Clinical, genomic and immune microenvironmental determinants of nivolumab response in head and neck squamous cell carcinoma

Abstract

Background: In view of improving biomarkers predicting the efficacy of immunotherapy for head and neck squamous cell carcinoma (R/M HNSCC), this multicenter retrospective study aimed to identify clinical, tumor microenvironmental, and genomic factors that are related to therapeutic response to the anti- Programmed cell death protein 1 (PD-1) antibody, nivolumab, in patients with R/M HNSCC.

Methods: The study compared 53 responders and 47 non-responders, analyzing formalin-fixed paraffin-embedded samples using 14-marker multiplex immunohistochemistry and targeted gene sequencing.

Results: Of 100 patients included, responders had significantly lower smoking and alcohol index, higher incidence of immune related adverse events, and higher PD-1 ligand (PD-L1) expression in immune cells as well as PD-L1 combined positive score (CPS) than non-responders. The frequency of natural killer cells was associated with nivolumab response in patients with prior cetuximab use, but not in cetuximab-naïve status. Age-stratified analysis showed nivolumab response was linked to high CPS and lymphoid-inflamed profiles in patients aged ≥ 65. In contrast, lower NLR in peripheral blood counts was associated with response in patients aged < 65. Notably, TP53 mutation-positive group had lower CPS and T cell densities, suggesting an immune-excluded microenvironment. Patients with altered tumor suppressor gene pathways, including TP53, CDKN2A, and SMAD4 mutations, had lower CPS, higher smoking index, and were associated with poor responses.

Conclusion: Nivolumab treatment efficacy in HNSCC is influenced by a combination of clinical factors, age, prior treatment, immune environmental characteristics, and gene mutation profiles.

Keywords: biomarker; head and neck squamous cell carcinoma; immune profile; mutation; nivolumab.

Figure 1

Clinicopathological characteristics related to nivolumab efficacy Baseline characteristics were stratified by the best overall responses to nivolumab in (A) smoking index (pack/year), (B) alcohol consumption index, (C) human papilloma-virus (HPV) status, and (D) presence of immune-related adverse events (irAEs) (CR, n = 14; PR, n = 39; SD, n = 2; PD, n = 45). Bars and whiskers represent median and interquartile ranges, respectively. Statistical differences were determined using the Kruskal-Wallis test.

Figure 2

Immune characterization based on 14-marker multiplex IHC comparing responders and non-responders for nivolumab in HNSCC. (A) Intratumoral cell densities (cells/mm²) of nine immune cell lineages in each case were quantified using multiplex IHC and image cytometry. A heatmap according to the color scale (upper right) is shown with a dendrogram of unsupervised hierarchical clustering (n = 95). Best overall response (BOR), baseline characteristics, the presence of immune-related adverse events (irAE), prior cetuximab (Cmab) usage, and PD-L1 expression by tumor proportion score (TPS) or combined positive score (CPS) are shown below. Tumor microenvironmental (TME) profiles of lymphoid-, myeloid-, and hypo-inflamed groups were determined by the cut-off values according to the previous study (see Methods). (B, C) PD-L1 expression scores by TPS (B) and CPS (C) were shown, comparing different nivolumab responses. (D) Immune cell densities were quantified, comparing responders and non-responders. (E) Immune profiles comparing subgroups are shown. Statistical differences were determined via the Chi-square test. (F) CD8⁺ T cell densities comparing different nivolumab responses. Bars and boxes/whiskers represent the median and interquartile ranges, respectively in (B–D, F). Statistical differences in (B–D, F) were determined via Kruskal-Wallis tests.

Figure 3

Stratification by prior cetuximab use and age in relation to the nivolumab response. (A, B) Stratification by the history of prior therapy revealed that the frequency of natural killer (NK) cells was associated with a nivolumab response in patients with prior cetuximab use (A) but not in a cetuximab-naïve status (B). (C, D) Neutrophil/lymphocyte ratios (NLR) in peripheral blood counts at nivolumab treatment were significantly lower in responders (mean of 4.96) than those in non-responders (mean of 10.46) in the group aged < 65 (C), but not in the group aged ≥ 65 (D). (E–H) A nivolumab response was significantly associated with high CPS (E) and lymphoid-inflamed profiles based on cell densities of nine immune cell lineages (G) in the group aged ≥ 65, but not in the group aged < 65 (F, H). Bars and whiskers represent the median and interquartile ranges, respectively. Statistical differences were determined via Kruskal-Wallis tests in (A–F) or Chi-square tests in (G, H).

Figure 4

Genomic profiles related to nivolumab outcomes and immune characteristics (A) An oncoplot of the top 27 mutated genes is shown. Alterations of cell signaling pathways of tumor suppressor genes (TSG), phosphatidylinositol-3 kinase (PI3K), NOTCH, and receptor tyrosine kinases (RTK) were curated by representative genetic alterations (see Supplementary Table 3 ). The upper bar plot indicates the number of intergenic somatic variants per patient while the right bar plot shows the number of variants per gene. Response and baseline characteristics are shown below the oncoplot. (B) Responder percentages, and the median values of combined positive score (CPS), total T cell density (cells/mm²), and ratios of CD45⁺ immune cells to tumor cells are shown, stratified by tumor mutation burden (TMB) (high ≥ 10mb, and low < 10mb), TP53 and PIK3CA mutations, and pathway alterations. (C–G) The violin plots present CPS, T cell density, ratios of CD45⁺ immune cells to tumor cells, and smoking index (pack/year), stratified by TP53 mutation status (C, D), and TSG pathway alterations (E, F). The bars represent the median and quartile. (H) Responders (R) and non-responders (NR) were stratified by the presence or absence of TSG pathway alterations in the group with CPS ≥ 20. Statistical differences were determined via Chi-square tests in (B, H) or Kruskal-Wallis tests in (B-G). *p < 0.05, **p < 0.01.