Microsatellite-stable (MSS) colorectal cancers are characterized by low mutation burden and limited immune-cell infiltration and thereby respond poorly to immunotherapy. Here, we report a case of metastatic MSS colorectal cancer with a robust anticancer immune response. The primary tumor was resected in 2012, and the patient received several cycles of chemotherapy until 2017. In 2018, the patient underwent a left hepatectomy to remove a new metastasis. Analysis of the metastatic tumor revealed a strong CD8+ T-cell response. A high frequency of CD8+ T cells coexpressed CD39 and CD103, a phenotype characteristic of tumor-reactive cells. Using whole-exome sequencing, we identified somatic mutations that generated peptides recognized by CD39+CD103+CD8+ T cells. The observed reactivity against the tumor was dominated by the response to a single mutation that emerged in the metastasis. Somatic mutations that were not immunogenic in the primary tumor led to robust CD8+ T-cell expansion later during disease progression. Our data suggest that the cytotoxic treatment regimen received by the patient might be responsible for this effect. Hence, the capacity of cytotoxic regimens to prime the immune system in colorectal cancer patients should be investigated further and might provide a rationale for combination with immunotherapy.
Research showing that the immune system can recognize and kill cancer cells led to a paradigm shift in the treatment of patients with cancer. Responses to immunotherapy are observed in patients with metastatic disease that progressed despite previous lines of treatment. Some of these responses are long-lasting, as opposed to the responses commonly elicited by existing targeted therapies. The benefits of immunotherapy are observed primarily in cancers with high numbers of nonsynonymous somatic mutations such as melanoma and mismatch-repair (MMR)–deficient colorectal cancer (1–3). This is because the mutations can lead to neoantigens that are tumor specific and can be recognized by the patient's own immune cells. In cancers with low tumor mutation burden (TMB), current immunotherapy approaches are much less effective, as illustrated by the lack of response to anti–PD-1 therapy in patients with MMR-proficient colorectal cancer (1). For patients with unresectable metastatic MMR-proficient colorectal cancer, the main treatment option remains chemotherapy and the median overall survival is 30.2 months (4).
Colorectal tumors display different levels of immune-cell infiltration, and the presence of CD8+ T cells at the invasive margin correlates with better overall survival (5). However, as in many other cancer types, the CD8+ T-cell infiltrate in colorectal cancer is heterogeneous, and we find that only a fraction of tumor-infiltrating CD8+ T cells can recognize tumor antigens. These tumor-reactive CD8+ T cells are enriched in cells that coexpress the molecules CD39 and CD103 [referred to here as double-positive (DP) CD8+; refs. 6, 7].
Here, we report the case of a patient with stage IV colorectal cancer and inoperable liver metastases at the time of diagnosis in 2012. The patient initially had a very limited response to chemotherapy, but eventually his immune system mounted a strong CD8+ T-cell response against the tumor, and he is now alive and disease free. The CD8+ T-cell response was principally directed against a single mutation that emerged in the metachronous colorectal liver metastasis (CRLM). Furthermore, tumor mutations that were not immunogenic in the primary tumor led to robust CD8+ T-cell expansion in the CRLM. Our data suggest that some cytotoxic regimens might be beneficial at priming the immune system to recognize tumor-specific mutations and that combining such regimens with checkpoints blockade might be a promising approach in the future.
Materials and Methods
Fresh peripheral blood and tumor samples were collected from 15 patients with colorectal cancer and CRLM undergoing surgical resection. All subjects signed written informed consent approved by the Providence Portland Medical Center Institutional Review Board (IRB protocol no. 06-108A), and the study was conducted in accordance with the ethical standards established by the Declaration of Helsinki. After pathologic exam, a portion of the resected specimens was immediately processed into single-cell suspension by mechanical dissociation and digested in RPMI-1640 medium (Gibco, cat. #11875119) using hyaluronidase (Sigma-Aldrich, cat. #H6254), collagenase (Sigma-Aldrich, cat. #C5138), and DNase (Roche, cat. #04536282001). Peripheral blood mononuclear cells (PBMC) were purified from whole blood collected at time of surgery over Ficoll-Paque PLUS (GE Healthcare, cat. #17144003) gradient. All samples were cryopreserved until further analysis.
