Evolution of the tumor immune landscape during treatment with tebentafusp, a T cell receptor-CD3 bispecific

Abstract

Metastatic uveal melanoma is an aggressive disease with poor outcome, which is refractory to immune checkpoint inhibitors. A T cell receptor (TCR)-based CD3 bispecific, tebentafusp, delivers clinical benefit in patients with metastatic uveal melanoma. Understanding the molecular basis for the anti-tumor activity of tebentafusp in an indication where checkpoint inhibitors are ineffective could aid in identification of other solid tumor indications where CD3 bispecifics may serve an unmet need. By analyzing tumor biopsies taken prior to treatment, early on-treatment, and at progression (NCT02570308), using RNA sequencing (RNA-seq) and immunohistochemistry (IHC), we show that expression of interferon-related genes in the tumor prior to treatment is associated with improved overall survival and tumor reduction on tebentafusp, that T cell recruitment occurs even in tumors with a low baseline level of T cell infiltration, and that durability of changes induced in the tumor microenvironment is key for survival duration.

Keywords: T cell engager; bispecific; gp100; immunotherapy; tumor immunology; tumor microenvironment; uveal melanoma.

Graphical abstract

Figure 1

Immune status of tumors prior to treatment with tebentafusp Tumor biopsies collected prior to treatment were analyzed by IHC. (A) Representative images from tumors categorized as deserted, excluded, and inflamed, stained for gp100 (red), CD3 (purple), and CD8 (purple). 100 μm scale bar shown. (B) Representative images of B cells within (circled) and in absence of (arrows) lymphoid aggregates (B cell, CD20, purple; T cell, CD3, yellow). 100 μm scale bar shown. (C) Plot of abundance of CD20⁺ cells in tumor stratified by presence/absence of lymphoid aggregates (21-fold difference, p = 1.4E−10). Median and interquartile range are indicated; n = 70. (D) Plot of abundance of CD3⁺ cells in tumor and peritumoral stroma, classifying T cell infiltration status as deserted, excluded, or inflamed; n = 146. (E) Kaplan-Meier plot of OS of patients stratified by T cell infiltration status at baseline. No significant difference between groups (p = 0.904, Cox regression); n = 96 inflamed, 12 excluded, 9 deserted. (F) Incidence of tumor reduction (sum of lesion diameters [SLD] at any on-treatment time point below pre-treatment SLD) stratified by T cell infiltration status. No significant difference (p = 0.802, Fisher’s exact test); n = 85 inflamed, 11 excluded, 8 deserted. (G) Boxplot of gp100 protein level in tumor cells, stratified by T cell infiltration status. Median and interquartile range are indicated.

Figure 2

Baseline expression of genes related to antigen presentation and interferon signaling is associated with outcome Gene expression in baseline tumor biopsies (n = 71) was analyzed by RNA-seq. (A) Gene Ontology and Reactome pathway analysis of genes associated with longer OS. (B and C) Kaplan-Meier plot of OS (HR = 0.3 [95% CI 0.18–0.56], p < 0.001) and (C) waterfall plot showing best reduction in SLD (OR = 0.13 [95% CI 0.03–0.47], p = 0.001), both stratified at median tumor UBA7 expression at baseline. (D) Kaplan-Meier plots of OS stratified by median expression in baseline tumor biopsy of CXCL9 (HR = 0.45 [95% CI 0.27–0.75], p = 0.002) and GBP1 (HR = 0.42 [95% CI 0.24–0.71], p = 0.001). (E) Kaplan-Meier plots of OS data from TCGA-UVM, stratified by median expression of CXCL9 (HR = 5.85 [95% CI 2.15–15.91], p < 0.0001) and GBP1 (HR = 4.38 [95% CI 1.62–11.82], p = 0.001). (F) Forest plot showing OS HR values and 95% confidence intervals for a set of genes reported to be associated with improved survival on CPIs. HLA-DRB4 not shown as expression level was below threshold for inclusion. (G) Kaplan-Meier plot of OS stratified by median PRAME gene expression in baseline tumor biopsy (HR = 2.0 [95% CI 1.2–3.4], p = 0.007).

