AUTHORS
Ivan Odak, PhD, Lennart Riemann, Dr.2, Rudra Prasad Dutta, PhD3, Xinping Xie4, Gvantsa Pantsulaia, MS5, Moah Sohn6, Darwin D'Souza7, Seunghee Kim-Schulze, PhD8, Edgar Gonzalez-Kozlova, PhD9, Kenneth B Hoehn10, Samir Parekh, MD11, Joshua D. Brody, MD12
AFFILIATIONS
1 Icahn School of Medicine at Mount Sinai, New York, New York, United States
2 Medizinische Hochschule Hannover, Hannover, Germany
3 Icahn School of Medicine at Mount Sinai, New York, New York, United States
4 Icahn School of Medicine at Mount Sinai, New York, New York, United States
5 Icahn School of Medicine at Mount Sinai, New York, New York, United States
6 Icahn School of Medicine at Mount Sinai, New York, New York, United States
7 Icahn School of Medicine at Mount Sinai, New York, New York, United States
8 Icahn School of Medicine at Mount Sinai, New York, New York, United States
9 Icahn School of Medicine at Mount Sinai, New York, New York, United States
10 Dartmouth Hitchcock Medical Center, Lebanon, New Hampshire, United States
11 Icahn School of Medicine at Mount Sinai, New York, New York, United States
12 Icahn School of Medicine at Mount Sinai, New York, New York, United States
ABSTRACT
CD19 and CD20 antigen (Ag) escape after CAR-T and bispecific antibody (bsAb) therapy are common causes of relapse in B-cell non-Hodgkin’s lymphomas (B-NHL). In the largest study of Ag escape after bsAb therapy, a notable 88% of relapsing patients with paired biopsies demonstrated CD20 loss [Brooks TR et al., Blood. 2025]. The need for paired biopsies largely precludes a deeper understanding of the mechanisms of Ag escape; a fraction of these Ag escape events occur due to protein-coding mutations, though many occur by unclear mechanisms.
We hypothesized that Ag-/lo (CD19-/lo or CD20-/lo) tumor cells which cause relapse, might be present at low levels prior to therapy, though these would be impossible to evaluate from paraffin-embedded biopsies. We used an exceptional resource- cryopreserved single-cells suspension biopsies to screen a panel of biopsy samples and revealed the presence of rare (0.1% - 2.5%) CD19-/lo and/or CD20-/lo negative cells in all B-NHL biopsy samples prior to any therapy. These cells may be impervious to targeted therapy and drive relapse. When tested in a killing assay with autologous T cells and anti-CD19 (blinatumomab) or CD20 (epcoritamab) bsAbs, only Ag-/lo target cells selectively evaded killing by the respective therapy.
Our next goal was to characterize rare Ag-/lo (CD19–/CD20–) lymphoma cells from biopsies. To enrich these rare cells, we flow-sorted and remixed the Ag-/lo and Ag+ lymphoma cells in near-to-equal ratios prior to downstream scRNAseq/CITEseq analysis. Our analysis confirmed the existence of unique transcriptomic signatures of CD20-/lo lymphoma cells and discrepancies between CD20 protein and its mRNA.
A distinct population of CD20-/lo cells with a plasma cell phenotype was present in every patient in our cohort. B-cell receptor sequencing revealed high clonality and single light chain restriction of the pPCs, confirming that these cells represented members of the malignant lymphoma clone rather than non-malignant plasma cells. Transcriptional program of pPCs largely resembled plasma cell programming, marked by increased IRF4 expression. Indeed, when we knocked out IRF4 from cell lines and primary FL cells, we observed an increase of CD20. Furthermore, we identified SLAMF7 expression on the surface of pPCs, and confirmed its targetability using elotuzumab, revealing a potential therapeutic vulnerability of pPCs. Whilst the data on CD20- relapsing patients following CD3xCD20 bsAb is only now emerging- we analyzed one publicly available dataset and revealed a strong upregulation of SLAMF7, SDC1 and transcription factors involved in PC differentiation in 2/4 CD20- relapsing patients.
However, pPCs made up a wide range (0.1%-77.2%, mean=35.1%) of CD20- cells, implying other mechanisms of CD20 loss. We identified MEF2B as another independent regulator of CD20 expression. MEF2B was downregulated in CD20-, non-pPC lymphoma cells, and CRISPR-mediated MEF2B knockout in lymphoma cell lines and primary patient biopsies decreased CD20 expression. These findings hint that MEF2B supports CD20 expression and that its loss can promote a CD20-negative state independently of the IRF4-driven pPC program.
Altered MS4A1 RNA processing has been described as a potential driver of CD20 loss in B-NHL. While we could find MS4A1 transcript in a fraction of CD20- cells, in 3/12 patients we observed a complete absence of V3 and V4- only alternatively spliced MS4A1 variants that have been demonstrated to efficiently translate to protein.
Together, our results show that CD20 loss in follicular lymphoma is not driven by a single adaptive mechanism. Instead, CD20-/lo malignant cells can exist before therapy through multiple biologically distinct processes, including IRF4-driven pseudo-plasma cell differentiation, reduced MEF2B activity, and alternative MS4A1 splicing. These pre-existing cells are highly clonal, survive CD20-directed bispecific antibody treatment, and become enriched following epcoritamab exposure.
Lastly, we provide a comprehensive overview of pre-therapy hallmarks of Ag-/lo tumor cells. These transcriptomic features may reveal the utility or futility of current combination strategies, e.g. if CD20- FL cells exhibit concurrent loss of multiple B cell antigens as has been observed in B-ALL, then current strategies (e.g. mosunetuzumab plus polatuzumab vedotin) may be ineffective.
DISCLOSURES
Ivan Odak, PhD: No relevant conflicts of interest to declare.