Response of B cells specific for polyomavirus-derived oncoprotein is predictive of Merkel cell carcinoma tumor control

Abstract

Merkel cell carcinomas (MCC) typically arise from clonal integration of the Merkel cell polyomavirus. Immunogenic viral oncoproteins then lead to tumorigenesis. Oncoprotein-specific T cells are essential for anti-MCC immunity, but it is unclear whether B cells promote tumor control. Here, we analyzed the frequency and phenotype of viral oncoprotein–specific and total B cells in blood samples from 47 patients with MCC and tumor samples from another 19 patients with MCC. The phenotype of blood B cells did not correlate with MCC patient outcomes. In contrast, all 11 patients with robust oncoprotein-specific antibody-secreting and/or germinal center B cells in tumors experienced long-term MCC control. In vitro, B cells engineered to be specific for viral oncoproteins increased the sensitivity of oncoprotein-specific CD4⁺ T cells by over 50-fold. Together, our findings suggest that cancer-specific B cells promote antitumor immunity via increased responses by T cells and that cancer-specific augmentation of B cells could be therapeutically relevant.

Introduction

Merkel cell carcinoma (MCC) is a rare aggressive skin cancer with a mortality rate of ~30% (1–4). MCC is usually driven by integration of Merkel cell polyomavirus (MCPyV) T-antigen (T-Ag) DNA into host chromosomes. This leads to constitutive expression of the viral small and truncated large T-Ag oncogenic proteins, which are responsible for tumorigenesis (5–7). T-Ag oncoproteins are immunogenic and targeted by T cells (8–11) and B cells (12,13). While studies have established the importance of MCPyV T-Ag–specific CD8⁺ T cells in MCC tumor control, the role that antibodies and B cells play in antitumor immunity is unknown.

Most patients with T-Ag–driven MCC have T-Ag–specific antibodies in blood that are largely undetectable in individuals without MCC (12,13). The level of T-Ag–specific antibodies at the time of diagnosis is correlated with tumor burden but is not predictive of MCC progression after tumor surgical excision and local radiation (12,13). In people that do not experience MCC progression, the level of T-Ag–specific antibodies declines rapidly, often becoming undetectable. In contrast, stable levels of T-Ag–specific antibodies or a brief decline followed by a rapid increase usually indicates MCC progression (12–14). These dynamics combined with the intracellular expression of T-Ag have led to the hypothesis that T-Ag–specific antibodies are produced in response to the presence of MCC but do not play a role in tumor control.

Addressing the function of tumor-specific B cells is challenging due to the diversity in tumor antigens which often varies from person-to-person within cancers of the same type (15,16). Studying tumor-specific B cells is further complicated by their low frequency in solid tumors (17–19). Animal studies have demonstrated that cancer-specific B cells can enhance antitumor immunity by promoting antigen presentation to CD4⁺ T cells, which in turn can recruit and enhance cancer-specific CD8⁺ T cells to the tumor microenvironment (20). Interestingly, tumors from Human Papillomavirus (HPV)⁺ head and neck cancer patients contain HPV-specific B cells (21,22), but whether or not these cells contribute to tumor control or associate with disease outcomes is not known.

Expression of T-Ag oncoproteins by MCC allows for the study of cancer-specific B cells across patients using T-Ag tetramers. Using this approach, we show herein that while T-Ag–specific B cells were found at increased frequencies in the blood of MCC patients compared to controls, the phenotype of B cells did not associate with disease outcome. In contrast, detection of T-Ag–specific antibody-secreting and/or germinal center B cells in tumor samples was predictive of extended progression-free survival after treatment, whereas low or undetectable levels of these cells predicted rapid MCC progression. Using antibodies and T cell receptors (TCR) identified from patient samples to engineer human T-Ag–specific B cells and CD4⁺ T cells, we further demonstrated highly efficient cognate antigen presentation, suggesting a mechanism by which local T-Ag–specific B cells could enhance T cell–mediated immunity. Together these findings show the predictive power of local T-Ag–specific B cells as biomarkers of MCC control and suggest that functional B cells promote antitumor immunity by facilitating antigen presentation to T cells.

MATERIALS AND METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contacts, Paul Nghiem (pnghiem@uw.edu) and/or Justin Taylor (justintaylor@virginia.edu).

Materials availability

Commercially available materials are listed in Supplementary Table S1. In-house made materials are available upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mouse Studies

Animal studies were conducted in accordance with protocols approved by the Fred Hutchinson Cancer Center Institutional Animal Care and Use Committee. Six- to ten-week-old female C57BL/6 mice were obtained from the Jackson Laboratory.

Human Studies

This study was conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent. Specific details of patient samples utilized, and corresponding institutional review board (IRB) protocols are described here.

Blood samples from 50 MCC patients were collected at time of disease diagnosis or disease recurrence and cryopreserved in liquid nitrogen. Patients provided informed consent for sample research use as part of an observational registry study focusing on MCC that was approved by the Fred Hutchinson Cancer Center (FHCC) IRB (IRB protocol 6585). Titers of T-Ag–specific antibodies in these blood samples were determined by the Clinical Immunology Lab at Department of Laboratory Medicine and Pathology at the University of Washington using the AMERK test (13). Patient samples were selected based on availability of frozen viable peripheral blood mononuclear cells (PBMCs) collected within 67 days of MCC diagnosis. In addition, patient samples were selected to be representative of the sex distribution observed in virally driven MCC. Patient samples were further selected to only include MCPyV-positive cases and those from immune competent patients. An additional 21 blood samples from age-matched, non-MCC controls were collected and cryopreserved in liquid nitrogen until use. Non-MCC donors provided informed consent for sample research use under protocols approved by the University of Washington IRB (IRB protocol STUDY00001399).

Tumor samples from 105 MCC patients were also collected at different disease stages as part of FHCC IRB protocol 6585. Of these 105 MCC tumor samples, 19 samples were digested after surgery to create single cell suspensions that were cryopreserved in liquid nitrogen. The additional 86 MCC tumors collected were formalin-fixed and parafilm-embedded for imaging. Tumor samples for single-cell experiments were selected based on availability of frozen viable tumor digests, number of cells available in sample (>10 million), and included only MCPyV-positive cases. Patient samples were primarily from males due to limited female tumor samples.

Recurrence risk of patients within the MCC PBMC cohort was calculated using an open-source MCC recurrence risk calculator (https://merkelcell.org/prognosis/recur/) (4). This calculator estimates the probability of recurrence from the time of diagnosis in patients with no prior presentation of MCC. Briefly, percent probability of recurrence for each patient following MCC diagnosis was calculated using known risk factors including age, sex, stage as defined by the American Joint Committee on Cancer, site of primary tumor, and immune status. In contrast, because some MCC patients in the single-cell RNA-sequencing (RNAseq) tumor cohort had prior disease recurrences, clinical recurrence risk in these patients was independently assessed by 3 blinded clinicians. Known risk factors used for this assessment included age, sex, MCC stage, presence of metastatic disease, prior recurrences, and immune status. Each patient was then categorized as having “Medium/Low” or “High” probability of MCC recurrence following surgical excision of analyzed tumor.

To assess long-term MCC outcomes in patients within the MCC PBMC cohort, patients were monitored for 3 years starting at the time of biopsy-proven MCC diagnosis. For progression-free survival analyses, an event consisted of a patient having a recurrence or progression of MCC following definitive treatment (surgery, radiation, or anti-PD1 immunotherapy alone or in combination). Patients with no MCC recurrences were censored following 3 years of follow-up.

To assess long-term MCC outcomes in patients with samples analyzed in the single-cell RNAseq tumor studies and in tumor imaging studies, patients were monitored for 4 and 3 years, respectively, starting at the time surgical excision of analyzed tumor. For progression-free survival analyses, an event consisted of a patient having a recurrence or progression of MCC following definitive treatment (surgery, radiation, or anti-PD1 immunotherapy alone or in combination). Patients with no MCC progression were censored following the monitoring period for each study.

METHOD DETAILS

Production of T-Ag, GST, and RSV F proteins

Production of the common T-Ag domain shared by both the small and large T-Ag isoforms, was performed as previously described (12). Briefly, pGEX4T3 plasmids (MilliporeSigma #GE28–9545-52) encoding the T-Ag fused to glutathione S-transferase (GST) or GST alone were expressed in Rosetta Escherichia coli (MilliporeSigma #70954) and harvested by centrifugation 15 h after induction. Pelleted bacteria were resuspended in 40 mM Tris pH 8.0, 200 mM NaCl, 1 mM EDTA, and 2 mM DTT and lysed by two passes through a Microfluidizer. Lysates were then diluted with an equal volume of glycerol and stored at −20°C. Pierce Glutathione Spin Columns (ThermoFisher #16105 or #16107) were used following the manufacturer’s instructions to purify GST and T-Ag. After purification, proteins were buffer exchanged by centrifugation using 10 kDa Amicon Ultra centrifugal filters (MilliporeSigma #UFC9010) into 1xDPBS, aliquoted at 40 μM concentrations, and run on SDS-PAGE to assess protein purification.

Use of the expression plasmid for production of His-tagged RSV prefusion F antigen was previously described (23). Briefly, 293F cells (ThermoFisher #R79007) were transfected at a density of 10⁶ cells/mL in Freestyle 293 media (ThermoFisher #12338026) using 1 mg/mL PEI Max (Polysciences #24765). Notably, 293F cells were passaged fewer than ten times and were not authenticated or tested for mycoplasma contamination prior to use. Transfected cells were cultured for 7 days with gentle shaking at 37 °C. Supernatant was collected by centrifuging cultures at 2,500×g for 30 minutes followed by filtration through a 0.2 μM filter. The clarified supernatant was incubated with Ni Sepharose beads (Cytiva #17526801) overnight at 4 °C, followed by washing with wash buffer containing 50 mM Tris, 300 mM NaCl, and 8 mM imidazole. His-tagged protein was eluted with an elution buffer containing 25 mM Tris, 150 mM NaCl, and 500 mM imidazole. The purified protein was run over a 10/300 Superose 6 size exclusion column (Cytiva #29091598). Fractions containing the trimeric F protein were pooled and concentrated using 50 kDa Amicon Ultra centrifugal filters (MilliporeSigma #UFC9050). The concentrated sample was stored in 50% glycerol at −20°C.

