Regulatory factor X 7 limits Myc activity during B cell activation and suppresses Myc-dependent lymphomagenesis

Rfx7 delays development of Bcl6-driven lymphoma

First, we asked whether deletion of Rfx7 would cause lymphoma. We analyzed 12- to 13-month-old mice lacking Rfx7 in the hematopoietic system (Cre-recombinase under the Vav1 promoter, referred to as VavRfx7^fl/fl) and did not observe increased cellularity of spleen and inguinal lymph nodes (iLNs), or splenic T and B cells as compared to their Rfx7^fl/fl counterparts (Extended Data Fig. 1a,b). Therefore, lack of Rfx7 alone does not promote lymphoproliferation in this timeframe.

As RFX7 mutations have been reported in DLBCL and are enriched numerically among cases harboring BCL6 alterations (Extended Data Fig. 1c), we next tested whether Rfx7 contributes to tumor suppression in the Iμ-Bcl6 mouse model, in which the proto-oncogene B cell leukemia/lymphoma 6 (Bcl6) is expressed under the immunoglobulin (Ig)M promoter mimicking a translocation common in this lymphoma²³. We thus crossed Iμ-Bcl6 with Cd19CreRfx7^fl/fl mice (in which Rfx7 deletion was confirmed by immunoblot; Extended Data Fig. 1d), resulting in mice with B cells expressing Bcl6 and lacking Rfx7 (hereafter called Iμ-Bcl6Cd19Rfx7^fl/fl). At a young age, Iμ-Bcl6Cd19Rfx7^fl/fl mice did not present significant differences as compared to Iμ-Bcl6Rfx7^fl/fl mice, but exhibited an increased incidence of splenomegaly at 10–13 months of age (Fig. 1a,b and Extended Data Fig. 1e,f). In agreement with the original characterization of this DLBCL model²³, spleens from Iμ-Bcl6Cd19Rfx7^fl/fl mice exhibited blastoid B cells with variably altered B cell marker expression, trends apparent—albeit less prominent—also in Iμ-Bcl6Rfx7^fl/fl mice (Fig. 1c and Extended Data Fig. 1g). Pathologic analysis of spleens from Iμ-Bcl6Rfx7^fl/fl and Iμ-Bcl6Cd19Rfx7^fl/fl mice at 7–8 months of age revealed benign hyperplasia in the former, whereas atypical hyperplasia and high-grade lymphoma were observed in the latter (Fig. 1d). Iμ-Bcl6Rfx7^fl/fl mice presented spleens with an altered splenic architecture, medium-to-large cell populations and mitotic figures (Fig. 1e). In contrast, spleens from Iμ-Bcl6Cd19Rfx7^fl/fl mice with B cell high-grade lymphoma exhibited loss of architecture, a prevailing population of large cells and typical or atypical mitotic figures; deposits of iron complexes (haemosiderin) were also observed (Fig. 1e). Complementarity-determining region 3 (CDR3) sequencing further revealed that one atypical hyperplasia presented a polyclonal repertoire, whereas the other one and the high-grade lymphomas exhibited mono- or oligo-clonal BCR repertoires with a dominant clone; mutations were identified in several of these clones (Supplementary Table 1a). This is in line with immunophenotypic analyses demonstrating expression of interferon regulatory factor 4 (Irf-4) and variable expression of peanut agglutinin (PNA), a GC B cell marker (Fig. 1f), indicative of GC or post-GC origin. Altogether, these data underscore the relevance of Rfx7 in suppressing Bcl6-driven lymphomagenesis.

Rfx7 limits Myc activity

To understand how Rfx7 regulates lymphomagenesis, we started by characterizing its expression and relevance for normal B cells. Rfx7 mRNA is detected in developing and splenic B cells, in which its expression at the protein level is confirmed by immunoblot (Fig. 2a,b). By analyzing Cd19Rfx7^fl/fl and control mice, we did not observe significant differences in absolute number and frequency of total, follicular (FO), marginal zone (MZ) or transitional B cells (Extended Data Fig. 2a,b). We thus investigated Rfx7-driven changes by RNA sequencing (RNA-seq) analyses of Rfx7^−/− and control resting FO B cells. Paired differential expression analysis revealed upregulated genes, including Asns and St6galnac, enzymes whose expression is associated with cancer progression^24–27 (Extended Data Fig. 2c and Supplementary Table 2). Among the downregulated genes in Rfx7^−/− FO B cells, we found the Rfx7 targets Mxd4 and Ccnd2, coding for an antagonist of Myc activity and a cyclin found to limit GC B cells entering the S-phase, respectively^3,5,28 (Extended Data Fig. 2c). Gene set enrichment analysis (GSEA) indicated ‘Myc targets’ as the top-upregulated signature (Fig. 2c).

