Disruption of splicing patterns due to mutations of genes coding splicing factors in tumors represents a potential source of tumor neoantigens, which would be both public (shared between patients) and tumor-specific (not expressed in normal tissues). In this study, we show that mutations of the splicing factor SF3B1 in uveal melanoma generate such immunogenic neoantigens. Memory CD8+ T cells specific for these neoantigens are preferentially found in 20% of patients with uveal melanoma bearing SF3B1-mutated tumors. Single-cell analyses of neoepitope-specific T cells from the blood identified large clonal T-cell expansions, with distinct effector transcription patterns. Some of these expanded T-cell receptors are also present in the corresponding tumors. CD8+ T-cell clones specific for the neoepitopes specifically recognize and kill SF3B1-mutated tumor cells, supporting the use of this new family of neoantigens as therapeutic targets.

Significance:

Mutations of the splicing factor SF3B1 in uveal melanoma generate shared neoantigens that are uniquely expressed by tumor cells, leading to recognition and killing by specific CD8 T cells. Mutations in splicing factors can be sources of new therapeutic strategies applicable to diverse tumors.

This article is highlighted in the In This Issue feature, p. 1861

Activation of the immune system by various immunomodulators can eradicate large solid tumors as well as the disseminated disease. Tumor destruction is most often related to the recognition by T cells of peptides derived from somatically mutated proteins expressed by cancerous cells and presented by MHC molecules. Indeed, clinical response to anti-checkpoint treatments is loosely correlated with the number of somatic mutations present in the tumor, suggesting that the neoepitope load expressed by the tumors is important to generate an efficient immune response (1). However, most of these neoepitopes correspond to passenger mutations that are different in each tumor and thus are unique to each patient. In the absence of spontaneous responses that can be therapeutically amplified (by immune checkpoint blockade, for example), inducing T cells to such epitopes requires personalized vaccines, which are costly and logistically complicated to implement (2). For these reasons, public (shared between individuals) epitopes deriving from germline-encoded antigens that are aberrantly expressed in tumors with limited expression in normal tissues, such as onco-testis antigens, are often used in vaccine strategies (3). However, many of these antigens are also expressed in the thymus, potentially leading to deletion of the high-avidity antigen-reactive T cells (4).

An alternative vaccine strategy would be to target neoepitopes resulting from recurrent cancer-specific mutations in splicing factors (5, 6). Such mutations lead to aberrant open reading frames transcribed specifically in tumor cells. The inclusion of intronic sequences in mature transcripts leads to frameshifts in the following exons or less often to an in-frame insertion of few codons, eventually encoding potential neoepitopes. When presented by the patient MHC alleles, the resulting peptides could generate neoepitopes common to all patients bearing both the mutation in the splicing factor and a given MHC allele, opening the way for generic therapeutic products adapted to common MHC haplotypes. Although proposed for several years (7), the clinical relevance of neoepitopes resulting from mutated splicing factors in tumors has not yet been demonstrated in humans.

One possible explanation is the technical difficulties in obtaining evidence for both the abnormal mRNA and the protein/peptide products. Splicing factor mutations lead to the addition or deletion of a variable number of nucleotides in hundreds of different intron/exon junctions that are specific to the particular mutated splicing factor and tissue. The abnormal mRNA products may not be easily detected because the resulting frameshifts that are present in two thirds of the alternative junctions often lead to premature translation termination and nonsense-mediated mRNA decay (NMD; ref. 8). The proteins or polypeptides resulting from splicing anomalies may be in low amounts, as the abnormal proteins are probably short-lived. However, from an immunologic perspective, the abnormal polypeptides generated by a premature stop codon are detected during the first-pass translation quality check by the ribosomes. These abnormal polypeptides represent defective ribosomal products (DRiP) that are preferentially loaded on the MHC class I molecules (9, 10). In the absence of a premature stop codon, the proteins harboring the amino-acid insertion may be misfolded and are also rapidly degraded and targeted to the MHC class I loading compartment (10). Both processes efficiently generate potential neoepitopes (10).

Whereas the detection of abnormal mRNA is relatively easy and sensitive, quantifying the amounts of the abnormal proteins or peptides is challenging and not sensitive. Biochemical methods require large batches of cells bearing the splicing factor mutations, and the polypeptides should be abundant, which is most often not the case, as seen above. Moreover, during in vitro culture, the cells bearing mutations in splicing factors may be less fit and at a growth disadvantage (11). As for mass spectrometry (MS)–based methods looking for either the abnormal proteins in the cells or the peptides loaded on MHC molecules, large numbers of cells are also a prerequisite and MS is 103 to 106 times less sensitive than T cell–based technologies. Indeed, CD8+ T cells can efficiently detect as few as 2 to 20 specific MHC:peptide complexes among the 105 to106 MHC molecules found on a mammalian cell. Thus, T-cell assays can be used to provide evidence of the presence of mutated splicing factor–related neoepitopes with exquisite sensitivity.

In vivo CD8 T-cell priming by such neoepitopes requires cross-presentation by professional antigen-presenting cells, which should capture the tumor cell debris, process the abnormal polypeptide, and transport the peptides to the MHC class I molecule loading compartment. Because of the allelic polymorphism of the molecules involved and potential peptide competition for loading, this process, which is not very efficient per se, is also highly variable from one patient to another (12, 13). Thus, the CD8 T cells may target different neoepitopes in patients bearing the same MHC class I allele(s). Moreover, expression of the neoepitopes in the absence of costimulation may lead to deletion of the specific T cells.

Because of the extreme diversity of the T-cell receptor (TCR) repertoire, it is expected that any individual harbors a naïve repertoire toward any MHC:peptide complex (14). MHC:peptide tetramer–based technologies are able to fish out these very low-frequency (10−5 to 10−6) CD8 T cells if a high enough number of T cells (>107) are studied (15). In patients bearing a tumor with a splicing factor mutation, some of the neoepitope-specific CD8 T cells should be primed and, if not subjected to peripheral tolerance by deletion, remain in the blood as effector/memory cells. Thus, the clinical relevance of splicing factor mutations in tumor immunology would be suggested by showing that some patients bearing tumors with such mutations harbor CD8 T cells recognizing the specific MHC:peptide complexes. The therapeutic potential of such neoepitopes would then be supported by the ability of neoepitope-specific CD8 T cells to directly recognize tumor cells.

Uveal melanoma is a rare disease (∼600 cases/year in France) with a dismal prognosis once metastatic, which occurs in more than 30% of cases and for which no therapy is currently available (16). Contrary to skin melanoma, uveal melanomas display very few somatic mutations (16 ± 4.0 per exome; ref. 17) and are accordingly resistant to anti-checkpoint immunomodulation. However, the characterization of the T-cell response in the tumor or the blood of patients with uveal melanoma evidenced oligoclonal T-cell expansion suggestive of an immune response toward unknown antigen(s) (18, 19). Twenty percent of the tumors harbor a mutation in the SF3B1 gene, generating more than 1,400 new splice junctions (20, 21). SF3B1 mutations induce an upstream shift of the splice acceptor sites, leading to inclusion of intronic sequences in the mRNA. The resulting additional amino acids and the frameshift that is often associated potentially generate a large number of public neoepitopes.

In this work, we show that among patients with metastatic uveal melanoma, some of those whose tumors harbored a mutated SF3B1 displayed oligoclonal memory CD8+ T cells with specificities for SF3B1mut-derived neoepitopes. Several of the corresponding TCRs could also be found in the tumor. SF3B1mut uveal melanoma cell lines were recognized and killed by neoepitope-specific T-cell clones, demonstrating that these neoepitopes are expressed by tumor cells in a way that can be recognized by CD8+ T cells.

