Peptide vaccine-treated, long-term surviving cancer patients harbor self-renewing tumor-specific CD8+ T cells

Abstract

The behaviors and fates of immune cells in cancer patients, such as dysfunction and stem-like states leading to memory formation in T cells, are in intense focus of investigation. Here we show, by post hoc analysis of peripheral blood lymphocytes of hepatocellular carcinoma patients previously undergoing vaccination with tumour-associated antigen-derived peptides in our clinical trials (registration numbers UMIN000003511, UMIN000004540, UMIN000005677, UMIN000003514 and UMIN000005678), that induced peptide-specific T cell responses may persist beyond 10 years following vaccination. Tracking TCR clonotypes at the single cell level reveals in two patients that peptide-specific long-lasting CD8⁺ T cells acquire an effector memory phenotype that associates with cell cycle-related genes (CCNA2 and CDK1), and are characterized by high expression of IL7R, SELL, and NOSIP along with a later stage promotion of the AP-1 transcription factor network (5 years or more past vaccination). We conclude that effective anti-tumor immunity is governed by potentially proliferative memory T cells, specific to cancer antigens.

Fig. 1. Peptide-specific immune responses induced by peptide vaccine treatments.

PBMCs were collected before and at 3–6 months after vaccination (pre and post, respectively) from each participant and tested to assess antigen-specific IFN-γ production by an ELISpot assay. Patients from the five clinical trials were tested. Representative images of the IFN-γ ELISpot assay of an hTERT-peptide vaccine patient (case A2) and an AFP-peptide vaccine patient (case D1) are shown. In the A2 experiment, PBMCs collected at two time points were examined and IFN-γ secretion in response to either no peptide, hTERT₄₆₁ peptide (test peptide) or CMV (cytomegalovirus) pp65₃₂₈ peptide (positive control) were displayed in the round membranes. Each colored spot was considered to be a cell that produced IFN-γ (a). In the D2 experiment, IFN-γ production against no peptide, AFP₃₅₇, AFP₄₀₃, or CMVpp65₃₂₈ peptide is presented (b). Immune responses measured by the IFN-γ ELISpot assay are plotted in X-Y graphs with annotations of the patient IDs. The X-axis represents the fold change of specific spots and the Y-axis indicates an increase of the specific spot number by comparing pre and post samples. Because positive induction of peptide-specific cytotoxic T lymphocytes (CTLs) was defined as both a more than 10 increase and more than two-fold increase of the specific spot number, dotted lines are drawn at X = 2 and Y = 10 to distinguish cases with positive CTL induction. Positive CTL inductions are highlighted in red and negatives are highlighted in blue. If both pre and post were 0, X was defined as 1. If only pre was 0, the X-value was offset out of X = 100 and if only post was 0, the X-value was placed next to X = 0.1 (c). The proportions of positive CTL induction in each peptide study are shown in bar graphs. Positive CTL induction (CTL + ) is filled in magenta and negative CTL induction (CTL–) is shown in cyan (d). Source data are provided as a Source Data file.

Fig. 2. Induction of HLA-A24:peptide tetramer-binding T cells and phenotypic features.

A representative example of flow plots is shown. The frequency of the CD8α⁺tetramer⁺ fraction was calculated (a). Both total CD8⁺ T cells and tetramer⁺CD8⁺ T cells were detected by CD45RA and CCR7. Four fractions of naïve, central memory (CM), effector memory (EM), and terminally differentiated effector memory with CD45RA (TEMRA) are shown (b). Tetramer positivity among CD8⁺ T cells was compared between pre and postvaccination in CTL( − ) and CTL( + ) of each vaccine study. Samples from the same donors are connected by solid lines. c. The results were combined in one graph. Samples from the same donors and tetramers are connected by solid lines. Mean values are shown by open bars. Two-way repeated-measures ANOVA followed by Bonferroni and Sidak’s multiple comparison post hoc test was carried out. *P < 0.05. d Frequencies of naïve, CM, EM, and TEMRA fractions among tetramer⁺CD8⁺ T cells are displayed in box-and-whisker plots. CTL( − ) (n = 26) and CTL( + ) (n = 16) of pre and post are shown. Statistical analysis was conducted using two-way ANOVA followed by Bonferroni and Sidak’s multiple comparison test to assess differences in composition among the different conditions of CTL induction and treatments. *P < 0.05; **P < 0.005; NS, not significant; error bar, maximum to minimum; horizontal lines of box, first quartile to third quartile; +, mean value (e). The frequencies of naïve, CM, EM, and TEMRA fractions among tetramer⁺CD8⁺ T cells of the hTERT study (n = 14) and the other three cohorts (n = 28) were analyzed separately. Two-way ANOVA followed by Bonferroni and Sidak’s multiple comparison test was carried out. *P < 0.05; **P < 0.005; NS, not significant; whiskers, maximum to minimum; horizontal lines of box, first quartile to third quartile; +, mean value (f). Expression of inhibitory receptors PD-1 and CTLA-4 by HLA-A24:hTERT₄₆₁ tetramer-binding CD8⁺ T cells was assessed. g The percentages of PD-1/CTLA-4 expression and overall survival (OS) period/progression-free survival (PFS) were plotted, then liner regression lines were calculated. r, Pearson’s r; *P < 0.05; **P < 0.005; NS, not significant (h, i) Source data are provided as a Source Data file.

Fig. 3. Persistent peptide-specific immune responses after vaccination.