Production of T-Ag and control tetramers

Production of T-Ag, GST, and RSV F tetramers was performed using previously described protocols (24,25). Briefly, purified T-Ag, GST, and RSV F proteins were biotinylated using the EZ-Link Sulfo-NHS-LC Biotinylation kit (ThermoFisher #A39257) according to manufacturer instructions. Following the biotinylation reaction, excess biotin was removed by centrifugation using 10 kDa Amicon Ultra centrifugal filters. Next, the molar ratio of biotin:protein was assessed by Western blot and selected to be ~1:1 to prevent over- or under-biotinylation of antigen that could reduce the downstream efficacy of protein tetramers. After optimization of biotinylation, T-Ag, GST, and RSV F tetramers were assembled by incubating biotinylated protein with either streptavidin-APC (Agilent #PJ27S) or streptavidin-PE (Agilent #PJRS25) at a 4:1 molar ratio of protein:streptavidin for 30 minutes at room temperature and in the dark. Unconjugated T-Ag and GST were removed by centrifugation using 100 kDa Amicon Ultra centrifugal filters (MilliporeSigma #UFC9100), whereas unconjugated RSV F was removed by centrifugation using a 300 K Nanosep centrifugal device (Cytiva #OD300C33). Tetramers were buffer exchanged into 1xDPBS by centrifugation using a 100 kDa Amicon Ultra centrifugal filter (MilliporeSigma UFC810096) and diluted to 1 μM in 50% glycerol in 1xDPBS for long-term storage at −20°C.

To make control tetramers, streptavidin-PE was conjugated to DyLight 594 (PE594) or DyLight 650 (PE650), and streptavidin-APC conjugated to DyLight 755 (APC755) using DyLight Antibody Labeling kits (ThermoFisher #46413, #62266, or #62279, respectively) following the manufacturer instructions. After removal of excess dye by centrifugation using 10 kDa Amicon Ultra centrifugal filters, biotinylated GST was incubated with streptavidin-PE594 or streptavidin-APC755 at a 4:1 ratio of protein:streptavidin for 30 minutes at room temperature and in the dark. Control tetramers were diluted to 1–3 μM in 50% glycerol diluted with 1xDPBS for long-term storage at −20°C.

Oligo-barcoded tetramers were assembled using TotalSeq-C streptavidin-PE (BioLegend, Supplementary Table S1). Briefly, biotinylated T-Ag, biotinylated GST-DL594, and biotinylated RSV F were incubated with TotalSeq-C streptavidin-PE containing uniquely identifying oligo tags (Supplementary Table S1) at a molar 4:1 ratio of protein:streptavidin for 30 minutes at room temperature and in the dark. Unconjugated T-Ag and GST were removed by centrifugation using 100 kDa Amicon Ultra centrifugal filters, whereas unconjugated RSV F was removed by centrifugation using a 300 K Nanosep centrifugal device.

Enrichment and analysis of tetramer-binding mouse B cells

Six- to ten-week-old C57BL/6J female mice were used for experiments. Mice were injected intraperitoneally with 50 μL of complete Freund’s adjuvant (CFA, InvivoGen #vac-cfa) emulsion containing 5 μg of T-Ag in 1xDPBS or 1xDPBS alone as control. Seven days post-vaccination, tetramer-binding B cells were enriched using a previously described protocol (24,25). Briefly, the spleen, and inguinal, axillary, brachial, cervical, mesenteric, and periaortic lymph nodes for each mouse were pooled and manually dissociated. Following dissociation, 10 mL of 1xDPBS containing 1% heat-inactivated newborn calf serum (ThermoFisher #26010074) was added, and tissue debris removed, the samples pelleted, and supernatant discarded. For staining of T-Ag–specific B cells, each sample was first incubated with 1 pmol of PE594 control tetramer, 3 pmol of APC755 control tetramer, and 2 μg of anti-CD16/32 clone 2.4G2 (BioXcell #BE0307) in 0.2 mL sorting buffer for 10 minutes on ice. Next, 1 pmol GST APC tetramer, 1 pmol T-Ag PE tetramer, Fixable Viability Dye eFluor 506 (ThermoFisher #65–0866-18), and fluorescent antibodies against cell-surface lineage markers CD3, F4/80, Gr-1, B220, CD38, and GL7 were added to samples followed by incubation on ice for 25 minutes. See Supplementary Table S1 for details of the antibodies used. After the incubation, ~10 mL of sorting buffer was added and the samples centrifuged at 300 × g for five minutes at 4°C and supernatant discarded. Cells were resuspended in 0.15 mL sorting buffer containing 25 μL of anti-APC microbeads (Miltenyi #130–090-855) and 25 μL of anti-PE microbeads (Miltenyi #130–048-801) and incubated for 30 minutes on ice. Next, 5 mL of sorting buffer was added to each sample and passed over a magnetized LS column (Miltenyi #130–042-401). The tube and column were washed once with 5 mL of sorting buffer and then removed from the magnetic field. Five mL of sorting buffer was pushed through the column with a plunger to elute column-bound cells. Another 5 mL of sorting buffer was plunged through the column to maximize recovery.

For intracellular staining, cells were fixed and permeabilized using the BD Cytofix/Cytoperm kit (ThermoFisher #554722) according to the manufacturer’s instructions. Following permeabilization, anti-mouse Ig Alexa Fluor 350 (ThermoFisher #A-11068) was added and samples incubated for 30 minutes on ice. After incubation, ~4 mL of sorting buffer was added and the samples centrifuged at 300×g for five minutes at 4°C and supernatant discarded. Samples were resuspended sorting buffer containing 20,000 AccuCheck Counting Beads (ThermoFisher #PCB100) analyzed by the BD Fortessa flow cytometer at the Fred Hutch Cancer Center’s Flow Core Facility. UltraComp eBeads Plus (ThermoFisher #01–333-42) or splenocytes were used as single stained controls. Compensation matrix was set up using the BD FACSDiva software. Analysis performed using FlowJo v.10.10.

Enrichment and sorting of T-Ag–specific B cells from human blood for single-cell B cell receptor (BCR) sequencing

Heparinized whole blood from MCC patients was processed at the Specimen Processing Lab (FHCC). PBMC were isolated by routine Ficoll density gradient centrifugation and cryopreserved in freezing medium containing 50% human serum (Akron Bio #AR1010–0100), 40% RPMI (ThermoFisher #22400), and 10% DMSO (MilliporeSigma #D2650). 5–10 × 10⁷ frozen PBMCs from MCC patients or non-MCC controls were thawed into DMEM (ThermoFisher #11965) with 10% heat-inactivated fetal bovine serum (Peak Serum #PS-FB3 or GeminiBio #100–500), 100 U/mL penicillin + 100 μg/mL streptomycin (ThermoFisher #15140122). Cells were centrifuged at 300×g for five minutes at 4°C, resuspended in 100 μL of ice-cold sorting buffer containing 1 pmol of GST PE650 tetramers and 2% rat serum (MilliporeSigma #S7648) and 2% mouse serum (MilliporeSigma #S7273), and incubated at on ice for 10 min. Next, 1 pmol of T-Ag PE tetramers were then added at a final concentration of 5 nM and incubated on ice for 25 min, followed by a 10 mL wash with ice-cold sorting buffer. Next, 25 μL each of anti-APC and anti-PE microbeads were added and incubated on ice for 30 minutes, after which 3 mL of sorting buffer was added, and the mixture was passed over a magnetized LS column. The column was washed once with 5 mL ice-cold sorting buffer and then removed from the magnetic field and 5 mL ice-cold sorting buffer was pushed through the unmagnetized column twice using a plunger to elute the bound cell fraction. Cells were centrifuged at 300×g for five minutes at 4°C, then incubated with 50 μL of sorting buffer containing a Fixable Viability Dye eFluor 506, and fluorescent antibodies against cell-surface lineage markers (CD3, CD14, CD16, and CD19) for 30 min on ice prior to washing and analysis on a BD FACS Aria sorter at FHCC’s Flow Core Facility. UltraComp eBeads Plus or PBMCs were used as single stained controls (see Supplementary Table S1 for details of the antibodies used). Compensation matrix and gating of tetramer-binding B cells was set up using the BD FACSDiva software. Tetramer-binding B cells were then individually sorted into empty 96-well PCR plates and immediately frozen.

BCR sequencing using single-cell reverse transcription PCR

Reverse transcription (RT) was directly performed on individual B cells sorted and frozen into empty 96-well PCR plates, as described previously (26). Briefly, after plates were thawed, 3 μL RT reaction mix consisting of 3 μL of 50 μM random hexamers (ThermoFisher #48190011), 0.8 μL of 25 mM deoxyribonucleotide triphosphates (dNTPs; ThermoFisher #N8080261), 1 μL (20 U) SuperScript IV Reverse transcriptase (ThermoFisher #18090200), 0.5 μL (20 U) RNaseOUT (ThermoFisher #10777019), 0.6 μL of 10% Igepal (MilliporeSigma #I8896), and 15 μL DEPC-treated water (ThermoFisher #750023) was added to each well containing a single sorted B cell and incubated at 50°C for 1 h. Following RT, 2 μL of cDNA was added to 19 μL PCR reaction mix so that the final reaction contained 0.2 μL (0.5 U) HotStarTaq Polymerase (Qiagen #203607), 0.075 μL of 50 μM 3′ reverse primers, 0.115 μL of 50 μM 5′ forward primers, 0.24 μL of 25 mM dNTPs, 1.9 μL of 10× buffer (Qiagen #203607), and 16.5 μL of DEPC-treated water. See Supplementary Table S1 for primer details. The PCR program was 50 cycles of 94°C for 30 s, 57°C for 30 s, and 72°C for 55 s, followed by 72°C for 10 min for heavy and kappa light chains. The PCR program was 50 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 55 s, followed by 72°C for 10 min for lambda light chains. After the first round of PCR, 2 μL of the PCR product was added to 19 μL of the second-round PCR reaction so that the final reaction contained 0.2 μL (0.5 U) HotStarTaq Polymerase, 0.075 μL of 50 μM 3′ reverse primers, 0.075 μL of 50 μM 5′ forward primers, 0.24 μL of 25 mM dNTPs, 1.9 μL 10× buffer, and 16.5 μL of DEPC-treated water. See Supplementary Table S1 for primer details. PCR programs were the same as for the first round of PCR. 4 μL of the PCR product was run on an agarose gel to confirm the presence of a ~500-bp heavy chain band or ~450-bp light chain band. 5 μL from the PCR reactions showing the presence of heavy or light chain amplicons was mixed with 2 μL of ExoSAP-IT (ThermoFisher #78201) and incubated at 37°C for 15 min followed by 80 °C for 15 min to hydrolyze excess primers and nucleotides. Hydrolyzed second-round PCR products were sequenced by Genewiz with the respective reverse primer used in the second-round PCR, and sequences were analyzed using IMGT/V-Quest to identify V, D, and J gene segments. In-Fusion cloning (Clontech #639650) was used to clone heavy chain VDJ and light chain VJ sequences were cloned into pTT3-derived expression vectors containing the human IgG1, IgK, or IgL constant regions obtained from Andrew McGuire FHCC. See Supplementary Table S2 for BCR sequences.

