Enhanced mutanome analysis towards the induction of neoepitope-reactive T-cell responses for personalized immunotherapy of pancreatic cancer

Abstract

Background: Personalized immunotherapy of pancreatic ductal adenocarcinoma (PDAC) through T-cell mediated targeting of tumor-specific, mutanome-encoded neoepitopes may offer new opportunities to combat this disease, in particular by countering recurrence after primary tumor resection. However, the sensitive and accurate calling of somatic mutations in PDAC tissue samples is compromised by the low tumor cell content. Moreover, the repertoire of immunogenic neoepitopes in PDAC is limited due to the low mutational load of the majority of these tumors.

Methods: We developed a workflow involving the combined analysis of next-generation DNA and RNA sequencing data from matched pairs of primary tumor samples and patient-derived xenograft models towards the enhanced detection of driver mutations as well as single nucleotide variants encoding potentially immunogenic T-cell neoepitopes. Subsequently, we immunized HLA/human T-cell receptor (TCR) locus-transgenic mice with synthetic peptides representing candidate neoepitopes, and molecularly cloned the genes encoding TCRs targeting these epitopes.

Result: Application of our pipeline resulted in the identification of greater numbers of non-synonymous mutations encoding candidate neoepitopes with increased confidence. Furthermore, we provide proof of concept for the successful isolation of HLA-restricted TCRs from humanized mice immunized with different neoepitopes, several of which would not have been selected based on mutanome analysis of PDAC tissue samples alone. These TCRs mediate specific T-cell reactivity against the tumor cells in which the corresponding mutations were identified.

Conclusion: Enhanced mutanome analysis and candidate neoepitope selection increase the likelihood of identifying therapeutically relevant neoepitopes, and thereby support the optimization of personalized immunotherapy for PDAC and other poorly immunogenic cancers.

Keywords: Adoptive cell therapy - ACT; Gastrointestinal Cancer; Immunotherapy; T cell Receptor - TCR; Vaccine.

Figure 1. Summary of the T-cell neoepitope identification pipeline. (a) Primary tumor exomes were analyzed using a human reference genome (GRCh37/hg19) while the PDX exomes were analyzed using a hybrid reference genome consisting of the human chromosomes (GRCh37/hg19) and mouse chromosomes (GRCm38), adapted to comprise the SNPs of the NOD/ShiLtJ mouse strain. Expression of mutated alleles was assessed from RNA-seq data using the same procedure. RNA data were also used to define the HLA type of each sample. (b, c) Exclusion of mouse reads from PDX exome data by the hybrid reference genome. (d) Immunohistochemistry of primary tumor samples for SMAD4, revealing absence of SMAD4 from the tumor cells of sample T34 (red arrowheads; the other, SMAD4-positive cells are part of the tumor stroma). Tumor T100 shows strong SMAD4 expression in the tumor cells (black arrow heads). See online supplemental figure S1a for further examples of SMAD4-positive and negative tumors. (e) Analysis of exome data of sample T34, as exemplified in the context of a highly conserved (96%) 400 bp region encompassing exons 6 and 7 of the SMAD4 tumor suppressor gene, which contains 17 inter-species SNPs (online supplemental figure S1b). Analysis of exome data of the primary tumor sample with the human reference genome resulted in the mapping of 321 reads without mismatches. (f) Analysis of PDX exome data with the human reference genome resulted in the mapping of 89 reads, the majority of which showed one or more mismatches due to inter-species SNPs, marked by colored, vertical lines. (g) Analysis of PDX exome data with the hybrid reference genome resulted in mapping of all but three of the aforementioned reads to the mouse genome, thereby revealing a homozygous deletion for sample T34 in this part of the SMAD4 locus. The remaining three reads are still mapped to the human genome because these fall within a sequence stretch that shares 100% homology between mouse and human (see online supplemental figure S1b). PDX, patient derived xenograft; RNA-seq, RNA sequencing; SNPs, single nucleotide polymorphisms; sSNVs, somatic single nucleotide variants, .