Selection of the Patients and Identification of SF3B1mut-Induced Epitopes

Because of the ≈50% HLA-A2 allele frequency in the European population, 13 HLA-A2+ patients with metastatic uveal melanoma (8 SF3B1mut and 5 SF3B1WT) and 4 HLA-A2+ healthy donors were studied. We first verified that the tumors were correctly classified by resequencing SF3B1 in tumors (Supplementary Table S1). RNA sequencing (RNA-seq) of the tumors allowed us to measure the proportion of SF3B1mut-modified intron–exon junctions as compared with the SF3B1WT tumors (Fig.1A). The pattern of junctions according to SF3B1 status reproduced our previous results (20). We then analyzed the pattern of the reads encompassing the different junctions according to the nucleotide distance toward the canonical splicing acceptor site (Fig.1B). The number of reads leading to in-frame junctions was 55% instead of the expected 33%, indicating a loss of out-of-frame mRNA species. The distances peaked at –18 nt and a clear excess of multiples of 3 nt distances was apparent, confirming that out-of-frame junctions were often subjected to NMD.

For the junctions corresponding to neoepitopes (#14, #17, #18, #26, #37) that will be studied further below, using an RT-PCR including one fluorescent primer followed by capillary electrophoresis, we quantified the wild-type (WT) and alternative junction mRNA in 2 HLA-A2+ patient-derived xenografts (PDX): 1 SF3B1mut and 1 SF3B1WT PDXs (Supplementary Table S1; Fig.1C). The alternative (bigger) junctions were clearly expressed in SF3B1mut tumors (3%–>50% of the total junctional mRNA) but were not found in SF3B1WT tumors. We also measured these WT and alternative mRNA junctions in SF3B1WT (MP41) and SF3B1mut (Mel202) uveal melanoma cell lines in the presence or absence of puromycin, a protein synthase inhibitor that inhibits NMD (22). After puromycin treatment, the mRNA corresponding to #14 and #17 alternative junctions increased from 14% to 48% and 8% to 14%, respectively, confirming that the premature stop codon found in these two junctions leads to NMD (Fig.1D).

The alternative junctions may give rise to proteins, which may however be unstable and difficult to detect using biochemical methods. As a proof of concept, we focused on the NET1 protein, in which an alternative junction is predicted to lead to an alternative transcript encoding a polypeptide generating epitope #26. An immunoprecipitating antibody for a domain upstream of the alternative coding-frame aberrant-splicing event was available (detailed in the Methods and Supplementary Methods) and used to immune-precipitate endogenous NET1 from two uveal melanoma cell lines (MP41 and Mel202; SF3B1WT and SF3B1mut, respectively) as well as two PDXs (PDX-MP41 and PDX-MM267; SF3B1WT and SF3B1mut, respectively). After trypsin digestion, the immunoprecipitated samples were analyzed by targeted MS. As shown by their retention time and MS fingerprints comparatively similar to the synthetic peptides, the two predicted peptides corresponding to the aberrant reading frame were readily detected in SF3B1mut cells (Mel202 and PDX-MM267) but absent in the SF3B1WT cells (MP41 and PDX-MP41), whereas the WT NET1 peptides were present in all samples (Fig.1E; Supplementary Fig. S1A–S1E). These findings obtained for NET1 demonstrate that aberrantly spliced transcripts in SF3B1mut cells can be translated into detectable aberrant protein products.

To test whether these polypeptides could be processed and loaded onto MHC molecules and detected by T cells, we first evaluated all the 9 amino-acid (AA) long peptides induced by SF3B1mut-neoreading frames for their avidity toward the most common MHC class I alleles according to a publicly available bioinformatics pipeline (23). We selected the 0.5% most avid HLA-A0201 restricted peptides (n = 43; Fig.1F; Supplementary Table S2) and generated the corresponding HLA-A2:peptide complexes using empty monomers (24). We verified their stability (Supplementary Fig. S2A) and discarded 4 peptides with low HLA-A2 binding, allowing us to make tetramers able to detect specific T cells for 39 HLA-A2 restricted SF3B1mut-related peptides.

Frequency and Phenotype of SF3B1mut-Related HLA Epitopes Reacting CD8 T Cells

In order to detect and characterize a potential immune response against the SF3B1mut-related neoepitopes, we stained blood CD8+ T cells from patients with uveal melanoma and healthy donors with tetramers labeled with two different fluorochromes to increase specificity and thereby sensitivity. We used as controls HLA-A2 tetramers loaded with pp65 from cytomegalovirus (CMV), with a flu peptide and with Melan-A, a melanocyte differentiation antigen (Fig.2A; Supplementary Fig. S2B). In two healthy donors, flu-specific CD8 T cells were clearly visible and displayed an effector (CD45RACCR7) or memory (CD45RACCR7+) phenotype (Supplementary Fig. S2B), whereas all HLA-A2:Melan-A tetramer+ (A2:Melan-A) CD8+ T cells were naïve (CD45RA+CCR7+). In a CMV+ healthy donor, a well-defined cluster of HLA-A2-CMV tetramer+ (A2:CMV) CD8+ T cells was present in the blood and displayed a similar effector memory phenotype (Supplementary Fig. S2B). In patients with uveal melanoma, a large proportion of A2:Melan-A CD8+ T cells displayed an effector/memory phenotype (Fig.2A) as previously reported (25), confirming that the immune system is stimulated by this melanoma differentiation antigen. We then analyzed the frequency and phenotype of CD8+ T cells specific for the SF3B1mut-related neoepitopes. The frequency of CD8+ T cells specific for the SF3B1mut-related epitope HLA-A2:peptide 37 (A2:37) was very high in both patients with uveal melanoma and healthy donors, similarly to A2:Melan-A (Fig.2A and B). Notably, these A2:37-specific CD8+ T cells were naïve in both healthy donors and SF3B1WT patients with uveal melanoma, but >40% of these cells were effector or memory in 7 out of 8 SF3B1mut patients with uveal melanoma (Fig.2A and C). These results indicate that the SF3B1mut-derived A2:37 epitope has directly or indirectly been recognized by the immune system in SF3B1mut patients with uveal melanoma.

The frequency of CD8+ T cells specific for the 39 HLA-A2 restricted SF3B1mut neoepitopes varied between <0.0001% and 0.3% in the patients with SF3B1mut tumors (Fig.2B). Notably, patient UM2 harbored increased frequencies of CD8 T cells specific for several epitopes, suggesting a coordinated immune response toward SF3B1mut-derived neoepitopes. Moreover, the frequency and/or the proportion of cells with an effector/memory phenotype in CD8+ T cells specific for several SF3B1mut-related neoepitopes appeared increased in some patients with SF3B1mut tumors in comparison with both healthy subjects and SF3B1WT patients with uveal melanoma (Fig.2C), suggesting an immune response toward several SF3B1mut-related neoepitopes.

Characterization of T Cells Specific for the SF3B1mut-Induced A2 Restricted Neoepitopes in Three Patients with Uveal Melanoma

