PD-1 Status in CD8+ T Cells Associates with Survival and Anti-PD-1 Therapeutic Outcomes in Head and Neck Cancer

Abstract

Improved understanding of expression of immune checkpoint receptors (ICR) on tumor-infiltrating lymphocytes (TIL) may facilitate more effective immunotherapy in head and neck cancer (HNC) patients. A higher frequency of PD-1⁺ TIL has been reported in human papillomavirus (HPV)⁺ HNC patients, despite the role of PD-1 in T-cell exhaustion. This discordance led us to hypothesize that the extent of PD-1 expression more accurately defines T-cell function and prognostic impact, because PD-1^high T cells may be more exhausted than PD-1^low T cells and may influence clinical outcome and response to anti-PD-1 immunotherapy. In this study, PD-1 expression was indeed upregulated on HNC patient TIL, and the frequency of these PD-1⁺ TIL was higher in HPV⁺ patients (P = 0.006), who nonetheless experienced significantly better clinical outcome. However, PD-1^high CD8⁺ TILs were more frequent in HPV^- patients and represented a more dysfunctional subset with compromised IFN-γ secretion. Moreover, HNC patients with higher frequencies of PD-1^high CD8⁺ TIL showed significantly worse disease-free survival and higher hazard ratio for recurrence (P < 0.001), while higher fractions of PD-1^low T cells associated with HPV positivity and better outcome. In a murine HPV⁺ HNC model, anti-PD-1 mAb therapy differentially modulated PD-1^high/low populations, and tumor rejection associated with loss of dysfunctional PD-1^high CD8⁺ T cells and a significant increase in PD-1^low TIL. Thus, the extent of PD-1 expression on CD8⁺ TIL provides a potential biomarker for anti-PD-1-based immunotherapy. Cancer Res; 77(22); 6353-64. ©2017 AACR.

Fig. 1. Expression of PD-1 and CD8A, CD8B is significantly higher in HPV⁺ than in HPV⁻ tumor specimens

(A) QRT-PCR from clinical HNC tumor samples (n=40) was performed for several ICR: PD-1, CTLA-4, LAG-3, TIM-3, BTLA and the Treg transcription factor FOXP3. As a reference, the fold change of average expression of each ICR from healthy donor pharynx mucosa tissues (n=4) was utilized. Immune checkpoint receptors show highly variable expression for HPV⁺ and HPV⁻ tumors in qPCR in relation to reference tissue. (B–D) Additionally, ICR mRNA expression data retrieved from TCGA was analyzed and compared to healthy donor mucosa. Expression of all ICRs and CD8A, CD8B and CD4 were higher in tumors as compared to normal TCGA reference mucosa. (E–F) The analysis of the TCGA database revealed that PD-1 was the only ICR that was significantly higher in HPV ⁺ tumors along with CD8 T cells markers: CD8A and CD8B. All ICRs highly correlated with CD8A expression regardless of HPV status. However, HPV⁺ tumors showed a higher correlation coefficient than HPV⁻ ones for all ICRs (G). PD-1 expression highly correlated with that of CD8A in both HPV⁺ and HPV⁻ tumors, CTLA-4 being the other ICR with a slightly higher correlation coefficient in HPV⁺ tumors (G, H).

Fig. 2. PD-1 is predominantly expressed on antigen experienced peripheral T_EM cells in both HPV⁺ and HPV⁻ HNC patients

(A) PBL from HNC patients were analyzed for co-expression of PD-1 and CTLA-4 and phenotypic markers: effector memory (T_EM: CCR7⁻CD45RA⁻) naive (CCR7⁺CD45RA⁺), central memory (T_CM: CCR7⁺CD45RA⁻), and terminal effectors (T_EMRA: CCR7⁻CD45RA⁺). (B) Comparison of phenotypic proportions between HPV⁺ and HPV⁻ T cells (one-way ANOVA, Bonferroni’s mult. comparison test, HPV⁺ n = 11, HPV⁻ n = 15). (C) Percentage of PD-1⁺ T cells were measured by flow cytometry depending on phenotype and HPV status (one-way ANOVA, Bonferroni’s multiple comparison test, HPV⁺ n = 11, HPV⁻ n = 15). (D) Comparison of single positive (PD-1 or CTLA-4) and double positive T cells in PBL (one-way ANOVA, Bonferroni’s multiple comparison test, HPV⁺ n = 20, HPV⁻ n = 22).