Figure 3

Tebentafusp induces significant tumor immune infiltration even in tumors with low initial T cell infiltration Paired tumor biopsies taken at baseline and day 16 of tebentafusp treatment were analyzed by immunohistochemistry for CD3 (n = 57 pairs) and CD8 (n = 56 pairs). (A) Boxplot showing abundance of CD3⁺ and CD8⁺ cells in tumor regions. Median and interquartile range are indicated. Increase at day 16 2.9-fold for CD3 (p < 0.0001) and 2.3-fold for CD8 (p < 0.0001). (B) Boxplot showing abundance of CD3⁺ cells in tumor by baseline T cell infiltration status (see Figure 1B). Median and interquartile range are indicated. Increase at day 16 16.9-fold for deserted (ns, p = 0.25), 6.8 for excluded (p = 0.016), and 2.5-fold for inflamed (p < 0.001). (C) Representative images showing CD3 (purple) at baseline and day 16, from patients with deserted, excluded, and inflamed tumors. Brown pigmentation is melanin. 50 μm scale bar shown.

Figure 4

Tebentafusp induces T cell activation at day 16 Tumor biopsies taken at day 16 of treatment were analyzed for gene expression by RNA-seq. Differential gene expression analysis was performed against paired baseline biopsies (n = 35 pairs). (A) Volcano plot indicating change in gene expression in tumor from baseline to day 16. Differentially expressed genes are shown in yellow. (B) GSEA plot of genes ranked by increase in expression in between baseline and day 16, using interferon alpha/beta signaling and interferon gamma signaling gene sets. (C) Fold change in expression of selected immune-related genes between baseline and day 16. Data are represented as median and standard error. (D) Boxplot of IL7R expression in tumor biopsy at baseline and day 16 (1.7-fold increase, p = 0.003). Median and interquartile range are indicated. Lines join paired biopsies. (E) Waterfall plot showing maximum tumor reduction stratified at median of expression of IL7R in tumor at day 16 (OR = 0.06 [95% CI 0–0.61], p = 0.007). (F) Kaplan-Meier plots of OS stratified at median of expression of IL7R in tumor at day 16 (HR = 0.47 [95% CI 0.23–0.98], p = 0.04).

Figure 5

Target expression is associated with greater T cell infiltration and activation on tebentafusp (A) Box plot of number of cells in tumor regions of paired biopsies expressing CD3 and CD8 (by IHC) at baseline and at day 16, stratified at lower quartile of baseline gp100 expression. Median and interquartile range are indicated. n = 50 pairs. (B) Volcano plots showing genes upregulated at day 16 compared to baseline in paired biopsies, stratified at lower quartile of baseline gp100 expression. n = 35 pairs. (C) Kaplan-Meier plot of OS stratified at lower quartile of baseline gp100 expression (HR = 0.79 [95% CI 0.49–1.29], not significant; p = 0.35).

Figure 6

At progression, higher expression of antigen presentation genes and T cell infiltration are associated with longer survival Tumor biopsies collected after radiological progression were analyzed by IHC (N = 18, except for CD20 where n = 17) and RNA-seq (n = 14). (A) Box plot of expression of APM gene signature in tumor biopsies at progression, stratified by duration of OS (5.6-fold higher in long OS, p = 0.014). Median and interquartile range are indicated. (B) Representative images of progression biopsies from short OS and long OS patients, stained for (left) HLA-A (purple) and melanoma triple (gp100, MART1, tyrosinase) (yellow), (middle) B2M (purple) and CD3 (yellow), and (right) CD20 (purple) and CD3 (yellow). 100 μm scale bar shown. (C) Kaplan-Meier plots of OS stratified by tumor cell membrane HLA-A staining (HR = 0.3, p = 0.038, 95% CI 0.075–1), by CD3 IHC (HR = 0.29, p = 0.025, 95% CI 0.09–0.91), and by CD20 IHC (HR = 0.26, p = 0.018, 95% CI 0.08–0.85). (D) Kaplan-Meier plot of OS stratified at median of PRF1 gene expression (HR = 0, p < 0.001, 95% CI undefined), GZMB1 gene expression (HR = 0.194, p = 0.015, 95% CI 0.05–0.81), mean expression of CD3D, CD3E, and CD3G (HR = 0, p < 0.001, 95% CI NA), and PMEL (gp100) gene expression (HR = 0.78, p = 0.69, 95% CI 0.24–2.6).