highlighted in blue and y ions in red.
Extended Data Fig. 6 mITGB1 CD4+ T cells are required for tumour rejection in response to ICT.
a, Comparison of total number of expressed missense mutations between ten different MCA-induced sarcomas and KP9025 cells. Mutations were defined by whole-exome sequencing and RNA sequencing, and mutational load is shown on a per cell basis. b, Comparison of predicted neoantigen MHC-I affinity values between KP9025 and MCA-induced sarcoma F244 for H-2Db (top) and H-2Kb (bottom). KP9025 cells were not predicted to express any MHC-I neoantigens. c, Rag2−/− mice were subcutaneously injected with 1 × 106 KP.mLAMA4, KP.mITGB1, KP.mLAMA4.mITGB1 or KP.mSB2.SIINFEKL cells. Representative data from one of two independent experiments are presented as tumour diameters from individual mice (n = 5 mice per group for KP.mLAMA4, KP.mITGB1 and KP.mLAMA4.mITGB1 and n = 3 mice for KP.mSB2.SIINFEKL group per experiment). d, Wild-type syngeneic 129S4 mice were injected subcutaneously with 1 × 106 KP.mLAMA4, KP.mITGB1 or KP.mLAMA4.mITGB1 cells and treated with anti-PD-1 (top) or anti-CTLA single agent ICT (bottom) on days 3, 6, and 9 after transplantation. Representative data from one of three independent experiments are shown as tumour diameters from individual mice (n = 5 in all groups per experiment). e, Survival curves for experiments in d and Fig. 2e (n = 15 in all groups).
Extended Data Fig. 7 Outgrowth of nonimmunogenic sarcoma cells expressing MHC-I neoantigens is not a result of cancer immunoediting.
a, Rag2−/− or wild-type 129S4 mice were injected with 1 × 106 KP9025 or KP.mLAMA4 cells and treated with anti-PD-1, anti-CTLA or anti-PD-1 + anti-CTLA4 on days 3, 6 and 9 after injection. Tumours were removed once they reached a maximum diameter of 20 mm in any direction and sarcoma cell lines were established ex vivo. Cell lines were stimulated with IFNγ to upregulate MHC-I and subsequently used to stimulate the mLAMA4-specific CD8+ 74.14 T cell clone. Secretion of IFNγ by T cells was measured using enzyme-linked immunosorbent assay (ELISA). Representative data from two independent experiments are represented as the average of two independent tumour samples in each group. b, Wild-type 129S4 mice were injected with 1 × 106 KP.mSB2.SIINFEKL cells and treated with anti-PD-1 + anti-CTLA4 combination ICT on days 3, 6 and 9 after injection. Tumours were removed as in a. Established ex vivo cell lines were cloned by limiting dilution and parental KP.mSB2.SIINFEKL cells or individual clones from outgrown tumours were used to stimulate the mSB2-specific C3 CD8+ T cell clone; production of IFNγ was quantified using ELISA. Representative data from four independent experiments are presented as average IFNγ concentration of eight individual clones ± s.e.m. Significance was determine using an unpaired, two-tailed t-test. c, Cell surface staining of SIINFEKL-H-2-Kb expressed by unstimulated or IFNγ-stimulated parental KP.mSB2.SIINFEKL cells or individual clones described in b. A representative histogram is shown. d, Quantification of mean ± s.e.m. SIINFEKL-H-2-Kb mean fluorescence intensity (MFI) from eight individual clones in c. NS, not significant. e, Survival curves of wild-type 129S4 mice injected subcutaneously with 1 × 106 KP.mSB2.SIINFEKL.mITGB1 cells. Mice were treated with control monoclonal antibodies or anti-PD-1 + anti-CTLA4 combination ICT on days 3, 6 and 9 after injection. n = 10 mice per group from two independent experiments. ****P = 1.5 × 10−5, Mantel–Cox test.
Extended Data Fig. 8 mITGB1-specific CD4+ T cells display an activated TH1 phenotype.
a, Whole TILs from KP.mLAMA4.mITGB1 tumours 12 days after transplantation were stained with mITGB1-I-Ab tetramers. Populations were previously gated on viable CD11b−CD4+ cells. Representative data from one of two independent experiments with five pooled tumours each are shown. b, mITGB1-I-Ab tetramer-negative and tetramer-positive cells described in a were analysed for expression of T-BET and FOXP3. Representative plots are shown. c, Quantification of two independent experiments in b as average per cent of tetramer-negative and tetramer-positive cells staining positive for the indicated protein. Tumour-bearing mice were treated with control monoclonal antibodies or anti-CTLA4 on days 3, 6, and 9 after transplantation where indicated. d, mITGB1-I-Ab tetramer-positive and tetramer-negative cells in a were analysed for expression of PD-1. Representative plots are shown. e, Quantification of two independent experiments described in d shown as average per cent of tetramer-negative and tetramer-positive cells staining positive for PD-1. f, mITGB1-I-Ab tetramer-positive cells in a were analysed for expression of the indicated proteins. Representative histograms from one of two independent experiments using pools of five tumours each are shown.