Figure 2. Increased sensitivity of mutation and T-cell epitope calling in PDX exomes. (a–c) Boxplots of the mutated allele frequency (MAF) of shared, tumor-unique somatic and PDX-unique somatic, non-synonymous SNVs for three selected datasets indicate that, for mutations detected in primary tumor as well as PDX, the MAF in the PDX is generally higher. See online supplemental figure S6 for all 15 data sets. (d) The number of somatic SNVs called from the exome data is in most cases higher in the PDX than in the corresponding primary tumor. (e) This difference is more prominent when only non-synonymous SNVs are considered. (f). Comparison of numbers of SNVs, nsSNVs and potential T-cell epitopes detected in primary tumor and PDX exomes. Shown are the data for tumor samples T15, T02 and T109. See online supplemental figure S8 for an overview on all eight HLA-A*02:01 positive tumor samples. (g) Graphic summary of the outcome of the mutation calling and neoepitope prediction workflow for the eight HLA-A*02:01 positive tumor samples involving, subsequently, nsSNV calling based on exome data, candidate neoepitope prediction with the NetMHC4.0 algorithm, verification of expression of the genes encoding aforementioned peptide sequences at the messenger RNA level, and selection of a set of 18 candidate neoepitopes based on a combination of most favorable parameters (primarily high predicted major histocompatibility complex binding affinity and expression of mutated allele) as detailed in online supplemental table S7. nsSNV, non-synonymous SNV; PDX, patient derived xenograft; SNV, single nucleotide variant.

Figure 3. Experimental analysis of HLA-binding of predicted neoepitopes. HLA-A*02:01 binding of peptides representing the potential T-cell neoepitopes (orange) and their wild-type counterparts (blue) was assessed at 25°C (a) to confirm overall binding and 37°C (b) to distinguish strong from weaker binders. Numbers refer to the peptides listed in online supplemental table S8, which also lists the well-defined HLA-A*02:01 restricted epitopes that were used as positive (A–C) and negative (D–G) controls. Plotted are the ratios of FACS-derived HLA-A*02:01 MFI values of peptide-loaded T2 cells versus untreated T2 cells. Asterisks indicate significantly increased binding of the mutant peptide compared with the wildtype version; the pound sign highlights candidates with an overall high affinity to HLA-A*02:01 irrespective of the mutation. FACS, fluorescent activated cell sorting; MFI, mean fluorescent intensity.

Figure 4. Generation of neoepitope-specific TCRs in HLA-A*02:01/human TCR locus transgenic mice. (a) Schematic overview of the workflow. Splenocytes from HLA-A*02:01 × human TCR-locus transgenic mice, repeatedly immunized with the synthetic peptide of interest, were cultured in the presence of the peptide concerned. Peptide-reactive CD8+T cells as identified by IFN-γ capture assay were sorted, after which the repertoire of TCR Vα and Vβ sequences was analyzed by 5’-RACE RT-PCR on the basis of RNA from these T cells. The most prominent Vα and Vβ sequences were cloned into retroviral expression vectors, upstream of the murine Cα and Cβ regions to allow for selective pairing and detection of cell-surface expression of the transgene encoded TCRs in primary human T cells. In case of identification of multiple Vα and/or Vβ sequences, all possible combinations were tested. The reactivity of efficiently expressed Vα/Vβ combinations against mutant and wild-type peptide epitopes was subsequently tested by means of IFN-γ ELISA. (b–e) Shown as an example is the isolation of the TCR against the 9-mer neoepitope FILIP1L_F109V as identified in tumor T102. See online supplemental figure S10–S12 for corresponding data for the other three primary neoepitopes. (b) Peptide-reactive T-cell population as detected in splenocyte culture of a mouse (ID #9654) repeatedly immunized with the FILIP1L_F109V peptide by means of IFN-γ capture assay in the presence of the relevant peptide. Shown are the cells in the CD3-positive gate. (c) TCR Vα and Vβ sequences identified by means of 5’-RACE RT-PCR in FACS-sorted T-cells. (d) Expression of the four different combinations of TCR alpha and beta chains identified on retroviral gene transduction into primary human PBMC. In this case, all four Valpha/Vbeta combinations were expressed and therefore tested for reactivity (d) Response of human T cells transduced with one of the four TCRα/β combinations (T9654a) against T2-cells loaded with titered concentrations of the FILIP1L_F109V 9-mer neoepitope of its wild-type counterpart. Neither of the other three TCR-α/β combinations mediated significant peptide epitope recognition to TCR-transduced T cells. FACS, fluoresence activated cell sorting; IFN, interferon; PBMC, peripheral blood mononuclear cells; RACE, rapid amplification of complementary DNA ends; TCR, T-cell receptor.

