A Novel Prognostic Model Based on Tumor Microenvironment Biology in Relapse Biopsies Predicts Post-Autologous Stem Cell Transplantation Outcomes in Classical Hodgkin Lymphoma

Abstract Introduction: Disease progression remains a significant clinical burden in classical Hodgkin lymphoma (CHL) with 25-30% of patients relapsing after first-line treatment. The current standard of care for relapsed CHL is high dose chemotherapy followed by autologous stem cell transplantation (ASCT). This secondary therapy only cures approximately 50% of patients, with virtually no reliable biomarkers to identify the patients in which this salvage treatment regimen fails. The specific aim of this study was to establish the extent of changes in tumor microenvironment (TME) composition between matched primary and relapse samples, and to build and validate a prognostic model for post-ASCT outcomes using relapse samples. Materials & Methods:NanoString digital gene expression profiling was used to ascertain the gene expression of 784 genes of interest from 245 biopsies sampled from 174 CHL patients. This cohort included 90 patients with single biopsies performed at first diagnosis (primary), 13 patients with single biopsies taken at relapse, and 71 patients with paired biopsies taken at first diagnosis and at relapse. All patients received ABVD as first-line treatment, and 151 patients went on to receive ASCT. The 784 genes of interest were selected based on previously reported associations with outcome in CHL and/or components of the TME. Spearman statistics were used for pairwise correlations and log-rank tests were used to assess survival differences. Bootstrap aggregation with concordance statistics (C-stat) was used to calculate the post-ASCT prognostic properties and a two-sample t-test was used to compare primary and relapse samples. Penalized elastic-net multivariate cox regression was used for model construction. An independent, similarly treated, cohort of 31 relapse biopsies was used for model validation. Results: Comparative gene expression analysis revealed that 17 of the 71 patients (24%) exhibited poor correlations between their paired primary and relapse samples (r2 < 0.75) - indicative of significant differences in their TME compositions. Amongst these differences was a striking inverse correlation between macrophage and B-cell gene expression pattern changes (r2 = -0.809). We validated these findings by using CD20 and CD163 immunohistochemistry confirming this inverse correlation of relative changes in macrophage and B cell content in the TME (r2= -0.645). Patients who exhibited poor gene expression correlations had inferior post-ASCT failure-free survival (FFS) (3-year: 38.5%) compared to patients with high gene expression correlations (3-year: 77%; P = 0.005). A comparative C-stat analysis of prognostic properties between primary and relapse samples demonstrated that relapse samples contain superior prognostic features for prediction of post-ASCT outcomes (relapse C-stat 0.785 ± 0.073 vs. primary C-stat 0.594 ± 0.079, P < 0.001). To this end, we developed a gene expression prognostic model using penalized Cox regression that was based on gene expression measurements in relapse samples (RHL30). RHL30 was able to risk stratify patients according to post-ASCT outcomes in the training cohort (5-year post-ASCT FFS: 23.8% in high-risk vs. 77.5% in low-risk; 5-year post-ASCT overall survival [OS]: 30.9% in high-risk vs. 85.4% in low-risk). This was validated in an independent cohort of 31 patients with relapsed CHL with a 5-year post-ASCT FFS of 37.5% in the high-risk vs. 70.1% in the low-risk groups (P = 0.017) and 5-year post-ASCT OS of 37.5% in the high-risk vs. 71.6% in the low-risk groups (P = 0.006). Conclusions: The TME gene expression profile differs significantly between matched primary and relapse CHL samples in a subset of patients with relapsed CHL. Gene expression measurements derived from relapse samples contain superior predictive properties for response to ASCT. To this end, we have developed a novel clinically applicable prognostic model (RHL30), derived from relapse samples, that identifies patients who have a high likelihood to benefit from ASCT (low-risk), and conversely a subgroup of patients who may benefit from additional or alternative therapeutic approaches (high-risk). Disclosures Connors: Seattle Genetics: Research Funding; Bristol Myers Squib: Research Funding; Millennium Takeda: Research Funding; NanoString Technologies: Research Funding; F Hoffmann-La Roche: Research Funding. Scott:NanoString Technologies: Patents & Royalties: named inventor on a patent for molecular subtyping of DLBCL that has been licensed to NanoString Technologies.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.