Research blood samples were collected at day −13, 42, 63 and 455 on Fred Hutchinson Cancer Research Center (FHCRC) protocol 1765 and tumor biopsies were obtained on the clinical trial protocol 6 days before and 17 days after initiation of treatment.

Whole exome sequencing (WES) of pretreatment tumor biopsy and peripheral blood mononuclear cell (PBMC) was performed and single nucleotide variants called by both Mutect11 and Strelka12 were filtered for a variant allele frequency (VAF) >0.2 (online supplemental figure S1). The 45 mutations with the highest level of mRNA expression (TPM >12) were selected for screening (online supplemental table S1). RNA sequencing was performed as described previously and in the online supplemental methods.13

PBMC were cultured for 13 days in the presence of 2 µg/mL peptide encompassing each neoantigen (27-mer with mutated amino acid at position 14, 80% pure, Elim Biopharma) or self-antigen (Peptivator, Milltenyi) in the presence of interleukin (IL)-21, IL-15, and IL-7 (Peprotech) as described previously.13 14 Reactivity of cultures with individual peptides was assayed by interferon (IFN)-γ ELISpot assay (10 µg/mL peptide, ELISpot Pro human IFN-γ, Mabtech).13 14 IFN-γ secretion assay used the IFN-γ secretion kit (Milltenyi). Sorted IFN-γ-positive cells were expanded in 96-well plates with allogeneic irradiated PBMC, phytohemagglutinin and IL-2 for 14–21 days as described.14

TCRVβ sequencing was performed on DNA samples using the human TCRB kit from Adaptive Biotechnologies, and data were analyzed using Adaptive software. Frequencies of TCR clonotypes were determined by amino acid sequence (online supplemental table S2). Clonotypes identified in the T cell cultures were presumed antigen specific if they met the following criteria: (1) expansion to >5 templates with specific stimulation and at least 10-fold enrichment in frequency over any other stimulation and (2) greater frequency of antigen-dependent IFN-γ secreting cells than either prestimulation sample or irrelevant antigen control culture, with at least five templates observed in the IFN-γ-enriched sample. CDR3 consensus sequences were generated using weblogo15 (weblogo.berkeley.edu).

A man aged 21 years presented with a lung nodule that was determined by core needle biopsy to be metastatic melanoma, BRAF V600E mutated. Staging showed multiple lung nodules and mediastinal lymphadenopathy, without a detectable primary skin lesion. The patient was enrolled in a phase I clinical trial combining nivolumab 360 mg and bempegaldesleukin 0.06 mg/kg16 intravenously every 3 weeks. A CT scan at 4.5 months showed a partial response by RECIST V.1.1 (figure 1A,B) and positron emission tomography showed no Fluorodeoxyglucose (FDG) avid disease after 13 months of therapy. The patient developed vitiligo 1 year following initiation of treatment and bempegaldesleukin was discontinued after 15 months due to recurrent grade 1 fatigue and myalgias. Nivolumab was discontinued after 2 years and the patient has not progressed after >3 years (figure 1A).

Whole transcriptome sequencing performed on pretreatment and on-treatment tumor tissue from a lung lesion that subsequently completely regressed showed a >10-fold increase in expression of multiple genes annotated as immune response genes (3.47-fold enrichment, p=1.3e-81), including those reflecting T cell infiltration (CD3, CD4 and CD8) and activation (IFN-γ, CD69, CD25, programmed cell death protein 1, Lag-3, Tim-3, granzyme, perforin and a decrease in lineage-specific melanocyte proteins tyrosinase, GP100, Mart1 and TRP2 and cancer testis antigens Mage A3, Mage A2 and PRAME (figure 1C,D, online supplemental table S2), consistent with nivolumab and bempegaldesleukin resulting in rapid immune infiltration and tumor regression.16 17