Antibodies and flow cytometry
The following fluorescent-labeled antibodies were used in various combinations: FITC and Alexa Fluor (AF) 700 anti-CD3 (UCHT1; cat. #300440 and 300424, respectively), brilliant violet (BV) 785 anti-CD4 (OKT4; cat. #317442), BV510 and BV711 anti-CD8 (RPA-T8; cat. #301048 and 301044, respectively), BV650 anti-CD25 (BV96, cat. #302634), BV510 anti-CD45 (2D1; cat. #368526), PerCP/Cy5.5 anti-CD45RA (HI100; cat. #304122), BV605 anti-CD69 (FN50; cat. #310938), AF647 anti-CD103 (Ber-ACT8; cat. #350209), BV421 anti-CD127 (A019D5; cat. #351310), PE anti-CD137 (4-1BB; 4B4-1; cat. #309804), PE/Dazzle 594 anti-CD152 (CTLA-4; BNI3; cat. #369616), and APC/Fire 750 anti-CD197 (CCR7; G0443H7; cat. #353246; all from BioLegend); BV650 anti-CD39 (TU66; cat. #563681), PE-Cy7 anti-CD279 (PD-1; EH12.1; cat. #561272), and AF488 anti–Ki-67 (B56; cat. #561165; all from BD Biosciences). After thawing, cryopreserved tumor digest and PBMCs were incubated with a fixable live/dead dye to distinguish viable cells (BioLegend, Zombie Yellow Fixable Viability kit, cat. #423104) and then stained with a combination of antibodies. Cell-surface staining was performed in FACS buffer. Intracellular staining was performed using the eBioscience Foxp3/Transcription Factor Staining Buffer set (eBioscience, cat. #00552300) according to the manufacturer's instructions. Stained cells were acquired on an LSRFortessa flow cytometer. Data were analyzed with FlowJo software v10.6 (Treestar). UMAP analysis was performed on concatenated tumor-infiltrating total CD3+ cells (5,000 cells/patient) isolated from 11 patients with CRLM using the UMAP plugin in the FlowJo software.
Cell sorting and T-cell expansion
Cryopreserved tumor cell suspensions were thawed and labeled for sorting as described in “Antibodies and flow cytometry.” Cells were sorted on a FACSAria II. CD8+ subsets from TILs were sorted as CD45+CD4−CD8+CD45RA−CCR7+/−CD39−CD103− (double negative, DN), CD45+CD4−CD8+CD45RA−CCR7+/−CD39−CD103+ (single positive, SP), and CD45+CD4−CD8+CD45RA−CCR7+/−CD39+CD103+ (DP). Sorted T cells were expanded in vitro with 1 μg/mL Phytohemagglutinin (PHA; Remel, cat. #R30852801) in the presence of irradiated (5,000 rad) allogeneic feeder cells (PBMC isolated from one healthy donor) and 10 ng/mL of recombinant human IL15 (BioLegend, cat. #570304). Cells were cultured in a 96-well round-bottom plate in RPMI-1640 supplemented with 2 mmol/L L-Glutamine (Gibco, cat. #25030081), 1% (vol/vol) nonessential amino acids (Gibco, cat. #11140050), 1% (vol/vol) sodium pyruvate (Gibco, cat. #11360070), penicillin (100 U/mL), streptomycin (100 μg/mL; Gibco, cat. #15140122), and 10% pooled human serum (in-house). T-cell lines were maintained in complete medium with IL15 until analysis. For TCR-sequencing analysis, sorted cell pellets were frozen until further processing.
