Calreticulin mutant myeloproliferative neoplasms induce MHC-I skewing, which can be overcome by an optimized peptide cancer vaccine

Abstract

The majority of JAK2^V617F-negative myeloproliferative neoplasms (MPNs) have disease-initiating frameshift mutations in calreticulin (CALR), resulting in a common carboxyl-terminal mutant fragment (CALR^MUT), representing an attractive source of neoantigens for cancer vaccines. However, studies have shown that CALR^MUT-specific T cells are rare in patients with CALR^MUT MPN for unknown reasons. We examined class I major histocompatibility complex (MHC-I) allele frequencies in patients with CALR^MUT MPN from two independent cohorts. We observed that MHC-I alleles that present CALR^MUT neoepitopes with high affinity are underrepresented in patients with CALR^MUT MPN. We speculated that this was due to an increased chance of immune-mediated tumor rejection by individuals expressing one of these MHC-I alleles such that the disease never clinically manifested. As a consequence of this MHC-I allele restriction, we reasoned that patients with CALR^MUT MPN would not efficiently respond to a CALR^MUT fragment cancer vaccine but would when immunized with a modified CALR^MUT heteroclitic peptide vaccine approach. We found that heteroclitic CALR^MUT peptides specifically designed for the MHC-I alleles of patients with CALR^MUT MPN efficiently elicited a CALR^MUT cross-reactive CD8⁺ T cell response in human peripheral blood samples but not to the matched weakly immunogenic CALR^MUT native peptides. We corroborated this effect in vivo in mice and observed that C57BL/6J mice can mount a CD8⁺ T cell response to the CALR^MUT fragment upon immunization with a CALR^MUT heteroclitic, but not native, peptide. Together, our data emphasize the therapeutic potential of heteroclitic peptide-based cancer vaccines in patients with CALR^MUT MPN.

Fig. 1.. MHC-I alleles with predicted binding to CALR^MUT-derived peptides are less frequent in CALR^MUT MPNs.

(A) Principal components analysis of MHC-I allele frequencies from patients with CALR^MUT MPN, patients with JAK2^V617F MPN, and U.S. Caucasian population. (B) Comparison of MHC-I allele frequencies from patients with CALR^MUT MPN, patients with JAK2^V617F MPN, and U.S. Caucasian population compared to each other in the NEUS cohort and (C) the Danish cohort. Frequencies are expressed as percentages. MHC-I alleles that are overrepresented in patients with CALR^MUT are in blue (NEUS cohort) and green (Danish cohort), whereas MHC-I alleles that are underrepresented in patients with CALR^MUT are in red (NEUS cohort) and pink (Danish cohort). For the NEUS cohort, only MHC-I alleles differentially expressed between patients with CALR^MUT MPN compared to both patients with JAK2^V617F MPN and U.S. Caucasian population were considered for further analysis. (D) Heatmap of predicted binding of each CALR^MUT-derived peptide to each MHC-I allele from (B) (left). MHC-I alleles that are underrepresented or overrepresented in patients with CALR^MUT MPN in both cohorts are noted with white or black circles, respectively. Actual MHC-I allele frequencies in CALR^MUT and JAK2^V617F MPN patient populations are also noted (right). (E) Cohort breakdown of CALR^MUT MPN MHC-I allele frequencies of individual institution consisting of the NEUS cohort for the six less frequent MHC-I alleles.

Fig. 2.. MHC-I bias is selective against the CALR^MUT fragment.

(A) Schema depicting calculation of patient:peptide score (PPS). Briefly, single peptides derived from longer sequences were applied to NetMHCpan v3 to predict class-I HLA binding against any given patient’s six possible class-I HLAs. In each patient-peptide combination, the strongest affinity was registered as the PPS. (B) Mean PPS of individual peptides derived from indicated protein sequence (9- to 10-mers) from patients with CALR^MUT and JAK2^V617F MPN and pseudopopulation created from expected frequencies of U.S. Caucasian population (NEUS only). Protein sequences are the 44–amino acid CALR^MUT sequence (CALR^MUT−44aa), the wild-type CALR sequence upstream of CALR^MUT−44aa (CALR^1–361), and the irrelevant foreign antigen neuraminidase (NA) from influenza. Also shown are peptides broadly subdivided from predicted possible binding (<10⁴ nM) and predicted nonbinding (>10⁴ nM) peptides. (C) Difference in mean PPS of each group from (B). The Student’s t test was performed to calculate significance. The horizontal bars represent the combination where the statistical test resulted in the noted P value.

Fig. 3.. Underrepresented MHC-I alleles potentiate response to CALR^MUT peptides.