Extended Data Fig. 9 CD4+ T cell help is required at the tumour site during primary and memory responses.
a, Rag2−/− mice were simultaneously injected with 1 × 106 KP.mLAMA4 and KP.mLAMA4.mITGB1 cells on opposite flanks. Representative data from one of two independent experiments are shown as individual tumour diameter (n = 3 in each experiment). b, Wild-type 129S4 mice were injected with 1 × 106 KP.mITGB1 cells and were treated with anti-PD-1 + anti-CTLA4 combination ICT on days 3, 6, and 9 after injection. Representative data from one of two individual experiments are shown as individual tumour diameters (n = 5 in all experiments). c, Wild-type 129S4 mice were simultaneously injected with 1 × 106 KP.mLAMA4 and KP.mLAMA4.mITGB1 cells on opposite flanks and treated as in b. Representative data from one of two individual experiments are shown as individual tumour diameters (n = 5 in all experiments). d, Wild-type 129S6 mice were injected subcutaneously with 2 × 106 T3 sarcoma cells and were treated with anti-PD-1 + anti-CTLA4 combination ICT on days 3, 6, and 9 after injection. Following tumour rejection and a 30-day recovery period, tumour-experienced mice were rechallenged with 2 × 106 T3 cells in the presence of control monoclonal antibody or CD4-depleting antibody, or with irrelevant sarcoma cells. Representative data from one of two independent experiments are shown as average tumour diameter ± s.e.m. (n = 5 in all groups per experiment). e, Wild-type 129S4 mice were injected subcutaneously with 1 × 106 KP.mLAMA4.mITGB1 cells followed by surgical resection 10 days after transplantation. After a 30-day recovery period, tumour-experienced mice were rechallenged with 1 × 106 KP9025, KP.mLAMA4.mITGB1, or KP.mLAMA4 cells. Representative data from one of two independent experiments are shown as average tumour diameter ± s.e.m. (n = 5 in all groups per experiment). ****P = 2 × 10−6 by two-way ANOVA with multiple comparisons and Bonferroni correction. f, Quantification of data from three independent experiments in Fig. 5c is shown as average number of spots ± s.e.m. (left) and average number of mITGB1-specific CD4+ cells ± s.e.m. (right). **P = 0.003, ****P = 7.2 × 10−5 (unpaired, two-tailed t-test). g, CD45+Ly6G−MHCII+CD64+CD25−CD11b+F4/80+ macrophages in TILs from mice bearing the indicated contralateral tumours were analysed for expression of iNOS 11 days after tumour transplant. Representative data from four independent experiments are shown. h, Quantification of iNOS+ macrophages from experiments in f as a per cent of total CD45+ cells. Data are shown as average ± s.e.m. of four independent experiments. *P = 0.03 by unpaired, two-tailed t-test. i, CD45+Ly6G−MHCII+CD64+CD25−CD11b+F4/80+ macrophages from the indicated contralateral tumours described were isolated 11 days after transplantation and analysed for expression of iNOS. Representative plots from two independent experiments are shown. j, Quantification of iNOS+ macrophages from two independent experiments in h is shown as average per cent of total CD45+ cells.
Extended Data Fig. 10 Gating strategies for multi-colour flow cytometry.
Gating strategies for multi-colour flow cytometry analysis of tumour-infiltrating macrophage (a) and T cell (b) populations.
Supplementary information
41586_2019_1671_MOESM1_ESM.pdf
Supplementary Table 1 Supplementary Table 1: Characteristics of the screened predicted MHC class II neoantigens. Sequences of mutant peptides identified in T3 sarcomas used in the screening experiments
41586_2019_1671_MOESM3_ESM.pdf
Supplementary Table 2 Supplementary Table 2: Expressed mutations in the KP9025 sarcoma cell line. The four non-synonymous point mutations expressed in the KP9025 sarcoma cell line are shown as the corresponding amino acid substitution
41586_2019_1671_MOESM4_ESM.xlsx
Supplementary Data 1 Supplementary Data 1: T3 nucleotide variant calls. This file contains all the single nucleotide variants determined to generate missense mutations for the methylcholanthrene (MCA)-induced T3 sarcoma cell line as detected by cDNA capture sequencing
41586_2019_1671_MOESM5_ESM.xls
Supplementary Data 2 Supplementary Data 2: KP9025 nucleotide variant calls. This file contains all the single nucleotide variants determined to generate missense mutations for the oncogene driven KP9025 sarcoma cell line as detected by cDNA capture sequencing
Source data
Rights and permissions
About this article
Cite this article
Alspach, E., Lussier, D.M., Miceli, A.P. et al. MHC-II neoantigens shape tumour immunity and response to immunotherapy. Nature 574, 696–701 (2019). https://doi.org/10.1038/s41586-019-1671-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41586-019-1671-8
This article is cited by
-
Regulatory T cell-mediated immunosuppression orchestrated by cancer: towards an immuno-genomic paradigm for precision medicine
Nature Reviews Clinical Oncology (2024)
-
Discovery of tumor-reactive T cell receptors by massively parallel library synthesis and screening
Nature Biotechnology (2024)
-
Class II HLA-DRB4 is a predictive biomarker for survival following immunotherapy in metastatic non-small cell lung cancer
Scientific Reports (2024)
-
The effect of human leukocyte antigen genotype on survival in advanced prostate cancer treated with primary androgen deprivation therapy: the KYUCOG-1401-A study
Prostate Cancer and Prostatic Diseases (2024)
-
Soluble lymphocyte activation gene-3 (sLAG3) and CD4/CD8 ratio dynamics as predictive biomarkers in patients undergoing immune checkpoint blockade for solid malignancies
British Journal of Cancer (2024)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.