PBMCs of CTL( + ) patients were collected at each timepoint and tested by the IFN-γ ELISpot assay. The experiments were performed in duplicate and mean numbers of specific spots are displayed in bar graphs per patient with experimental values (cross marks). The CTL( + ) cases among the hTERT participants were examined. The upper graph presents the 1-year result and the bottom graph shows the 5-year result. ND, not done because of unavailability of patient PBMC samples for a reason other than patient death. error bar, + s.e.m. a CTL( + ) cases of the SART2 peptide vaccine study are also shown. ‡PBMCs were unavailable because of patient death. ND, not done. bar, mean number of spots; error bar, + s.e.m. b Four patients with positive CTL induction in the SART3 trial were. ‡, PBMCs were unavailable because of patient death; ND, not done; bar, mean number of spots; error bar, + s.e.m. c For CTL( + ) participants in the AFP trial, ELISpot assays using AFP357 and AFP403 were separately carried out and shown in graphs. ‡PBMCs were unavailable because of patient death. bar, mean number of spots; error bar, + s.e.m. d ELISpot results of CTL( + ) patients from the MRP3 study are shown. Because no patient survived for 5 years, only one graph is shown. Because of the low cell number, a preculture experiment was not performed using E6 PBMCs. ‡PBMCs were unavailable because of patient death. ND, not done. bar, mean number of spots; error bar, + s.e.m. e At the 10-year timepoint after vaccination, there were only five patients who had survived: four patients in the hTERT cohort and one patient in the AFP study. These were all CTL( + ) cases. Immune responses against the peptides are shown in a graph. bar, mean number of spots; error bar, + s.e.m. f Source data are provided as a Source Data file.

Fig. 4. Peptide-specific T cell receptor repertoires over time.

A2 from the hTERT trial and D3 from the AFP trial were analyzed. AFP₃₅₇ was used to analyze the TCR repertoire in D3 in this experiment. The repertoires were presented in pie charts. The size of each pie indicates the frequency of the clone in the total obtained repertoire. The same pies and the same colors represent identical TCR pairs including TRAV/TRAJ and TRBV/TRBJ/TRBD as well as CDR3αβ sequences (a). Three TCR pairs obtained from A2 were tested for their function by ⁵¹Cr release assay. The experiments were done in triplicates. % ⁵¹Cr release were shown in the graphs. Error bar, +/− s.e.m. b. Three TCR pairs obtained from D3 were tested by the luciferase-based killing assay. Mean viabilities from two independent experiments were combined and presented in a bar graph. Each experiment was performed with triplicates. One-way ANOVA was carried out followed by Bonferroni and Sidak’s multiple comparison test. ***P < 0.0005; **P < 0.005; ns, not significant; error bar, +/− s.e.m. c Cytotoxicity induced by the D3.14 of D3 is also shown using the HepG2 hepatoma cell line (HLA-A24⁺AFP⁺hTERT⁺) and fluorescence microscopic time-lapse imaging. A CMVpp65₃₂₈-specific TCR was used as a negative control TCR (see Supplementary Table 4). Time-lapse snapshots are presented (0, 125, 250 min.). This experiment was duplicated and produced a similar result (Source data). d Percentages of dead cells were plotted on a graph over time (minutes) (e). Cytotoxicity induced by the A2.57 and A2.80 of A3 are tested as well using the HepG2 hepatoma cell line and the CMVpp65₃₂₈-specific TCR as a negative control. Time-lapse snapshots are presented (0, 250, 300 min.). f Percentages of dead cells were plotted on a graph over time (minutes) These experiments were duplicated and produced similar results as shown in the Source data. g Source data are provided as a Source Data file.

Fig. 5. Transcriptome landscape of peptide-specific T cells.

Unstimulated PBMCs of patient A2 at each timepoint were stained with HLA-A24:hTERT₄₆₁ tetramer along with an anti-CD8α antibody and 7-AAD. Flow plots are presented with frequencies (%) of tetramer⁺ cells among CD8⁺ cells (a). Unstimulated PBMCs of patient D3 at each timepoint were stained with HLA-A24:AFP₃₅₇ tetramer along with the anti-CD8α antibody and 7-AAD. Flow plots are presented with frequencies (%) of tetramer⁺ cells among CD8⁺ cells (b). To proceed with such small frequencies of tetramer⁺ cells, the 1-year tetramer⁺ cells were enriched and mixed to prepare the 1-year enriched mixture. A 5-year enriched sample was also prepared. These samples were applied to the scRNA-seq analysis pipeline and the obtained transcriptome matrixes are presented in the t-SNE plots with clustering (see Supplementary Figure 3) (c). Single-cell VDJ-seq data were imported into the scRNAseq in accordance with their UMIs. Three hTERT-specific TCR sequences (A2.112, A2.80, and A2.57) and two AFP-specific TCR sequences (D3.14 and D3.16) were retrieved and visualized on the t-SNE plots. Cells that met the criteria of CD3E log2 fold change >0 and CD8A log2 fold change >0 are highlighted in light green (d). Cells that met the criteria of CD3E log2 fold change >0 and CD8A log2 fold change >0 were extracted and underwent the t-SNE dimension reduction followed by the k-means clustering again. Each plot generated three clusters (left and middle). The 1-year data and 5-year data were combined and subject to the t-SNE dimension reduction with the cluster marks held over (right) (e). Peptide-specific T cells were overlayed on the new plots (f). Differentially expressed genes were extracted by comparing the newly generated cluster (CL) 2 with the others of the 1-year t-SNE and then shown on a volcano plot (g). Differentially expressed genes were extracted by comparing newly generated cluster 2 with the others of the 5-year t-SNE and then shown on a volcano plot (h). Source data are provided as a Source Data file.