We first focused on A2:37-specific CD8+ T cells from patient UM1 as both the frequency and the proportion of cells with an effector/memory phenotype was increased (Fig.2C). We isolated A2:37-specific CD8+ T cells from the blood of patient UM1 to analyze their transcriptome and TCR repertoire by single-cell RNA sequencing (scRNA-seq) coupled to TCR analysis. After quality control and filtering, 3213 A2:37-specific CD8+ T cells could be divided into 7 expression clusters (Fig.3A; Supplementary Fig. S3A; Supplementary Table S3). Cluster #1 (n = 355 cells) expressed SELL and LEF1, characteristic of naïve CD8+ T cells (Fig.3B), and probably corresponds to the 8% naïve A2:37-specific T cells found in this patient by flow cytometry (Fig.2A). Cluster #2 (n = 1,416) displayed a cytotoxic (GZMK/H; Fig.3B) and a central memory (CCR7) phenotype (Supplementary Fig. S3A). Cluster #3 (n = 995) included cells expressing ZNF683 (HOBIT) and ITGB1 (CD49a), both associated with tissue residency and expressed XCL1 and XCL2, two chemokines secreted by CD8+ T cells to attract dendritic cells (Fig.3B). Cluster #4 (n = 235) shared many features with clusters #2 and #3 but also expressed FOS, JUNB, CD69, NR4A1, NR4A2, TNF, and IFNG, indicating TCR activation (Fig.3B; Supplementary Fig. S3A; Supplementary Table S3). Cluster #5 (n = 110) expressed CCR4 (implicated in homing to tissues), TNFRSF4 (CD134), implicated in T-cell survival and helper function, as well as intermediate level of SELL and CCR7, but very low levels of cytotoxic or chemokine molecules, compatible with a circulating helperlike function. Cluster #6 (n = 56) expressed CCR9, CCR6, KLRC1, and CCL5 (Fig.3B). Cluster #7 is small (n = 46) and expressed TCRγ genes and may encompass γδ T cells fished out by the HLA-A2:37 tetramer due to the intrinsic cross-reactivity of TCRs. Among the 2,780 cells in which at least one TCRα and/or TCRβ chain was retrieved after filtering out doublets and artifacts, one clone was strikingly expanded and represented ∼80% (2,259) of the cells (Supplementary Table S4). Interestingly, this clone represented most if not all of the cells in the neighboring clusters #2 through #4 (Fig.3C), suggesting that clonal circulating cells specific for a given epitope may display distinct but related functional and trafficking features. The second and third most abundant clones (50 and 40 cells) represented most of the circulating helperlike cluster #5 cells, whereas the fourth clone (15 cells) was found in the CCR6/9 cluster #6 (Fig.3C). The TCRs retrieved only once (singletons) encompassed both naïve cluster #1 cells (n = 265) and various effector memory clusters (n = 94; Fig.3D). Altogether, these results indicate that T-cell clones specific for one given tumor neoepitope display several differentiation patterns associated with the expression of distinct TCRs. Moreover, for the most abundant clonotype #1, cells were either cytotoxic (cluster #2) or tissue resident (cluster #3) with a further portion being activated (cluster #4).

Using FACS of oligo-barcoded (Total-Seq) tetramers followed by 5′ transcriptome and TCR single-cell 10X technology, we analyzed the transcriptome and TCR repertoire of the blood CD8+ T cells specific for A2:18 in patient UM3 and 4 SF3B1mut-related specificities (A2:14, A2:17, A2:26, and A2:37) in patient UM2, three of which were strikingly increased in the blood of this patient (Fig.2B and C). After quality control and filtering, 3,231 cells could be divided into 7 clusters (Fig.3E), whose transcriptome patterns corresponded to various effector/memory clusters and one naïve subset (#1). The CD8+ T cells specific for UM2 A2:14, A2:17, A2:26, and UM3-A2:18 displayed transcriptome patterns corresponding to various types of effector or memory subsets with very few naïve cells (Fig.3E and F; Supplementary Fig. S3A–S3C). In contrast, the A2:37 specificity encompassed about 50% naïve and 50% effector cells (Fig.3E and F) in agreement with the cytometry data (Fig.2C). Repertoire analysis demonstrated large TCR expansions, making up to 94% of the T cells for a given specificity (Fig.3G; Supplementary Fig. S3D; Supplementary Table S4). As in patient UM1, each clonotype expressed a particular transcriptome pattern, displayed by the corresponding colors in Fig.3F and Supplementary Fig. S3D. The number of expanded T-cell clones was higher for UM2-A2:37, but their size was smaller. For UM2-A2:37, we also observed a large number of nonrecurrent TCRs (singletons), 20% of them expressing various effector/memory transcriptome patterns and 80% being naïve (Fig.3F). Altogether, expanded T-cell clones specific for 5 SF3B1mut-related specificities expressing distinct effector/memory transcriptome patterns were found in the blood of the 3 patients indicating previous contact with antigen.

T Cells Specific for SF3B1mut-Induced Neoepitopes Are Also Found in the Tumor

To determine in patient UM1 whether A2:37-specific T cells were present in the tumor, we amplified all the TCRα and TCRβ transcripts in UM1 liver metastasis RNA and deep sequenced them (Supplementary Table S5). Fifteen A2:37-specific TCRs found in the blood were also detected in the tumor as indicated by an asterisk (*) in Fig.3D. Notably, 5 out of 14 (36%) recurrent TCRs in the blood, including the most abundant one, #1, were found in the tumor (Supplementary Table S6). The 10 TCR singletons found in the tumor belonged to the effector memory clusters. These results indicate that some of the most expanded T-cell clones and singletons from blood are found in the uveal melanoma liver metastasis. Interestingly, the clonally expanded blood T cells whose TCR was found in the tumor belonged to 6 out of 7 of the transcriptional clusters, indicating various differentiation patterns. These results suggest that the effector/memory A2:37-specific T cells found in blood, some of which are highly amplified, may represent an ongoing antitumor immune response at the tumor site. In patient UM2, although their clonal size was much smaller than in patient UM1, only TCRs corresponding to A2:37 specificities were found in the tumor (Supplementary Table S6). Most of these TCRs belonged to effector or memory clusters with various transcriptome patterns (Fig.3F; Supplementary Fig. S3D). Interestingly, in patient UM2 harboring increased frequency of memory CD8+ T cells specific for several SF3B1mut-related neoepitopes, an unusual lymphoid infiltrate was observed in the periphery of the liver metastasis (Fig.3H). Thus, some of the expanded clonotypes found in the blood were also observed in the tumors of two patients harboring increased frequency of memory SF3B1mut neoepitope-specific CD8+ T cells, suggesting an active immune response toward SF3B1mut-related neoepitopes at the tumor site.

SF3B1mut-Induced Neoepitopes on Tumor Cells Are Recognized by Specific CD8 T cells

To determine whether the SF3B1mut-induced neoepitopes were presented on the surface of the tumor cells themselves, we generated tools to analyze the direct interaction between the neoepitope-specific CD8 T cells and tumor cells. We transduced the available HLA-A2negSF3B1mut uveal melanoma cell line Mel202 with an HLA-A2 expression vector. We also generated an isogenic negative control by inserting a DEGRON sequence into the SF3B1-mutated allele using CRISPR/Cas9 technology, normalizing to some extent the splicing pattern (11). The resulting SF3B1WT cell line was also transduced with HLA-A2 (Supplementary Fig. S4A and S4B). In parallel, we generated T-cell clones for 5 SF3B1mut neoepitopes (A2:14,A2:17; A2:18, A2:26, and A2:37) by direct FACS assisted single-cell cloning of tetramer+ CD8+ T cells from both healthy donors (HD) and patients with uveal melanoma, followed by expansion and verification of the specificity using the relevant tetramers (Supplementary Fig. S4C). Twenty-three clones were obtained (Supplementary Table S7). The T-cell clone functionality was verified by stimulating them with HLA-A2 transfected K562 loaded with the relevant peptide followed by assessment of IFNγ, granzyme B (GzmB), and TNF release (Supplementary Fig. S4D). For each specificity, the clones with the highest sensitivity (activated by the lowest peptide concentration) were selected for the following functional assays.

SF3B1mut and SF3B1WT HLA-A2+ or HLA-A2 Mel202 as well as the SF3B1WT HLA-A2+ MP41 uveal melanoma cell lines were used to stimulate the T-cell clones. Addition of the relevant peptide was used as positive control. An A2:18-specific T-cell clone from an HD was specifically activated by SF3B1mut and not by SF3B1WT Mel202 uveal melanoma cells as seen by CD25 upregulation (Fig.4A). Similarly, this clone was activated (CD25) by a SF3B1mut but not SF3B1WT HLA-A2+ tumors from uveal melanoma PDX (Fig.4B and C). CD107a expression indicating degranulation and suggesting cytotoxicity was also upregulated after SF3B1mut but not SF3B1WT HLA-A2+ uveal melanoma PDX stimulation (Fig.4D). Similar results were observed with 4 additional clones specific for 4 other specificities: HD:A2:14, HD-A2:17, UM2-A2:26, and UM1-A2:37 (Supplementary Fig. S5A–S5D).

