Tumor-Infiltrating Merkel Cell Polyomavirus-Specific T Cells Are Diverse and Associated with Improved Patient Survival

Merkel cell carcinoma (MCC) is a highly aggressive skin cancer associated with UV exposure, advanced age, immune suppression, and in approximately 80% of cases, the Merkel cell polyomavirus (MCPyV; refs. 1–3). MCC incidence is approximately 2,000 cases per year in the United States (4, 5). It has not been possible to effectively treat advanced disease, leading to a 5-year survival rate of 0 to 18% for patients with distant metastatic disease and a median survival of only 9.6 months from diagnosis of initial metastasis to death (5–7). Although approximately half of patients initially respond to chemotherapy, responses are not durable, with a median time to progression of only 94 days (8), highlighting the need for improved therapies to treat MCC.

MCPyV is a prevalent, chronic virus that normally does not cause disease. Rarely, it clonally integrates into the host chromosomal DNA, and when paired with UV-induced mutations, can cause MCC (1, 9, 10). Virus-positive MCC is characterized by persistent expression of two MCPyV oncoproteins (the small T- and truncated large T-antigens; ref. 10), which share a common N-terminus (common T-Ag). Multiple studies have linked the immune system with the incidence and prognosis of MCC. Although <10% of MCC patients have systemic immune suppression, these patients have significantly poorer MCC-specific survival (11, 12). The adaptive immune system can recognize and mount protective responses to MCPyV. Specifically, increased intratumoral CD3⁺ cell counts are an independent prognostic factor of increased MCC-specific survival (13). Robust intratumoral CD8⁺ lymphocyte infiltration, though present in <20% of MCCs, has been associated in two independent cohorts with 100% MCC-specific survival, independent of tumor stage at diagnosis (14, 15). Recent clinical trials of T cell–activating therapies, such as PD-1 axis blockade, have shown impressive initial response rates and durability (16), which provide further impetus for the study of T cell–based therapies for MCC.

CD8⁺ T cells recognizing at least 17 unique epitopes of the persistently expressed T-antigens of MCPyV can be isolated and tracked in blood and tumors from MCC patients using HLA class I tetramers (refs. 17–19 and our unpublished observations). MCPyV-specific CD8⁺ T cells have been harnessed for adoptive T-cell therapy and yet resulted in a durable response in just one of four patients treated (20, 21). One strategy to increase the efficacy of adoptive T-cell therapy, and/or offer T cell–based therapies to patients who lack endogenous MCPyV-specific T cells, would be to engineer T cells to express highly avid MCPyV-specific T-cell receptors (TCR).

To better characterize the MCPyV-specific CD8⁺ TCR repertoire and measure correlations between TCR clonotype repertoire, intratumoral infiltration, and patient outcomes, we studied primary CD8⁺ T cells that recognize the epitope KLLEIAPNC (derived from the MCPyV common T-Ag), restricted to human leukocyte antigen (HLA) A*02:01, an HLA class I type present in ∼50% of our patient cohort (hereafter, KLL-specific T cells). Similar to other infections and malignancies (22–26), we hypothesized that diversity or functional avidity of the TCR repertoire recognizing this epitope may be correlated with the effectiveness of the T-cell response to MCC in vivo.

Using next-generation TCRβ sequencing, we found significant genetic diversity among TCRs recognizing the KLL epitope. Using paired blood and tumor specimens, we can now extend previous findings of positive correlations between MCC tumor T-cell infiltration and effector gene signatures (14) to the level of tumor antigen-specific CD8⁺ T cells. In addition, patients with greater KLL-specific clonotype diversity in their tumors have significantly improved MCC-specific survival. We studied the functional avidity of CD8⁺ clones expressing unique KLL-specific TCRs, and found that clones generally expressed a narrow range of functional avidities that were largely similar within each patient. Only 5 of 28 clonotypes tested from one of four patients recognized a MCPyV⁺ MCC cell line. Tumor cell recognition by T-cell clones correlated with high functional avidity. These studies have elucidated the genetic diversity of CD8⁺ T cells restricted to KLL and support continued investigation of methods to increase intratumoral CD8⁺ T-cell infiltration. The avid T-cell clones we identified could potentially lend their effector function through transgenic TCR therapy for the ∼80% of HLA-A*02⁺ MCC patients who lack endogenous KLL-specific T cells.