Formalin-fixed, paraffin-embedded (FFPE) blocks were cut in serial section of 4 μm thickness and placed onto Superfrost Ultra Plus Adhesion slides (Fisher Scientific). Deparaffinization was performed according to standard protocols. Tissue sections were stained for multiplex immunohistochemistry (mIHC) on a Leica-Bond RX autostainer (Leica Biosystems). Samples were evaluated using two different multiplex immunofluorescence staining panels (for details, see Supplementary Table S1). Antigen–antibody binding was visualized with the TSA-Opal reagents (Akoya Biosciences). Antigen retrieval was performed between antibody detection to prevent cross-reactivity. Tissue slides were incubated with DAPI as counterstain and coverslipped with VectaShield mounting media (Vector Labs, cat. #H-1000-10). Control tissue samples were stained for each marker separately.
mIHC image acquisition and analysis
Digital mIHC images were acquired with the Vectra Polaris Automated Quantitative Pathology Imaging System (Akoya Biosciences). Whole tissue slides were scanned at 20× magnification and visualized with Phenochart software (Akoya Biosciences) to identify the regions of interest (ROI). Between 13 and 20 ROIs were selected from each tissue sample for mIHC analysis with InForm 2.4 software (Akoya Biosciences). Cell phenotypes were enumerated in the stroma and tumor compartment for the images. Each tissue slide was analyzed in individual InForm projects.
Consecutive tissue slides for each case were stained by conventional hematoxylin and eosin method to ensure the presence of tumor and evaluate fixation quality. Tissue slides were digitally scanned with the Leica SCN400F platform at 20× and magnified at 200× to 400× for immune infiltration evaluation.
Whole-exome sequencing and RNA sequencing
Genomic DNA and total RNA from patient CRI 3280 were purified from 5-μm FFPE sections on an automated QiaCube instrument using the AllPrep DNA/RNA kit (Qiagen, cat. #80234) according to the manufacturer's instructions. Corresponding normal DNA for germline exome testing was purified from the patient's CD8+ T cells, generated as described in “Cell sorting and T-cell expansion.” DNA and RNA were quantified using a Qubit fluorometer (Thermo Fisher).
Whole-exome sequencing (WES) for tumor and germline specimens was performed on 200 ng of purified DNA as follows: DNA was prepared into tagged sequencing libraries using Kapa HyperPlus library preparation reagents (KAPA HyperPlus, cat. #7962428001; KAPA UDI Adapter Kit, cat. #8861919702; KAPA HiFi HotStart Library Amplification Kit with Primer Mix, cat. #7958986001 all from Roche), and exome hybrid capture was performed using the xGen Exome Research Panel kit (IDT, cat. #1056115). Captured library pools were normalized and loaded onto a HiSeq 4000 sequencer (Illumina) for next-generation sequencing. WES reads were aligned to Genome Reference Consortium Human Build 37 (hg19) followed by GATK preprocessing. Four independent single-nucleotide mutation callers (Varscan 2.3.6, SomaticSniper 220.127.116.11, Mutect 1.1.7 and Strelka 1.0.15) were used to call somatic nonsynonymous single-nucleotide variants and two Insertion-Deletion callers (Strelka 1.0.15 and Varscan 2.3.6) were used to identify somatic nonsynonymous In-Dels. Identified mutations were filtered according to the following criteria: minimum coverage of 10 reads, greater than 5% variant allele frequency (VAF) and called by two or more callers. Somatic mutations that passed the filters were further annotated with the 1000 Genomes Project, Exome Aggregation Consortium (ExAC), and Catalogue of Somatic Mutations in Cancer (COSMIC) databases using Annovar. SNPeff was used to predict variant functional effect. Every nonsynonymous mutation was used to build putative neoepitopes of 25-mer amino acids.
An mRNA sequencing library was also prepared from FFPE tissues using RNA Access Library Preparation reagents (IDT for Illumina – TruSeq RNA UD Indexes, cat. #20022371; TruSeq RNA Library Prep for Enrichment, cat. #20020189; TruSeq RNA Enrichment, cat. #20020490; Illumina Exome Panel, cat. #20020183 all from Illumina) according to the manufacturer's instructions. Libraries were pooled and sequenced at a depth of 25 to 50 million reads on a HiSeq 4000 sequencer (Illumina). RNA alignment was performed using STAR, followed by GATK preprocessing best practices workflow. Fragments per kb per million mapped reads (FPKM) values were calculated using Cufflinks. FPKM levels were used to assess expression of candidate mutations.
