Sex-specific cytokine signatures as predictors of anti-PD1 therapy response in non-small cell lung cancer

Abstract

Background: The introduction of immune checkpoint inhibitors (ICI) as first-line therapy in the treatment of non-small cell lung cancer has dramatically improved response rates. However, more than half of NSCLC patients receiving ICI fail to have a durable response to treatment and therefore the identification of circulating biomarkers to improve patient stratification is required. Cytokines and chemokines are critical mediators of immune responses, affecting tumor progression and immune evasion mechanisms. Thus, profiling circulating cytokines is particularly important, as these signaling molecules may provide valuable insights into predicting response and resistance to ICI.

Methods: Twenty-four circulating chemokines and cytokines were profiled in NSCLC patient plasma collected either prior to treatment or while on-treatment with anti-PD1 therapy and correlated to treatment response as well as to progression-free survival (PFS) and overall survival (OS). Sex-disparities in correlations of cytokines to response and survival was analyzed.

Results: Regardless of sex, baseline levels of CCL5/RANTES were associated with anti-PD1 treatment response, while CXCL5 was associated with response in males and CXCL10 was elevated in female responders to anti-PD1 treatment. VEGF and CD40L were associated with short PFS and OS, while CCL5 and CXCL5 were correlated to longer PFS and OS. Sex disparities in baseline cytokine levels were also observed. CCL5 was significantly correlated to PFS and OS in females but not males, and CXCL10 was found to be predictive of longer OS in females only. VEGF was found to be a better predictor of response t to anti-PD1 in females, while CXCL12 was found to be associated with short PFS and OS in males but not females. Uniform Manifold Approximation and Projection (UMAP) dimension reduction method and k-means clustering analysis identified a cluster of male patients with short PFS characterized by elevated baseline levels of VEGF, CCL4, CCL5, CCL20, and CXCL2.

Conclusions: Plasma cytokine levels can be useful biomarkers for predicting response to anti-PD1 therapy in NSCLC patients. However, the data presented in this study demonstrate that sex needs to be considered as an important variable in biomarker studies in immuno-oncology due to sex disparities in correlations of cytokines to anti-PD1 treatment response.

Keywords: CXCL10; CXCL12; NSCLC; chemokine; cytokine; immune checkpoint inhibitor; sex disparity.

Figure 1

(A) Table describing the clinicopathological characteristics of the study. Kaplan-meier plots demonstrating the correlation of PFS with clinicopathological characteristics including (B) PDL1 tumor proportion score (TPS) <50% versus >50, (C) treatment with either pembrolizumab monotherapy (Pembro) versus in combination with chemotherapy (Pembro + Chemo), and (D) male versus female. Logrank p values are shown. PDL1, Programmed death-ligand 1; TPS, tumor proportion score; PR, partial response.

Figure 2

Baseline expression levels of (A) CCL5, (B) CXCL5, and (C) CXCL10 in responders and non-responders from the entire patient cohort, in male versus female patients, in male responders versus male non-responders, and in female responders versus female non-responders. The table shows means ± SEM and p values (Mann-Whitney test; *p<0.05). Resp., responders; Non-Resp., non-responders; n.s., not significant.

Figure 3

(A) Expression of plasma cytokines post-treatment with anti-PD1 therapy shown as fold change over baseline. The table shows means ± SEM and p values (Mann-Whitney test; * p<0.05). Grouped dot plots showing fold change in expression for (B) CXCL10 and (C) IL-6. Resp., responders; Non-Resp., non-responders.

Figure 4

Forest plots showing (A) univariate Cox regression analysis and (B) multivariate cox regression analysis, adjusted for age and sex, for progression-free survival. Kaplan-meier plots showing percent progression-free survival for (C) VEGF and (D) CXCL5 are shown. Logrank p values are shown. *p<0.05, **p<0.01.

Figure 5

Forest plots showing univariate Cox proportional hazard analysis for males and females for both PFS and OS for (A) CCL5, (B) VEGF, (C) CXCL10, and (D) CXCL12. *p<0.05, **p<0.01, ***p<0.001.

Figure 6

Kaplan-meier plots demonstration the percentage of PFS and OS for both male and female patients for (A) CCL5, (B) VEGF, (C) CXCL10, and (D) CXCL12. Logrank p values are shown. *p<0.05, **p<0.01.

Figure 7

(A) UMAP of clustering of all patients in the cohort was used to identify 3 K-clusters of that were then correlated to progression-free survival. (B) number of male and female patients at risk in each cluster with median PFS indicated. Separate UMAP clustering (2 k-clusters) was performed for both males (C) and females (D). (E) Normalized mean expression of most important principal components (CCL20, CCL4, CCL5, CXL2, IL-6 and VEGF) are shown for male patients. M, male; F, female; norm., normalized (normalized to reduce batch variability between assays). *p<0.05, **p<0.01, ****p<0.0001.