Monoclonal antibody production

Secreted T-Ag–specific and control IgG1 antibodies were produced by Genscript, WuXi, or in house, as described previously (26). For in-house generation, 10⁶ cells/mL of 293F cells were co-transfected with heavy and light chain expression plasmids at a ratio of 1:1 in Freestyle 293 media (ThermoFisher) using 1 mg/mL PEI Max (Polysciences #24765). Transfected cells were cultured for 7 days with gentle shaking at 37 °C. Supernatant was collected by centrifuging cultures at 2,500×g for 15 minutes followed by filtration through a 0.2 μM filter. Clarified supernatants were then incubated with Protein A agarose (GoldBio #P-400–50) followed by washing with IgG-binding buffer (GoldBio). Antibodies were eluted with IgG Elution Buffer (ThermoFisher #21028) into a neutralization buffer containing 1 M Tris-base pH 9.0. Purified antibody was concentrated and buffer exchanged into 1xDPBS using 50 kDa Amicon Ultra centrifugal filters.

Bio-Layer Interferometry

Bio-layer interferometry (BLI) assays were performed on the Octet.Red instrument (ForteBio) at room temperature with shaking at 500 rpm. To assess binding of monoclonal antibodies to T-Ag, streptavidin capture sensors (ForteBio #18–5020) were loaded in kinetics buffer (1xDPBS with 0.01% heat-inactivated bovine serum albumin, 0.02% Tween 20, and 0.005% NaN₃, pH 7.4) containing 1 μM biotinylated GST as control or biotinylated T-Ag for 2.5 minutes. After loading, the baseline signal was recorded for 1 minute in kinetics buffer. The sensors were then immersed in kinetics buffer containing 40 μg/mL of purified monoclonal antibody for a five-minute association step followed by immersion in kinetics buffer for a five-minute dissociation phase. The binding curves were generated after subtracting the background signal from each analyte-containing well using a negative control mAb at each time point. Curve fitting was performed using a 1:1 binding model and ForteBio Octet data analysis software release 9.0.

Spectral flow cytometry of B cells in MCC patient blood

PBMC from MCC patients were analyzed by spectral flow cytometry. Briefly, frozen PBMC were thawed into DMEM with 10% heat-inactivated fetal bovine serum and 100 U/mL penicillin + 100 μg/mL streptomycin. Cells were centrifuged and resuspended in 50 μL of ice-cold sorting buffer. GST APC755 was added at a final concentration of 1.25 pmol in the presence of 2% rat serum (MilliporeSigma #S7648) and 2% mouse serum (MilliporeSigma #S7273), and incubated at room temperature for 10 min. Cells were then incubated with 50 μL of sorting buffer Fixable Viability Dye eFluor 506, 0.5 pmol of T-Ag APC tetramer, and fluorescent antibodies against cell-surface lineage markers (CD3, CD14, CD16, CD19, CD20, CD10, CD27, CD11c, CD71, IgM, IgD, and IgG) for 30 min on ice prior to washing and analysis on a Cytek Aurora spectral analyzer at the University of Washington, Department of Immunology’s Cell Analysis Facility. See Supplementary Table S1 for details of the antibodies used. Antibody capture beads (ThermoFisher #A10497) or cells were used to compensate each fluorophore in the experiment. Spectral unmixing was performed using SpectroFlo software. Analysis was performed using FlowJo v.10.10.

Tumor Cellular Indexing of Transcriptomics and Epitopes by Sequencing (CITEseq) sample preparation

Fresh MCC tumor specimens from needle cores, punch biopsies, or surgical excisions were processed into single-cell digests by mincing them into small pieces with sterile forceps and scissors, followed by incubation in 20 mL of digestion medium composed of RPMI plus 0.002 g DNase (Worthington Biochemical #LS002139), 0.008 g collagenase (Worthington Biochemical #NC991993), and 0.002 g hyaluronidase (Worthington Biochemical #LS005477) in a 10-cm dish at 37°C with frequent, gentle swirling. After 3 hours of digestion, cells were strained through a 70 μm filter, centrifuged, resuspended in freezing medium containing 50% human serum, 40% RPMI, and 10% DMSO and stored in liquid nitrogen.

Frozen cells were thawed at 37°C, followed by dropwise addition of 1 mL complete media (RPMI, 10% fetal bovine serum, 100 U/mL penicillin + 100 μg/mL streptomycin). Four equal volume additions of complete media were added dropwise with gentle mixing in between additions for a total volume of 32 mL. Thawed cells were washed twice with 4°C 1xDPBS and transferred to 5 mL FACS tubes (ThermoFisher #352054). Fixable Viability Dye eFluor 506 was added, followed by a blocking buffer to bring samples to 0.5% BSA, 5% TruStain FcX buffer (BioLegend #422302), 100 nM dasatinib (Selleck Chem #S1021), and oligo-PE-GSTDL594 and APCDL755-GST control tetramers. Samples were then incubated on ice for 10 min followed by the following reagents in order: tetramers (T-Ag PE-oligo tetramer, T-Ag APC tetramer, RSV F PE-oligo tetramer, and RSV F APC tetramer), oligo-hashtag antibodies to identify sample origin in subsequent pooling steps, fluorochrome-labeled antibodies, and oligo-labeled antibodies. Cells were then incubated on ice for 30 min and washed three times with sorting buffer. Supplementary Table S1 contains details of the antibodies used. Cells were then sorted on an Aria II Cell sorter (BD Biosciences). After exclusion of cell debris, dead cells, PE/APC tetramer-binding CD19⁺CD3⁻CD56⁻PE594⁻APC755⁻ B cells, tetramer-negative B cells CD19⁺CD3⁻CD56⁻PE594⁻APC755⁻ B cells, CD3⁺CD19⁻CD56⁻ T cells, and CD56⁺CD3⁻CD19⁻ cells were sorted into cold complete media, pooled, and immediately prepared for sequencing. For most experiments, tetramer⁺CD19⁺ B cells were sorted into a tube with CD3⁺ T cells and kept separate from Tetramer⁻CD19⁺ B cells, which were sorted into a tube with CD56⁺ MCC cells.

Single cell CITEseq

Single-cell suspensions sorted from tumors and brought to a concentration of 700–1,200 cells/mL were loaded into microfluidic chip K (10X Genomics #1000286) and run through a 10X Genomics Chromium controller to obtain Gel Beads-in-Emulsion. Resulting cell suspensions then went through a library preparation process for single-cell RNAseq along with paired V(D)J-seq for BCR (10X Genomics #1000253) and TCR (10X Genomics #1000252) clonotypes using the 5’ transcriptome kit with feature barcoding (10X Genomics #1000263, #1000190, #1000215, #1000250, #1000541, #1000248) per manufacturer guidelines. The cDNA library was enriched for 1,056 human immunology panel genes (Supplementary Table S3) (10X Genomics #1000246) and sequenced using an Illumina NovaSeq instrument with 2 × 92 base pair paired-end reads aiming for an average of 20,000 reads/cell.

Single-cell CITEseq data analysis

Raw sequencing reads were aligned to the hg38 genome using Cell Ranger v.8.0.1. Filtered counts matrices of transcripts and feature barcoding counts were loaded into a SingleCellExperiment object for further analyses in R (v.4.1.2). Sample hash deconvolution was performed using DropletUtils (v.1.14.2). Doublets were detected using scds (v.1.10.0) and hash deconvolution and subsequently removed.

Cells from different runs were then integrated using the mutual nearest neighbor method though the batchelor package (v1.10.0). Uniform manifold approximation and projection (UMAP) dimensionality reduction was performed using the integrated values. Clustering was performed using the integrated transcript values reads through the walktrap algorithm on a nearest neighbor graph (scran v.1.22.1). Numbers of clusters was varied by scaling the number of nearest neighbors (k) during graph construction followed by analysis via clustree (v.0.5.0). Clusters were then labeled as major cell lineages of T cells, B cells, myeloid cells, endothelial cells, dendritic cells, and tumor cells through expression of key genes including CD79A, CLEC4C, COL3A1, NCAM1, CD14, and CD3E. Cluster labels were then validated by investigating the portion of cluster with productive BCR or TCR rearrangements. Cell lineages were then isolated in silico and dimensionality reduction and B cells re-clustered as above. Clusters were labeled based on expression of lineage markers.

To identify T-Ag–specific B cells in tumors, we analyzed T-Ag oligo reads and control antigen oligo reads in B cells. T-Ag–specific CD19⁺ B cells were defined as cells with T-Ag oligo reads at least one log higher than background and fewer than 20 reads from GST or RSV F control antigens. Additionally, eight cell families in which fewer than 50% of clones met this threshold were classified as non-specific. Sequences from tested antibodies were obtained from BCR sequencing analyses and are reported in Supplementary Table S4.

BCR sequencing analysis

Raw output files were demultiplexed and processed using CellRanger v.8.0.1 software (10X Genomics). For each donor, publicly available reference sequences from the IMGT/V-QUEST reference directory at https://www.imgt.org/ were used to identify V, D, and J gene segments. Next, data were processed and analyzed using the Immcantation Framework (http://immcantation.org) with Change-O v.1.0.2. For each 10X dataset, the filtered_contig.fasta file was annotated using IgBlast v.1.16 with the related donor-specific VJ genes database. To generate adaptive immune receptor repertoire (AIRR) rearrangement data, the filtered_contig_annotations.csv file was used, and only productive sequences were kept. The heavy and light chain sequences were separated in two files. The threshold for trimming the hierarchical clustering of clones was determined by the SHazaM module for determining distance to nearest neighbor. With the Change-O DefineClones function, clones were assigned based on IGHV genes, IGHJ gene and junction distance calculated by SHazaM (distance 0.14). The generated clone-pass file was verified and corrected using the Change-O light_cluster function, based on the analysis of the light chain partners associated with the heavy chain clone. Independent clone-pass files were generated for each 10X run. For downstream analysis, all clone-pass files from were combined and re-clustered all together. Germlines were reconstructed using the Change-O CreateGermlines function. To obtain the final AIRR format file containing paired information on the same row, we used a Java script to process and filter the sequences. Only the heavy chains paired with one IGK and/or one IGL were filtered in for downstream analysis. Number and frequency of IgH mutations for each B cell were obtained using the SHazaM observedMutations function. Clonal families were visualized using the R package ggalluvial.