Tandem minigene construction and in vitro transcription of RNA
TMGs were constructed as previously described (8, 9). For each nonsynonymous variant identified by WES, we constructed a “minigene,” consisting of the mutant amino acid flanked by 12 amino acids of the wild-type (WT) protein sequence. Up to 16 minigenes were concatenated to generate a tandem minigene (TMG) construct. TMG constructs were codon optimized and subcloned into pcDNA3.1 + CEF-MHC-1-v2 (kindly provided by Dr. Eric Tran; performed by GenScript). The pcDNA3.1 + CEF-MHC-1-v2 vector is based on the pcDNA3.1 vector (Thermo Fisher Scientific, cat. #V79020) modified to contain a repeat of the beta-globin 3′ UTR sequence and polyA tail (∼120 As) after the TMG sequence. These modifications have been found to stabilize in vitro–transcribed RNA (IVT; ref. 10). One microgram of linearized (NotI; New England Biolabs) plasmid DNA was used as a template to generate IVT TMG RNA using the Mmessage Mmachine T7 Ultra kit (Life Technologies, cat. #AMB13455) as instructed by the manufacturer. RNA was cleaned up (Zymo Research, cat. #R1017) and concentration was measured using a NanoDrop spectrophotometer prior to use in transfections.
Peptide synthesis and peptide pulsing
Potential 25-mer neoantigen sequences contained in TMG1, TMG2, and TMG3 were converted into FASTA format and run through NetMHCpan 4.0 Server (Technical University of Denmark). NetMHCpan generates 8- to 11-mer peptides from the 25-mer neoantigen sequence and predicts binding affinity to patient-specific MHC class I molecules. Peptides predicted to have a binding affinity ≤0.5 nmol/L were considered candidates for further evaluation. Those peptides were synthesized by GenScript USA Inc. and subsequently tested for reactivity against patient CD8+ T-cell subsets. Crude peptides were used for the initial in vitro screening of TMG-reactive DP CD8+ T cells. To validate reactivities, selected HPLC-purified mutant peptides and their WT counterparts were purchased from GenScript. TMG-reactive DP CD8+ T cells were directly pulsed with relevant peptides at the indicated concentrations.
Transfection of TMG RNA
Cryopreserved expanded DN CD8+ T cells isolated from patient CRI 3280 were used as APCs for the reactivity experiments. APCs were labeled with Cell Proliferation Dye eFluor 450 (eBioscience, cat. #65084285), washed, and resuspended in Opti-MEM (Life Technologies, cat. #31985062) at 20 × 106 cells/mL. TMG RNA (4 μg) was added to a 2 mm gap electroporation cuvette, followed by 50 μL of APCs. Cells were electroporated at 250 V for 5 ms for one pulse, using a BTX ECM 830 Square Wave Electroporation System (Harvard Bioscience Inc.). Electroporated cells were incubated for 4 hours before coculture with CD8+ T-cell subsets.
T-cell reactivity assay
T cells were thawed and cultured in RPMI-1640 medium with 10% pooled human serum supplemented with 10 ng/mL IL15 the day before the coculture with target cells. Before each coculture, target cells and T cells were washed and the medium was replaced with cytokine-free medium. Typically, equal volumes (100 μL) of T cells, and target cells were mixed together in an ELISPOT plate (Millipore, MAIPSWU10). 1 × 105 T cells were coincubated with TMG-electroporated APCs. Alternatively, for peptide reactivity, T cells were directly incubated with the corresponding peptides. All cocultures were performed in the absence of exogenous cytokines. For all assays, plate-bound anti-CD3 Ab (OKT3, cat. #317326, 1 μg/mL; BioLegend) was used as a positive control. Media, DMSO (Corning, cat. #25950CQC) were used as negative controls. Cell-surface T-cell activation was assessed by 4-1BB upregulation by flow cytometry 18 hours after stimulation. Briefly, cocultured cells were pelleted and labeled with a viability dye, followed by CD3, CD8, 4-1BB, PD-1, and CD25 cell-surface staining (as described in “Antibodies and flow cytometry”). Cells were washed and resuspended in staining buffer and acquired on a BD LSRII flow cytometer. Data were analyzed using FlowJo software. IFNγ secretion was measured by ELISPOT. Cells were stimulated overnight in anti-human IFNγ-coated plates (Millipore, cat. #MAIPSWU10). For positive control, wells were coated with anti-CD3. Plates were developed as previously described (6) and spots were quantified using an ImmunoSpot plate reader and associated software (Cellular Technology Limited). For the peptide reactivity experiments, TMG-reactive DP CD8+ T cells were enriched by repeating the coculture of DP CD8+ TILs with TMG RNA-electroporated APCs and sorting 4-1BB+CD25+CD8+ T cells for TMG1, TMG2, and TMG3. Following a 2-week polyclonal stimulation with 1 μg/mL PHA in the presence of irradiated (5,000 rad) allogeneic feeder cells (PBMC) and 10 ng/mL of recombinant human IL15, TMG-reactive DP CD8+ T cells were incubated with single peptide candidates and positive responses were identified by 4-1BB upregulation as indicated above.
