Chemoimmunotherapy‐associated CD20 (MS4A1) resistance mutations in Waldenström macroglobulinemia: Epitope disruption independent of expression with implications for CD20‐directed cellular therapies

CD20, encoded by MS4A1, is one of the earliest and most extensively used immunotherapeutic targets in B‐cell malignancies. In Waldenström macroglobulinemia (WM), rituximab administered alone or with chemotherapy or BTK inhibitors is a mainstay of frontline therapy. ¹ Acquired resistance to rituximab is common, though the mechanisms underlying resistance remain poorly understood. This knowledge gap is particularly relevant as novel CD20 therapies (e.g., T‐cell engagers [TCE], chimeric antigen receptor [CAR] T‐cell therapy) are entering clinical trials for previously treated WM patients. ² , ³ , ⁴ , ⁵ , ⁶ We therefore performed a large multi‐omic analysis to characterize MS4A1 alterations in previously treated WM.

We performed whole‐exome sequencing (WES) in CD19+ sorted bone marrow mononuclear cells with matched CD19‐depleted peripheral blood mononuclear cells as germline control in 247 previously treated patients with WM. MS4A1 mutations (MS4A1 ^MUT ) were structurally mapped onto the CD20 crystal structure (PDB: 6Y90). Functional impact was evaluated using in silico prediction tools with predefined deleteriousness thresholds for point mutations (CADD > 20) and splice‐site mutations (Splice AI > 0.7). RNA sequencing with rRNA depletion was performed to assess CD20 transcript levels, isoforms, and mutation expression. Flow cytometry (L27 clone) and CD20 immunohistochemistry (IHC) (L26 clone) were used to assess CD20 expression and localization (Supporting Information Methods).

We restricted our analysis to 164 patients who had a >5% MYD88 ^L265P variant allele fraction (VAF) to ensure adequate tumor content (median VAF, 36%; range, 5.4%–98%). The median age for these patients was 65 (37–89 years) and 61% were male. The median time from diagnosis and first therapy to study sample was 6.2 (95% CI 4.3–7.1) and 4.4 (95% CI 2.7–5.8) years, respectively. Patients received a median of 1.5 prior lines (range 1–9); 63 (38%) received CD20‐containing CIT, 81 (49%) anti‐CD20 without chemotherapy, and 20 (12%) no prior anti‐CD20 therapy. At the time of study biopsy, 95/164 (58%) patients were progressing. A total of 9 (5%) patients harbored somatic MS4A1 ^MUT (Supporting Information S1: Table 1).

MS4A1 ^MUT was observed exclusively in patients who were previously exposed to CD20‐containing CIT (9/63; 14%) and was absent among those treated with anti‐CD20 antibody without chemotherapy (0/81) or never exposed to an anti‐CD20 antibody (0/20; P < 0.001 for three‐way comparison; Figure 1A). However, compared with patients treated with anti‐CD20 without chemotherapy, those receiving CIT had more prior therapy lines (median, 3 vs. 1; P < 0.001) and were more often progressing at biopsy (73% vs. 48%; P = 0.008) (Supporting Information S1: Table 2).

Figure 1.

Clinical, structural, and functional characterization of MS4A1 mutations. (A) Distribution of MS4A1 mutations according to prior therapy exposure, showing restriction to patients treated with CD20‐containing chemoimmunotherapy (CIT). (B) Mapping of MS4A1 mutations onto the CD20 crystal structure (PDB: 6Y90) with residue number on the x‐axis and average B‐factor (Ų) on the y‐axis, reflecting local structural flexibility; mutations are color‐coded by type (truncating, missense/in‐frame, splice‐site), and the rituximab‐binding epitope in the second extracellular loop (ECL2) is highlighted in green. Truncating variants cluster within transmembrane domains upstream of this region (in black). (C) Frequency of CD20‐negative clonal B‐cell populations by flow cytometry (L27 clone) according to MS4A1 status. (D) Distribution of CD20‐positive cells (total bone marrow) by flow cytometry demonstrating reduced surface expression in MS4A1‐mutated cases. (E) CD20 immunohistochemistry (L26 clone) showing heterogeneous membranous expression and aberrant intracellular localization. CD20 expression by immunohistochemistry tracked with MS4A1 clonality (MS4A1/MYD88 VAF), with fully clonal alterations associated with loss of membranous staining, subclonal alterations with partial expression, and low‐ratio alterations with preserved expression.