Multiplex immunohistochemistry of MCC tumors

Tumor microarray (TMA) blocks were constructed on the fully automated 3DHistech TMA Grand Master instrument. Two 1.0 mm diameter tumor cores and two tumor margin cores were sampled per case. The Grand Master TMA software overlapped an annotated digital slide with the image of the corresponding donor block and cores were removed from the blocks using the Grand Master TMA instrument. These cores were then relocated to a recipient block in a precise alignment. Subsequently, 4 μm sections were cut from the constructed blocks and were stained on a Leica BOND Rx autostainer using the Akoya Opal Multiplex IHC assay (Akoya Biosciences #NEL871001KT) with the following changes: Additional high stringency washes were performed after the secondary antibody and Opal fluor applications using high-salt TBST (0.05M Tris, 0.3M NaCl, and 0.1% Tween-20, pH 7.2–7.6). TCT was used as the blocking buffer (0.05M Tris, 0.15M NaCl, 0.25% Casein, 0.1% Tween 20, pH 7.6 +/− 0.1). Primary antibodies specific for CD3, CD19, CD20, CD56, CD138, and Ki67 were incubated for 1 hour at room temperature. See Supplementary Table S1 for details of the antibodies used. Tissues were counterstained with DAPI to identify the nuclei.

Slides were mounted with ProLong Gold (ThermoFisher #P36930) and cured for 24 hours at room temperature in the dark before image acquisition at 20x magnification on the Akoya PhenoImager HT Automated Imaging System. Images were spectrally unmixed using Akoya inForm software.

Multiplex immunohistochemistry image analysis

HALO software (Indica Labs) High-Plex FL module was used to analyze each tumor microarray slide obtained from MCC tumor tissues described above. Annotations of tumor and stroma sections were performed by a licensed pathologist. Cells were then identified based on the DAPI nuclear stain, and mean pixel fluorescence intensity was measured in applicable compartments for each cell. To account for variability across the tumor microarray, cell detection algorithms were optimized individually for each tumor microarray core to ensure accuracy. While many cores ultimately shared the same algorithm parameters, each core was initially assessed independently to determine the most appropriate settings using real-time tuning. The final algorithm parameters for each core were then applied accordingly. Two layers of quality control were performed: 1) comparing positive cell signals to positive object data results on at least six different cores, and 2) comparing cell counts to slide images reviewed by a pathologist. Summary data were exported to CSV files, and cell counts were merged with corresponding patient metadata using STATA 16. GraphPad PRISM 9 and R version 4.2.1 (R Foundation for Statistical Computing) were used to generate plots and compare patient treatment groups.

TCR sequencing analysis

Single-cell gene expression libraries were processed using Cell Ranger (10x Genomics) to generate filtered feature-barcode matrices. V(D)J TCR libraries were processed using Cell Ranger vdj to assemble TCR contigs and assign clonotypes. Downstream analyses were performed in R using Seurat for transcriptomic processing and scRepertoire for TCR integration. High-confidence productive TCR contigs from filtered_contig_annotations.csv were retained. Cells were assigned clonotypes based on identical CDR3 amino acid sequences of paired TRA/TRB. Cells with multiple productive TRB chains were flagged as potential doublets and excluded from clonotype-based analyses. Clonal expansion was quantified as the number of cells per clonotype, and repertoire diversity metrics (Shannon entropy, Simpson index) and gene usage frequencies were computed per sample and compared across conditions.

Lentiviral expression in primary CD4⁺ T cells

TCR sequences were synthesized in a PRRL vector modified with six start codon a promoter point mutations and cysteines were introduced to mediate TCR pairing (27) and sequence verified by Twist Bioscience. The constructs contained codon-optimized DNA fragments of TRBV-CDR3-TRBJ-TRBC followed by a P2A skip sequence and TRAV-CDR3-TRAJ-TRAC cloned into PRRL-SIN. Lentivirus was produced freshly for each experiment. For this, Lenti-X 293T cells (Takara #632180) were cultured according to the manufacturers recommendations were transiently transfected with the TCR vectors and psPAX2 (Addgene #12260) and pCMV-VSV-G (Addgene #8454) packaging plasmids. Notably, Lenti-X 293T cells were split fewer than ten times but not authenticated or screened for mycoplasma before use. Two days later, lentiviral supernatant was harvested from Lenti-X cultures, filtered using 0.45 μm polyethersulfone syringe filters (MilliporeSigma #SLHPR33RS), and added to stimulated CD4⁺ T cells in a 48-well tissue culture plate.

For transduction of T cells, CD4⁺ T cells from control PBMC (Stemcell #70025, Bloodworks Northwest custom order) were thawed and stimulated with anti-CD3/anti-CD28 dynabeads (ThermoFisher #11132D) at a 3:1 bead:cell ratio in stimulation media (RPMI supplemented with 10% human serum (BloodWorks Northwest lot #W141622870003), 50 mM beta-mercaptoethanol (ThermoFisher #21985023), 100 U/mL penicillin + 100 μg/mL streptomycin, 4 mM L-glutamine, 50 U/mL IL-2 (ThermoFisher #200–02), and 5 ng/mL IL-7 (ThermoFisher #200–07)) for 2 days. On day 2, dynabeads were removed and cells were nucleofected using a Lonza 4D nucleofector in 100 mL of buffer P3 using program EH-115. TCR was knocked out using CRISPR targeting the first exon of the TRBC and TRAC alone, as previously described (27). For this, 40 mmol/L of the gRNA AGAGTCTCTCAGCTGGTACA (IDT, custom order) was mixed with an equal volume of 24 mmol/L Cas9 (IDT #1081060) and 1/20th volume of 400 mmol/L Cas9 electroporation enhancer (IDT #1075916) and incubated at room temperature for 15 minutes prior to nucleofection. Cells were allowed to rest for 4 hours in media prior to lentiviral transduction. Polybrene (MilliporeSigma #TR-1003) was added to a final concentration of 4.4 mg/mL, and cells were centrifuged at 800 × g and 32°C for 90 minutes. Sixteen hours later, viral supernatant was replaced with fresh stimulation media. Half of the media was removed and replaced with fresh stimulation media every 48 to 72 hours. Transduction efficiency into TCR knockout T cells was determined one week after nucleofection by measuring CD3 expression using anti-human CD3 APC (BioLegend #317317) in each transduced sample compared to TCR knockout control T cells.

CRISPR engineering of B cells

B cells were engineered to express T-Ag–specific antibodies, as described previously (28). For this assay, engineering medium comprised IMDM (ThermoFisher #31980030) supplemented with 10% heat-inactivated bovine serum, 100 U/mL penicillin + 100 μg/mL streptomycin, except in antibiotic free steps as noted. Control PBMCs (Stemcell #70025, Bloodworks Northwest custom order) from a donor expressing the DR4-DQ8 haplotype able to present T-Ag peptide to the F5 TCR were thawed and B cells isolated by negative selection using the Human B Cell Isolation Kit II (Miltenyi #130–091-151) according to the manufacturer’s recommendations. Isolated B cells were resuspended at 0.5–1.0×10⁶ cells/mL in stimulation media, which consisted of medium described above supplemented with 100 ng/mL MEGACD40L (Enzo Life Sciences #ALX-522–110-C010), 50 ng/mL recombinant IL-2 (BioLegend #589104), 50 ng/mL IL-10 (Shenandoah Biotech #100–83), 10 ng/mL IL-15 (Shenandoah Biotech #100–86), and 1 μg/mL CpG ODN 2006 (InvivoGen #tlrl-2006), and incubated. After 48 hours, cells were electroporated using the ThermoFisher Neon Transfection System. Cas9 (ThermoFisher #A36499 or IDT #1081060) and huIgH296 sgRNA GUCUCAGGAGCGGUGUCUGU (Synthego CRISPRevolution sgRNA EZ Kit, custom order) were precomplexed at a 1 to 2 molar ratio in Neon Buffer T for 20 minutes at room temperature. Cells were washed with 1xDPBS and resuspended to 2.5×10⁷ cells/mL in Neon Buffer T containing 12 μg of precomplexed gRNA/Cas9 per 10⁶ cells. The cell + huIgH₂₉₆ gRNA + Cas9 mixture was electroporated with one 20 millisecond pulse at 1750V and immediately plated into stimulation media as described above, without antibiotics. After 30 minutes, Adeno-associated virus (AAV) was added to a final concentration of up to 20% culture volume amounting to a MOI of 10⁵–10⁶ genome copies per cell and incubated for 2–4 hours. AAV6 or AAV-DJ were produced by VectorBuilder and contained an equimolar concentration of antibody constructs corresponding to 1G04, 2H04, and 1B09. Each construct contained DNA encoding for a version of antibody in which the full antibody light chain would be expressed physically linked to the heavy chain VDJ with a linker (29) containing Strep-tagII (30–32). Cells were next transferred to a larger culture dish to allow for further expansion for two days. For secondary expansion, every four days B cells were passaged onto irradiated (80 gy) NIH 3T3-CD40L feeder cells in engineering medium containing 5 μg/mL human recombinant insulin (MilliporeSigma #I2643), 50 μg/mL transferrin (MilliporeSigma #T8158), 50 ng/mL human IL-2 (BioLegend #589104), 20 ng/mL human IL-21 (Shenandoah Biotech #100–90), and 10 ng/mL human IL-15 (Shenandoah Biotech #100–86). Antibody expression was confirmed based on binding to strep-tactin PE (IBA Lifesciences #6–5000-001) and T-Ag APC tetramer by flow cytometry.

Stimulation of T-Ag–specific CD4⁺ T cells by B cells

Engineered or control B cells from the same donor as T cells were isolated from thawed PBMC by positive selection using human CD19 microbeads (Miltenyi #130–050-301) according to the manufacturer’s instructions. Isolated B cells were cultured with irradiated (5,000Gy) NIH 3T3 cells expressing human CD40L (3T3-CD40L) in engineering medium supplemented with 200 U/ml human IL-4 (ThermoFisher #200–04). 3T3-CD40L cells were received from Brian Till (FHCC) in ~2019 and split fewer than ten times but not authenticated or screened for mycoplasma before use. B cells were harvested and restimulated with 3T3-CD40L every 7 days and supplemented with fresh medium and cytokines every 3 days. For tumor lysate preparation, WaGa MCC tumor cells were counted and resuspended at 5×10⁶/mL in serum free RPMI media. The cell suspension was frozen/thawed 5 times alternating between three minutes submerged in liquid nitrogen and three minutes in a 37 °C water bath. The solution was then centrifuged for 5 minutes at 1,000 × g to remove debris and remaining cells. Supernatant was then transferred to a fresh tube to be used as tumor lysate. 25 μL lysate was added to co-cultures for a total volume of 200 μL per well. 10⁵ B cells were mixed with 50,000 transduced T cells in the presence or absence of 0.1 or 5 mg/mL of T-Ag peptide pools (Supplementary Table S5), 0.1–10 mg/mL of purified T-Ag, or crude lysate derived from WaGa MCC cells for 24 hours prior to detection of IFNγ in culture supernatant using the using an ELISA kit (ThermoFisher #88–7316-88). Alternatively, anti-human OX40 APC (BioLegend #350008), anti-human CD69 PE-CF594 (ThermoFisher #562617), anti-human CD3 FITC (ThermoFisher #561806), and anti-human CD4 BUV805 (ThermoFisher #612888) was used to measure expression of OX40 and CD69 using flow cytometry.