This study was approved by the Fred Hutchinson Cancer Research Center (FHCRC) Institutional Review Board and conducted according to Declaration of Helsinki principles. Informed consent was received from all participants. Subjects were HLA class I typed via polymerase chain reaction (PCR) at Bloodworks Northwest (Seattle, WA). All samples were clinically annotated with long-term patient follow-up data.

Heparinized blood was obtained from MCC patients, and peripheral blood mononuclear cells (PBMC) were cryopreserved after routine Ficoll preparation at a dedicated specimen processing facility at FHCRC.

When available, fresh MCC tumor material from core and/or punch biopsy samples were processed and TIL cultured for 2 weeks before analysis as described (17). For excised tumors of larger volume (>1 cm³), the remaining tissue was digested as described (18), and single-cell suspensions were cryopreserved.

At least 2 million PBMCs or TILs were stained with PE-conjugated monoclonal antibody (mAb) to CD8 (Clone 3B5, Life Technologies), A*02/KLL-APC tetramer (Immune Monitoring Lab, FHCRC), and 7-AAD viability dye (BioLegend). Tetramer⁺, CD8^high cells were sorted via FACSAriaII (BD) and flash frozen [average of 710 cells from PBMC (n = 9), 5,776 cells from TIL (n = 5), range 350–8,000 and 1,844–12,799, respectively]. Samples were submitted to Adaptive Biotechnologies for genomic DNA extraction, TRB sequencing, and normalization. All TRB sequences detected in ≥2 cells (estimated number of genomes ≥ 2) were categorized as tetramer⁺ clonotypes.

Primary tumors were used for analysis, except when patients presented with unknown primaries and nodal disease (n = 2), primaries with limited material but abundant nodal disease available for analysis (n = 1), or metastatic disease (n = 1). Tumor samples consisted of molecular curls of 25 μm from formalin-fixed, paraffin embedded (FFPE) tissue blocks (n = 10), nodal tumor digest frozen ex vivo (n = 1), or flash frozen core biopsy of a metastatic lesion (n = 1). Samples were submitted to Adaptive Biotechnologies as described above.

For tetramer-sorted cells, Shannon entropy was calculated on the estimated number of genomes (≥2) of all productive TRB and normalized by dividing by the log₂ of unique productive sequences in each sample. Clonality was calculated as 1 – normalized entropy. For whole tumors, clonality was calculated in the same method, using all TRB sequences in the sample to calculate normalized entropy.

KLL-specific clonotypes within tumors (n = 12 tumors) were identified based on TCRβ CDR3 amino acid sequences from the tetramer-sorted samples. The frequency of all KLL-specific T cells within each tumor is reported as the cumulative percentage of productive sequencing reads of clonotypes detected in both the tetramer-sorted sample and the tumor.

FFPE-embedded tumor tissue was stained (Experimental Histopathology at FHCRC) and slides scored by a dermatopathologist who was blinded to patient characteristics. Samples were stained with anti-CD8 (Dako, clone 144B at 1:100) and intratumoral CD8⁺ T cells (completely surrounded by tumor without neighboring stroma) on a scale from 0 (absent CD8⁺ cells) to 5 (>732 intratumoral CD8⁺ cells/mm²) as described by Paulson et al. (14). In addition, tumors were stained with anti-MHC class I (ref. 27; MBL, clone EMR8-5) and CM2B4 to measure MCPyV T-antigen expression (ref. 28; Santa Cruz Biotechnology, 1:50). Tumors were stained with anti-CD4 (Cell Marque clone SP35, 1:25) and anti-FoxP3 (eBiosciences clone FJK-16s, 1:25) and reported as the number of positive cells/mm².

Statistical analyses were performed on Stata software version 14.0 for Macintosh (StataCorp) and Prism 6 for Mac OS X (Graph Pad Software, Inc). MCC-specific survival is defined as the interval from the diagnostic biopsy date to death by MCC. Recurrence-free survival was defined as the interval from the diagnostic biopsy date to the date of MCC recurrence, last follow-up or death by MCC. Log-rank analysis was performed, and a P value of .05 was considered statistically significant. Kaplan–Meier survival curves were created to visualize MCC-specific survival and recurrence-free survival data; groupings of patients were based on the percentage of tetramer⁺ T cells in the tumor (higher = 1.9%–18%, n = 9 vs. lower = 0%–0.14%, n = 2) as well as number of T-cell clonotypes (many = 5–108, n = 7; vs. few = 0–3, n = 4) were selected a priori. Patients who were alive at the last time of follow-up were censored on their last day of follow-up and patients who died of unknown causes were censored on their day of death.

