Immune Checkpoint Blockade Enhances Shared Neoantigen-Induced T-cell Immunity Directed against Mutated Calreticulin in Myeloproliferative Neoplasms

T cells from patients with MPN are exhausted and blockade of checkpoint receptors restores mut-CALR–specific T-cell immunity in vitro. A, Representative flow cytometric analyses showing PD-1 and CTLA4 expression in peripheral blood T cells from patients with CALR⁺ MPN and HDs. B, Summary of flow data for cell-surface expression of checkpoint receptors, listed on the left, in HD and MPN T cells (n = 9 and 8, respectively). Each cell corresponds to one HD or patient with MPN. The color intensity indicates the % expression for each checkpoint receptor as gated under live, CD3⁺ cells. Statistical significance of MPN versus HD for each checkpoint receptor was evaluated by t test. PD-1: ***, P = 0.0002; CTLA4: **, P = 0.0037; LAG3: ***, P = 0.0004; TIM3: *, P = 0.0217; TIM4: **, P = 0.0051; TIGIT: ns, P = 0.2433; B7-H3: *, P = 0.0101; B7-H4: ns, P = 0.0615; VISTA: ns, P = 0.99. Quantification of PD-1 (C) and CTLA4-expressing (D) cells within CD8⁺ and CD4⁺ T-cell subsets (n = 13 for HD and MPN). Each square represents one subject. Data were pooled from 3 independent experiments. Statistical significance was evaluated by t test; HD versus MPN: CD8⁺PD-1⁺ **, P = 0.0014; CD4⁺PD-1⁺ *, P = 0.0106; CD8⁺CTLA4⁺ **, P = 0.0024; CD4⁺CTLA4⁺ *, P = 0.0207. PBMCs from CALR⁺ MPN patients were stimulated in vitro with pooled mut-CALR in the absence or presence of mAbs blocking PD-1 or CTLA-4 (10 μg/mL). Representative IFNγ ELISPOT images (E) and summary of ELISPOT results (F) generated in PBMCs from 18 patients with CALR⁺ MPN. Each data point represents one patient with MPN. The change in spot numbers was displayed as fold change by dividing the number of spots formed after OLP pool stimulation to the number of spots formed after MOG stimulation. Horizontal lines indicate the median. Statistical significance for changes at population level was evaluated by Wilcoxon signed rank test. Isotype versus α-PD-1: P = 0.3465, isotype versus α-CTLA4: 0.4171. In addition, statistical significance was evaluated for each subject by t test by comparing isotype versus checkpoint blockade. Three subjects who showed significant response to checkpoint blockade were denoted. *, P = 0.0121; **, P = 0.0045; ***, P = 0.0005.

Pembrolizumab administration rescues mut-CALR–specific immunity. A, Schema of a phase I/II clinical trial to assess the efficacy, safety, and tolerability of pembrolizumab in patients with chronic phase myelofibrosis. The FDA-approved dose of 200 mg pembrolizumab is administered via intravenous infusion every 3 weeks. Nine patients will be enrolled in the first stage of the Simon two-stage design, and 15 in the second stage. A treatment cycle is 3 weeks and the core study period is 6 cycles. PBMCs were collected from a patient with CALR⁺ MPN receiving pembrolizumab, before (baseline) and after 2 and 6 cycles of treatment, T1 and T2, respectively. The frequency of T cells was analyzed by flow cytometry (B and C; CD3⁺ cells) or by TCRseq (D; number of cells expressing TCR/number of total nucleated cells). E, Clonality of T cells was calculated as 1 – Pielou's evenness. Changes in the abundance of unique TCR Vβ sequences were analyzed using the ImmunoSEQ platform: F and G, T1 versus baseline (F) and T2 versus baseline (G). Only clones with a cumulative abundance of 10 or above were included in the analysis. The binomial method was used to calculate P values. FDRs were controlled by the Benjamini–Hochberg method. The differential abundance of clones was considered significant (red and blue circles) when P value was equal to or less than 0.01. H, PBMCs collected at T1 were stimulated in vitro with either WT or mut-CALR OLPs, alone or in the presence of PD-1–blocking antibodies. IFNγ production by expanded T cells was measured by ELISPOT. Statistical significance was evaluated by t-test, comparing MOG and peptide stimulation for each group: a-PD-1 *, P = 0.0168; CEFT at baseline ****, P < 0.0001; at T1 **, P = 0.0016; at T2 ****, P < 0.0001. I, Two clones that were significantly expanded in peripheral blood upon pembrolizumab treatment were also expanded in in vitro cultures upon stimulation with mut-CALR OLPs when PD-1 was blocked. TCR Vβ rearrangements for each clone are indicated next to frequency graphs.

Healthy donors provide mut-CALR reactive T cells. PBMCs collected from healthy donors were stimulated in vitro with pooled WT or mut-CALR OLPs. Stimulation with a CEFT pool was used as control. Representative IFNγ ELISPOT images (A) and summary of ELISPOT results (B) generated in PBMCs from 16 healthy donors. The spot numbers were calculated by subtracting the number of spots formed after MOG stimulation from the number of spots formed after OLP pool stimulation and negative values were set to zero. P value was calculated by the Wilcoxon signed-rank test. ***, P = 0.001. Horizontal lines indicate the median. C, Percentage of CD137⁺IFNγ⁺ T cells, assessed by flow cytometry. Each bar represents an individual donor. P values were calculated by the Wilcoxon signed-rank test. MOG versus WT: *, P = 0.0144; WT versus mutant: **, P = 0.007. D, Representative flow plots showing IFNγ production by CD8⁺ and CD4⁺ T cells upon priming with MOG or mut-CALR OLPs. E, Quantitative summary of frequencies of T-cell subsets producing IFNγ (n = 15). P values were calculated by the Wilcoxon signed-rank test. *, P = 0.0413; ****, P < 0.0001. Each data point represents one healthy donor. F, Naïve (CD45RO⁻CD45RA⁺CCR7⁺) and memory (CD45RO⁺ CD45RA⁻) T cells and APCs (CD3⁻) were isolated by FACS from HD PBMCs (n = 2). APCs pulsed with MOG or mut-CALR OLPs were cocultured with naïve or memory T cells. IFNγ production by each T-cell population was measured by flow cytometry. P values were calculated by t-test. **, P = 0.078; *, P = 0.0365.

T cells recognize multiple epitopes in mut-CALR. PBMCs from 3 HDs, donors 1, 2, and 3, were stimulated in vitro with individual OLPs, and IFNγ production by CD8⁺ and CD4⁺ T cells was measured by flow cytometry. HLA alleles of individuals were identified by sequence-based genotyping. Binding affinities of donors’ alleles to epitopes within the OLPs that induced IFNγ production were predicted by IEDB's recommended algorithm. ANN, artificial neural network (NetMHC 4.0); SMM, stabilized matrix method; NN-align, NetMHCII 2.2; SMM-align, NetMHCII 1.1.