All nine patients with MS4A1 ^MUT progressed after anti‐CD20‐containing CIT. They had more prior therapy lines (median, 3 vs. 1; P = 0.04), more anti‐CD20‐containing therapy lines (3 vs. 1; P < 0.001), and more frequent TP53 alterations (33% vs. 10%; P = 0.04) than patients without MS4A1 mutations (MS4A1 ^WT ).

A sensitivity analysis restricted to patients with MYD88 ^L265P VAF > 10% (n = 140), among whom all MS4A1 ^MUT cases were included, yielded identical results. MS4A1 ^MUT remained associated with prior CD20‐containing CIT (9/54; 17%; P < 0.001) and with a greater number of prior anti‐CD20‐containing regimens (median, 3 vs. 1; P < 0.001).

A total of 20 MS4A1 ^MUT with a median of 2 (range 1–4) per patient were detected by WES among the nine patients carrying these variants, with only one recurrent variant (p.Lys50Ter) observed in two patients. Seven of nine (78%) patients harbored truncating mutations within transmembrane domains 1 (TM1) and 3 (TM3), which were upstream of the rituximab epitope in extracellular‐loop‐2 (ECL2; amino acids 168‐175), and predicted to eliminate rituximab's binding site (Figure 1B). As for the remaining 2 patients, one (#4) harbored two TM1 variants within a structurally stable α‐helix (B‐factor <50 Ų), while the other (#7) carried a splice‐site mutation, all predicted to be deleterious (Table 1).

Table 1.

Clinicogenomic characteristics of patients with CD20 (MS4A1) mutations.

Patient	Prior therapies	DNA change	Protein change	Mutation type	Location	CD20 DNA VAF	CD20 RNA VAF	MYD88 DNA VAF	Clonality (CD20/MYD88)*	Flow CD20 (Clone)	Flow CD20 (Total BMBx)	IHC CD20 (Membrane)
Patient 1	1 L Chlorambucil; 2 L Rituximab; 3 L Fludarabine rituximab	c.197dup	p.L69Sfs*72	FS*	TM1	0.021	0.01	0.432	5.6%	Negative	3%	—
		c.385_424del	p.S129Kfs*6	FS*	TM3	0.008	Not seen
		c.552C>A	p.Y184*	NS	ECL2	0.01	0.006
		c.573+1G>C	p.X191_splice	Splice (Splice AI 0.85)	Intron (E6‐7)	0.024	0.222
Patient 2	1 L Rituximab; 2 L Bendamustine rituximab	c.147dup	p.K50*	FS*	TM1	0.045	0.023	0.269	16.7%	Negative	0%	—
Patient 3	1 L Cladribine; 2 L Thalidomide rituximab; 3 L imatinib; 4 L Cyclophosphamide prednisone rituximab	c.185T>C	p.F62S	Missense (CADD 27.5)	TM1	0.036	0.019	0.226	25.2%	Positive	3%	Subclonal (50%)
		c.209T>C	p.L70P	Missense (CADD 26.7)	TM1	0.057	0.039
		c.362C>A	p.S121*	NS*	TM3	0.017	0.004
		c.367_369dup	p.S123dup	INS (in frame)	TM3	0.053	0.043
Patient 4	1 L Cyclophosphamide; 2 L Rituximab; 3 L Sildenafil; 4 L Imatinib; 5 L Everolimus; 6 L Cyclophosphamide dexamethasone rituximab; 7 L Pomalidomide rituximab	c.159+1dup	—	FS	TM1	0.426	0.247	0.429	99.3%	Negative	1%	Negative
		c.179G>T	p.G60V	Missense (CADD 28.8)	TM1	0.081	0.207
Patient 5	1 L Rituximab; 2 L Fludarabine cyclophosphamide rituximab; 3 L Bortezomib	c.147dup	p.K50*	FS*	TM1	0.015	Not seen (Only 1 CD20MUT read)	0.381	3.9%	Positive	12%	—
Patient 6	1 L Bortezomib dexamethasone rituximab; 2 L Bendamustine rituximab	c.155del	p.L52Wfs*7	FS*	TM1	0.262	0.373	0.253	100%	Negative	2%	Negative (*intranuclear)
Patient 7	1 L Cyclophosphamide dexamethasone rituximab	c.336+1G>T	p.X112_splice	Splice (Splice AI 0.79)	Intron (E5‐6)	0.03	0.13	0.655	4.6%	Positive	48%	Positive (100%)
Patient 8	1 L Fludarabine; 2 L Rituximab; 3 L Cyclophosphamide hydroxydaunorubicin vincristine prednisone rituximab; 4 L Bortezomib dexamethasone rituximab	c.166C>T	p.Q56*	NS*	TM1	0.046	0.0098	0.633	34.4%	Negative	1%	Subclonal (30%)
		c.173T>G	p.M58R	Missense (CADD 27)	TM1	0.217	0.366
		c.358_363del	p.N120_S121del	DEL (in frame)	TM3	0.167	0.221
		c.568_573+43del	p.X190_splice	Splice (Splice AI not assessable)	Intron (E6‐7)	0.121	Not quantifiable
Patient 9	1 L Cyclophosphamide rituximab	c.427_445del	p.I143Wfs*4	FS*	TM3	0.428	0.313	0.45	95.1%	Negative	1%	Negative
		c.527del	p.N176Tfs*21	FS	ECL2	0.365	0.438

