AUTHORS
Carla Guarinos, PhD, Jessica Voetsch, PhD2, Alicia Benítez Rondán3, Matthew E. Stokes4, Cristina Guillén-Mendoza5, Berengere De Moucheron6, Vincent Ribrag, MD7, Franck Morschhauser, phd8, Jennifer Kimberly Lue, M.D.9, Marcel Nijland, MD, PhD10, David Qualls11, Neil Bence, PhD12, Alberto Risueño13, Noelia Purroy, MD, PhD14, Soraya Carrancio, PhD15
AFFILIATIONS
1 Bristol Myers Squibb, Seville, Spain
2 Bristol Myers Squibb, Lawrence Township, New Jersey, United States
3 Bristol Myers Squibb, Seville, Spain
4 Bristol Myers Squibb, Madison, New Jersey, United States
5 Bristol Myers Squibb, Seville, Spain
6 Bristol Myers Squibb, Seville, Spain
7 Gustave Roussy, Villejuif, France
8 Lille University Hospital, Lille, France
9 Memorial Sloan Kettering Cancer Center, New York, New York, United States
10 University Medical Center Groningen, Groningen, Netherlands
11 Dana-Farber Cancer Institute, Boston, Massachusetts, United States
12 Bristol Myers Squibb, San Diego, California, United States
13 Bristol Myers Squibb, Seville, Spain
14 Bristol Myers Squibb, Cambridge, Massachusetts, United States
15 Bristol Myers Squibb, San Diego, California, United States
ABSTRACT
Background
R/R LBCL has poor outcomes despite availability of standard and novel therapies. BCL6, a master regulator of germinal-center biology (GCB), is frequently dysregulated in LBCL and represents a validated oncogenic dependency. BMS-986458 is a first-in-class, oral, cereblon-dependent BCL6 degrader that induces both tumor-intrinsic and immunomodulatory effects (Groocock, et al. ASH 2024). We investigated baseline molecular and tumor microenvironment characteristics associated with clinical activity of BMS-986458 monotherapy in pts enrolled in the dose escalation portion of the phase 1/2 CA123-1000 study (NCT06090539).
Methods
Pts with R/R LBCL who received ≥ 2 prior lines of therapy and treated with BMS-986458 monotherapy (N=31). Baseline FFPE biopsies were profiled by immunohistochemistry (IHC; BCL6 H-score), RNA sequencing (BCL6 and CD20 expression, COO [gneSeqCOO], DZsig [dark zone signature; BCCA DLBCL90], tumor microenvironment [TME] deconvolution using Tumor Microenvironment Signature [TME26] scoring, single cell tumor-intrinsic DLBCL archetypes), and whole-genome sequencing; circulating tumor DNA (ctDNA) was quantified by CAPP-seq. Responders were defined as those achieving complete (CMR) or partial metabolic response (PMR). Gene-expression counts were voom-normalized; cell-type proportions were estimated with ADAPTs (DLBCL-tailored signature matrix); signature enrichment was assessed by ssGSEA; between-group comparisons used Wilcoxon tests with Benjamini-Hochberg correction.
Results
At the 5 June 2026 cutoff, 31 pts (23 with DLBCL, 8 with HGBL) were evaluable. Pts were heavily pretreated, with a median of 4 prior lines of therapy (range 2–13); 84% had received prior CAR-T therapy, 77% prior bi-/tri-specific antibodies, and 58% were refractory to their most recent line. Objective response rate (ORR) was 53%, including 3/30 CMR (10%) and 13/30 PMR (43%). Clinical activity was observed in multiple high-risk subgroups, including: transformed LBCL (87.5%, 7/8), HGBL (57.1%, 4/7), and previously treated with CAR-T cells (87.5%, 14/16), bi-/tri-specific antibodies (68.8%, 11/16), and patients exposed to both CAR-T and bi-/tri-specifics (56.3%, 9/16). Response rates were similar regardless of prior exposure to T-cell redirecting therapies.
Notably, baseline BCL6 expression was not associated with clinical outcome. Responses were seen regardless of BCL6 expression in tumor B cells measured by either IHC or RNAseq, including in BCL6-negative tumors (defined as <30% nuclear staining by IHC; ORR 66.7% in BCL6-negative vs 46.7% in BCL6-positive). Protein and RNA BCL6 expression were correlated (Spearman R=0.65, p=2.2×10-5). Similarly, no significant difference in baseline CD20 expression by RNAseq was observed between responders and non-responders, p=0.72).
RNAseq analyses demonstrated activity across molecularly diverse disease subsets. ORR was independent of COO, with responses in GCB, ABC, and unclassified DLBCL (69.2%, 50%, and 66.7%, respectively; p=0.83). A response was seen in the single DZsig-positive pt, supporting potential activity in biologically high-risk disease. Responses occurred in tumors regardless of TME status (69% vs 50% in high vs low TME26 scores; p=0.412), suggesting activity despite biologic features frequently associated with resistance to T-cell-dependent therapies.
Reductions in ctDNA were associated with clinical response. GCB pts showed a deeper ctDNA reduction than ABC pts at Cycle 2 Day 1 (0.7-fold greater; p=0.11), although this difference was no longer evident by Cycle 3 Day 1 (p=1.0). Baseline ctDNA spanned a wide range in both responders and non-responders (median 1.8 vs 2.0 log(MMPM); p=0.54), suggesting that molecular tumor burden did not predict response.
No significant associations were observed between clinical outcome and genomic alterations, malignant-cell abundance, or DLBCL archetype; archetypes segregated according to COO but did not correlate with outcome.
Conclusions
BMS-986458 demonstrated clinical activity in heavily pretreated R/R LBCL, including multiple high-risk subgroups such as HGBL and pts previously exposed to CAR-T and T cell engaging therapies. Importantly, responses were observed regardless of baseline BCL6 expression and across the molecularly diverse subsets analyzed. These findings support BCL6 degradation as a broadly active therapeutic agent in LBCL.
DISCLOSURES
Carla Guarinos, PhD: BMS - Current equity holder in private company; BMS - Current Employment