DNA preparation and TCR sequencing
Deep sequencing of the variable V-J or V-D-J regions of TCRβ genes was performed on genomic DNA of sorted T-cell populations. DNA was extracted from ex vivo TIL CD8+ T-cell subsets or from in vitro–expanded TIL CD8+ T cells (DNeasy Blood and Tissue Kit, Qiagen, cat. #69504). For TCR repertoire analysis of the primary tumor and metachronous liver metastasis, DNA extracted from FFPE tissue sections was shipped to Adaptive Biotechnologies for sequencing. The TCRβ CDR3 regions were sequenced and mapped (ImmunoSEQ, Adaptive Biotech). Coverage per sample was >10×. Only data from productive rearrangements were extracted from the ImmunoSEQ Analyzer platform for further analysis. The circos plot was created with the circlize R package (11) and depicts TCR repertoire similarity between different populations of cells. The ribbons connect a highlighted population with other populations that contain shared productive nucleotide sequences. Connections between the non-highlighted populations are not depicted. The width of the ribbon at each end is the proportion of total sequences in that population that are shared with the connected population.
TCR-sequencing data have been deposited in the ImmuneACCESS database (Adaptive Biotechnologies) under https://doi.org/10.21417/VR2021CIR and https://clients.adaptivebiotech.com/pub/rajamanickam-2021-cir. The somatic mutations contained in TMG1, TMG2, and TMG3 as well as the sequences of the predicted peptides are available in Supplementary Table S2. Underlying WES sequencing data are not publicly available due to HIPAA protection of the patient's germline sequencing data. We do not have patient consent to release read-level RNA-sequencing (RNA-seq) data containing private/rare variants. Any inquiries for accessing these data (including RNA-seq data) should be directed to firstname.lastname@example.org and we will grant access to the deidentified data sets for research purposes.
A 53-year-old man (referred to as patient CRI 3280) was diagnosed with obstructing MMR-proficient metastatic colon cancer in February 2012. He underwent a laparoscopic assisted anterior resection with en bloc partial bladder resection and diverting loop ileostomy on March 6, 2012. Final pathologic staging was T4bN2aM1. At the time, the liver metastases were deemed inoperable. After surgery, the patient was treated with FOLFOX chemotherapy [a combination of folinic acid, 5-fluorouracil (5-FU), and oxaliplatin] and Zometa from April 9, 2012 to October 22, 2012, followed by 5-FU from October 22, 2012, to December 3, 2012. Imaging in January 2013 demonstrated a mixed response to chemotherapy: The left-side liver lesion had decreased in size, but multiple right-side lesions had doubled in size during the 3-month period off FOLFOX chemotherapy. The patient underwent transarterial radioembolization with yttrium-90 (Y-90) to the right hepatic lobe on February 13, 2013, and then resumed chemotherapy with FOLFIRI regimen (a combination of folinic acid, fluorouracil, and irinotecan hydrochloride) plus the anti-VEGF monoclonal antibody bevacizumab from February 25, 2013, to August 19, 2013. After a left hip pathologic fracture was biopsied and confirmed to be metastatic adenocarcinoma, the patient was treated with palliative radiation, 3,000 cGy in 10 fractions, he then resumed FOLFIRI chemotherapy until the end of 2014. The patient received Zometa on and off between 2015 and early 2016. Because of disease progression, he was treated with irinotecan chemotherapy plus the anti-EGFR monoclonal antibody panitumumab from January 7, 2015, to June 21, 2017. Chemotherapy was held for hip replacement in June 2015. The patient had stable disease at this time and was given a chemotherapy holiday starting in June 2017. In 2018, his carcinoembryonic antigen (CEA) levels began to rise slowly, and the only evidence of disease progression was a slowly growing mass in the left hepatic lobe. PET/CT confirmed the only site of hypermetabolic activity was in the left hepatic lobe, which led to a left hepatectomy on September 21, 2018. Pathology revealed diffuse liver involvement by moderately differentiated adenocarcinoma. Repeat imaging in the Spring of 2019 showed no evidence of recurrent disease, and ongoing decreasing size of the last visible hepatic lesions. The patient remained disease free as of August 2020.