Note: DNA and protein changes are reported using HGVS nomenclature. Mutation types include frameshift (FS), nonsense (NS), missense, in‐frame insertion (INS), in‐frame deletion (DEL), and splice‐site (Splice); CADD (Combined Annotation Dependent Depletion) scores are shown for missense variants, with scores >20 considered deleterious, and SpliceAI scores are shown for splice variants, with scores >0.7 considered deleterious. Locations are annotated by CD20 structure: transmembrane domains (TM1‐TM4), extracellular loops (ECL1‐ECL2), and intronic splice regions. CD20 DNA and RNA VAF denote variant allele fraction by whole‐exome and RNA sequencing, respectively (“Not seen” indicates absence in RNA). MYD88 DNA VAF corresponds to the L265P variant and serves as a tumor clonality reference. Clonality (CD20/MYD88)* was calculated as the ratio of the highest CD20 DNA VAF to MYD88 DNA VAF ×100, estimating relative clonal burden. Flow CD20 indicates expression by flow cytometry (L27; extracellular loop 2 epitope‐targeting clone), with percentage of positive B cells in bone marrow biopsy (BMBx); IHC CD20 (L26; c‐terminal epitope‐targeting clone) reflects protein expression and localization (membranous or aberrant). Prior therapies are listed sequentially (Line 1 [L1], L2, L3). Rituximab is bolded in red to indicate rituximab exposure within 6 months prior to sample collection.

RNA sequencing, available for 137/164 (84%) patients including all MS4A1 ^MUT cases, confirmed 18/20 (90%) DNA‐level MS4A1 variants (Table 1). To assess allelic co‐occurrence, we evaluated three patients with paired mutations within the same read span; no reads contained both mutations, supporting distinct allelic events.

MS4A1 ^MUT was not associated with significant differences in CD20 transcript abundance (P = 0.13) (Supporting Information S1: Figure 1A) nor with differential isoform expression (Supporting Information S1: Figure 1B). There were no differences in CD20 expression in patients treated with and without CIT (Supporting Information S1: Figures 1C,D). In an exploratory analysis comparing patients with upper versus lower quartiles of CD20 expression, no clinical variables were associated with CD20 expression, except for bone marrow involvement, which was higher in the upper quartile (P = 0.05) (Supporting Information S1: Table 3).

Flow cytometry data assessing CD20 expression in the light‐chain‐restricted clonal B‐cell population were available in 140/164 (85%) patients. Among MS4A1 ^MUT , absence of surface CD20 was observed in 6/9 (67%) cases versus 5/131 (4%) in MS4A1 ^WT (P < 0.001) (Figure 1C). Accordingly, MS4A1 ^MUT had lower total CD20‐positive cells (median 2% [range 0%–48%] vs. 15% [range 0%–82%]) compared to MS4A1 ^WT (P = 0.01) (Figure 1D).

