An Integrated Spatial, Codex, and Genomic Analysis Predicts Responsiveness and Survival in the Phase II Combination of Pembrolizumab and Romidepsin in PTCL

Background Peripheral T-cell lymphomas (PTCLs) are a diverse group of aggressive non-Hodgkin lymphomas characterized by a high relapse rate and poor prognosis in the relapsed/refractory (r/r) setting. Despite advances in treatment, there remains a critical need for new therapeutic strategies. This study investigates the combination of Pembrolizumab, an immune checkpoint inhibitor, and Romidepsin, a histone deacetylase inhibitor, based on the hypothesis that PTCLs harbor mutations in epigenetic modifier genes that may impair immunogenicity and promote immune escape. The combination aims to synergistically prime the immune system, potentially enhancing anti-tumor responses. Here, we present updated results from our phase I/II study (NCT03278782), including survival analysis and integrated spatial and genomic data, to correlate the efficacy of this novel therapeutic approach in r/r PTCL patients. Methods We enrolled 38 patients (pts) between February 2018 and April 2022. Treatment consisted of Pembrolizumab (200 mg, day 1) and Romidepsin (14 mg/m², days 1 and 8) in 21-day cycles, with a maximum of 35 cycles. The primary endpoint was overall response (OR: CR+PR) using Lugano Revised Response Criteria, with secondary endpoints of progression-free survival (PFS), overall survival (OS), and exploratory studies. Correlative studies were performed in 26 pt samples to delineate components of the tumor microenvironment (TME) contributing to primary or acquired resistance to immune checkpoint blockade. These included CODEX analysis of 122 phenotypes using 33 different markers, whole exome sequencing (WES), RNA-seq, targeted next-generation sequencing (NGS), and whole slide image (WSI) analysis. Survival analysis was conducted using R and Python with a p-value threshold of ≤0.05. Results Patient characteristics (n=38) included: median age 67, 57.9% male, 86.8% >60 years, 68.4% with elevated LDH, 63.2% at Stage III/IV. Common ≥grade 3 adverse events were infections (n=11) and thrombocytopenia (n=10). Three patients discontinued due to immune-related adverse events (iRAEs), while two continued after steroid treatment. Response rates varied by PTCL subtype, with best responses for the TfH/AITL subtype showing CR 57.1%, PR 28.6%, ORR 85.7%. At a median follow-up of 40 months (Mo), median OS was 21.32 months (TfH vs. PTCL: 65 vs. 16.6 Mo, p=0.024). OS rates were: 1-year 67.9%, 2-year 48.7%, 3-year 36.7%. Median PFS was 3.6 Mo (TfH vs. PTCL-NOS: NR vs. 1.23 Mo, p=0.013), with PFS rates: 1-year 42.1%, 2-year 36.8%, 3-year 34.0%. CODEX analysis revealed higher levels of cytotoxic T cells (CD3+CD8+), macrophages (CD68+), and proliferating B cell subset (CD20+Ki67+) in responders. Non-responders showed higher levels of CD4+PD1+ T cells, suggesting T cell exhaustion. Spatial analysis showed CTLs were generally closest to Tregs across most lymphoma subtypes, with subtype-specific variations in cellular distances and clustering coefficients. RNA-seq analysis of 10 samples revealed distinct clustering of responders and non-responders in PCA and heatmap analyses of the top 500, 800, and 1000 variable genes. NGS showed heterogeneity in genetic alterations with top genes including TET2, DNMT3A, RHOA, IDH2, PLCG1, CD28, FYN, and STAT3. Finally, WSI analysis (n=27) using machine learning and deep learning techniques identified patch clusters that correlated with outcomes. Conclusions The combination of Romidepsin and Pembrolizumab leads to high response rates and prolonged remissions in r/r TCL, particularly in the TFH subtype. The exploratory analysis revealed a clear separation of responders and non-responders based on WES, CODEX and spatial analysis. WES and RNA-seq show significant genetic and transcriptomic heterogeneity among the samples, with some patterns potentially associated with treatment response. CODEX displayed a more diverse and active immune profile in responders, including higher levels of cytotoxic T cells, B cells, and macrophages, with subtype-specific variations in the TME. Through single-cell analysis, we were able to simultaneously detect various biomarkers in neoplastic cells and the immune microenvironment while considering spatial relationships. Further integration of biomarker for treatment response and resistance mechanisms will be presented at the annual meeting.