PBMCs or lymphocytes from tumor digest were stained and sorted as described above into T-cell medium (TCM) containing RPMI, 8% human serum, 200 nmol/L l-glutamine and 100 U/mL penicillin–streptomycin, and cloned at 0.25 to 3 cells per well with allogeneic irradiated feeders, IL2 (Hemagen Diagnostics) and PHA (Remel) as described (29) with addition of rIL15 (20 ng/mL; R&D Systems) after day 2. After 2 weeks, microcultures with visible growth were screened for specificity via tetramer; TCRβ variable chain (TCRVβ) expression was assessed by staining clones with fluorescent mAbs to TCRVβ (IOTest Beta Mark, Beckman Coulter). Wells selected for screening, expansion, and TCR analysis came from plates with <37% of cultures having visual growth, yielding a 95% chance of clonality per the Poisson distribution (30). Cultures with tetramer⁺ cells, reactivity to peptide, and dissimilar TCRVβ chains were further expanded with IL2 and the OKT3 mAb clone to CD3 (Miltenyi Biotec) as described (17), plus rIL15 (20 ng/mL). Prior to harvesting RNA for TCR analysis, cultures were held at least 2 weeks to minimize persistent feeder cell-derived RNA.

Clones were stained with an HLA-A*02:01/KLL tetramer containing D227K/T228A mutations in HLA-A*02:01, using methods as above. These mutations abrogate HLA class I:CD8 binding to identify clones expressing TCRs with the ability to bind independent of CD8 stabilization, which may indicate high TCR avidity (31, 32).

Simultaneous sequencing of TCRα and TCRβ repertoires was performed as described (33). Briefly, total RNA was isolated from clonally expanded populations using Qiagen RNeasy Plus, followed by One Step RT/PCR (Qiagen) primed with multiplexed TCR primers. This reaction was used as template with a second set of nested TCRα and TCRβ primers, followed by PCR to add barcoding and paired end primers. Templates were purified using AMPure (Agencourt Biosciences) then normalized prior to running on Illumina MiSeq v2-300. Pairs of 150 nucleotide sequences were merged into contigs using PandaSeq (34). Merged sequences were then separated according to inline barcodes identifying the plate and well of origin, generating one file of derived sequences for each clone of interest. Files for each clone were processed with MiXCR (35) to identify and quantify clonotypes and assign VDJ allele usage. Cultures in which the dominant TCRβ nucleotide sequence was present at <97% of productive sequence reads were classified as possibly polyclonal and excluded from further analysis.

T-cell clones were tested for specificity and functional avidity via cytokine release assays.

Secreted IFNγ was measured after coincubating 2×10⁴ clonal KLL-specific T cells with 5×10⁴ T2 cells [ATCC, 2015, validated by ATCC via short-tandem repeat (STR) analysis] plus antigenic peptide at log₁₀ dilutions to final concentration from 10⁻⁶ to 10⁻¹² molar in 200 μL TCM for 36 hours. Due to possible oxidation and dimerization of cysteine residues in the antigenic KLLEIAPNC peptide, the homolog KLLEIAPNA was used to allow for efficient HLA class I presentation; similar substitution has been shown to not alter recognition of the HLA–peptide complex by TCRs raised against the native peptide (36). IFNγ in cell culture supernatants was assayed by ELISA according to the manufacturer's recommendations (Human IFN gamma ELISA Ready-SET-Go Kit; Affymetrix). To estimate EC₅₀ (the amount of peptide leading to 50% of maximum IFNγ secretion), IFNγ secretion by each T-cell clone was analyzed via nonlinear regression using Prism version 6.0 (GraphPad). In addition, IFNγ release by KLL-specific clonotypes was measured after incubation with three MCPyV⁺, HLA-A*02⁺ MCC cell lines [WaGa and MKL-2; (gift of Dr. Becker, German Cancer Research Center, 2015); authenticated by Becker lab via STR analysis in 2014, and MS-1 (gift of Dr. Shuda, University of Pittsburg, 2015); validated by STR and deposited into European Collection of Authenticated Cell Cultures]. Cell lines were early passage and authenticated with independent STR profiling (ATCC). Cell lines were stimulated with IFNβ (Betaseron, BayerHealthCare; 3,000 U/mL) for 24 hours to induce expression of HLA class I, followed by 24 hours of culture after IFNβ washout. 2×10⁴ clonal KLL-specific T cells were incubated with 5×10⁴ cells from each MCC cell line, ± IFNβ treatment, and incubated for 36 hours. Supernatants were assayed by ELISA as described above.