After a transient downregulation, Rfx7 was detected 18 h postactivation with CpG (Fig. 2b), which homogeneously stimulates B cells. RNA-seq data from Rfx7^−/− versus control FO B cells triggered with CpG for 18 h confirmed the altered expression of Mxd4 and St6galnac 1 and 2 (Extended Data Fig. 2d and Supplementary Table 2), and the increase in selected metabolic pathways, including ‘Myc targets’ (Fig. 2d). Notably, the pathways ‘E2F targets’ and ‘G2M checkpoint,’ linked to cell cycle, were augmented strongly in FO B cells lacking Rfx7 (Fig. 2d). As there is no evidence so far supporting a transcriptional repressive role for Rfx7^3,5, we hypothesized the upregulation of gene expression in Rfx7-deficient B cells to be indirect. We thus compared the predicted Rfx7 binding sites in the promoter of the top genes upregulated in Rfx7^−/− activated FO B cells from the E2F signature²⁹ to those of nonmodulated genes and of known RFX7 targets^3,5. We observed a significant enrichment of putative Rfx7 binding sites in the promoters of known target genes, whereas such enrichment was absent in those from the E2F signature (Fig. 2e). Although such in silico analysis is merely indicative, the fact that the top upregulated genes of the E2F signature have not been identified as RFX7 targets^3,5,30 and are not downregulated in DLBCL-derived U-2932 cells overexpressing RFX7 (Extended Data Fig. 2e) corroborate the hypothesis that their upregulation is an indirect effect of Rfx7 deletion.

We next stimulated control and Rfx7^−/− B lymphocytes in vitro using CpG or anti-IgM in the presence of interleukin (IL)-4, which more accurately reflects antigen-dependent activation and also stimulates Rfx7 expression by 18 h (Extended Data Fig. 2f) for immunoblot analysis. Upon activation, Rfx7^−/− B cells exhibited increased expression of Myc and its targets identified through transcriptomic analyses, such as Cyclin A2 (Ccna2), the cell division cycle protein 20 (Cdc20), lactate dehydrogenase A (Ldha)³¹ and ribosomal protein L11 (Rpl11)—another Myc-regulated protein³² (Fig. 2f). Collectively, these results demonstrate that Rfx7-deficient B cells engage Myc activity more efficiently than control B cells in response to BCR-dependent and -independent stimuli.

Rfx7 limits B cell activation in vitro and in vivo

Given the key role played by Myc in B cell blasting and activation^19–21, we aimed to assess its involvement downstream of Rfx7 in vitro and in vivo. Nonstimulated Rfx7^−/− FO B cells exhibited an increased frequency of blasts, a feature that became more noticeable upon in vitro stimulation (Fig. 3a). This was accompanied by an increased proliferation, as measured by the frequency of divided cells following anti-IgM/IL-4 and of cells in each division after CpG stimulation (Extended Data Fig. 3a). We thus tested the effects of homozygous or heterozygous Myc deletion^19,33; Cd19Rfx7^fl/flMyc^fl/fl or Cd19Rfx7^fl/flMyc^fl/wt—as well as Cd19Myc^fl/fl and Cd19Myc^fl/wt—mice presented normal spleen cellularity, B cell percentage and number, with increased MZ and decreased—or tendentially decreased—FO and early transitional B cells (Extended Data Fig. 2a,b). Upon activation, B cells from Cd19Rfx7^fl/flMyc^fl/fl mice showed lower expression of Myc targets as compared to Rfx7^−/− B cells and markedly reduced Myc levels, confirming its deletion (Extended Data Fig. 3b). It can also be appreciated that, although the effects of Rfx7-deficiency on Myc targets are consistent with previous data, those on Myc levels are not observed, suggesting that Rfx7 acts on Myc activity rather than Myc levels (Extended Data Fig. 3b). Notably, whereas FO B cells from Cd19Rfx7^fl/flMyc^fl/fl mice were almost unresponsive to CpG or anti-IgM stimulation, FO B cells from Cd19Rfx7^fl/flMyc^fl/wt mice showed normalized blasting (Fig. 3a), in agreement with the role of Myc in regulating lymphocyte granularity upon activation and substantiating its role in the enhanced activation of Rfx7-deficient B cells³⁴.