Fig. 3. HPV⁺ HNSCC patient TIL demonstrate higher overall expression of PD-1

The frequency of CTLA-4⁺and PD-1⁺ cells was compared by flow cytometry in CD8⁺ TIL and PBL from HPV⁺ and HPV⁻ patients. (A) Representative gating and stratification and sorting of TIL by extent of PD-1 level of expression. (B) Comparison of CTLA-4⁺ and PD-1⁺ in PBMC and TIL of HPV⁺ and HPV⁻ (matched-pair Wilcoxon test for comparison of TIL vs. PBL, Mann-Whitney test for comparison HPV⁺ vs. HPV⁻ TIL, HPV⁺ n = 10, HPV⁻ n = 16).

Fig. 4. HPV⁻ tumors contain significantly more PD-1^high CD8⁺ TIL that manifest an exhaustion phenotype and produce little IFN-γ

(A) The range from highest detected PD-1 to the negative control (PD-1^negative) was divided into the three subgroups with low (PD-1^low), intermediate (PD-1^int) and high (PD-1^high) PD-1 expressing cells. (B) Flow cytometric staining for CD3⁺, CD8⁺, and granzyme B on representative TIL (n=3, two HPV⁺ and one HPV⁻), gated on PD-1^negative, PD-1^low, PD-1^int and PD-1^high (n = 3, one-way ANOVA, p = 0.03). (C) ELISPOT analysis for IFN-γ secretion per 100,000 cells of CD8⁺ TIL (n=3, two HPV⁺ and one HPV⁻, distinct tumor samples from B), sorted into PD-1^negative, PD-1^low+int and PD-1^high expressing subsets (n = 3, one-way ANOVA p = 0.01). (D, E) Box and Whiskers plots for PD-1^high (D) and PD-1^low (E) fractions of HPV⁺ and HPV⁻ patients (Mann-Whitney comparison, HPV⁺ n = 20, HPV⁻ n = 36).

Fig. 5. PD-1^high TIL are associated with worse DFS whereas high levels of PD-1^low TIL portend better clinical outcome

Kaplan-Meier curves portray DFS for HNC patients (HPV⁺ n=20, HPV⁻ n=36) in relation to PD-1^high vs. PD-1^low fractions of CD8⁺ TIL. Tertiles for PD-1^high fractions were distinguished and survival plotted as shown (A, ranges of lower tertile 0–1.75; middle tertile 1.75–8.7; upper tertile 8.7–51). Cox proportional hazards models were used to investigate the relative risk and models checked for linearity and adequacy of the proportional hazards assumptions. (B) Hazard ratio = 2.25 (95% CI = 1.46 – 3.15), p < .0001. Tertiles for PD-1^low fractions were distinguished and survival plotted as shown (C, ranges of lower tertile 9.3–49.5; middle tertile 49.5–78.5; upper tertile 78.5–99.4). (D) Hazard ratio = 0.19 (95% CI = .07 – .49, p < .0006).

Fig. 6. Dysfunctional PD-1^high CD8⁺ T cells are dramatically reduced in the treatment response group to anti-PD-1 plus RT targeted therapy

(A) 33 mice with positive tumor growth were randomized into two treatment groups (n= 16 for isotype + RT, and n=17 for anti-PD-1 Ab + RT). Treatment response was analyzed by measurement of the tumor volume on different days. A comparison of radiotherapy alone (isotype) and radiotherapy in combination with anti-PD-1 Ab treatment was performed. By modeling the fixed and random coefficients of a polynomial regression model, we conclude that the slopes between two treatment groups differ and that RT + anti-PD-1 Ab treated tumors grew more slowly than RT + isotype treated tumors. Data represent observed mean with bootstrap 95% CI (p = .0039). Anti-PD-1 Ab or isotype control (3mg/kg body weight) was administered at days 1, 4, 11, 15; radiotherapy was fractionated in 2 Gy ×10 days (on day 4,5,6,7,8,11,12,13,14,15). (B, C) PD-1 expression of isolated splenic CD3⁺ PD-1⁺ lymphocytes, PD-1^high (B) and PD-1^int/low fractions (C) of RT + isotype vs. RT + anti-PD-1 Ab were compared with non-parametric Mann-Whitney test, data are representative of +/− SEM.