Clone HD-A2:18 also secreted more GzmB after stimulation by SF3B1mut cells in comparison with SF3B1WT Mel202 UM cells in an HLA-A2–dependent manner (Fig.4E), suggesting possible cytotoxic activity. We therefore examined whether SF3B1mut Mel202 cells would be directly killed by clone HD-A2:18. After 4 hours with the clone, a small but significant increase in dead (7AADpos) tumor cells was observed for HLA-A2+SF3B1mut Mel202 cells as compared with SF3B1WT HLA-A2+ (MP41) or HLA-A2SF3B1mut Mel202 UM cells (Fig.4F and G), indicating specific and HLA-A2–dependent killing. These results demonstrate that the 5 SF3B1mut neoepitopes for which a memory T-cell response was observed in the peripheral blood of patients with uveal melanoma can be directly recognized on tumor cells by CD8 T cells.

Importantly, the HD-A2:18 clone specifically killed SF3B1mut HLA-A2+ Mel202 UM cells and not the SF3B1WT isogenic cells, whereas an irrelevant A2-CMV clone did not (Fig.4H; Supplementary Fig. S5E). Increasing the effector:target ratio increased the cytotoxic activity (Supplementary Fig. S5F). Finally, 2 T-cell clones from 2 patients with metastatic uveal melanoma (UM1-A2:26 and UM1-A2:37) killed SF3B1mut HLA-A2+ Mel202 uveal melanoma cells and not the SF3B1WT isogenic cells (Fig.4I). Thus, SF3B1mut-related neoepitopes stimulate the immune system of patients with uveal melanoma at the metastatic stage to generate circulating CD8+ T cells that are specific for the tumor neoepitopes and can directly recognize and kill tumor cells.

In this work, we show that splicing alterations caused by mutations in the splicing factor gene SF3B1 in uveal melanoma tumors generate MHC class I restricted tumor neoepitopes that are detected by the patient CD8 T cells. Expanded T cells specific for these antigens are found at the tumor site. These neoepitopes are expressed by tumor cells and can be directly recognized by specific CD8 T cells able to kill the SF3B1mut uveal melanoma cells.

To our knowledge, this work is the first experimental demonstration of an immune response against public neoepitopes generated by an abnormal splicing process, a hypothesis suggested some time ago (5, 26). The frameshift or the short AA stretch included in the mature proteins induced by the mutated splicing factors lead to DRiPS that are preferentially degraded and targeted to the MHC I peptide loading compartment (9, 10). It remains to be determined whether the alternative splicing junctions leading to NMD generated higher levels of MHC:peptide complexes at the surface of the tumor cells than the abnormal proteins harboring a few additional AAs. Immunopeptidomics would provide an unbiased agnostic catalog of the most abundant SF3B1mut-related peptides presented by the MHC molecules of uveal melanoma cells.

Here, we studied mutations of the SF3B1 splicing factor in uveal melanoma. Interestingly, SF3B1 mutations are also found in a wide range of malignancies (7), including hemopathies (27, 28), carcinomas, and other melanomas (29). Further work will determine whether neoepitopes are also generated and seen by the immune system in these cancers. It is probable that a core of epitopes will be shared, while some neoepitopes will be specific for each cancer type. Finally, the same questions pertain to the mutations of the other splicing factors (SRSF2, U2AF1, ZRSR2, etc.) that are frequently found in tumors (30) or to the splicing anomalies related to the oncogenic process itself (for an example, see ref. 31).

Here, we identify a relevant immune response toward MHC class I restricted epitopes. It is probable that MHC II-restricted restricted epitopes are also generated and presented by antigen-presenting cells to induce neoepitope-specific CD4+ T cells. The latter cells may correspond to the chronically activated CD4+ T cells we previously described in patients with cancer (19). However, the neoepitope prediction algorithms are not performing well for MHC II-restricted antigens, and MHC II tetramers are still very challenging to produce. Thus, analysis of the CD4+ T-cell response will require new high-throughput technologies, such as reporter antigen-presenting cells expressing a library of potential epitopes as chimeric TCR-MHC II-peptide molecules (32). This is an important issue because CD4 T-cell responses are required for both efficient priming and strong effector activity by CD8 T cells (33). Moreover, CD4 T cells can be either directly cytotoxic (34) or recruit innate effector cells with antitumor activity in the tumor microenvironment (35). This is particularly important when the tumor cells lose MHC I expression.

Importantly, an immune response was detected only for a few of the in silico predicted neoepitopes, and only some patients displayed an immune response. This is not surprising given the polymorphism of the molecules involved in peptide processing, transport, and loading. Moreover, a potential competition for MHC alleles may occur. If the natural immune response against SF3B1mut-related epitopes was still functional at the time of metastasis diagnosis, some efficacy of the anti-checkpoint inhibitors (ICI) would be expected in 20% to 30% of metastatic uveal melanomas. However, metastatic uveal melanomas are usually resistant to ICI monotherapy (36), although ICI associations may be more active (37). This resistance to ICIs could be related to peripheral deletion of the neoepitope-specific CD8 T cells during the slow growth of the SF3B1mut tumors, which usually recurs a long time after treatment of the primary tumors (16). Even though neoepitope-specific CD8 T-cell clones could be derived from patients and healthy donors, the exact functionality and TCR avidity of the neoepitope-specific CD8 T cells remain to be determined in patients as compared to healthy donors.

Since the splicing factor SF3B1 is only mutated in the tumor cells and modifies the splicing pattern in more than 1,000 junctions, the neoepitopes are tumor-specific and numerous. Being germline encoded, these neoepitopes are shared across most patients according to their particular HLA haplotype. Given the high frequency of the HLA-A2 allele in Europe, by characterizing SF3B1mut-related epitopes presented by other prevalent HLA alleles, it can be envisioned that a limited (15–20) number of public neoepitopes would enable treatment of almost all patients. Effective vaccinations using long peptides or DNA- or RNA-based vaccines would be relatively easy to implement (2). Neoepitopes are also attractive targets for adoptive transfer therapies relying either on T cells transduced with specific TCRs or on soluble bispecific reagents redirecting the activity of effector T cells toward neoepitopes expressing tumor cells with antibodies or affinity matured TCR, similarly to what is proposed for the Melan-A or gp100 HLA-A2 epitopes (38, 39). The same approach could be used in the other cancers bearing splicing factor mutations, extending the potential therapeutic importance of our findings.

Human Samples

Blood and leukaphereses from HDs were provided by the Etablissement Français du Sang. Leukaphereses from patients with metastatic uveal melanoma were obtained from peptide vaccine trials CP-99-03 and IC-2004-01 before treatment (Supplementary Table S1). Written informed consent had been obtained from the patients. The studies were conducted in accordance with Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule and had been approved by an institutional review board. We selected the patients who had more than 30 frozen vials of peripheral blood mononuclear cells (PBMC) and for whom liver metastasis RNA was available. Cells were stored in liquid nitrogen until the time of analysis. RNA from the confirmatory diagnostic biopsy of the uveal melanoma liver metastasis performed at inclusion was used for SF3B1 mutation confirmation and TCR sequencing. SF3B1 was sequenced as previously described (20). This study was conducted in accordance with the declaration of Helsinki and was approved by the Internal Review Board of Institut Curie. All patients were informed that pathologic specimens might be used for research purposes.