Given the relevance of Myc and the increased RFX7 transcript levels in GC B cells^19–21 (Fig. 2a), we next assessed the role of Rfx7 following immunization of Cd19Rfx7^fl/fl and Cd19Rfx7^wt/wt mice. Lack of Rfx7 led to increased frequency and number of GC B cells (Fig. 3b and Extended Data Fig. 3c). Immunofluorescence staining with the GL-7 antibody revealed a trend towards larger GC area in immunized Cd19Rfx7^fl/fl mice, as compared to Cd19Rfx7^wt/wt mice (Fig. 3c and Extended Data Fig. 3d). As compared to Rfx7^+/+ GC B cells, their Rfx7^−/− counterparts expressed more homogenously the proliferative marker Ki67 and presented increased side scatter (SSC), indicative of cell content granularity (Fig. 3b and Extended Data Fig. 3c). Moreover, they did not present significant differences in the relative abundance of dark zone (DZ) and light zone (LZ) GC subsets (Extended Data Fig. 3c). Notably, immunized Cγ1CreRfx7^fl/fl mice³⁵, further crossed onto a Cre-reporter (YFP) strain, showed an increased proportion of YFP-positivity among total and GC B cells (with a tendency to have a DZ phenotype) as compared to control mice, indicating that the effects of Rfx7 on activated and GC B cells were recapitulated when its deletion was coupled to antigenic stimulation (Fig. 3d). Immunized Cd19Rfx7^fl/fl mice also showed increased frequency and number of PBs, but not of plasma cells (PCs; Fig. 3b). As with GC B cells, Rfx7^−/− PBs and PCs presented increased SSC (Fig. 3b). Cd19Rfx7^fl/flMyc^fl/wt mice showed normalized GC B cells, PB percentages and SSC in these subsets (Fig. 3b), highlighting the role of Rfx7 in restricting Myc activity.

Rfx7 delays development of p53-loss-driven lymphoma

As TP53 (the human gene encoding p53) aberrations are found in 30–70% of BL and DLBCL cases presenting RFX7 mutations¹⁴ (Supplementary Table 3 and co-occurrence analyses in Extended Data Fig. 4a) and a TP53–RFX7 signaling axis was reported in other cancer types⁵, we evaluated the contribution of Rfx7-loss and Myc (haplo-)insufficiency in a lymphoma model with B cell-specific deletion of p53 (encoded by the Trp53 gene, referred to here as p53), which causes IgM⁺ MZ expansion progressing towards DLBCL^36,37. At a young age, Cd19p53^fl/fl, Cd19p53^fl/flRfx7^fl/fl, Cd19p53^fl/flRfx7^fl/flMyc^fl/fl and control mice did not present substantial differences in spleen size and B cell percentage except for MZ B cells, which were increased mildly in Cd19p53^fl/flRfx7^fl/fl as compared to Cd19p53^fl/fl mice and further increased by Myc ablation (Extended Data Fig. 4b). We thus monitored these mice over time for tumor development. Mice that were asymptomatic until the endpoint (300 days) are displayed by black or white symbols (Fig. 4a,e); mice that were euthanized at the appearance of disease symptoms before the endpoint are instead represented by red symbols (Fig. 4a,e). As compared to Cd19p53^fl/fl mice, Cd19p53^fl/flRfx7^fl/fl mice exhibited enlarged spleens and two out of eight cases of symptomatic disease (Fig. 4a and Extended Data Fig. 4c). Histologic examination of spleens from asymptomatic Cd19p53^fl/flRfx7^fl/fl mice with moderately enlarged spleens already revealed loss of architecture and medium-to-large cell populations with vesicular and enlarged nuclei, congruent with a high-grade lymphoma (Fig. 4b).