Characterization of the immune infiltrate
We performed an in-depth analysis of infiltrating T cells by flow cytometry using fresh tissue from the CRLM surgically removed from patient CRI 3280 on September 21, 2018. Because of the patient's unusual clinical course and successful response to therapy, we compared the tumor immune-cell infiltrates from patient CRI 3280′s CRLM and other CRLMs collected previously. Following UMAP analysis, patient CRI 3280 displayed a very distinct T-cell infiltrate (Fig. 1A). The majority of immune cells infiltrating his tumor were CD3+ T cells, and the frequency of CD8+ T cells was much higher than in all other CRLMs we analyzed (Fig. 1B and C). The CD8+ TILs in patient CRI 3280′s CRLM contained a high frequency of cells coexpressing CD39 and CD103, cell-surface markers expressed by tumor-reactive CD8+ T cells (6). The frequency of CD39+CD103+ (DP) CD8+ TILs was 36.1%, whereas the mean frequency of these cells in 14 independent CRLM samples analyzed was 6.25%, suggesting that patient CRI 3280′s CRLM might be more immunogenic. This hypothesis was supported by high expression levels of PD1, CTLA-4, and low levels of IL7R (CD127) on both CD39−CD103+ (SP) CD8+ and DP CD8+ TILs (Fig. 1D). Both subsets also displayed increased frequencies of cells expressing 4-1BB and Ki-67, signs of a recent antigen encounter in the tumor microenvironment (TME).
Sequencing of the CDR3 regions of the TRB genes revealed that SP CD8+ and DP CD8+ TILs were more oligoclonal than CD39− CD103− (DN) CD8+ TILs (Fig. 1E). The most expanded clonotype in DP CD8+ TILs was also the dominant clonotype in the SP CD8+ TILs (Fig. 1F), and it represented one third of all CD8+ TILs, which is highly unusual. DP CD8+ TILs shared a number of other clonotypes with SP CD8+ TILs. In contrast, there was little overlap with the repertoire of the DN CD8+ TILs. The CRLM from patient CRI 3280 was also characterized by a high frequency of FOXP3+ regulatory T cells (Treg) in the CD4+ T-cell compartment, which was much higher than detected in the other CRLMs analyzed (Supplementary Fig. S1A–S1C). No differences were observed between the different CRLM patients in regard to the expression of activation markers at the surface of tumor-infiltrating CD4+ Th cells, but these cells from patient CRI 3280 were more proliferative, as demonstrated by a higher frequency of Ki-67+ cells (Supplementary Fig. S1D).
To determine whether the primary colorectal cancer tumor resected from patient CRI 3280 in 2012 was highly infiltrated with T cells, we stained FFPE tissue sections from the primary tumor by mIHC and compared it with the CRLM resected in 2018. Our results indicate that even though a significant number of CD3+ cells were present in the primary tumor, their frequency was highly increased in the CRLM (Fig. 2A–C). Moreover, more T cells were found infiltrating the tumor itself in the CRLM sample. We also observed a shift in the distribution of CD4+ Th cells (CD3+CD8−FOXP3−), Tregs (CD3+CD8−FOXP3+), and CD8+ T cells (CD3+CD8+FOXP3−), with a sharp rise in CD8+ T cells in the CRLM as compared with the primary tumor. To further dissect the composition of CD8+ T cells in both tumors and their capacity to proliferate in situ, we stained tissue sections for CD103 and Ki-67. We found that half of the CD8+ T cells in the CRLM were infiltrating the tumor (Fig. 2D–F). Moreover, the majority of those CD8+ T cells in the tumor epithelium expressed CD103 and Ki-67 (Fig. 2G and H). In contrast, CD8+ T cells were mostly present in the stroma in the primary tumor and much fewer cells expressed CD103. Collectively, these results support the notion that there was a strong immune response in the CRLM that was absent in the primary tumor.