Cell Lines

PDX-MM267 and PDX-MP41 are PDXs established at Institut Curie (40). MP41 is a cell line derived from the PDX-MP41 (41). A Degron-KI system was used to generate isogenic cell lines from Mel202, a uveal melanoma cell line mutated for SF3B1 (c.R625G) as described in ref. 11. In short, a Degron sequence was inserted by CRISPR/Cas9 5′ to the start codon of the mutated SF3B1 allele. MP41 is SF3B1WT HLA-A2+, whereas Mel-202 is SF3B1mut HLA-A2. An expression vector for HLA-A2 kindly provided by O. Schwartz (42) was stably integrated in Mel202 cell lines and validated by FACS analysis for HLA-A2 membrane expression using BB7.2-FITC antibody (BD).

RNA-seq, Analysis, and Neoepitope Prediction

RNA was isolated from fresh tumor samples using a CsCl cushion as described in ref. 40 and then quantified with Qubit RNA HS assay kit (Thermo Fisher Scientific). RNA-seq libraries were constructed using the TruSeq Stranded mRNA Sample Preparation Kit (Illumina) and sequenced on an Illumina HiSeq 2500 platform using 100-bp paired-end sequencing. An average depth of global sequence coverage of 111 million and a median coverage of 75 million were attained. Differential junctions using alternative acceptors were identified as previously described (20) comparing the 8 SF3B1mut tumors with 5 SF3B1WT tumors. Sequences of aberrant and corresponding normal transcripts were extracted using ANNOVAR and ENSEMBL databases (43). NetMHCpan v4, an artificial neural network-based algorithm, was used to predict MHC class I affinity for splicing anomalies derived peptides (23). Only nonamer peptides with strong affinity for HLA-A*02:01 (rank < 0.5% compared to a set of 400,000 random peptides) were retained for this study. Sequencing data have been deposited in and are available from the European Genome-phenome Archive database under number EGAS00001005226.

Analysis of Alternative Intron–Exon Junction Expression and NMD

RNA extracted from tumor cell lines or xenograft was reverse-transcribed as previously described (40). Primer pairs (one of which was fluorescently labeled) encompassing intron–exon junctions of interest (Supplementary Table S8) were used. The PCR product was cleaned up, denaturated, and run on a capillary electrophoresis ABI 3500xl apparatus. To analyze the affect of NMD on mRNA levels of selected junctions, uveal melanoma cell lines were cultured without or with Puromycin (200 μg/mL) for 6 hours before RNA extraction.

Analysis of Aberrant Peptides by MS

Endogenous NET1 was immunoprecipitated from protein extracts using antibody targeting the N-terminus part of NET1 (sc-271941; 2 μg per 200 μg of cell extract), as described in ref. 44 with modifications detailed in Supplementary Methods.

Peptidtic samples were analyzed using an Orbitrap Exploris 480 mass spectrometer (Thermo Scientific) coupled to an RSLCnano system (Ultimate 3000, Thermo Scientific). Details of the LC separation and the targeted MS analysis are described in Supplementary Methods.

MHC/Peptide Complex Generation

Recombinant HLA-A*02:01 molecules (24) were purchased from immunAware (Copenhagen, Denmark) as easYmers (catalog #1002–1). All peptides were synthesized at > 95% purity (Synpeptide) and tested for HLA-A2 monomer avidity following immunAware bead-based recommended assay and ELISA (45). Briefly, for each tetramer, MHC/peptide complex (100 μmol/L) was combined for 1 hour at room temperature with fluorescent streptavidin (BioLegend) or oligo-tagged streptavidin (BioLegend) for a single-cell experiment. Tetramers were stored at 4°C for a maximum of 3 months.

FACS Analysis and Antibodies

PBMC were thawed in CO2-independent medium (GIBCO) and incubated for 30 minutes in culture medium containing 50 nmol/L dasatinib (46) to improve tetramer staining. CD8+ T cells were enriched using a human CD8 T-cell enrichment kit (Stemcell) according to manufacturer instructions. Dead cells were stained with live/dead aqua (Invitrogen). For tetramer staining, tetramers for each specificity were labelled separately with 2 different fluorochromes in order to combine 10 different tetramer/peptide complexes in the same experiment and to decrease the noise related to nonspecific binding (47). Briefly, cells were incubated for 20 minutes with each tetramer complex in brilliant stain buffer (BD), and then cells were stained for 20 minutes with indicated antibodies [CD3-BUV737, CD8-BUV395 (BD), CCR7-BV421 (BioLegend), CD45RA FITC (Miltenyi Biotec), CD25 PE-CF594 (BD), CD8 FITC, CD3 Alexa fluor 700]. Cells were then washed and analyzed in a LSR Fortessa cytometer (BD).

T-cell Clone Generation and Cell Culture

After tetramer staining, double tetramer positive CD8+ single cells were FACS sorted into 96 well plates containing 1:1 AIM-V/RPMI medium supplemented with 5% human serum, 100 U/mL penicillin, and 100 μg/mL streptomycin in the presence of 2 × 105 irradiated (50 Gy) allogenic feeder cells. Cells were stimulated with human IL2 (Novartis; 3000 UI/mL) and anti-CD3 (OKT3; 30 ng/mL). Starting on day 5, half of media was replaced with a 1:1 mixture of AIM-V/RPMI containing IL2 (3000 IU/mL) every 3 days. When lymphocyte growth was evident, clones were transferred into T25 flasks. The clones were restimulated every 3 weeks using the same media containing IL2, OKT3, and irradiated feeders. After each cycle of clone amplification, each clone was tested for tetramer binding by cytometry and their capacity to respond to peptide stimulation using IFNγ and GzmB secretion (BD) and/or intracellular IFNγ staining (eBioscience). cDNA from each clone was amplified by PCR using primers for TRAV, TRBV, and constant regions (48), the PCR products sequenced and the resulting sequences analyzed using IMGT/V-QUEST (49).

Clone Activation, Cytokine Release, and Killing Assay

T-cell clones were cocultured at a 1:1 ratio with the indicated cell line pulsed or not with peptide (15 μmol/L to 0.3 pmol/L) in AIM-V/RPMI 10% fetal calf serum (FCS) medium for 18 hours at 37°C. Activation was measured by CD69, CD25, and CD107a staining, and cytokine secretion was analyzed in supernatant using cytometric beads array kits (BD) according to the manufacturer's instruction. Killing assay was performed using an xCELLigence real time cell analysis (RTCA) system. Tumor cell lines were plated onto an E-plate 16 (Cat: 05469813001) at 1 and 2 × 104 cells per well for SF3B1WT and SF3B1mut Mel202 cells, respectively. Twenty hours later, peptide (0.1 μmol/L) was added to the control wells. Two hours later, the wells into which peptide had been added were washed four times. Then, clone T cells at a 10/1 E:T ratio, if not indicated otherwise, were added to each well, and data acquisition lasted for the next 24 hours. For 7AAD experiments, target cells were incubated with specific T-cell clones during 4 hours. For the positive control, the cells were incubated with 1 μmol/L of corresponding peptide. The cells were then stained with 7AAD final concentration (0.5 μg/mL; cat: 559763 from BD) and antibodies (CD45-BV711, CD8-FITC). Cells were then washed and analyzed in an LSR Fortessa cytometer.

Short-Term Culture of PDXs

Cell suspensions obtained by mechanical dissociation of PDX explants were transferred to the top of gradient media Lymphoprep (StemCell) and centrifuged for 20 minutes at 700 g. The cell layer was treated with a dead cell removal kit (Milteny Biotech). Viable cells were cultured in RPMI 20%, FCS 1%, 100 U/mL penicillin, and 100 μg/mL streptomycin.