Figure 5. TCR-mediated T-cell reactivity against cognate peptide antigen and autologous pancreatic ductal adenocarcinoma tumor cells. Human T cells (T222 cell line, see Materials and methods) were transiently transfected with in vitro transcribed messenger RNA encoding the indicated TCR-α/β pairs, and subsequently incubated with the following antigen presenting cells: peptide-loaded T2 cells (left panels), autologous HLA-A*02:01+antigen presenting cells transduced with multi-epitope gene constructs encoding a string-of-beads arrangement of either the mutant or the wild-type peptide sequences of interest (middle panels), or T15 and T102 tumor cells (right panels). T-cell responses were measured by means of IFN-γ ELISpot assays as further detailed below. Spot counts represent the numbers of T cells that secreted IFN-γ as captured by anti-IFN-γ antibodies adhered at the bottom of the 96-well plates (see Materials and methods). (a, b, c) TCRs raised against the FARP1_V785I, and TTC39A_S54R neoepitopes identified in the T15 tumor sample. (d) TCR raised against the FILIP1L_F109V 9-mer neoepitope identified in the T102 tumor sample. (e) TCR raised against the RABL6_K164E neoepitope identified in the T109 sample. The T2 cells were pre-pulsed with synthetic peptides at a concentration of 5 µg/mL. Each TCR was tested against their cognate neoepitope peptide (T2_mut) and the wild-type counterpart (T2_wt). In the case of the multi-epitope gene constructs, the epitopes of interest, flanked on each side by 10 amino acids of natural sequence context, were incorporated into a string of beads arrangement that was inserted between the luminal and transmembrane of the LAMP1 protein to allow for efficient processing into major histocompatibility complex. The HLA-A*02:01 MART-1 epitope was added at the C-terminal end of the multi-epitope arrangement to serve as internal positive control through its recognition by DMF5 TCR-transduced T cells (online supplemental figure S14). In these experiments, multi-epitope gene constructs were transfected into T222 cells, which express HLA-A*02:01, to create a fully autologous T-cell/antigen presenting cell setting. Antitumor reactivity was measured against IFN-γ-pretreated (T15+IFN; T102+IFN) and non-treated (T15; T102) tumor cells, as well as in the presence of pan-HLA class I antibody W6/32 (block). Mock-transduced T-cells and stimulation with PMA/ionomycin served as negative and positive controls, respectively. Responses were measured after 24 hours co-cultivation by means of IFN-γ ELISpot. Each TCR was tested in at least two independent experiments with similar outcome. ELIspot, enzyme-linked immunosorbent spot; IFN, interferon; PMA, phorbol-12-myristate-13acetate; TCR, T-cell receptor.

Figure 6. TCR-mediated T-cell reactivity against naturally processed antigen and limiting antigen concentrations. Human T cells were transduced with the indicated TCRs as in figure 5 and incubated with (a) HLA-A*02:01+antigen presenting cells transiently RNA-transfected with full-length gene constructs encoding the K164E-mutated or wild-type RABL6 antigen (b) T-2 cells loaded with indicated concentrations of the cognate peptide epitope or its wild-type counterpart. Responses by T cells transduced with the indicated TCR were measured after 24 hours co-cultivation by means of IFN-γ ELISpot. Each TCR was tested in two independent experiments with similar outcome. Data obtained show good correspondence with initial TCR screenings as depicted in figure 4e and online supplemental figures S10–12. (c) Models for binding of the indicated wild-type (purple) and mutant (ochre) peptides to the peptide binding groove of HLA-A*02:01, as generated using PANDORA (version 2.0.0b2) with n_loop_models=1000 and loop_refinement=slow. The best model was defined using molpdf score and visualized using Mol* Viewer (https://molstar.org/viewer/). The primary sequences of the neoepitopes shown are as follows (substituted amino acid underlined). FARP1_V785I: FLFNDILLYT, FILIP1L_F109V: ALLEAQYGV, RABL6_K164E: YILRELPEV, TTC39A_S54R: RMYHSLTYA. ELIspot, enzyme-linked immunosorbent spot; IFN, interferon; TCR, T-cell receptor.