Recent rituximab exposure (≤6 months), previously reported to be associated with CD20 masking when using overlapping diagnostic antibodies such as L27, ⁷ was in fact associated with higher flow‐CD20 negativity in our cohort (17% vs. 5%; P = 0.03). To account for potential rituximab masking, we excluded recently exposed patients, after which MS4A1 ^MUT remained associated with absent flow CD20 (4/7 [57%] vs. 0/71 [0%]; P < 0.001) (Supporting Information S1: Figure 2A) and lower expression (median 3% [1–48] vs. 15% [0–80]; P = 0.03) (Supporting Information S1: Figure 2B).

IHC for CD20 was available in 6/9 (67%) MS4A1 ^MUT cases and showed heterogeneous expression and/or localization. Among these cases, 3/6 lacked membranous CD20 (including one with a minor 5% aberrant nuclear population), 2/6 showed subclonal membranous positivity (30% and 50% of LPL cells), and 1/6 retained normal membranous expression (100%).

To evaluate whether CD20 detectability reflected clonal burden, clonality was defined as the ratio of the highest MS4A1 ^MUT ‐VAF to MYD88 ^L265P ‐VAF (MS4A1/MYD88). CD20 expression aligned directly with this metric, with 3/3 CD20‐negative cases examined having full clonal expression (95‐100% MS4A1/MYD88), while for 2/2 cases with subclonal expression, intermediate MS4A1/MYD88 expression ratios (25%–35%) were observed. For the single CD20‐positive case, a low ratio (5%) of a single splice‐mutation was observed (Figure 1E and Supporting Information S1: Figure 3).

We report on the first dedicated study evaluating MS4A1 ^MUT (CD20) in previously treated WM. MS4A1 ^MUT was relatively common among patients exposed to anti‐CD20‐based CIT, occurring in about 1 of 7 such patients, and was absent in anti‐CD20‐ and chemotherapy‐naïve individuals. This pattern supports an acquired origin of MS4A1 ^MUT following rituximab‐based therapy, potentially facilitated by chemotherapy‐related genomic stress ⁸ , ⁹ and/or repeated anti‐CD20 exposure. ¹⁰

Prior studies investigating rituximab resistance in lymphomas have largely focused on CD20 expression levels. ¹¹ , ¹² In contrast, our in silico analyses of mutations in patients progressing after CIT suggest that the MS4A1 ^MUT may confer functional resistance independent of RNA expression, either through epitope loss caused by truncating mutations upstream of the rituximab binding site or through destabilization of transmembrane domains required for proper CD20 surface localization. Consistent with this model, flow cytometry using the L27 antibody, which recognizes an epitope overlapping rituximab at the membrane level, ¹³ demonstrated marked, though not universal, enrichment of CD20 negativity within light‐chain‐restricted clonal B‐cell populations harboring MS4A1 ^MUT . Correspondingly, IHC using the non‐rituximab overlapping antibody L26 that targets the C‐terminal domain ¹³ revealed a clonal relationship with MS4A1 ^MUT , including at least one case of aberrant CD20 localization consistent with intracellular accumulation of truncated or misfolded CD20.

This is particularly relevant as CD20 is a key target of TCE and CAR‐T‐cell therapies entering clinical evaluation in WM. Epcoritamab, the first prospectively evaluated TCE in WM (NCT06510491), is predicted to bind a conformational epitope spanning regions of CD20 ECL1 (~aa 72–80) and ECL2 (~aa 148–160). ¹⁴ Similarly, mosunetuzumab and glofitamab (CD20/CD3 TCEs) bind epitopes overlapping rituximab (including the critical ECL2 170‐ANPSE‐174 region) and obinutuzumab (ECL2 172–178), respectively. ¹³ , ¹⁵ , ¹⁶ In our cohort, several truncating mutations were located upstream of these regions, predicting impaired binding of these therapies. Likewise, a recent study in non‐Hodgkin lymphoma showed that truncating mutations confer mosunetuzumab resistance, including in cases with preserved membranous CD20 expression. ¹⁷ These findings could extend to upcoming CD20‐directed therapies with reported use in WM, including MB‐106 (third‐generation CD20‐CAR‐T) and plamotamab (CD20‐TCE), ⁴ , ⁵ yet functional validation in WM is lacking.