The oncogenic BRAF V600E mutation and 279 additional single nucleotide variants were identified as candidate neoantigens by WES of pretreatment tumor and germline DNA (online supplemental table S1). T cell responses were evaluated to 45 candidates with VAF >0.2 and expression >12 TPM. PBMCs were stimulated with a single pool of two 20 amino acid peptides encompassing each of the mutations and reactivity evaluated by IFN-γ ELISpot (figure 2A). Following stimulation and prior to ELISpot assay, CD8⁺ T cells were enriched by positive immunomagnetic selection and analyzed separately from CD8⁺ depleted PBMC. Candidate mutations that showed qualitative responses over baseline with crude mutated peptides (online supplemental figure S2A–C) were tested for IFN-γ production in response to 80% purified 27-mer peptides corresponding to the mutant and wild-type sequences in a confirmatory assay. We identified CD8⁺ T cell responses to mutations in SLC39A14 (p<0.0001 vs negative control, p=0.0096 vs wild-type) and to NACC1 (p=0.0075 vs negative control, p=0.09 vs wild-type) (figure 2B). We also identified CD4⁺ T cell responses to the same mutation in SLC39A14 (p<0.0001 vs negative control and p=0.0004 vs wild-type) and a mutation in CAPG (p<0.0001 vs negative control and p=0.052 vs the wild-type peptide, figure 2C).

We then used TCRVβ deep sequencing of PBMC that had been stimulated with individual peptides to identify putative antigen-specific T cell clonotypes that expand to one antigen and not others.10 TCR seq is robust in this assay to identify TCR clones present at high enough frequencies to be evenly distributed across parallel cultures. A limitation with this method is that clones present in only one culture could expand non-specifically. Indeed, it was common for some T cell clones to expand without antigen-specific stimulation in one but not in multiple other parallel cultures (online supplemental figure S3). As a second method of establishing antigen specificity, we tested whether restimulating the cultures with peptide and selecting IFN-γ secreting T cells by fluorescence activated cell sorting would further enrich antigen-specific cells (figure 2D) but not non-specifically expanded cells. Many T cells in these cultures secreted IFN-γ even without antigen restimulation (figure 2D), therefore CD4⁺ and CD8⁺ IFN-γ secreting cells were also sorted following control stimulations without peptide. With this combination of assays, we assigned the TCRVβ of putative antigen-specific cells to be those that 1) expanded with one antigen and no other antigens and 2) were enriched in IFN-γ secreting cells with specific antigen stimulation to a greater degree than with control stimulation without antigen.

We then applied these methods to identifying putative neoantigen-specific TCRVβ clones in stimulations of post-treatment PBMC in separate cultures with purified peptides containing the mutations in CAPG, SLC39A14 and NACC1. We identified 16 CD8⁺ clonotypes specific for SLC39A14, 8 CD4⁺ clonotypes specific for SLC39A14, 16 CD8⁺ clonotypes specific for NACC1 and a single CD4⁺ clonotype specific for CAPG (figure 2E, online supplemental table S3). Many of these TCR clonotypes were expanded in only one out of two independent stimulations of 5 million PBMC (seven CD8⁺ clones and all CD4⁺ clones specific for SLC39A14, and nine CD8⁺ clones specific for NACC1, figure 2E) indicating the low frequency of many of these clonotypes capable of expanding with peptide stimulation.