Based on these, we probed the relationship between p53 and Rfx7. Both proteins were expressed by in vitro activated B cells without exhibiting obvious interdependence, in line with p53 not being altered in our RNA-seq data and Rfx7 not being reported as p53 target in B cells³⁸ (Extended Data Fig. 4d,e). We next assessed the mRNA levels of p53 and Rfx7 target genes (based on published^3,5,38 and our own data; Extended Data Fig. 4e). Whereas the Rfx7 targets—in particular Mxd4 and Tsc22d1, reported to have tumor suppressor functions^39–41—were decreased in Rfx7-deficient B cells but not in p53^−/− B cells, the p53 targets Ddit4, Serpine2 and Psrc1 were variably modulated by Rfx7 (Ddit4 is also a direct Rfx7 target; Extended Data Fig. 4f). Moreover, following CpG stimulation, the enhanced blasting caused by Rfx7 deletion was similar on a p53^−/− and p53^+/+ background (Extended Data Fig. 4g). These findings suggest that RFX7 might contribute to the p53 signature, while regulating additional p53-independent pathways.

Myc and AID contribute to p53^−/−Rfx7^−/− B cell lymphoma

Although Cd19p53^fl/flRfx7^fl/flMyc^fl/wt mice also exhibited enlarged spleens and symptomatic disease, Cd19p53^fl/flRfx7^fl/flMyc^fl/fl mice were protected from lymphoma (Fig. 4a and Extended Data Fig. 4c). Flow cytometry analyses revealed the presence of B cell populations with variably altered CD19/B220 staining in Cd19p53^fl/flRfx7^fl/fl and Cd19p53^fl/flRfx7^fl/flMyc^fl/wt mice (Fig. 4c). Such populations exhibited marked blastoid morphology and were mostly positive for IgM (Fig. 4c–e). In 4 out of 12 cases, lymphoma cells were IgM⁻IgD⁻, with one positive for kappa light chain, compatibly with a switched PB phenotype (Fig. 4c,d). Only one Cd19p53^fl/flRfx7^fl/fl mouse exhibited a mix of two distinct Ig expression patterns, suggesting the existence of at least two clones (Fig. 4c; empty triangular symbols in Fig. 4a,e). Cd19p53^fl/fl mice showed no or mild alterations of CD19/B220 staining, with few mice presenting with IgM⁺ blastoid cells in the spleen but negligibly in iLN and bone marrow (BM) (Fig. 4c,e). Cd19p53^fl/flRfx7^fl/fl and Cd19p53^fl/flRfx7^fl/flMyc^fl/wt mice exhibited instead spreading of blasts to LNs and BM, a phenomenon associated with symptomatic disease (Fig. 4e, left). When focusing on mice euthanized at endpoint, Myc heterozygosity trended towards or significantly reduced the frequency of blastoid B cells in spleen, iLN and BM (Fig. 4e, right). These data indicate that loss of Rfx7 promotes lymphomagenesis whereas Myc deletion blocks it and Myc heterozygosity displays a protective effect on nonsymptomatic disease.

Myc deregulation occurs commonly due to chromosome translocations caused by the aberrant activity of AID (encoded by the Aicda gene)^22,42–44. We therefore asked whether AID contributes to lymphomagenesis of B cells on a p53^−/−Rfx7^−/− background. At a young age, Cd19p53^fl/flRfx7^fl/flAicda^−/− and control mice did not present substantial differences in spleen size and B cell percentage except for FO B cells, which were decreased mildly (Extended Data Fig. 4h). Notably, Cd19p53^fl/flRfx7^fl/flAicda^−/− mice were asymptomatic until the endpoint; six out of seven mice presented negligible alterations of the B cell compartment, whereas one mouse exhibited a mildly enlarged spleen with IgM⁺ blastoid cells not spreading to iLNs and BM (Fig. 4a,c,e and Extended Data Fig. 4c). This indicates that AID is not absolutely required for lymphomagenesis driven by p53 and Rfx7 loss. Yet, protection from advanced disease provided by its deficiency reveals the importance of B cell activation and AID activity in the trajectory promoted by Rfx7 loss.

p53^−/−Rfx7^−/− lymphoma cells exhibit largely PB and PC features

Molecular phenotyping revealed that lymphoma cells on a p53^−/−Rfx7^−/− background—especially from symptomatic mice—exhibited high levels of Ki67, paired box 5 (Pax5)⁴⁵ and Myc expression, indicating a clear deregulation of Myc even in diseased Cd19p53^fl/flRfx7^fl/flMyc^fl/wt mice (Fig. 5a,b and Extended Data Fig. 5a). As expected, Myc was negligible in Cd19p53^fl/flRfx7^fl/flMyc^fl/fl mice, whereas it was expressed at levels comparable to p53^−/− B cells in most Cd19p53^fl/flRfx7^fl/flAicda^−/− mice (Fig. 5a,b and Extended Data Fig. 5a). This suggests that AID promotes deregulation of pro-tumorigenic genes in p53^−/−Rfx7^−/− B cells.