Single-Cell Experiments

Thawed PBMCs from patient UM1 were stained with PE and APC tetramers loaded with peptide 37, then tetramer-positive cells were positively enriched using anti-APC and anti-PE microbeads (Miltenyi Biotech) and stained with CD3 A700 and CD8 FITC and finally with DAPI. The positive fraction was sorted in a FACS ARIA (BD). To combine 5 tetramer positive populations from 2 donors (14, 17, 26, and 37 for UM2 and 18 for UM3), the tetramers were prepared using 5 different TotalSeq-PE streptavidins (BioLegend) and 1 classic fluorochrome-streptavidin (APC, PE-CF594, PE-Cy5, PE-Cy7); PE-CF594 was used for UM3, who had only one population sorted. The PBMCs were stained with the 4 pairs of tetramers for patient UM2 and 1 pair of tetramers for patient UM3, enriched with anti-PE microbeads (Milteny Biotech), stained with CD3 A700 and CD8 FITC and DAPI, and sorted separately. The cells were then counted, mixed, and loaded onto a Chromium controller using a Chromium Next GEM Single Cell V(D)J reagent kit with feature barcoding technology according to manufacturer's instructions.

Single-Cell RNA-seq Processing

Single-cell expression was analyzed using the Cell Ranger single-cell Software Suite (v3.0.2, 10x Genomics; ref. 50) to perform quality control, sample demultiplexing, barcode processing, and single-cell 5′ gene counting. Sequencing reads were aligned to the GRCh38 human reference genome. Further analysis was performed in R (v3.5.1) using the Seurat package (v3.1.1; 51). Cells were then filtered out when expressing fewer than 500 genes for UM1 and fewer than 1,000 genes for UM2/UM3 because this sample was of lower quality. Cells were also filtered out when expressing more than 10% mitochondrial genes, indicative of potential cell death or stress. Samples were then filtered for contaminating cells using classic markers. Notably, CD19 was used to remove B cells, MAFB was used to remove myeloid cells, and CD3D/E/G and CD8A/B were used as positive controls. Altogether, 3,441 cells were kept for UM1 and 3,231 were kept for UM2 and UM3. For each sample, the gene-cell-barcode matrix of the samples was then normalized to a total of 1 × 104 molecules. TotalSeq values were normalized according to the CLR method implemented in Seurat. The top 2,000 variable features were identified using the “vst” method from Seurat. For UM2/UM3 samples, the doublets were removed, leveraging the TotalSeq information. Since the TotalSeq features were bimodal, we first binarized the TotalSeq features. The expression threshold was defined as 1 for UM2-A2:26 and 1.2 for the rest of the specificities. Cells were then labeled cells as doublets if they were expressing more than 1 TotalSeq above the expression threshold. In all, 490 cells out of the 3,213 (15%) were excluded after removing TotalSeq doublets. The data have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE169610.

Dimension Reduction and Unsupervised Clustering

The top 30 principal components (PC) were computed and UMAP was performed using the top 30 PCs of the normalized matrix. Clusters were identified using the FindNeighbors and FindClusters function in Seurat with a resolution parameter of 0.4 for UM1 and 0.35 for UM2/UM3 and using the first 30 PCs. To choose the optimal number of clusters and prevent overclustering, clustree analysis was performed using the clustree package (52). Unique cluster-specific genes were identified by running the Seurat FindAllMarkers function using Wilcoxon test.

Single-Cell TCR-Seq Data Processing and Analysis

TCR-seq data for each sample was processed using Cell Ranger software with the command “cellranger vdj” using the human reference genome GRCh38. Because of dropouts, both TCRα and TCRβ are not always sequenced in a given T cell. Thus, as a T cell can express up to two TCRα chains and one TCRβ chain, it is easy to artificially split true T-cell clones into two different clonotypes. To the contrary, incompletely sequenced doublets can mistakenly lead to the creation of artifactual clonotypes. Because our data sets encompassed very large clonal expansions and the dropout and the number of cells loaded on the chip was high for patient UM2/3 data sets, we manually curated all the recurrent clonotypes to exclude doublets and merge clonotypes. We merged clonotypes using the same TCRα or TCRβ chains. We excluded from downstream analysis all “cells” made of nonattributed TCRα or TCRβ chains that would be associated with a TCRα or TCRβ chains belonging to a clonal expansion. TotalSeq features were also used to exclude doublets.

TCR-Seq

Reverse transcription of tumor RNA was performed using random hexamers and SuperScript IV according to manufacturer instruction (Thermo Fisher Scientific). cDNAs were cleaned using Agencourt RNAclean XP Kit (Beckman Coulter). A combination of Vα and Vβ specific primers slightly modified from ref. 48 (Supplementary Table S9) was used in 2 seminested PCR steps followed by a barcoding step. The first PCR reaction was performed separately for alpha and beta TCRs using multiplex Vα and Vβ primer associated with constant TCRα (TRAC) and TCRβ (TRBC) region primers. Each primer was used at 0.2 μmol/L each (95°C 3 minutes) and 22 cycles (90°C 30 seconds, 58°C 30 seconds, 72°C 30 seconds). cDNAs were cleaned using Agencourt AMPure XP Kit. In the second step, two distinct seminested PCR multiplex for Vα and Vβ reactions were performed 95°C 3 minutes followed by 35 cycles (90°C 30 seconds, 63°C 30 seconds, 72°C 30 seconds). Barcoding and incorporation of the sequencing primers for Paired-end Illumina sequencing was performed with PE1_CS1 forward primer and PE2_barcode_CS2 reverse primer (Fluidigm) at 400 nmol/L using Platinium Taq DNA Polymerase High Fidelity (Thermo Fisher Scientific). PCR products were sequenced using Miseq V3 PE-300 kit (Illumina). The data have been deposited in the GEO database under accession number GSE169610.

J. Bigot reports a patent for EP20305477 pending. A.I. Lalanne reports a patent for 20305477.0 pending. A. Houy reports personal fees from Institut Curie during the conduct of the study; in addition, A. Houy has a patent for 20305477.0 pending. S. Dayot reports personal fees from Institut Curie during the conduct of the study. O. Ganier reports personal fees from Institut Curie during the conduct of the study. T. Popova reports grants from Canceropole Ile de France during the conduct of the study. V. Masson reports other support from Institut Curie during the conduct of the study. D. Loew reports other support from Institut Curie during the conduct of the study. S. Amigorena reports grants and personal fees from Mnemo Therapeutics, personal fees from Biomunex Pharmaceuticals, personal fees from Innate Pharma, and personal fees from Light Chain Bioscience outside the submitted work. M. Rodrigues reports grants and non-financial support from MSD, grants from BMS, personal fees from GSK, and personal fees from AstraZeneca outside the submitted work. M. Stern reports grants and personal fees from INSERM, grants from Institut Curie, grants from Ligue Nationale Contre le Cancer, grants from SIRIC, grants from UMCURE H2020, and grants from INCA during the conduct of the study; in addition, M. Stern has a patent for 20305477.0 pending. O. Lantz reports grants from European union H2020 program, grants from Canceropole Ile de france, grants from SIRIC Curie, grants from ANR ICGEX, grants from CIC Biotherapy IGR-CURIE, and grants from ARC FONDATION during the conduct of the study; personal fees from HIFIBIO, personal fees from ENARA, grants from Biomunex, personal fees from Biomunex, and personal fees from ANSM outside the submitted work. In addition, O. Lantz has a patent European Patent Application number EP20305477, filed on May 12, 2020, and entitled “NEOANTIGENIC EPITOPES ASSOCIATED WITH SF3B1 MUTATIONS” pending. No disclosures were reported by the other authors.

J. Bigot: Investigation, methodology. A.I. Lalanne: Investigation, visualization, methodology. F. Lucibello: Investigation, visualization, methodology. P. Gueguen: Software, investigation, visualization, methodology. A. Houy: Data curation, software, visualization. S. Dayot: Investigation, visualization, methodology, writing–review and editing. O. Ganier: Investigation, writing–original draft. J. Gilet: Data curation, software. J. Tosello: Formal analysis. F. Nemati: Methodology. G. Pierron: Methodology. J.J. Waterfall: Software, supervision. R. Barnhill: Investigation. S. Gardrat: Investigation. S. Piperno-Neumann: Data curation. T. Popova: Methodology. V. Masson: Investigation. D. Loew: Supervision, investigation.P. Mariani: Investigation. N. Cassoux: Investigation. S. Amigorena: Supervision. M. Rodrigues: Investigation. S. Alsafadi: Methodology. M.-H. Stern: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, writing–review and editing. O. Lantz: Conceptualization, formal analysis, supervision, funding acquisition, validation, visualization, writing–original draft, writing–review and editing.