Quantification and statistical analysis

The statistical tests applied were two-sided unless specified otherwise. T tests were used to compare differences between two groups unless otherwise noted. When comparing more than two groups, the nonparametric Kruskal–Wallis test was used. Fisher’s exact test was used to evaluate differences between two categorical variables. The significance levels for Kaplan–Meier analyses were calculated using a two-sided log rank test. Grubb’s test was used for determination of outliers. GraphPad Prism 9 and R version 4.2+ (R Foundation for Statistical Computing) were used to generate plots and compare patient treatment groups.

Data availability

The expression data in this study are publicly available in Gene Expression Omnibus (GEO) at GSE301498. Any requests for the raw data will be reviewed by the corresponding authors to ensure patient confidentiality is maintained. If possible, the data will be shared under a material transfer agreement. This paper does not report original code, however, code utilized can be found in https://github.com/JJTaylorLab/scRNAseq-human-MCC-Bcell.

Results

Validation of T-Ag protein tetramers to detect T-Ag–specific B cells.

To identify T-Ag–specific B cells using flow cytometry, fluorescent T-Ag tetramers were used in combination with control tetramers (24,25) to exclude B cells specific for streptavidin, the fluorochrome, and the GST purification tag added to T-Ag. We used the common domain shared by the small and large T-Ag isoforms since this domain is rarely mutated from patient-to-patient and most T-Ag–specific serum antibodies appear to target this region (12,33). In contrast, the large T-Ag varies from patient-to-patient as a protein ranging from 228–778 amino acids depending upon the location of truncation (33–36). Therefore, use of a tetramer containing only the “common” T-Ag domain (37,38) allowed direct comparison of B cells binding the same antigen across cohorts. As demonstrated previously (24,25), rare tetramer-binding B cells were more easily assessed when enriched using anti-fluorochrome microbeads before analysis. To validate the T-Ag tetramer, we immunized mice with T-Ag in CFA and found a robust expansion of T-Ag tetramer–binding live B cells compared to control animals injected with CFA alone (Figure 1a, b). B220^LOWIg⁺⁺ antibody-secreting cells and GL7⁺CD38^–B220⁺Ig⁺ germinal center B cells were also present within this expanded population in mice injected with T-Ag in CFA but absent in control animals (Figure 1c).

Figure 1. Detection of T-Ag–specific B cells in patient blood using T-Ag tetramers.

A. Schematic representation of two experiments containing two mice per group used to validate identification of T-Ag–specific mouse B cells using T-Ag tetramers. B. Representative flow cytometry of live CD19⁺ B220⁺ CD3^– Gr-1^– F4/80^– B cells binding T-Ag PE tetramers but not GST APC control tetramers, PE594 control tetramers or APC755 control tetramers in samples from mice seven days after subcutaneous injection of 5 μg of T-Ag in CFA in the base of the tail with compared to control mice injected with CFA alone. Each sample was enriched with anti-PE and anti-APC microbeads prior to analysis and both the PE/APC-enriched and -depleted fractions are displayed. C. Representative flow cytometry of T antigen tetramer-binding B220^LOW Ig⁺⁺ antibody-secreting cells and GL7⁺ CD38^– B220⁺ Ig⁺ germinal center B cells from mice seven days after subcutaneous injection of 5 μg of T-Ag in CFA in the base of the tail compared to uninjected control mice. D. Schematic representation of an experiment used to validate identification of T-Ag-specific human B cells using T-Ag tetramers. E. Representative flow cytometry of live CD19⁺ CD3^– CD14^– CD16^– B cells binding T-Ag PE tetramers but not GST PE650 control tetramers in blood samples from MCC patients and controls with and without enrichment with anti-PE microbeads prior to analysis. The displayed plots contain pooled cells from three individuals with or without MCC from the experiment in which cells were sorted for paired heavy and light chain sequencing from single cells. F, G. Antibodies from ten T-Ag tetramer-binding human B cells were cloned and assessed for binding to T-Ag using Bio-Layer Interferometry (BLI). Five tested antibodies derived from patient samples and five from controls. Area under the curve (AUC) of the BLI shift for each antibody. Representative of two similar experiments.

We next assessed T-Ag tetramer–binding CD19⁺ B cells from the blood of patients with progressive T-Ag–driven MCC and control individuals with no history of MCC (Figure 1d, e, Supplementary Table S6). Single cells were sorted into individual wells of a 96-well plate and paired heavy and light chain sequences were determined using RT-PCR (Supplementary Table S2). From these sequences, five antibodies were produced from control samples and five from MCC samples. In total, 8/10 antibodies were confirmed to bind T-Ag using BLI, indicating a high specificity of the tetramer-based approach (Figure 1f, g).

Frequency of antigen-experienced B cells in the blood of female patients near the time of diagnosis associates with MCC progression.

After confirming that tetramer-binding B cells were T-Ag-specific, we analyzed blood cells from a prospective cohort of 47 T-Ag–driven MCC patients collected within 67 days of diagnosis and shortly before or after treatment initiation (Supplementary Table S7). This cohort included 23 patients that experienced MCC progression within three years following our analysis, and 24 stage- and age-matched patients with non-progressive disease (Supplementary Figure S2a). Detailed patient histories were collected to account for confounding variables and a previously published clinical recurrence risk calculator (4) was used to match the two patient cohorts (Supplementary Figure S2b). In agreement with published data from larger cohorts (12,13), the level of T-Ag-specific antibodies in blood prior to treatment was not associated with MCC progression after treatment (Supplementary Figure S2c). Also in agreement with published data (12,13), the level of T-Ag–specific antibodies in blood diminished in most patients that did not experience MCC progression, whereas patients experiencing progressive MCC exhibited sustained or rising titers (Supplementary Figure S2d). Combined, the alignment of blood antibody data with previous work indicates the suitability of this prospective MCC patient cohort for analysis of B cells.

T-Ag–specific B cells represented ~0.09% of B cells in blood from MCC patients, which was increased ~10-fold compared to control samples from individuals with no history of MCC (Supplementary Figure S2e). While the frequency of T-Ag–specific B cells ranged from 0.04–0.16% from patient-to-patient, there was no association with the frequency of T-Ag–specific B cells and MCC progression (Supplementary Figure S2e). There was also no association between the frequency of T-Ag–specific B cells with the total level of T-Ag–specific antibodies in the blood (Supplementary Figure S3a).

We next considered whether different subtypes of B cells associated with disease outcome and stratified the cohorts by sex because phenotypes are affected by estrogen levels (39). We first assessed total CD27⁺⁺CD20^– antibody-secreting cells, which varied from 0.1–19% of B cells in blood and frequencies did not associate with outcome (Supplementary Figures S2f, S3b). T-Ag–specific antibody-secreting cells in blood were usually below the limit of detection, and frequencies did not associate with outcome (Supplementary Figure S3c). In contrast, IgM^–IgD^– isotype-switched cells ranged from 3–54% of B cells and higher frequencies associated with MCC progression in female patients, but not in male patients (Supplementary Fig S2f, g). These associations were maintained across several isotype-switched cell subtypes including CD27⁺ memory B cells and CD11c⁺ memory B cells (Supplementary Figure S2g). The % of B cells that were isotype-switched and expressed CD71, which indicates recent activation (21), was increased in both male and female MCC patients but did not significantly associate with progression (Supplementary Figure S2g). We further observed that the frequency of T-Ag–specific B cells within each of these subtypes was low and did not associate with MCC outcome, even when stratified by sex (Supplementary Figure S2h, S3c). Together, our results demonstrate that while the level of T-Ag–specific antibodies or T-Ag–specific B cells did not associate with MCC outcome, higher frequencies of total circulating isotype-switched B cells prior to treatment associated with worse outcomes in female patients following MCC treatment.

T-Ag–specific B cells are present in MCC tumors.

The association between isotype-switched B cells in the blood with MCC progression for some patients led us to consider whether stronger associations would be identified studying B cells from tumor samples. For this we assessed a cohort of 19 patients with T-Ag–driven MCC from which single-cell tumor digests were cryopreserved (Figure 2a). Detailed patient histories were accounted for to identify adverse clinical risk factors at the time of tumor removal that may confound survival analyses (Supplementary Figure S4, Supplementary Table S8). Sixteen of the 19 tumor samples came from male patients, limiting stratification by sex. Using known recurrence risk factors (40), eleven patients from this cohort were considered to have low or medium risk for MCC progression, whereas eight patients were considered high risk. While this risk calculator has been able to predict risk in larger cohorts, risk assignments for this cohort did not associate with outcome (Figure 2b). T-Ag–specific antibodies at time of surgery also did not predict outcomes and were sustained or increased in patients with progressive MCC (Figure 2c, d). Together, these results highlight the need for additional biomarkers for more accurate MCC outcome prediction.

Figure 2. Identification and analysis of T-Ag–specific B cells from T-Ag–driven MCC patient tumor samples.