(A) IFN-γ ELISpot of PBMCs from healthy donors that had either at least one (black circles) or zero (white circles) underrepresented MHC-I allele expanded with a peptide pool (15-mers) derived from the entire CALR^MUT fragment and restimulated after 10 days with either the irrelevant peptide (MOG) or the same CALR^MUT fragment peptide pool. (B) Reactivity of CALRLong36-specific CD8⁺ T cells derived from an HLA-B*07:02⁺ patient with post–essential thrombocythemia myelofibrosis. Cells restimulated with different CALR^MUT peptides in an IFN-γ ELISpot, (C) intracellular IFN-γ and TNFα cytokine staining on the CALRLong36-specific CD8⁺ T cells stimulated with CALRLong36 (D) target cell lysis of wild-type or HLA-B*07:02–transduced T2 cells pulsed with different CALR^MUT peptides.

Fig. 4.. Human CD8⁺ T cells activated with heteroclitic CALR9p2 peptides cross-react with CALR9p2 peptide.

(A) Predicted mean PPS in patients with CALR^MUT MPN of all single amino acid substitution of CALR9p2. (B) Predicted affinity of all single amino acid substitution of CALR9p2 to HLA-A*02:01. Seven heteroclitic peptides were chosen for further testing and are identified here only by their amino acid substitution in the CALR9p2 peptide. (C) Binding affinity of CALR9p2 and indicated heteroclitic peptide to the most common MHC-I in the CALR^MUT MPN patient cohorts. Shadowed area indicates predicted binding affinity range of 5000 to 500 nM. (D) Percent IFN-γ⁺ CD8⁺ T cells after primary in vitro stimulation of PBMCs with CALR9p2 or CALR9p2 heteroclitic peptides (noted on x axis) followed by secondary restimulation with either control (MOG, median represented by dotted blue line) peptide, CALR9p2, or initial CALR9p2 heteroclitic peptide (noted by color of bar). (E) Summary of responding donor PBMCs to CALR9p2 or CALR9p2 heteroclitic peptides.

Fig. 5.. CALR^MUT sequence is not immunogenic in C57BL/6 J mice.

(A) The predicted binding affinity of CALR^MUT-derived peptides (8- to 10-mers shown) against all available murine MHC-I alleles. The strongest binding peptide CALR9p2 is identified in red. (B) H-2K^b stabilization assay using TAP-deficient RMA/S cells was performed for CALR9p2 in the presence and absence of serum. The chicken ovalbumin (OVA)–derived peptide SIINFEKL was used as a positive control. (C) Timeline of DNA immunization schedule and CD8⁺ T cell collection for the experiment in (D). (D) IFN-γ ELISpot depicting secondary reactivity of CD8⁺ T cells isolated from draining lymph nodes of mice DNA immunized with full-length CALR^WT, CALR^MUT, and OVA. (E) Timeline of peptide immunization and CD8⁺ T cell collection for the experiment in (F). (F) IFN-γ ELISpot depicting secondary reactivity of CD8⁺ T cells isolated from draining lymph nodes of mice peptide immunized with adjuvant and DMSO, CALR9p2, or SIINFEKL. Data shown represent results from one repeat of experiments performed at least three times.

Fig. 6.. Heteroclitic CALR9p2 peptide vaccine elicits cross-reactive CD8⁺ T cell response against CALR9p2 and controls tumor growth in mice.

(A) Predicted binding affinity to H-2K^b of all single amino acid substitution variants of CALR9p2. Top predicted peptide CALR9p2(T5F) is shown. (B) Cartoon of the expected effect of T5F substitution in CALR9p2 peptide conformational binding into H-2K^b. Known dominant anchor sites and minor anchor sites are depicted in red and blue, respectively. (C) MHC-I stabilization assay using TAP-deficient RMA/S cells was performed for CALR9p2 and CALR9p2(T5F) in the absence of serum. SIINFEKL was used as a positive control. Representative results from one repeat of an experiment performed at least three times. (D) IFN-γ ELISpot depicting secondary reactivity of CD8⁺ T cells isolated from draining lymph nodes of mice peptide immunized with adjuvant and DMSO, CALR9p2, or CALR9p2(T5F). Representative results from one repeat of an experiment performed at least three times. The Student’s t test was performed to calculate significance. (E) Killing assay of peptide-pulsed B16 cells by CD8⁺ T cells isolated from peptide-immunized mice. Representative results from one repeat of an experiment performed at least three times. The Student’s t test was performed to calculate statistical significance. (F) Timeline of peptide immunization prior to tumor implantation for prophylactic vaccine. (G) RMA/S^pER-CALR9p2 tumor growth over time after prophylactic peptide immunization for individual tumors or (H) average growth until the second mouse reached the end point of a tumor size of 400 mm². (I) Survival of mice after the prophylactic vaccine. (J) Timeline of peptide immunization and tumor implantation for therapeutic vaccine and in combination with anti–PD-1 therapy. ip, intraperitoneally. (K) Tumor growth over time after therapeutic vaccine for individual tumors or (L) averaged tumor growth until the second mouse reached the end point. (M) Survival of mice after the therapeutic vaccine, with or without anti–PD-1 therapy. Data from tumor growth experiments represent results from one repeat of experiments performed twice. Significance for tumor growth experiments was calculated by performing a Student’s t test on the area under the curve of each tumor. Significance for survival was calculated by performing a log-rank test.