We acknowledge the help of C. Guerin, S. Grondin, and A. Viguier from the flow cytometry core at Institut Curie. We thank Professor Caroline Robert for insightful comments. We thank the ICGex NGS platform of the Institut Curie (S. Lameiras, S. Baulande, M. Bohec) for technical help with single-cell RNA-seq experiments. ICGex is supported by the grants ANR10EQPX03 (Equipex) and ANR10INBS0908 (France Génomique Consortium) from the Agence Nationale de la Recherche (“Investissements d'Avenir” program), by the Canceropole Ile-de-France, and by the SiRIC Curie program. The Laboratoire de Spectrométrie de Masse Protéomique is supported by “Région Ile-de-France” and Fondation pour la Recherche Médicale grants (D. Loew). This work has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 667787 (UM Cure 2020 project). P. Gueguen has received a fellowship from the Ligue Nationale Contre le Cancer and F. Lucibello from Institut Curie. This work was supported by the Institut National de la Santé et de la Recherche Médicale, Institut Curie, ANR, ARC fondation, INCA, Ligue Nationale Contre le Cancer (M.-H. Stern's team is “Equipe Labellisée par la Ligue Contre le Cancer”) and Labex DCBIOL, Cancéropole Ile de France émergence, SIRIC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Rizvi
NA
,
Hellmann
MD
,
Snyder
A
,
Kvistborg
P
,
Makarov
V
,
Havel
JJ
et al
.
Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer
.
Science
2015
;
348
:
124
8
.
2.
Sahin
U
,
Tureci
O
.
Personalized vaccines for cancer immunotherapy
.
Science
2018
;
359
:
1355
60
.
3.
Coulie
PG
,
Van den Eynde
BJ
,
van der Bruggen
P
,
Boon
T
.
Tumour antigens recognized by T lymphocytes: at the core of cancer immunotherapy
.
Nat Rev Cancer
2014
;
14
:
135
46
.
4.
Gotter
J
,
Brors
B
,
Hergenhahn
M
,
Kyewski
B
.
Medullary epithelial cells of the human thymus express a highly diverse selection of tissue-specific genes colocalized in chromosomal clusters
.
J Exp Med
2004
;
199
:
155
66
.
5.
Slansky
JE
,
Spellman
PT
.
Alternative splicing in tumors - a path to immunogenicity?
N Engl J Med
2019
;
380
:
877
80
.
6.
Obeng
EA
,
Stewart
C
,
Abdel-Wahab
O
.
Altered RNA processing in cancer pathogenesis and therapy
.
Cancer Discov
2019
;
9
:
1493
510
.
7.
Wang
RF
,
Wang
HY
.
Immune targets and neoantigens for cancer immunotherapy and precision medicine
.
Cell Res
2017
;
27
:
11
37
.
8.
Kurosaki
T
,
Popp
MW
,
Maquat
LE
.
Quality and quantity control of gene expression by nonsense-mediated mRNA decay
.
Nat Rev Mol Cell Biol
2019
;
20
:
406
20
.
9.
Apcher
S
,
Millot
G
,
Daskalogianni
C
,
Scherl
A
,
Manoury
B
,
Fahraeus
R
.
Translation of pre-spliced RNAs in the nuclear compartment generates peptides for the MHC class I pathway
.
Proc Natl Acad Sci U S A
2013
;
110
:
17951
6
.
10.
Yewdell
JW
,
Dersh
D
,
Fahraeus
R
.
Peptide channeling: the key to MHC class I immunosurveillance?
Trends Cell Biol
2019
;
29
:
929
39
.
11.
Zhou
Q
,
Derti
A
,
Ruddy
D
,
Rakiec
D
,
Kao
I
,
Lira
M
et al
.
A chemical genetics approach for the functional assessment of novel cancer genes
.
Cancer Res
2015
;
75
:
1949
58
.
12.
Tussey
LG
,
Rowland-Jones
S
,
Zheng
TS
,
Androlewicz
MJ
,
Cresswell
P
,
Frelinger
JA
et al
.
Different MHC class I alleles compete for presentation of overlapping viral epitopes
.
Immunity
1995
;
3
:
65
77
.
13.
Tourdot
S
,
Gould
KG
.
Competition between MHC class I alleles for cell surface expression alters CTL responses to influenza A virus
.
J Immunol
2002
;
169
:
5615
21
.
14.
Legoux
F
,
Debeaupuis
E
,
Echasserieau
K
,
De La Salle
H
,
Saulquin
X
,
Bonneville
M
.
Impact of TCR reactivity and HLA phenotype on naive CD8 T cell frequency in humans
.
J Immunol
2010
;
184
:
6731
8
.
15.
Legoux
FP
,
Moon
JJ
.
Peptide: MHC tetramer-based enrichment of epitope-specific T cells
.
J Vis Exp
2012
;
68
:
e4420
.
16.
Rodrigues
M
,
Koning
L
,
Coupland
SE
,
Jochemsen
AG
,
Marais
R
,
Stern
MH
et al
.
So close, yet so far: discrepancies between uveal and other melanomas. a position paper from UM cure 2020
.
Cancers
2019
;
11
:
1032
.
17.
Rodrigues
M
,
Mobuchon
L
,
Houy
A
,
Alsafadi
S
,
Baulande
S
,
Mariani
O
et al
.
Evolutionary routes in metastatic uveal melanomas depend on MBD4 alterations
.
Clin Cancer Res
2019
;
25
:
5513
24
.
18.
Durante
MA
,
Rodriguez
DA
,
Kurtenbach
S
,
Kuznetsov
JN
,
Sanchez
MI
,
Decatur
CL
et al
.
Single-cell analysis reveals new evolutionary complexity in uveal melanoma
.
Nat Commun
2020
;
11
:
496
.
19.
Peguillet
I
,
Milder
M
,
Louis
D
,
Vincent-Salomon
A
,
Dorval
T
,
Piperno-Neumann
S
et al
.
High numbers of differentiated effector CD4 T cells are found in patients with cancer and correlate with clinical response after neoadjuvant therapy of breast cancer
.
Cancer Res
2014
;
74
:
2204
16
.
20.
Alsafadi
S
,
Houy
A
,
Battistella
A
,
Popova
T
,
Wassef
M
,
Henry
E
et al
.
Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage
.
Nat Commun
2016
;
7
:
10615
.
21.
Darman
RB
,
Seiler
M
,
Agrawal
AA
,
Lim
KH
,
Peng
S
,
Aird
D
et al
.
Cancer-associated SF3B1 hotspot mutations induce cryptic 3′ splice site selection through use of a different branch point
.
Cell Rep
2015
;
13
:
1033
45
.
22.
Lei
XD
,
Chapman
B
,
Hankinson
O
.
Loss of cyp1a1 messenger rna expression due to nonsense-mediated decay
.
Mol Pharmacol
2001
;
60
:
388
93
.
23.
Jurtz
V
,
Paul
S
,
Andreatta
M
,
Marcatili
P
,
Peters
B
,
Nielsen
M
.
NetMHCpan-4.0: improved peptide-MHC class I interaction predictions integrating eluted ligand and peptide binding affinity data
.
J Immunol
2017
;
199
:
3360
8
.
24.
Leisner
C
,
Loeth
N
,
Lamberth
K
,
Justesen
S
,
Sylvester-Hvid
C
,
Schmidt
EG
et al
.
One-pot, mix-and-read peptide-MHC tetramers