A. Schematic representation of single-cell RNAseq experiments assessing B cells, T cells and MCC cells from skin tumor and LN tumor samples. B. Kaplan-Meier plots displaying percent progression-free survival for patients during the four-year monitoring period following tumor removal surgery stratified based on clinical risk scoring of data available at the time of surgery. C. Level of T-Ag–specific antibodies in the blood of MCC tumor cohort patients near the time of diagnosis stratified based upon whether patients experienced MCC progression (n=8) or not (n=11) during the four-year monitoring period. D. Longitudinal analysis of T-Ag–specific antibodies in the blood of MCC tumor cohort patients. Two patients from the non-progressive group for which antibody levels were not available for at least a year after diagnosis were excluded. E. Representative flow cytometry of CD3⁺ T cells, CD56⁺ MCC cells, and CD19⁺ B cells that bound pooled T-Ag and RSV F APC and PE-oligo tetramers but not APC755 and PE594-oligo control tetramers and B cells of unknown specificity that did not bind any tetramers. F. Pooled data from seven experiments showing the % of CD19⁺ B cells, % of CD3⁺ T cells, and % of CD56⁺ MCC cells in skin tumor (n=5), and LN tumor (n=14) samples from MCC patients determined by flow cytometry. G. UMAP clustering of CITEseq data of cells pooled from all skin and LN tumor samples after alignment and normalization with highlighted clusters of T cells (yellow), B cells (blue), MCC cells (pink), and contaminating cells (red, green). H. Computational strategy to identify T-Ag tetramer-binding B cells. I, J. Antibodies from thirteen T-Ag tetramer-binding B cells and ten from control B cells were cloned and assessed for binding to T-Ag using BLI. Representative of two similar experiments. K. Pooled data from seven experiments displaying the % of CD19⁺ B cells specific for T-Ag in MCC skin and LN tumor samples compared to the frequency found in blood. The bars in C, F, and K represent the mean, and p values calculated using a non-parametric Mann-Whitney test for C and F, and a non-parametric Dunn’s multiple comparison test for K are displayed when p<0.05.

To study T-Ag–specific B cells in MCC tumors, we performed paired Cellular Indexing of Transcriptomics and Epitopes by Sequencing (CITEseq) (Supplementary Tables S1, S3) and antibody IgH + IgL sequencing on five MCC skin tumor samples and fourteen MCC lymph node (LN) tumor samples. The presence of MCC in the LN samples was first detected during biopsy and later confirmed during single-cell sorting based on the increased presence of CD56⁺ cells (Figure 2e, f). Of note, while CD56 is typically used as a marker of NK cells and some subsets of T cells, NK cells are rarely found in MCC tumors making it a reliable marker for the identification of MCC when paired with CD3 (9). From these tumor samples we FACS-purified CD3⁺ T cells, CD56⁺ MCC cells, and CD19⁺ B cells that bound T-Ag-tetramers or RSV F antigen-tetramers but not control tetramers, and B cells of unknown specificity that did not bind tetramers (Figure 2e). Frequencies of CD19⁺ B cells were lower in MCC skin tumor samples when compared to blood or LN tumor samples (Figure 2f). In contrast, the frequencies of CD3⁺ T cells and CD56⁺ MCC cells were similar across both tumor sample types (Figure 2f). While the frequencies of B cells, T cells, and MCC cells varied greatly from sample-to-sample, associations with MCC progression were not detected (Supplementary Figure S5a).

After standard processing and quality control metrics, we used UMAP (41) of single-cell transcriptomic data from 1,015 immune-associated genes (Supplementary Table S3), which identified clusters of MCC cells, T cells, and B cells (Figures 2g, Supplementary Figures S5b, S5c). Additional cell clusters of myeloid, dendritic, and endothelial cells were identified due to contamination during cell sorting but accounted for less than 1% of the total cells sequenced (Supplementary Figures S5b, S5c). Within the population of B cells, we identified cells that bound T-Ag tetramer-oligo but not RSV F or GST control tetramers (Figure 2h). We validated the specificity of our approach by confirming T-Ag binding of 87% (12/14) of antibodies derived from T-Ag tetramer-oligo⁺ B cells by BLI, whereas none of the 10 antibodies cloned from tetramer^– B cells bound T-Ag (Figure 2i, j, Supplementary Table S4). Together, we analyzed 941 T-Ag–specific B cells, 36,641 B cells of unknown specificity, 58,708 CD3⁺ T cells, and 33,573 CD56⁺ MCC cells from MCC tumor samples. Within the cohort, 0.01–4% of B cells were T-Ag–specific in the 16 of 19 samples where these cells were detected (Figure 2k). Despite this high variability, no association between the frequency of T-Ag–specific CD19⁺ B cells and disease progression was detected (Supplementary Figure S5d).

T-Ag–specific antibody-secreting cells in MCC tumor samples are associated with MCC control.

To assess subtypes of B cells in patient tumor samples, B cells were subclustered and populations corresponding to antibody-secreting cells, germinal center B cells, memory B cells, and naïve B cells were identified (Figure 3a, b). B cells that did not cluster with these canonical subsets were also present and did not appear to be activated B cells since CD71 expression was not detected (Figure 3a, b). All analyzed subtypes were observed in T-Ag–specific B cells and the much more numerous B cells of unknown specificity, with highly varied frequency distributions across the cohort (Figure 3c). We determined that two LN tumor samples concurrently contained malignant B cells (Figure 3a, b), Mantle Cell Lymphoma (MCL) or Chronic Lymphocytic Leukemia (CLL). For these, MCL and CLL cells were excluded to focus only on non-malignant B cells for subtype analyses.

Figure 3. Phenotypes of T-Ag–specific B cells in MCC tumors.

A. UMAP clustering of single cell RNA expression by B cells from pooled skin (n=5) and LN (n=14) tumor samples with clusters corresponding to different B cell subtypes highlighted with different colors. B. Violin plots showing expression of select genes used to identify B cell subtypes. C. Stacked bar graphs displaying the average % of T-Ag–specific and B cells of unknown specificity of each subtype found in all samples, or LN or skin tumor samples. D. Stacked bar graphs displaying the % of T-Ag–specific and B cells of unknown specificity of each subtype found in individual patients. The numbers to the right of each stacked bar indicate the number of cells assessed in that stacked bar. *Indicates samples in which CLL or MCL cells were excluded from subset analysis. E. % of total (top) and T-Ag–specific (bottom) B cells that were memory in skin and LN tumor samples. The bars indicate median for each tissue type. F. % of T-Ag-specific B cells that were memory versus % of total B cells that were memory in each skin (top) and lymph node (bottom) tumor samples. Correlation p value was determined using nonparametric Spearman’s test. G. Kaplan-Meier plots displaying % progression-free survival for patients during the monitoring period following analysis divided into groups with a % of total (top) and T-Ag–specific (bottom) B cells that were memory above (blue) or at/below (red) the median. The p value nearing significance generated using Mantel-Cox Log-rank test.

Since increased frequencies of isotype-switched B cells in the blood associated with MCC progression for some patients, we next assessed memory B cells in tumor samples as a population able to exit the tissue and enter the blood. Memory B cells were detected in every tumor sample and ranged from 6–82% of B cells (Figure 3d). T-Ag–specific memory B cells exhibited similar variability but were not detected in three LN tumor samples and one skin tumor sample (Figure 3d). In LN tumor samples, the frequency of T-Ag–specific memory B cells correlated with the total frequency of memory B cells (Figure 3e). Higher frequency of T-Ag–specific or total memory B cells in tumor samples did not associate with MCC progression (Figure 3f). Since MCC skin lesion samples were limited, these analyses were conducted with LN tumor samples alone or combined with skin tumor samples (Figure 3f). Moreover, lower frequency of T-Ag–specific memory B cells in LN tumor samples trended towards an association with MCC progression but did not achieve statistical significance (Figure 3g).

Antibody-secreting cells also exhibited high sample-to-sample variability, accounting for 10–87% and 1–85% of CD19⁺ cells in skin and LN tumor samples, respectively (Figure 4a). Variability was similar within the T-Ag–specific B cells, except that T-Ag–specific antibody-secreting cells were not detected in one skin tumor sample and half of the fourteen LN tumor samples (Figure 4a). Unlike memory B cells (Figure 3e), higher frequency of T-Ag–specific antibody-secreting cells did not correlate with frequencies of total antibody-secreting cells in MCC tumor samples (Figure 4b). Similarly, the frequency of T-Ag–specific antibody-secreting cells in MCC tumor samples did not correlate with levels of T-Ag–specific antibodies in the blood at the time of tumor removal (Figure 4c). The lack of correlation between levels of antibody in blood and frequency of detected antibody-secreting cells did not appear to be the result of IgG being detected in blood and antibody-secreting cells of all isotypes being interrogated since nearly all antibody-secreting cells detected were IgG⁺ (Figure 4d, Supplementary Figure S6a). Combined, these data suggest that the antibody-secreting cell response detected in tumor samples is distinct from what can be inferred from analysis of serum antibodies in the blood.

Figure 4. Lower frequency of T-Ag–specific antibody-secreting cells in MCC tumors associates with progressive disease.

A. Data from single-cell RNAseq experiments showing the % of total (top) and T-Ag–specific (bottom) antibody-secreting cells in skin and LN tumor samples. Bars indicate medians. B. % of T-Ag–specific cells that were antibody-secreting versus the % of all B cells that were antibody-secreting in skin (top) and LN (bottom) tumor samples. C. Level of T-Ag–specific antibodies in the blood of MCC patients near the time of surgery versus the % of T-Ag-specific or total B cells that were antibody-secreting in skin tumor (top) and LN (bottom) tumor samples. D. % of B cells of total (top) or T-Ag–specific B cells of the listed B cell subtype that expressed IGHM/IGHD or class-switched to IGHA or an IGHG. E. Kaplan-Meier plots displaying % progression-free survival for patients during the monitoring period divided into groups with a % of total (top) and T-Ag–specific (bottom) B cells that were antibody-secreting above (blue) or at/below (red) the median. Significant p values (p<0.05) generated using Mantel-Cox Log-rank test are displayed. F. Select multiplex immunohistochemistry (mIHC) images of skin and LN tumor samples determined to contain high or low frequencies of CD19⁺ CD138⁺ antibody-secreting cells. G. % of CD19⁺ cells that were CD138⁺ antibody-secreting cells within the CD56⁺ tumor tissue or adjacent non-tumor stroma detected using mIHC of LN (n=23) and skin tumor samples (n=63). Bars indicate medians. H. Kaplan-Meier plots displaying % progression-free survival for patients during a three-year monitoring period following analysis stratified based on % of B cells that were antibody-secreting above (blue) or at/below (red) the median within the tumor tissue (top) and adjacent non-tumor stroma (bottom) determined by mIHC.