T-cell clones were incubated with antigen-presenting cells transiently transfected with plasmids encoding HLA-A*02:01 and GFP-truncated Large T-Ag (tLTAg) fusion protein (pDEST103-GFP-tLTAg). pDEST103-GFP-tLTAg was created using Gateway recombination cloning technology (Gateway) to insert tLTAg from pCMV-MCV156 (37) into pDEST103-GFP. A total of 3×10⁴ COS-7 cells (ATCC, CRL-1651, 2005) were plated in flat-bottom 96-well plates in DMEM + 10% FBS, 200 nmol/L L-glutamine and 100 U/mL penicillin–streptomycin. Twenty-four hours later, wells were transfected using FuGENE HD (Promega) at a 6:1 ratio of transfection reagent to DNA with 25 ng HLA-A*02:01 and limiting dilution of pDEST103-GFP-tLTAg (25-0.08 ng) plus irrelevant DNA (pcDNA-6/myc-His C, Gateway) to a total of 25 ng. 48 hours after transfection, 10⁴ viable KLL-specific T cells in TCM were added to target wells in duplicate. After 36 hours, supernatants were assayed by ELISA for IFNγ secretion, and EC₅₀ was calculated as above. Transfected COS-7 cells were harvested at 48 and 72 hours after transfection to quantitate transfection efficiency by flow cytometry.

HLA-A*02 is a prevalent HLA-type present in ∼55% of our MCC cohort (n = 97 low-resolution HLA class I typed patients; HLA-A*02:01 is the dominant A02 allele). We detected A*02-restricted T-cell responses in MCC patients to an epitope of the common T-Ag (aa 15–23) in 14% of PBMC (10 of 69) and 21% of cultured TIL (5 of 24; TIL were expanded with mitogen/cytokine for 2 weeks; ref. 17) from HLA-A*02⁺ patients. No tetramer⁺ cells were detected in PBMCs from healthy HLA-matched controls (0 of 15; Fig. 1). Among HLA-A*02⁺ patients, neither MCC-specific survival nor recurrence-free survival was significantly different between patients with or without detectable KLL-specific tetramer⁺ T cells (P = 0.593 and 0.643; data not shown). We believe that the detected KLL-specific T cells were primed by MCPyV due to the limited homology between this epitope and the homologous region of other polyomaviruses known to infect humans (Supplementary Table S1). Moreover, this epitope is predicted to bind to HLA-A*02 ∼100× better than these homologous peptides (IC₅₀ for the KLL MCPyV peptide is 299 nmol/L vs. 6,950–25,799 nmol/L for all other homologs as determined by the Immune Epitope Database; ref. 38).

Twelve patients had robust populations of KLL tetramer⁺ cells (>0.04% of CD8⁺ T cells) in their PBMCs and/or cultured TILs. Patient demographics, relevant disease metrics, and frequency of tetramer⁺ populations are summarized in Supplementary Table S2. All patients were Caucasian, with a median age of 65 (range, 50–77). The patients presented at varying stages of disease. Some developed progressive disease, and others showed no evidence of disease after definitive treatment during a median follow-up period of 2.7 years (range, 1.1–6.0) years.

We sequenced the complementarity determining region 3 (CDR3) region of TRB of KLL tetramer–sorted cells from PBMC (n = 9) and/or TIL (n = 5) from 12 patients (Fig. 2). Out of 397 unique TRB sequences, only one public TCRβ clonotype was detected and shared between just two patients. This clonotype dominated the KLL-specific repertoire of these patients (59.1% or 21.5% of KLL-specific TRB sequencing reads). Complete TCRβ sequence results for each patient, in order of decreasing frequency, are in Supplementary Table S3. Paired KLL tetramer⁺ T cells from both PBMCs and TILs were available for two patients (boxed). The diversity of the tetramer⁺TRB repertoire varied greatly between patients. The overall TRB diversity in a sample was not correlated with the frequency of tetramer⁺ T cells among total CD8⁺ cells in PBMCs (Fig. 1). We determined the clonality of each tetramer⁺ sample from PBMC (range, 0–1 with a completely clonal sample = 1; see Materials and Methods for details) and found no significant difference in MCC-specific survival or recurrence-free survival between patients with a less clonal (clonality < 0.3, n = 6) or more clonal (clonality >0.3, n = 3) KLL-specific repertoire in their PBMCs (Supplementary Fig. S2, P = 0.52 and 0.81 by the log-rank test).