.
PLoS One
2008
;
3
:
e1678
.
25.
D'Souza
S
,
Rimoldi
D
,
Lienard
D
,
Lejeune
F
,
Cerottini
JC
,
Romero
P
.
Circulating Melan-A/Mart-1 specific cytolytic T lymphocyte precursors in HLA-A2+ melanoma patients have a memory phenotype
.
Int J Cancer
1998
;
78
:
699
706
.
26.
Frankiw
L
,
Baltimore
D
,
Li
G
.
Alternative mRNA splicing in cancer immunotherapy
.
Nat Rev Immunol
2019
;
19
:
675
87
.
27.
Yoshida
K
,
Sanada
M
,
Shiraishi
Y
,
Nowak
D
,
Nagata
Y
,
Yamamoto
R
et al
.
Frequent pathway mutations of splicing machinery in myelodysplasia
.
Nature
2011
;
478
:
64
9
.
28.
Wang
L
,
Lawrence
MS
,
Wan
Y
,
Stojanov
P
,
Sougnez
C
,
Stevenson
K
et al
.
SF3B1 and other novel cancer genes in chronic lymphocytic leukemia
.
N Engl J Med
2011
;
365
:
2497
506
.
29.
Kahles
A
,
Lehmann
KV
,
Toussaint
NC
,
Huser
M
,
Stark
SG
,
Sachsenberg
T
et al
.
Comprehensive analysis of alternative splicing across tumors from 8,705 patients
.
Cancer Cell
2018
;
34
:
211
24
.
30.
Seiler
M
,
Peng
S
,
Agrawal
AA
,
Palacino
J
,
Teng
T
,
Zhu
P
et al
.
Somatic mutational landscape of splicing factor genes and their functional consequences across 33 cancer types
.
Cell Rep
2018
;
23
:
282
96
.
31.
Misquitta-Ali
CM
,
Cheng
E
,
O'Hanlon
D
,
Liu
N
,
McGlade
CJ
,
Tsao
MS
et al
.
Global profiling and molecular characterization of alternative splicing events misregulated in lung cancer
.
Mol Cell Biol
2011
;
31
:
138
50
.
32.
Kisielow
J
,
Obermair
FJ
,
Kopf
M
.
Deciphering CD4(+) T cell specificity using novel MHC-TCR chimeric receptors
.
Nat Immunol
2019
;
20
:
652
62
.
33.
Guerder
S
,
Matzinger
P
.
Activation versus tolerance: a decision made by T helper cells
.
Cold Spring Harb Symp Quant Biol
1989
;
54
:
799
805
.
34.
Quezada
SA
,
Simpson
TR
,
Peggs
KS
,
Merghoub
T
,
Vider
J
,
Fan
X
et al
.
Tumor-reactive CD4(+) T cells develop cytotoxic activity and eradicate large established melanoma after transfer into lymphopenic hosts
.
J Exp Med
2010
;
207
:
637
50
.
35.
Corthay
A
,
Skovseth
DK
,
Lundin
KU
,
Rosjo
E
,
Omholt
H
,
Hofgaard
PO
et al
.
Primary antitumor immune response mediated by CD4+ T cells
.
Immunity
2005
;
22
:
371
83
.
36.
Najjar
YG
,
Navrazhina
K
,
Ding
F
,
Bhatia
R
,
Tsai
K
,
Abbate
K
et al
.
Ipilimumab plus nivolumab for patients with metastatic uveal melanoma: a multicenter, retrospective study
.
J Immunother Cancer
2020
;
8
:
e000331
.
37.
Piulats
JM
,
Espinosa
E
,
de la Cruz Merino
L
,
Varela
M
,
Alonso Carrion
L
,
Martin-Algarra
S
et al
.
Nivolumab plus ipilimumab for treatment-naive metastatic uveal melanoma: an open-label, multicenter, phase II trial by the Spanish multidisciplinary melanoma group (GEM-1402)
.
J Clin Oncol
2021
;
39
:
586
98
.
38.
Liddy
N
,
Bossi
G
,
Adams
KJ
,
Lissina
A
,
Mahon
TM
,
Hassan
NJ
et al
.
Monoclonal TCR-redirected tumor cell killing
.
Nat Med
2012
;
18
:
980
7
.
39.
Harper
J
,
Adams
KJ
,
Bossi
G
,
Wright
DE
,
Stacey
AR
,
Bedke
N
et al
.
An approved in vitro approach to preclinical safety and efficacy evaluation of engineered T cell receptor anti-CD3 bispecific (ImmTAC) molecules
.
PLoS One
2018
;
13
:
e0205491
.
40.
Nemati
F
,
Sastre-Garau
X
,
Laurent
C
,
Couturier
J
,
Mariani
P
,
Desjardins
L
et al
.
Establishment and characterization of a panel of human uveal melanoma xenografts derived from primary and/or metastatic tumors
.
Clin Cancer Res
2010
;
16
:
2352
62
.
41.
Amirouchene-Angelozzi
N
,
Nemati
F
,
Gentien
D
,
Nicolas
A
,
Dumont
A
,
Carita
G
et al
.
Establishment of novel cell lines recapitulating the genetic landscape of uveal melanoma and preclinical validation of mTOR as a therapeutic target
.
Mol Oncol
2014
;
8
:
1508
20
.
42.
Le Gall
S
,
Erdtmann
L
,
Benichou
S
,
Berlioz-Torrent
C
,
Liu
L
,
Benarous
R
et al
.
Nef interacts with the mu subunit of clathrin adaptor complexes and reveals a cryptic sorting signal in MHC I molecules
.
Immunity
1998
;
8
:
483
95
.
43.
Wang
K
,
Li
M
,
Hakonarson
H
.
ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data
.
Nucleic Acids Res
2010
;
38
:
e164
.
44.
Kweh
F
,
Zheng
M
,
Kurenova
E
,
Wallace
M
,
Golubovskaya
V
,
Cance
WG
.
Neurofibromin physically interacts with the N-terminal domain of focal adhesion kinase
.
Mol Carcinog
2009
;
48
:
1005
17
.
45.
Rodenko
B
,
Toebes
M
,
Hadrup
SR
,
van Esch
WJ
,
Molenaar
AM
,
Schumacher
TN
et al
.
Generation of peptide-MHC class I complexes through UV-mediated ligand exchange
.
Nat Protoc
2006
;
1
:
1120
32
.
46.
Lissina
A
,
Ladell
K
,
Skowera
A
,
Clement
M
,
Edwards
E
,
Seggewiss
R
et al
.
Protein kinase inhibitors substantially improve the physical detection of T-cells with peptide-MHC tetramers
.
J Immunol Methods
2009
;
340
:
11
24
.
47.
Hadrup
SR
,
Schumacher
TN
.
MHC-based detection of antigen-specific CD8+ T cell responses
.
Cancer Immunol Immunother
2010
;
59
:
1425
33
.
48.
Han
A
,
Glanville
J
,
Hansmann
L
,
Davis
MM
.
Linking T-cell receptor sequence to functional phenotype at the single-cell level
.
Nat Biotechnol
2014
;
32
:
684
92
.
49.
Giudicelli
V
,
Chaume
D
,
Lefranc
MP
.
IMGT/GENE-DB: a comprehensive database for human and mouse immunoglobulin and T cell receptor genes
.
Nucleic Acids Res
2005
;
33
:
D256
61
.
50.
Zheng
C
,
Zheng
L
,
Yoo
JK
,
Guo
H
,
Zhang
Y
,
Guo
X
et al
.
Landscape of infiltrating T cells in liver cancer revealed by single-cell sequencing
.
Cell
2017
;
169
:
1342
56
.
51.
Stuart
T
,
Butler
A
,
Hoffman
P
,
Hafemeister
C
,
Papalexi
E
,
Mauck
WM
3rd
et al
.
Comprehensive integration of single-cell data
.
Cell
2019
;
177
:
1888
902
.
52.
Zappia
L
,
Oshlack
A
.
Clustering trees: a visualization for evaluating clusterings at multiple resolutions
.
Gigascience
2018
;
7
:
giy083
.

Supplementary data