We next considered whether antibody-secreting cell frequencies in tumor samples was associated with MCC outcome after treatment. Since MCC skin lesion samples were limited, we conducted separate analyses either with LN tumor samples alone or combined with skin tumor samples. We found the presence of detectable T-Ag–specific antibody-secreting cells in LN tumor samples was associated with extended progression-free survival following tumor removal (Figure 4e). Similar results were found when skin tumor samples were included based upon whether the frequency of T-Ag–specific antibody-secreting cells was above or below the median found in skin tumors (Figure 4e). In contrast, a higher frequency of total antibody-secreting cells in skin or LN tumors was not associated with MCC outcome (Figure 4e). To further probe relationships between antibody-secreting cells and MCC progression, we used multiplex immunohistochemistry to assess an independent cohort of 23 MCC LN tumor and 63 MCC skin tumor sections (Figure 4f, Supplementary Figures S6b, S7, and S8, Supplementary Table S9). To differentiate between tumor mass and adjacent non-tumor tissue stroma, a trained pathologist blindly annotated tumor sections prior to cell segmentation and phenotyping. As observed by CITEseq, the frequency of CD138⁺ antibody-secreting cells among CD19⁺ cells in MCC tumors varied widely but did not differ between regions of tumor mass and adjacent non-tumor tissue stroma (Figure 4f, g). Similarly, the frequency of antibody-secreting cells out of total B cells in tumor mass or adjacent non-tumor tissue stroma did not associate with MCC outcomes for skin or LN tumor samples (Figure 4h). Together, our results indicate that while a high frequency of antibody-secreting cells in MCC tumor samples is not associated with MCC outcome, the absence of detectable T-Ag–specific cells within this population is associated with MCC progression.

Germinal center response in MCC tumor samples associates with MCC control.

The presence of tertiary lymphoid structures able to support germinal centers in tumors is associated with better outcomes for many solid cancers (17–19). We identified T-Ag–specific germinal center B cells in eight of the fourteen LN tumor samples but failed to detect any in skin tumor samples (Figure 5a). Few germinal center B cells were detected in skin tumor samples overall (Figure 5a), in agreement with previous work demonstrating limited development of organized tertiary lymphoid structures in MCC skin tumor samples (42–44). Among the LN tumor samples, higher frequencies of T-Ag–specific germinal center B cells trended with higher frequencies of total germinal center B cells (Figure 5b), even though T-Ag–specific cells made up less than 1% of germinal center B cells when detected (Figure 5c). The presence of detected T-Ag–specific germinal center B cells in LN tumor samples was predictive of prolonged progression-free survival, whereas absence of these cells was associated with rapid disease progression (Figure 5d). Higher frequencies of total germinal center B cells in LN tumor samples also was associated with progression-free survival (Figure 5d).

Figure 5: Lower frequency of germinal center B cells and T_FH cells in MCC tumor samples associates with progressive disease.

A. Data from single-cell RNAseq experiments showing the % of total (top) and T-Ag–specific (bottom) B cells that were germinal center phenotype in skin and LN tumor samples. Bars indicate medians. B. % of T-Ag–specific B cells that were germinal center phenotype versus the % of all B cells that were germinal center phenotype in LN tumor samples. Correlation p value determined using nonparametric Spearman’s test. C. % of T-Ag–specific cells within the total population of germinal center B cells in LN tumor samples. Bars indicate medians. D. Kaplan-Meier plots displaying % progression-free survival for patients during the four-year monitoring period following analysis divided into groups with a % of total B cells (left) that were germinal center phenotype above (blue) or at/below (red) the median and whether T-Ag–specific (right) germinal center phenotype cells were detected (violet) or below the limit of detection (orange). Displayed p values generated using Mantel-Cox Log-rank test. E. UMAP clustering of single-cell RNA expression by CD4⁺ T cells from pooled skin (n=5) and LN (n=14) tumor samples with clusters corresponding to subtypes highlighted with different colors. F. Violin plots showing the expression of select genes used to identify T_FH cells. G. % of CD4⁺ T cells that were in each cluster within skin and LN tumor samples. Bars indicate median and significant p value generated using a non-parametric Mann-Whitney test is displayed. H. Kaplan-Meier plots displaying % progression-free survival for patients during the four-year monitoring period following analysis divided into groups with a % of CD4⁺ T cells that were in each cluster above (blue) or at/below (red) the median. Significant p values generated using Mantel-Cox Log-rank test are displayed. I. % of total (left) and T-Ag–specific (right) B cells that were germinal center phenotype versus the % of CD4⁺ T cells that were T_FH in LN tumor samples. Correlation p values determined using nonparametric Spearman’s test.

Because robust germinal center responses are dependent on signals from follicular helper CD4⁺ T cells (T_FH), we assessed whether this population of T cells associated with MCC outcome. Subclustering of CD4⁺ T cells revealed a population of cells with a T_FH signature (27) that included increased expression of BCL6, PDCD1, CXCL13, and TCF7 transcripts (Figure 5e, f, Supplementary Figure S9a). CD4⁺ T cells in this cluster were detected in both skin and LN tumor samples but were more abundant in LN (Figure 5g, Supplementary Figure S9b, Supplementary Table S10). Like the germinal center B cells, frequencies of T_FH cells above the median in LN tumor samples were also predictive of progression-free survival, whereas patients whose tumors had frequencies of T_FH cells at or below the median were more likely to have rapid MCC progression (Figure 5h). We also observed that tumor samples with more T_FH cells tended to have higher frequencies of T-Ag–specific and total germinal center B cells (Figure 5i). In contrast, none of the five other clusters of CD4⁺ T cells associated with disease outcome (Figure 5g, h). Likewise, the frequencies of the five non-T_FH clusters did not correlate with the frequency of germinal center B cells (Supplementary Figure S9c).

T-Ag–specific B cells potently activate T-Ag–specific CD4⁺ T cells.

To confirm the presence of T-Ag–specific T_FH cells in MCC tumors, we examined paired TCRα/β CDR3 sequences from CD4⁺ T cells from three patient tumors. This analysis revealed that 31% (408/1,315) of these cells shared identical paired TCRα/β CDR3 sequences with at least one other cell, indicating clonal expansion (Figure 6a, Supplementary Table S11). TCRs from 69 clonal families containing CXCL13⁺ members were expressed in primary CD4⁺ T cells derived from healthy donors using a previously described lentiviral transduction approach (27) (Figure 6b). Transduced T cells were next assessed for T-Ag specificity by incubation with autologous B cells in the presence and absence of T-Ag peptides. This assay revealed that 23% (16/69) of tested clonal TCRs secreted IFNγ in response to pools of T-Ag peptides (Figure 6c, Supplementary Figure S10, Supplementary Table S11), confirming specificity for T-Ag.

Figure 6. T-Ag–specific engineered B cells potently stimulate T-Ag–specific engineered CD4⁺ T cells.

A. TCRs expressed by all CD4⁺ T cells (left) and identical TCRs expressed by multiple cells (right) from three LN and skin tumor samples linked to cell subtype. B. Representative flow cytometry of the expression of lentiviral-expressed T-Ag–specific TCR in CD4⁺ from blood detected as the expression of CD3 on the cell surface in cells lacking endogenous TCR due to CRISPR knockout using gRNA targeting TCRα and TCRβ. C. Eighteen clonal T_FH TCRs from MCC tumor samples were expressed via lentivirus in CD4⁺ cells from blood and incubated with sample-matched B cells in the absence or presence of 0.1 or 5 μg/mL of T-Ag peptide pools for three days prior to assessment of supernatant for secreted IFNγ. Representative of two similar experiments. D. Representative flow cytometry of Strep-Tactin and T-Ag binding to human B cells engineered to express a T-Ag–specific antibodies using CRISPR/Cas9. Antibody constructs encoding 1G04, 2H04, and 1B09 from containing the full antibody light chains physically linked to the heavy chain VDJs with a linker (29) containing Strep-tagII (30–32) were pooled prior to engineering. E. Representative flow cytometry and (F) quantitation of CD69 and OX40 expression by lentiviral-transduced T-Ag–specific F5 TCR CD4⁺ T cells expressing following 72-hour co-culture with CRISPR-engineered T-Ag–specific oligoclonal B cells or control B cells in the presence or absence of various concentrations of T-Ag. G. Quantitation of CD69 and OX40 expression by lentiviral-transduced T-Ag–specific F5 TCR CD4⁺ T cells expressing the were co-cultured for 24 hours with CRISPR-engineered T-Ag–specific oligoclonal B cells or control B cells in the presence or absence of MCC cell line lysate. Bars represent the mean. E-G representative of two similar experiments. ****p<0.0001 ****p<0.001 determined by Tukey’s multiple comparison test in F and G.

We next considered whether T-Ag–specific B cells could more potently stimulate T-Ag–specific CD4⁺ T cells compared to B cells of unknown specificity. For this, we cultured CD4⁺ T cells transduced with the T-Ag–specific F5 TCR, with CRISPR-engineered T-Ag–specific B cells generated using a previously described approach (28). For these assays we used a cocktail of constructs encoding the three T-Ag–specific antibodies 1G04, 2H04, and 1B09 (Figure 1F) to simulate an oligoclonal population of B cells. Three DNA constructs encoding the full 1G04, 2H04, and 1B09 antibody light chains physically linked to the heavy chain VDJs (29) were produced and pooled prior to engineering and antibody expression was validated based upon binding T-Ag tetramer and Strep-Tactin (Figure 6d), which binds the Strep-tagII (30–32) included in the linker. These assays revealed that a 100-fold higher concentration of purified T-Ag was needed for control B cells to induce similar expression of CD69 and OX40 on F5⁺ CD4⁺ T cells as T-Ag–specific engineered B cells (Figure 6e, f). T-Ag–specific engineered B cells could also induce CD69 and OX40 expression by T-Ag–specific F5 TCR CD4⁺ T cells in response to MCC cell lysate from a tumor cell line, which was not found with control B cells (Figure 6g). Combined, these data indicate that T-Ag-specific B cells can potently activate T-Ag–specific CD4⁺ T cells.

Antibody-secreting cells and memory B cells in tumor samples are largely germinal center–derived.

Our data demonstrated that the T-Ag–specific germinal center B cells and antibody-secreting cells in MCC tumors were associated with better progression-free survival (Figure 4e, 5d). Since antibody-secreting cell differentiation can occur through germinal center–dependent and –independent pathways, we analyzed paired antibody heavy and light chain sequences and clonal dynamics of B cells in MCC tumor samples for evidence of somatic hypermutation in a germinal center.

To assess somatic hypermutation in B cells from MCC tumors, we determined the mutation frequency in IgH sequences with respect to germline encoded sequences. Most T-Ag–specific antibody-secreting cells and germinal center B cells had IgH mutations at similar levels to B cells of unknown specificity with these phenotypes (Figure 7a). Somatic hypermutation was also present in T-Ag–specific memory B cells found in tumor samples, even in samples where germinal center B cells were not detected (Figure 7a). Similarly, IgH mutation frequencies were similar in B cells from skin and LN tumor samples, independently of T-Ag specificity (Figure 7b). Moreover, the frequency of somatic hypermutation was not associated with MCC progression (Figure 7c).