Archival tumor samples were analyzed from 11 of 12 patients; tumor from w750 was unavailable. When possible, primary tumors were acquired (n = 6). For four patients with an unknown primary who presented with nodal disease, lymph nodes were analyzed. Primary tumor from w878 had insufficient material for study, and we therefore analyzed a metastasis corresponding to the time of PBMC collection. The primary tumor sample from w782 was small, and therefore to ensure adequate sampling, we also analyzed a nodal tumor present at time of diagnosis from w782. Tumors were assessed via immunohistochemistry (IHC) for viral status, HLA-I expression, and CD8⁺, CD4⁺, and FoxP3⁺ T-cell infiltration (Supplementary Fig. S3A). All patients were robustly positive for MCPyV Large T-Ag protein by IHC. CD8⁺ cells were more predominant than CD4⁺ or FoxP3⁺ T cells in the majority of samples. TRB CDR3 of all T cells in each tumor sample were sequenced and total unique TCRβ clonotypes/tumor were plotted in Supplementary Fig. S3A (n = 12; range, 16–41, with 645 unique clonotypes/tumor).

We measured whether having a greater number of total T cells was associated with a survival benefit. A priori, patients were binned by whether their tumors had many infiltrating T cells (≥0.8 T cells/ng tumor DNA, n = 7) or few T cells < 0.3 T cells/ng tumor DNA, n = 3); these two groups of patients had no survival difference (Supplementary Fig. S3B, P = 0.59 by the log-rank test). In addition, we calculated the TRB clonality of each tumor analyzed. Increased clonality of the immune infiltrate within tumors is thought to represent an enrichment of cancer antigen-specific T cells and has been associated with improved response to immunotherapy (39). However, MCC-specific survival or recurrence-free survival was similar in patients whose tumors had a less clonal repertoire (clonality <0.1, n = 7) to those with a more clonal repertoire (clonality > 0.1, n = 4; Supplementary Fig. S4A and S4B, P = 0.50 and 0.64 by log-rank test).

We next assessed how frequently KLL-specific T cells infiltrated MCC tumors. KLL-specific clonotypes within tumors were identified by determining the intersection between TCRβ CDR3 amino-acid sequences in the tetramer-sorted sample (from Fig. 2) and whole tumor samples from each patient. KLL-specific T cells constituted between 0 and 18% of the T-cell repertoire of each tumor based on the total number of T-cell genomes sequenced (n = 12, mean 6.3%, SD = 5.8; Supplementary Fig. S5A). Tumors contained between 0 and 108 unique KLL-specific TCRβ clonotypes (mean = 19.4, SD = 32; Supplementary Fig. S5B). The rank (based on frequency) of each KLL-specific clonotype within each tumor was plotted; individual clonotypes ranged between being the most prevalent clonotype to rare within each autologous tumor. KLL-specific clonotypes appeared to be more abundant (based on total percentage of all KLL-specific T cells in tumor) and predominant (based on percentage of individual KLL-specific clonotypes) in patients that were alive at last follow-up (Fig. 3). Patients were binned a priori based on the percentage of tumor with KLL-specific T cells. MCC-specific survival was significantly increased for patients who had a higher (1.9%–18%; n = 7) versus lower (0–0.14%; n = 2) percentage of KLL-associated T cells in tumor (Fig. 4A, P = 0.0009 by the log-rank test).

In addition, we asked whether the number of unique KLL-specific TCRβ CDR3 clonotypes infiltrating tumors was associated with survival. Indeed, patients who had more unique KLL-specific clonotypes in their tumors (5–108 clonotypes, n = 7 patients) had a significant survival advantage, compared with patients with few KLL-specific clonotypes (0–3, n = 4; Fig. 4B, P = 0.0051). When patients were separated into those who did and did not have a recurrence, the frequency of KLL-specific T cells was higher in tumors from patients without disease recurrence (median 10.4%) compared with patients who did recur (median of 3.2%, Fig. 4C, P = 0.11). In addition, the diversity of unique KLL-specific clonotypes was significantly higher in patients who did not have recurrences (median of 38 clonotypes) compared with patients who did (median of 2 clonotypes; Fig. 4D, P = 0.02). Collectively, these data show a significant survival advantage for patients whose tumors contain a higher relative percentage or a greater clonotypic diversity of KLL-specific T cells.