Current National Comprehensive Cancer Network (NCCN) guidelines recommend repeat biopsy with CD20 assessment by IHC or flow cytometry prior to CD20‐directed TCEs to confirm target expression. ¹⁸ However, our data suggest that CD20‐directed therapy resistance is driven by MS4A1 ^MUT , with IHC/flow negativity reflecting epitope disruption rather than protein downregulation (all CD20‐negative cases occurred with MS4A1 mutations or rituximab masking). Because most CD20 mutations are subclonal, IHC and flow may miss small CD20‐negative subclones that could expand under CD20‐directed pressure. Moreover, CD20 status may be affected by variability in diagnostic antibodies used across centers.

Limitations of our study include the lack of paired pretreatment samples to confirm the acquired nature of MS4A1 ^MUT , limited IHC availability precluding evaluation of alternative rituximab resistance mechanisms in MS4A1 ^WT, including undetected mutations in those with low tumor burden and CD20‐independent resistance mechanisms such as CD55/CD47. ¹⁹ , ²⁰ Additionally, the retrospective design with imbalanced treatment exposures precluded distinguishing the relative contributions of CIT and repeated rituximab exposure to MS4A1 ^MUT development.

Still, we identify MS4A1 ^MUT as a potential resistance mechanism, adding to evidence of chemotherapy‐ and rituximab‐associated genomic damage and providing a rationale for prospective evaluation of their role in patient selection for CD20‐targeted TCE/CAR‐T therapies, earlier use before mutation development, and exploration of alternative targets beyond CD20.

AUTHOR CONTRIBUTIONS

Alberto Guijosa: Conceptualization; methodology; data curation; investigation; formal analysis; writing—original draft; writing—review & editing. Nickolas Tsakmaklis: Conceptualization; methodology; formal analysis; data curation; investigation; writing—original draft; writing—review & editing. Sirano Dhe‐Paganon: Investigation; methodology; formal analysis; writing—review & editing. Filip Garbicz: Writing— review & editing; investigation; methodology; formal analysis; data curation. Ruben Carrasco: Investigation; data curation; formal analysis; writing—review & editing; methodology. John Hatcher: Methodology; investigation; formal analysis; writing—review & editing. Andres Ramirez‐Gamero: Writing—review & editing; investigation; data curation. Dominic Pizzarella: Writing—review & editing; investigation; data curation. Amanda Kofides: Writing—review & editing; data curation; investigation. Shirong Liu: Writing—review & editing; investigation; data curation. Hao Sun: Writing—review & editing; data curation; investigation. Xia Liu: Writing—review & editing; investigation; data curation. María L. Guerrera: Writing—review & editing; investigation; data curation. Andrew R. Branagan: Writing—review & editing; data curation; investigation. Gottfried von Keudell: Writing—review & editing; investigation; data curation. Shayna R. Sarosiek: Writing—review & editing; investigation; data curation. Jorge J. Castillo: Writing—review & editing; investigation; data curation; supervision. Zachary R. Hunter: Conceptualization; investigation; formal analysis; data curation; supervision; writing—review & editing; funding acquisition. Steven P. Treon: Writing—review & editing; writing—original draft; conceptualization; investigation; methodology; supervision; formal analysis; funding acquisition; data curation; project administration.

CONFLICT OF INTEREST STATEMENT

S.P.T. received research funding and/or consulting fees from Abbvie/Pharmacyclics Inc., Janssen Oncology Inc., Beigene Inc., Eli Lilly Pharmaceuticals, Bristol Myers Squibb, and Ono Pharmaceuticals. J.J.C. received research funds or consulting fees from AbbVie, BeOne, Cellectar, Johnson & Johnson, LOXO, Nurix, Pharmacyclics, and Schrodinger. S.R.S. received research funding and/or consulting fees from ADC Therapeutics, Beigene, Cellectar, and SOBI. A.R.B. received research funds and/or honoraria from Adaptive, BeiGene, CSL Behring, Genzyme, Karyopharm, Pharmacyclics, and Sanofi. J.H. receives consulting fees from Arbella Therapeutics. All other authors declare no conflicts of interest.

ETHICS STATEMENT

This study was approved by the Dana‐Farber Cancer Institute's Institutional Review Board (IRB) and was conducted in accordance with the Declaration of Helsinki.

FUNDING

This work utilized an Illumina NovaSeq X Plus that was purchased with funding from a National Institutes of Health SIG grant 1S10OD036228‐01.

Supporting information

Supporting File 1.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.