Single-Nucleus Transcriptomics of FFPE Specimens from Mucosa-Associated Lymphoid Tissue Lymphoma Patients Reveal Tumor Heterogeneity and Changes in the Tumor Microenvironment during Relapse

Background: Marginal zone lymphomas (MZLs) are indolent B-cell lymphomas originating from the marginal zone of secondary lymphoid follicles. They are categorized into three subtypes based on sites of origin: extranodal MZL (EMZL), splenic MZL, and nodal MZL. EMZL, also known as mucosa-associated lymphoid tissue (MALT) lymphoma, is the most prevalent subtype and can primarily occur in various sites. Despite generally favorable prognosis, patients with MALT lymphoma may still experience poor outcome due to relapse, progression, or aggressive transformation. However, the biological characteristics that underlie the different origins and clinical evolutions of MALT lymphoma remain poorly understood. Methods: We collected formalin-fixed paraffin-embedded (FFPE) tissue samples of MALT lymphomas from a series of sites including the ocular adnexa, stomach, intestine, lung, liver, salivary gland, nasal sinuse, and lymph node. Using 10x Genomics' Flex technology, single-nucleus RNA sequencing was initially performed on 16 MALT lymphoma samples from 15 patients, which included 11 treatment-naive and 5 relapsed samples, with one patient having matched samples of both. Additionally, bulk RNA sequencing was performed on 89 FFPE samples from 81 patients. Results: After quality control, a total of 81,299 cells were obtained. Following preliminary clustering, 50,080 malignant B cells were identified. Using non-negative matrix factorization, gene expression programs were extracted separately from the lymphoma cells of each sample. By examining the consistency of different expression programs within and between samples, a total of 49 robust gene expression programs were selected. Further clustering yielded 6 meta-programs (MPs) based on program similarities. MP1 and MP5 correspond to the G1/S and G2/M cell cycles (PCNA, HELLS; MKI67, AURKB), respectively. MP2 corresponds to stress (HERPUD1, ZFP36), MP3 to B cell activation (TNFRSF13B, CD74), MP4 to immune modulation (CCL22, CCL17), and MP6 to the MAPK pathway (MAP4K1, MAPK8IP3). Lymphoma cells from relapsed samples exhibited significantly higher signature scores in MP1 and MP5 compared to treatment-naive samples, reflecting more active lymphoma cell proliferation. Additionally, lymphoma cells in relapsed samples also had significantly higher MP4 signature scores. Pseudotime analysis of lymphoma cells from matched samples suggested a trajectory of differentiation towards more active cell proliferation or higher immune modulation during relapse. Considering that CCL22 and CCL17 can specifically recruit regulatory T cells by binding to CCR4, we explored the role of MP4 and Tregs in relapsed samples. By clustering 8,235 T cells, CD4+ T and CD8+ cell subtypes were identified, showing a higher proportion of Treg cells in relapsed samples. Further analysis of T cell characteristics revealed stronger immunosuppressive signals in relapse samples, with CD4+ T cells showing higher exhaustion and CD8+ T cells showing weaker cytotoxicity and higher exhaustion. Conclusions: Our study firstly described the tumor microenvironment characteristics of MALT by multi-site single-nucleus and bulk RNA sequencing, revealing the complex mechanisms of internal cell proliferation and external immunosuppressive signals that comprehensively contribute to MALT lymphoma relapse.