To gain insight into functional differences of unique KLL-specific TCRs, we generated KLL-specific T-cell clones from MCC patients' PBMCs (n = 4) and/or ex vivo tumor digest (n = 1) by sorting KLL-tetramer⁺ cells followed by limiting dilution cloning. Diversity of the TCRVβ of several KLL-reactive clones per patient was studied with fluorescent anti-TCRVβ antibodies via flow cytometry, and clones encompassing the spectrum of TCRVβ usage were expanded for further study. We determined the V, J, and CDR3 sequences of both TCRα and β chains for 120 clones and identified 71 monoclonal cultures, 42 of which were comprised of distinct TCRs, recognizing this epitope among four patients (Table 1). Among many private TCRα chains sequenced, one public TCRα chain using TRAV12-1*01 and encoding the CDR3 “CVLNNNDMRF” was found among clones from three of four patients.

To investigate functional differences among MCPyV-specific T-cell clones, we measured secretion of IFNγ, a canonical T-cell cytokine, after stimulation with T2 target cells pulsed with limiting dilution of an alanine-substituted variant of the peptide (KLLEIAPNA; this peptide is antigenic but less susceptible to oxidation, allowing direct comparison of T-cell clones to each other; see Materials and Methods for details). Clones displayed narrow ranges of intrapatient variability for functional avidity (Table 1; Fig. 5A). Concordant results were obtained in a separate but analogous assay using targets transfected with limiting dilution of plasmid encoding truncated Large T-Ag (Table 1; Fig. 5B). Importantly, patients with improved MCC-specific survival had more functionally avid T-cell clonotypes (P < 0.05). To further interrogate the effector function of these clonotypes, we tested the ability of 28 unique KLL-specific clonotypes to recognize the MCPyV⁺, HLA-A*02⁺ MCC cell lines (WaGa, MS-1 and MKL-2) ± IFNβ treatment. Five unique clonotypes secreted IFNγ when incubated with MS-1; this response was generally lower than that with T2 cells pulsed with a maximal concentration of peptide. No clones recognized WaGa or MKL-2 (Table 1 and Fig. 5C). Reactive clones were derived from patient w678, who had the most functionally avid clonotypes in the IFNγ release assay. Lastly, we compared the ability of KLL-specific clonotypes to bind both wild-type and “CD8-independent” tetramers that contain mutations in HLA-A*02:01 to abrogate CD8 stabilization of the TCR:pMHC interaction, which may select for more avid TCRs (31, 32). More functionally avid clonotypes in the IFNγ assays (Fig. 5A and B) often bound both wild-type (WT) and CD8-independent tetramers equivalently; however, there were many clonotypes that did not follow this trend (Table 1 and Fig. 5D). Indeed, when clones from each patient were binned by whether they bound the CD8-independent tetramer “equally” or “lower,” the mean EC₅₀ of these two groups did not differ significantly (P = 0.57 for w678 by the Mann–Whitney test, P = 0.30 for w830, insufficient data for w830 and w683). No significant correlations between clonotype avidity and enrichment within tumors were identified.

The purpose of our study was to characterize the TCR repertoire restricted to a naturally processed epitope of MCPyV in the context of the prevalent HLA-A*02 allele, and to assess whether differences in either the breadth or avidity of the TCRs correlated with the effectiveness of the T-cell response in vivo. We identified KLL tetramer⁺ CD8⁺ T cells in a minority of HLA-matched MCC patients. Presence of circulating KLL-specific T cells was not associated with differences in survival or recurrence among HLA-A*02⁺. The TCRβ repertoires of the KLL-specific T cells were strikingly diverse. When we examined the T cells within tumors, higher frequencies of KLL-specific T cells, as identified by their signature TRB CDR3 sequences, correlated with a significant MCC-specific survival advantage. These findings identify the diversity of the CD8⁺ T-cell response to MCC and suggest therapeutic maneuvers to boost tumor immunity. Overall, our findings provide a rationale for active or passive immunization to increase MCPyV-specific CD8⁺ T-cell diversity and avidity, and for manipulations of the immunosuppressive tumor microenvironment to promote the infiltration by these T cells.

