Sequential antigen loss and branching evolution in lymphoma after CD19- and CD20-targeted T-cell-redirecting therapy

Abstract

CD19 chimeric antigen receptor (CAR) T cells and CD20 targeting T-cell-engaging bispecific antibodies (bispecs) have been approved in B-cell non-Hodgkin lymphoma lately, heralding a new clinical setting in which patients are treated with both approaches, sequentially. The aim of our study was to investigate the selective pressure of CD19- and CD20-directed therapy on the clonal architecture in lymphoma. Using a broad analytical pipeline on 28 longitudinally collected specimen from 7 patients, we identified truncating mutations in the gene encoding CD20 conferring antigen loss in 80% of patients relapsing from CD20 bispecs. Pronounced T-cell exhaustion was identified in cases with progressive disease and retained CD20 expression. We also confirmed CD19 loss after CAR T-cell therapy and reported the case of sequential CD19 and CD20 loss. We observed branching evolution with re-emergence of CD20+ subclones at later time points and spatial heterogeneity for CD20 expression in response to targeted therapy. Our results highlight immunotherapy as not only an evolutionary bottleneck selecting for antigen loss variants but also complex evolutionary pathways underlying disease progression from these novel therapies.

Graphical abstract

Figure 1.

Treatment history and sampling of studied patients. (A) Twenty-eight samples from 7 patients treated with CD20 bispecs, CD19 CAR T cells, or both sequentially were obtained using ultrasound-guided biopsy. The samples were studied using a comprehensive analytical panel including scRNA-seq, bulk RNA-seq, WES, flow cytometry, and IHC. The analytical panel was adjusted according to the amount and quality of available material. The median of prior lines of therapy before CD20 bispecs and CD19 CAR T cells was 4 (range, 3-6) and 4 (range, 2-7), respectively. (B) Table indicating whether therapy relapse from CD19- and CD20-targeting T-cell–redirecting therapies or tafasitamab (CD19 mAb) was associated with antigen loss (antigen negative) or antigen retainment (antigen positive). Forward-slash indicates that a patient was not treated with the respective therapy or response was not assessable. Allo Tx, allogeneic stem cell transplantation; Auto Tx, autologous stem cell transplantation; BEAM, carmustine, etoposide, cytarabine, and melphalan; Benda, bendamustine; CHOEP, CHOP + etoposide; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; Cis, cisplatin; Dexa, dexamethasone; DHAP, dexamethasone, high-dose cytarabine, and cisplatin; FCM, fludarabine, cyclophosphamide, and mitoxantrone; GemOx, gemcitabine-oxaliplatin; HD, high-dose; ICE, ifosfamide, carboplatin, and etoposide; Ifo, ifosfamide; MTX, methotrexate; O, obinutuzumab; Pola, polatuzumab vedotin; Pred, prednisone; R, rituximab; Rev, lenalidomide (Revlimid); Tafa, tafasitamab; Vin, vincristine.

Figure 2.

WES and bulk RNA-seq reveal genomic mechanisms of CD20 loss. (A) CD79A and CD20 protein expression determined using IHC on FFPE biopsy sections. The images show staining from a lesion located to the left axilla of patient 3 obtained shortly after the start of CD20 bispec treatment (sample 3.1) and at relapse (sample 3.2), which is representative of the CD20⁻ relapses. (B) Schematic illustration showing CD20 loss was accompanied by genomic alterations, namely an RNA fusion between NHLRC3- NXT1P1 intergenic region and MS4A1 in patient 1 (positions refer to hg19), an intronic T>C mutation leading to cryptic splicing and frameshift in patient 2 (exon5:c.573+2T>C; allelic frequency, 82.10%-100%; tumor content, 80%), a nucleotide deletion with frameshift in patient 3 (exon3:c.212delT:p.M71Rfs∗12; allelic frequency, 16.60%-20.60%; tumor content, 80%), and a biallelic deletion of MS4A1 in patient 6.

Figure 3.