Identification of tumor neoantigens recognized by infiltrating CD8+ TILs
To identify the tumor-specific mutations recognized by DP CD8+ TILs, we analyzed the mutanome of the CRLM and compared it with the mutanome of the primary tumor using WES and RNA-seq. We identified 156 mutations in the CRLM and 117 mutations in the primary tumor, 51 of which were shared. Somatic mutations in the CRLM with a VAF ≥5%, identified by at least two callers and expressed at an FPKM ≥0.5, were chosen for T cell–reactivity analysis. Using the TMG approach, we synthesized six TMGs encompassing a total of 97 mutations. As illustrated by 4-1BB upregulation, reactivity against TMG1, TMG2, and TMG3 was detected in DP CD8+ TILs (Fig. 3A). Lower frequencies of cells reactive against TMG1, TMG2, and TMG3 were also detected in SP CD8+ TILs but not in DN CD8+ TILs. To identify which of the mutations present in the three TMGs were recognized by DP CD8+ TILs, we cultured in vitro–expanded TMG-reactive CD8+ T cells with single peptide candidates, and positive responses were identified by 4-1BB upregulation. Our screening resulted in the identification of five putative neoantigens: XRCC5I213L and PGM2L1E435K from TMG1, PPP2R1AR183W from TMG2, and TBC1D10BK778T and EIF3II108M from TMG3 (Fig. 3B). Reactivity against the first four mutations was confirmed by titrating HPLC-purified WT and mutated peptides (Fig. 3C). In contrast, both the WT and the mutated peptides for EIF3I were recognized by DP CD8+ T cells, suggesting potential self-antigen reactivity.
We next sought to determine the frequency of the neoantigen-reactive TCR clonotypes in the DP CD8+ TILs ex vivo. The TCR clonotype recognizing the TBC1D10BK778T mutation corresponded to the dominant TCR clonotype found in both the DP CD8+ and the SP CD8+ TILs ex vivo. In addition, the TCR clonotypes recognizing mutation XRCC5I213L and mutation PPP2R1AR183W ranked fourth and seventh, respectively, in the DP CD8+ TILs. We also compared the TCR repertoire of the DP CD8+ TILs with the repertoire of the CRLM and of the primary tumor (FFPE tissue; Fig. 4A). As expected, there was a strong overlap between the DP CD8+ TILs and the CRLM repertoire (201 out of 706 shared sequences), with the dominant clonotypes in the DP CD8+ TILs being also among the most expanded clonotypes in the CRLM repertoire (Fig. 4B). In contrast, little overlap was observed with the primary tumor where only 22 clonotypes were shared. The dominant clonotypes recognizing mutations, TBC1D10BK778T, XRCC5I213L, and PPP2R1AR183W, were ranked first, second, and fourth, respectively, in the CRLM (Fig. 4C). Those clonotypes were not detected in the primary tumor, with the exception of a clonotype recognizing PPP2R1AR183W, which ranked 155th. Our data strongly suggest that the five somatic mutations against which we detected T-cell reactivity were highly immunogenic and led to robust clonal expansion in the CD8+ T-cell compartment of the CRLM. In particular, the mutations in XRCC5I213L, PPP2R1AR183W, and PGM2L1E435K were already present in the primary tumor but did not lead to a strong immune response at that stage of the disease. The lack of response was not due to a change in the level of expression of those mutations in the tumor, as shown by very similar FPKM values in the primary tumor and CRLM (see Supplementary Table S3).