This study is unique in its focus on the TCR repertoire specific for particular epitopes of MCPyV, using a high-throughput TCR sequencing approach to study tumor antigen–specific TCRs, which has revealed diversity that may have been missed using earlier methods. TCR sequencing was accompanied by the generation of KLL-specific clones and functional avidity characterization in a physiologically relevant system, allowing for paired analysis of unique TCRs and matched T-cell clones. Similar approaches have been utilized in other virally associated cancers to select the “best” TCRs for transgenic T-cell therapy (40), a therapeutic modality that has merit in MCC.

The frequency of KLL-specific T cells among A*02⁺ MCC patients was lower than expected (only 14% of A*02⁺ MCC patients have detectable KLL-specific T cells), given our finding that T cells restricted to an HLA-A*2402 epitope were found in 64% of HLA-matched PBMCs (18). These findings could be the result of poor processing and presentation of the KLL epitope (for instance, the A*2402-restricted epitope is 2-fold more avid for cognate HLA compared with the KLL epitope by Immune Epitope Database; ref. 36), which would prevent patients from being able to mount an immune response. Indeed, this hypothesis is supported by our finding that functionally avid KLL-specific clonotypes were only able to recognize one of three tested MCPyV⁺ A*02⁺ cell lines, even after upregulation of MHC-I on all three cell lines.

Alternatively, there may be adequate antigen presentation and yet patients are unable to mount their own endogenous effector response to this epitope (due to local immunosuppression or a host of other factors). These patients in particular may benefit from therapy with T cells expressing transgenic TCRs (tTCR) restricted to this epitope. The TCR sequences generated from this study could be used to create useful reagents (i.e., tTCR clones) to detect and quantify KLLEIAPNC on the surface of primary MCCs to help distinguish between these hypotheses.

In this in-depth examination of MCPyV-specific TCRs, we found that out of 397 KLL-specific TCRs detected among 12 patients, the vast majority (396) were private, with only one public TCR observed across two individuals. Public TCRs have been observed in multiple species in response to many viral infections and tumor antigens (41), and might be expected to be particularly prevalent in the response to a DNA virus thought to have low mutational capacity such as MCPyV. However, that idea is discordant with our observation in this study of a predominantly private repertoire, although the small sample size of our study (n = 12) is a limitation. Increased diversity of an antigen-specific T-cell response led to improved outcomes in models of chronic infections such as CMV (42) and herpes (43). In our study, the diversity of the circulating KLL-specific T-cell repertoire and MCC outcomes were not associated; however, within tumors an increased number of KLL-specific clonotypes was associated with improved MCC-specific survival. Although these studies elucidate TCR diversity restricted to a single epitope of MCPyV, we now have validated tetramers for five other peptide/HLA combinations (unpublished observations) and can replicate these studies to assess whether this striking diversity in immune response is specific to the KLL epitope or more broadly observed in the CD8⁺ T-cell response to MCPyV.

Our finding that a higher frequency of KLL-specific T cells within primary tumors is associated with a significant MCC-specific survival advantage builds on previously published work that CD8⁺ infiltration into tumors is associated with improved survival (14, 15), but is the first to confirm the importance of MCPyV-specific T cells in the infiltrate. In contrast, the presence of detectable circulating KLL-specific T cells was not associated with improved MCC-specific survival. Therefore, efforts should be focused on improving the tumor homing and infiltration of both endogenous and therapeutic MCPyV-specific T cells. Several candidate agents in are currently in preclinical development. In a mouse model of ovarian cancer, modification of histone methylation was associated with restored secretion of Th-1 cytokines by tumor cells, promoting T-cell infiltration and improved outcomes with T cell–based therapies such as anti–PD-L1 (44).

In parallel to other infections and malignancies (22–26), we hypothesized that increased avidity of the TCR repertoire may be correlated with the effectiveness of the T-cell response in vivo. Forty-two distinct clonotypes recognizing this epitope were identified by creating clones from PBMCs or tumor of four patients. Clones generally expressed a narrow intrapatient range of functional avidities with more avid clones isolated from patients with better MCC-specific outcomes. We detected specific IFNγ responses by the most avid clones from a patient with a favorable outcome, but only with one of three HLA-appropriate MCPyV cell lines. Antigen presentation by MCC cell lines is expected to be relatively minimal, because in vivo these cells survived by their ability to escape immune detection. There are numerous known mechanisms by which such immune evasion could be minimized, such as proteasome, TAP, and HLA class I/β₂-microglobulin expression and function. We hypothesize that variation in these factors in vivo may correlate with T-cell infiltration and immune checkpoint inhibitor responses and are investigating these areas in ongoing research.