Exhausted effector T cells in patients with CD20⁺ relapse from CD20 bispecs. (A) CD79A and CD20 protein expression determined using IHC on FFPE biopsy sections. The images show staining from patient 2 before CD20 bispec treatment (sample 2.1) and at relapse (sample 2.4), demonstrating relapse with retained CD20. (B) Schematic illustration showing the strategy used to compare scRNA-seq T cells from samples obtained during CD20 bispec treatment, including tumor-free or responsive lesions (n = 2, sample 3.3 and 7.1), CD20⁻ relapse (n = 1, sample 3.2) and CD20⁺ relapse (n = 3; sample 2.4, 2.5, and 7.2). T cells were combined into a joint embedding, and exhaustion was interrogated based on an established exhaustion gene signature within CD8 subpopulations. The exhaustion scores within CD8 effector T cells were compared between patients who experienced CD20⁺ relapse (patients 2 and 7) and the patient who did not experience CD20⁺ relapse (patient 3). (C) Joint UMAP embedding of 6381 T cells from 6 samples. T cells are color-coded based on identified subpopulation. (D) Heat map showing T-cell subset and state gene signature scores across subpopulations, underlying the subpopulation annotation in panel C. (E) Cells are color-coded based on their exhaustion signature score, which was computed across CD8 subpopulations. The gene list underlying the exhaustion score can be found in supplemental Table 3. (F) Box plot depicting the exhaustion scores within CD8 effector T cells across samples. Statistical significance was assessed using a pairwise, 2-sided Wilcoxon rank-sum test against combined samples 3.2 and 3.3 (∗∗∗P < 2.22 × 10⁻¹⁶). Center line indicates the median, box limits indicate the upper and lower quantiles, and whiskers indicate the 1.5× interquartile range. (G) Bar plot demonstrating proportions of T-cell subpopulations across samples. eff, effector; IFN, interferon; NK, natural killer; NKT, natural killer T cell; Tfh, T follicular helper; Th, T helper; Treg, regulatory T cell; UMAP, uniform manifold approximation and projection.

Figure 4.

Sequential CD19 and CD20 loss and branching evolution in lymphoma under the pressure of T-cell–redirecting immunotherapy. (A) Schematic timeline of the treatment history and sampling from patient 1 around the 4 sampling time points. Samples 1.2 and 1.4 were both obtained from a lymph node lesion at the left clavicula during CD20 bispecs and 80 days later, respectively. (B) CD79A, CD19, and CD20 protein expression determined using IHC on FFPE biopsy sections of sample 1.2 showing CD19/CD20 double-negative tumor after CD19 CAR T and CD20 bispec treatment, demonstrating sequential antigen loss. (C) Log-normalized MS4A1 expression color-coded and projected on the UMAP representation of sample 1.4 B cells/malignant cells demonstrates the recurrence of MS4A1-positive cells at a single-cell transcriptomic level. (D) Bar plot showing the proportion of MS4A1 expressing cells, determined by at least 1 count of MS4A1. (E) CD79A and CD20 protein expression determined using IHC on FFPE biopsy sections of sample 1.4 with a partially positive CD20 stain, validating CD20 recurrence on protein level. (F) Schematic timeline of the treatment history and sampling of patient 2 around the 4 sampling time points of samples that were subjected to WES (2.2, 2.4 or 2.5, 2.7, and 2.8). Sample ID, site, CD19 status, and CD20 status are indicated. (G) Mock phylogenetic tree constructed from SciClone clusters (indicated by colors) based on WES. The number at the start edge indicates the number of aberrations in the most recent common ancestor, and numbers on branches indicate the amount of acquired mutations. Length of branches does not indicate the extent of differences between subclones. Samples corresponding to subclone branches are indicated; see supplemental Figure 10 for details. IN, inguinal.

Figure 5.

Spatial heterogeneity of CD20 loss. (A) Schematic illustration showing the spatial response heterogeneity in patient 3 to CD20-targeted bispec therapy. (B) UMAP representation of single-cell transcriptomes (25 824 cells) from samples 3.2 and 3.3 colored by sample. Dashed outlines delineate cell type clusters. The schematic was partially created using BioRender.com.