We show here that the immune system of a patient with MMR-proficient advanced metastatic colorectal cancer can mount an efficient response against the tumor and that this is concomitant with complete control of the disease. This antitumor immune response, which was not detectable in the primary tumor was strongly detected in the metachronous CRLM even though the TMB did not significantly change and the DNA-repair machinery was unaffected. We find that somatic mutations shared between the primary tumor and the CRLM led to clonal T-cell expansion only later in the disease course. Our data suggest that the observed antitumor response played a role in the patient's clinical response; however, additional work is necessary to validate that point.
This work is consistent with previous studies showing that neoantigen-specific T cells can in some cases be isolated from MMR-proficient colorectal cancers, but that such cells are usually more difficult to detect than in cancers with high TMB such as MMR-deficient colorectal cancers (7, 12, 13). The strong antitumor immune response defined here could be linked to the treatment regimen received and its effect on boosting the antitumor immune response. Among the drugs received, oxaliplatin, which was part of the FOLFOX regimen, is a known inducer of immunologic cell death, a form of regulated cell death that can elicit an immune response against dead cell–associated antigens in the absence of any adjuvant (14–16). The positive effect of the FOLFOX regimen on the immune system has been illustrated in two recent studies, where it increased both the quantity and quality of tumor antigen–specific CD8+ T cells in mouse models of colorectal cancer (17, 18). In addition, another study has shown that Y-90 radioembolization in patients with hepatocellular carcinoma led to local and systemic immune activation (19). Thus, the chemotherapy regimens and radioembolization received by the patient might have played a role in disease control not only by their direct cytotoxic effect on proliferating tumor cells, but also by releasing tumor neoantigens from dying cells and priming the immune system to recognize and kill cancer cells. Soluble factors secreted in the TME might have also facilitated tumor recognition and CD8+ T-cell activation by increasing MHC class I expression and tumor-antigen presentation (Supplementary Fig. S2). Finally, studies have shown that blockade of VEGF, here using bevacizumab, can positively regulate the immune response in part by improving CD8+ T-cell functionality (20). Altogether, the immunostimulatory effects of the treatments received could explain the late expansion of T-cell clonotypes recognizing mutations shared between the primary tumor and the metachronous CRLM and the expansion of T-cell clonotypes recognizing new mutations. Our study indicates that T-cell priming plays an important role in inducing potent antitumor immune responses and implies that the lack of immune response against a tumor might not necessarily be due to the absence of neoantigen-reactive T cells but rather due to suboptimal T-cell priming (12, 13). In conclusion, we believe that the impact of cytotoxic treatment regimens on the immune-cell composition and activation in the TME of colorectal cancer patients should be investigated further and might provide a rationale for combining such treatments with immunotherapy in metastatic colorectal cancer patients with strong treatment-induced immune activation.
B.A. Fox reports nonfinancial support from Akoya Biosciences during the conduct of the study, as well as personal fees from AstraZeneca/Definiens, Boehinger Ingelheim, PrimeVax, Turnstone, UbiVac, and Ultivue, other support from Incyte, and grants from Bristol-Myers Squibb, OncoSec, Macrogenics, NanoString, Shimadzu, and Viralytics/Merck outside the submitted work. E. Tran reports personal fees from PACT Pharma and Genocea Biosciences outside the submitted work. No disclosures were reported by the other authors.
V. Rajamanickam: Data curation, software, visualization, writing–review and editing. C. Ballesteros-Merino: Investigation, visualization, writing–review and editing. K. Samson: Investigation. D. Ross: Software, visualization. B. Bernard: Data curation, software. B.A. Fox: Resources, writing–review and editing. E. Tran: Resources, writing–review and editing. P. Newell: Conceptualization, resources, writing–review and editing. T. Duhen: Conceptualization, supervision, funding acquisition, investigation, visualization, writing–original draft, writing–review and editing.
The authors thank R. Tamakawa and M. Schmidt for assistance with consenting patients for this study; R. Duhen for reviewing the manuscript and discussions; and R. Rattray and M.B. Campbell for TCR sequencing, WES, and RNA-seq performed at the Providence Molecular Genomics Laboratory. This work was supported by the Providence Portland Medical Foundation.