To confirm specificity, neoantigen-reactive T cell lines were obtained by expanding small numbers (3–30) of IFN-γ secreting T cells after stimulation with mutated NACC1 and SLC39A14 peptides. T cell lines were then restimulated with peptide to identify those containing >90% antigen-specific cells (figure 3A). TCRVβ deep sequencing of these lines was performed and demonstrated that many were highly oligoclonal (figure 3B–D). TCRVβ clonotypes present at a level of >10% in these cultures were presumed to be antigen specific, which identified four CD4⁺ T cell clones specific for SLC39A14, seven CD8⁺ T cell clones specific for SLC39A14 and four CD8⁺ T cell clones specific for NACC1 (figure 3B–D, online supplemental table S3). Five of these 15 TCR clonotypes from this analysis overlapped with those previously identified, but multiple additional clonotypes were identified in cultures for each antigen (online supplemental table S3). To assess neoantigen specificity, oligoclonal CD8⁺ T cell lines derived by stimulation with SLC39A14 and NACC1 mutant peptides and CD4⁺ T cell lines derived by stimulation with the SLC39A14 mutant peptide were tested for reactivity with the mutant and wild-type peptides at various concentrations. Two different CD8⁺ T cell lines showed preferential activation with a 9-mer peptide containing the SLC39A14 mutation predicted to bind HLA-A:02, relative to the corresponding wild-type peptide (figure 3E), and two different CD8⁺ T cell lines reacted preferentially with a 10-mer mutated NACC1 mutant peptide predicted to bind HLA B15 (figure 3F). Similarly, two different CD4⁺ T cell lines were reactive with a 27-mer peptide containing the mutation in SLC39A14 and not the wild-type peptide (figure 3G). There was sequence similarity between TCRVβ hypervariable CDR3 sequences between independent clonotypes, such as nine different NACC1 mutation-specific CDR3 sequences of length 12 (figure 3H). These data show that the expanded T cells were high avidity and neoantigen-specific.

In total, there were 54 TCRVβ clonotypes identified to be neoantigen-specific (online supplemental table S3). To determine the localization of neoantigen-specific T cell clones during treatment, we performed TCRVβ sequencing of blood and tumor samples and found that 32 of 54 neoantigen-specific clones were detectable in the pretreatment tumor, which in total made up 0.71% of the TCRVβ templates in the pretreatment tumor (figure 4A, online supplemental table S2). In contrast, only four of these clones were detected in unstimulated pretreatment PBMC, making up 0.005% of TCRVβ sequences in the PBMC (150-fold enrichment in tumor, p<0.0001). Furthermore, only six of these clones could be detected in pretreatment PBMC after peptide stimulation of 5 million PBMC (online supplemental table S3).

Neoantigen-specific T cell clones expanded in frequency in the peripheral blood in the first 50 days following treatment (fivefold, p<0.0001) and contracted in the blood 13 months after treatment (fivefold, p<0.0001, figure 4A). RNA sequencing of a tumor sample after treatment demonstrated a prominent increase T cell transcripts (figure 1C), and the cumulative frequency of intratumoral neoantigen-specific TCRVβ increased modestly (30%, p<0.0001, figure 4A), although the overall clonal diversity was similar prior to and on treatment (online supplemental figure S4). The findings that neoantigen-specific T cell clones were enriched in the tumor and expanded in the blood held true for neoantigen-specific CD8⁺ T cells (p<0.0001 for enrichment in tumor, expansion in the blood following treatment and contraction following tumor regression) and for neoantigen-specific CD4⁺ T cells (p<0.0001 for enrichment in the tumor, p=0.1 for expansion in the blood post-treatment and p=0.01 for contraction in the blood post-treatment, figure 4B). Taken together, these data support the methodology for identifying CD4⁺ and CD8⁺ T cells that recognize tumor neoantigens, and demonstrate such clones are present in pretreatment tumor, rare in the peripheral blood before treatment and transiently expand in the blood during treatment.

We also assessed localization and expansion of T cell clones specific for self-antigens using similar stimulation and IFN-γ secretion assays with overlapping long peptide pools encompassing lineage-specific (Tyrosinase, GP100, Mart1, TRP2) and cancer testis (MAGE A3, PRAME) self-antigens that were expressed in the tumor by RNA-seq. We identified CD8⁺ (n=2) and CD4⁺ (n=1) clonotypes specific for GP100; CD4⁺ (n=3) and CD8⁺ (n=2) clonotypes specific for MAGE A3; CD4⁺ (n=7) and CD8⁺ (n=1) clonotypes specific for PRAME; CD8⁺ (n=2) clonotypes specific for TRP2 and CD4⁺ (n=6) and CD8⁺ (n=12) clonotypes specific for tyrosinase (figure 4C, online supplemental table S3). We then compared self-antigen and neoantigen-specific clonotypes identified by the same method from the same blood sample and found, in contrast to neoantigen-specific T cell clones, self-antigen-specific T cell clones did not expand in the PBMC or localize to the tumor (figure 4D).