Figure 7: Analysis of somatic hypermutation and clonality indicates many antibody-secreting cells and memory B cells in tumors are germinal center derived.

A. Average % of mutated IgH nucleotides per patient for cells within the listed subpopulations of B cells of unknown specificities (top) and T-Ag–specific B cells (bottom). Height of bars represents median. B. Average % of mutated IgH nucleotides for B cells of unknown specificities (top) and T-Ag–specific B cells (bottom) found in skin and LN tumor samples. C. Kaplan-Meier plots displaying % progression-free survival for patients during the monitoring period following analysis divided into groups with a % mutation rate above (blue) or at/below (red) the median. D. Assessment of clonality within the B cells found in skin and LN tumor samples using Simpson Diversity Index after outlier removal (red data points on top figure) using Grubb’s test. E. Stacked clonal lineage alluvial plot with lines linking individual cells within clonal families (left) with the B cell subtype (right) of that cell within B cells of unknown specificities. Each clonal family is displayed as an individual bar stacked in alternating black and white sized corresponding to the number of cells detected in that clonal family. F. Stacked clonal lineage plots (top) and heatmaps (bottom) of all T-Ag–specific clonal families from LN and skin tumor samples, and plots divided into individual patients. The columns in the heat maps correspond to individual clonal families. G. Summary of findings from analyses of B cells and T cells with phenotypes with significant associations with MCC control highlighted in blue. H. Kaplan-Meier plots displaying % progression-free survival for patients during the monitoring period following analysis divided into a group with any of the B or T cell characteristics associated with MCC control from G (blue), versus the group that did not have any of those characteristics. The bars in B, and D represent the median. The p values calculated using a non-parametric Mann-Whitney test for C and D, a non-parametric Dunn’s multiple comparison test for A and Mantel-Cox Log-rank test were displayed when p<0.05.

While somatic hypermutation strongly suggested that T-Ag–specific antibody-secreting cells and memory B cells were derived from germinal centers, we sought to confirm this by analysis of clonal families containing germinal center and non-germinal center members. Analysis of paired IgH + IgL sequences revealed that B cells from skin tumor samples were more clonally related than B cells from LN tumor samples (Figure 7d). These slight differences were not driven by outliers in the data set since significance was maintained when two outliers were excluded (p=0.03). In total, we identified 4,119 expanded B cells of unknown specificity amongst 589 clonal families spread across 17 MCC patient tumor samples (Figure 7e). T-Ag–specific clonal families were only found in 7 of the 19 tumor samples and amounted to 150 expanded cells amongst 42 clonal families (Figure 7f, Supplementary Table S12). We observed that six of the seven samples with detectable clonal families were from patients who did not progress, with the smallest clonal family being derived from the LN tumor sample from patient Z1513, who experienced MCC progression (Figure 7f, Supplementary Figure S4).

The three LN tumor samples with the most abundant T-Ag–specific clonal families contained mutated antibody-secreting cells and germinal center B cells within the same clonal family (Figure 7f). In fact, 10 of the 18 T-Ag–specific clonal families containing an antibody-secreting B cell also contained a germinal center B cell (Figure 7f). These results demonstrated that T-Ag–specific antibody-secreting cells and memory B cells can derive from germinal centers found in LN tumor samples.

As a final analysis, we compiled the analyses of B cells and T cells to determine if combinations of parameters better predicted disease outcome (Figure 7g). When assessed together, we found that tumor samples from all eight MCC patients in the cohort that that later experienced MCC progression contained low frequencies of T-Ag–specific germinal center B cells, low frequencies of T-Ag–specific antibody-secreting cells, undetectable T-Ag–specific clonal families, low frequencies of total germinal center B cells, and low frequencies of T_FH cells (Figure 7g, h). In contrast, all eleven MCC patients had higher frequencies of at least two of these five readouts (Figure 7g, h).

Discussion

Increasing evidence suggests an important role for B cells in immune control of tumors. Here we directly analyzed B cells specific for the T-Ag oncoprotein during MCC in search of biomarkers able to predict MCC progression. Similar to analyses of T-Ag–specific antibodies in the blood (12,13), analysis of T-Ag–specific B cells in the blood near the time of diagnosis did not yield associations with disease progression. However, increased frequencies of isotype-switched total B cells in the blood of female patients did associate with disease progression. It is intriguing to speculate that these links are related to the role estrogen plays in mediating increased levels of isotype-switched B cells in the blood (39).

We also found increased isotype-switched CD71⁺ B cells in the blood of both male and female MCC patients. This is not surprising given that CD71 is a marker of B cells recently activated by antigen in response to cancer and vaccination (21). It is unclear what antigen these activated CD71⁺ B cells are recognizing since the vast majority did not bind the portion of T-Ag used here. Some of these activated CD71⁺ B cells could be specific for regions of large T-Ag that are not conserved from patient-to-patient. Alternatively, these activated CD71⁺ B cells could be specific for self-antigens since self-antigen–specific antibodies are found in many cancers (45,46).

Findings from populations of B cells in the blood were intriguing but not robust enough to be clinically actionable biomarkers, prompting assessment of B cells from 19 MCC patient tumor samples, which revealed several predictive biomarkers of MCC progression. Due to the rarity of MCC and the greater prevalence of this disease in males, nearly all tumor samples available for single-cell suspension analysis over the last decade were from male patients, which limited our ability to stratify these tumor findings by sex. We found that ongoing germinal center responses and high frequencies of T-Ag–specific antibody-secreting cells in tumor samples predicted control of MCC after treatment with remarkable accuracy. Given the strong predictive ability of the phenotype of T-Ag–specific B cells in tumors, it was surprising that the level of T-Ag–specific antibodies in blood at the time of treatment was not predictive of outcome. We hypothesize that this difference is reflective of responses of T-Ag–specific B cells we are unable to measure, such as responses in the weeks prior to tumor surgical excision, or responses in anatomical locations we are unable to assess, such as sites of disease spread or unaffected draining lymph nodes. This hypothesis is supported by our data indicating the presence of somatic hypermutation in most T-Ag–specific B cells in tumors even when germinal center responses are not detected.

Studies in other cancers analyzing the presence of antibody-secreting cells in tumors have often reported positive, negative, and no associations with disease outcome (47). Our work studying the overall antibody-secreting cell responses using flow cytometry and immunofluorescence did not find an association with MCC progression. However, both techniques revealed high sample-to-sample variability in the frequency of antibody-secreting cells in MCC tumor samples. This high variability could be the result of technical and/or biological reasons. From a technical standpoint, the tumor sections represent only a small fraction of the full tumors, which could by chance happen to include regions that have a frequency of antibody-secreting cells that is not reflective of the entire tissue. Likewise, single-cell RNAseq analysis may also result in sample bias as while a larger fraction of the tissue is made into a single-cell suspension, only a portion of the tissue is taken for processing and cells in some regions may be more easily removed from tissues than others. Taken together, we think that these two approaches are complimentary, suggesting that much of the variability reflects true biological differences.

While the study of the total antibody-secreting cells in MCC tumor samples did not reveal associations with outcome, by studying T-Ag–specific B cells we found a strong association between low frequencies of T-Ag–specific antibody-secreting cells and MCC progression. These data raise additional questions about the specificities and differentiation pathways of the more numerous antibody-secreting cells that are not T-Ag specific, which we will explore in future work.

While we found several promising biomarkers of progressive MCC, more work is needed to understand the mechanism. Enhanced responses of T-Ag–specific germinal center B cell and antibody-secreting cells in MCC tumor samples may reflect of an overall more robust immune response resulting in more tumor killing. In this scenario, B cells are responding to the presence of more T-Ag released from killed MCC cells. However, if more cell-free T-Ag is present, the levels of T-Ag–specific antibodies in blood may also be expected to be increased, which was not what we observed. The association of T_FH cell frequencies with MCC control, but not other subsets of CD4⁺ T cells, also suggests a role for B cells. If increased responses of B cell are merely a reflection of an overall better tumor-killing immune response, a specific increase in T_FH cells would not be required as numerous populations of CD4⁺ T cells would be expected to be increased if merely a reflection of overall immune potency.

Although these findings identify phenotypes of T-Ag–specific B cells with strong predictive value in our cohort, we anticipate that the technical complexity of assessing antigen-specific B cells will limit use as a standard-of-care prognostic tool. Future work will evaluate the clinical use of the assessment of total germinal center B cells in LN tumor sections.

We hypothesize that T-Ag–specific B cells help enhance the killing of tumors by T-Ag–specific T cells. In support of this, we show that B cells engineered to be T-Ag–specific can potently activate CD4⁺ T cells expressing a T-Ag–specific TCR derived from a patient tumor sample. While it is unlikely that CD4⁺ T cells directly kill MCC due to the lack of MHC class II expression by MCC, CD4⁺ T cells can enhance the antitumor function CD8⁺ T cells (20,48,49). Likewise, data from a mouse model has suggested that of T_FH cells by B cells enhances IL-21 production, which in turn enhances tumor killing by CD8⁺ T cells (20). Newly generated mouse models of MCC (50) will likely help to determine the role of T-Ag–specific B cells in MCC tumor control. These findings would support the development of new MCC immunotherapies that incorporate stimulation or adoptive transfer of engineered T-Ag–specific B cells to potentiate the response of tumor-specific CD4⁺ and CD8⁺ T cells and improve durable clinical outcomes.

Synopsis.

The link between tumor-specific B cells and cancer outcomes remains poorly defined. The authors show that tumor-associated B cells specific for viral oncoproteins expressed in MCC predict disease control, establishing their potential as active participants in tumor immunity.

Abbreviations:

APC

allophycocyanin

APC755

APC-DyLight 755

ASC

antibody-secreting cell

AUC

area under the curve

BLI

Bio-Layer Interferometry

CFA

complete Freund’s adjuvant

CITEseq

Cellular Indexing of Transcriptomics and Epitopes by Sequencing

CLL

Chronic Lymphocytic Leukemia

GST

glutathione S-transferase

HPV

Human Papillomavirus

Ig

immunoglobulin

IgH

antibody heavy chain gene

IgL

antibody light chain gene

LN

lymph node

MCC

Merkel cell carcinoma

MCL

Mantle Cell Lymphoma

MCPyV

Merkel cell polyomavirus

mIHC

multiplex immunohistochemistry

PBMC

peripheral blood mononuclear cells

PE

R-phycoerythrin

PE594

PE-DyLight 594

PE650

PE-DyLight 650

RSV

Respiratory Syncytial Virus

T-Ag

T-antigen

TCR

T cell receptor

T_FH

follicular helper CD4⁺ T cell

TMA

tumor microarray

UMAP

uniform manifold approximation and projection