In a trial of anti–PD-1 (pembrolizumab) in patients with metastatic MCC (16), objective responses were observed in 56% of patients overall and in 62% of patients with MCPyV-positive tumors. Among the 38% of patients with MCPyV⁺ MCCs that had stable or progressive disease (16), it is possible that these nonresponders, similar to the patients in this study with poor survival outcomes, lacked high-avidity and/or tumor-infiltrating MCPyV-specific T cells. Analyses similar to those outlined in this study may therefore elucidate prognostic markers for response to PD-1 axis blockade.

There are several limitations to our study. Our sample size was limited by the number of subjects with this rare cancer in our research cohort with detectable KLL-tetramer⁺ T cells (n = 12). Because there are no durable treatments for advanced disease, patients received a variety of therapies, including chemotherapy, radiation, and immunotherapies, that were not standardized between patients. All patients in this study who received no further therapy since definitive excision of their presenting lesion (n = 4) are currently alive with no evidence of disease (median follow-up time of 2.9 years; average recurrence for most MCC patients is 9 months), supporting the importance of their immune system in fighting MCC. Another limitation is that we studied only a single MCPyV epitope. It is almost certain that MCPyV-specific T cells recognizing other epitopes besides KLL contributed to the antitumor immune response. However, KLL-specific T cells were among the top 10 most frequent clonotypes within tumors of 7 of 9 patients studied, strongly suggesting that these T cells are a predominant factor in the effector response to MCC.

In summary, MCPyV-specific CD8⁺ T cells are detectable ex vivo in a substantial portion of HLA-A*02⁺ MCC patients and have considerable TCR diversity that corresponds to several orders of magnitude of functional avidity. Our hypothesis that infiltration of MCPyV-specific T cells leads to superior tumor control is supported by our findings of increased MCC-specific survival in patients with a higher frequency of intratumoral KLL-tetramer⁺ T cells. Our findings support further investigation of agents that improve T-cell tumor homing and infiltration, as well as use of avid TCRs for transgenic T-cell therapy in advanced MCC.

N.J. Miller, C.D. Church, D.M. Koelle, and P. Nghiem are co-inventors on a patent application filed by their employer, University of Washington, concerning TCR sequences. P. Nghiem reports receiving a commercial research grant from EMD Serono and Bristol-Myers Squibb, and is a consultant/advisory board member for EMD Serono. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N.J. Miller, D.M. Koelle, P. Nghiem

Development of methodology: N.J. Miller, L. Dong, D. Crispin, I. Gavvovidis, M. McIntosh, T. Blankenstein, D.M. Koelle

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.J. Miller, C.D. Church, M. Shinohara

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.J. Miller, C.D. Church, D. Crispin, M.P. Fitzgibbon, M. Shinohara, G. Willimsky, M. McIntosh, P. Nghiem, K. Lachance

Writing, review, and/or revision of the manuscript: N.J. Miller, C.D. Church, D. Crispin, I. Gavvovidis, G. Willimsky, M. McIntosh, T. Blankenstein, D.M. Koelle, P. Nghiem

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Jing, P. Nghiem

Study supervision: P. Nghiem

We thank the patients who made this research possible. In addition, we thank Dr. Juergen Becker for MCC cell lines WaGa and MKL-2 and Dr. Masahiro Shuda for the MCC cell line MS-1.

Supported by NIH grants (F30-CA192475-02, K24-CA139052, R01-CA162522-05, R01-CA 176841, and R01-AI094109), Adaptive Biotechnologies Young Investigator Award, Kelsey Dickson Team Science Courage Research Team Award, Prostate Cancer Foundation Award#15CHAS04, ARCS Fellowship, Environmental Pathology/Toxicology Training Grant T32ES007032-36, NIH Center for Aids Research (CFAR) AI027757, National Cancer Institute (NCI) Cancer Target Discovery and Development (CTD2) U01 CA176270, the David & Rosalind Bloom Endowment for MCC Research, the Michael Piepkorn Endowment Fund, the UW MCC Patient Gift Fund, and the Deutsche Forschungsgemeinschaft (SFB TR36).

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.