We next assessed the capacity of lymphoma cells on a p53^−/−Rfx7^−/− background from asymptomatic and symptomatic diseased mice to proliferate in vitro in the absence of specific growth factors. Even though Cd19p53^fl/flRfx7^fl/fl and Cd19p53^fl/flRfx7^fl/flMyc^fl/wt mice presented a limited number of symptomatic lymphomas, most of them grew in vitro, something that was not observed for nonsymptomatic disease (Extended Data Fig. 5c). The established ‘cell lines,’ which were transplantable in immunocompetent syngeneic mice, lacked Rfx7 and p53 expression, but exhibited high Myc and variable AID levels (Extended Data Fig. 5c,d). In line with the presence of genotoxic stress, we detected γH2AX positivity and polymerase chain reaction (PCR) products indicative of chromosome translocations between the IgH and Myc loci in four out of five cell lines (Fig. 5c and Extended Data Fig. 5d,e), supporting the involvement of AID in the progression of the accelerated lymphomas on the p53^−/−Rfx7^−/− background.

These lines were mono-clonal based on CDR3-sequencing and largely retained the original phenotype, with IgM⁺ tumors giving rise to both IgM⁺ and IgM⁻ fractions (Extended Data Fig. 5f and Supplementary Table 1b). Whereas ex vivo only few tumors presented CD138 positivity, the cell lines were all positive for this dynamically expressed PB/PC marker⁴⁶ (Extended Data Fig. 5b,f). Notably, B cells from Cd19p53^fl/flRfx7^fl/fl and Cd19p53^fl/flRfx7^fl/flMyc^fl/wt were largely negative for Bcl6 and positive for Irf-4 and—particularly in the symptomatic cases—for B lymphocyte-induced maturation protein-1 (Blimp-1) (Fig. 5a,b and Extended Data Fig. 5b). Irf-4 and Blimp-1 were expressed also by the derived cell lines, with one line (line 5) presenting lower Irf-4 and Blimp-1 levels and positive for Bcl6 and CD95, compatible with a GC origin (Fig. 5d,e). These results underscore how lymphoma cells on a p53^−/−Rfx7^−/− background combine aggressive proliferative features with markers associated to GC or PB/PC differentiation^47,48.

RFX7 nonsense mutations cause loss of function

Finally, we aimed to better understand the impact of RFX7 alterations in human lymphoid malignancies. We analyzed overall survival (OS) and progression free survival (PFS) relative to RFX7 levels (which are not altered in different DLBCL subtypes, Fig. 6a) in DLBCL. Notably, high RFX7 expression associated with better prognosis in two DLBCL cohorts of 301 and 35 patients^49,50 (Fig. 6a and Extended Data Fig. 6a), suggesting that RFX7 transcript abundance impacts on disease outcomes. Multivariate Cox analysis for OS and PFS demonstrated that the RFX7 classification is independent of the cell of origin (hazard ratio (HR) = 1.55 (low-expression group) and P = 0.01 for OS, HR = 1.46 and P = 0.02 for PFS) for the larger cohort (HR = 3.07 and P = 0.06 for OS in the smaller cohort). Although the analyzed biopsies are not restricted to B cells, FO B cells, together with T cells, represent major LN populations and express high levels of RFX7 transcripts (Extended Data Fig. 6b). Moreover, reanalysis of single-cell RNA-seq data⁵¹ shows that RFX7 transcript levels are stable in T and B cells from tonsil, follicular lymphoma—an indolent lymphoma which can transform into DLBCL—and DLBCL, with the exception of DLBCL B cells, which express reduced levels (Extended Data Fig. 6c). These data, which are in line with previous observations⁶, hint to a significant impact of DLBCL cells on RFX7 mRNA levels in the biopsies used to predict disease outcomes (Fig. 6a and Extended Data Fig. 6a).

We next screened the mutational landscape of RFX7 in B cell malignancies, by reviewing studies on BL^16,52, DLBCL^{11,12,50,53,54}, CLL¹⁷ and a small cohort of primary mediastinal large B cell lymphoma⁵⁵, in which we found three RFX7 mutations corresponding to 8% of patients (Supplementary Table 3). An array of nonsense mutations and frameshifts—distributed over the central portion of the protein, a region encoded largely by the last exon—represented a substantial fraction of RFX7 alterations (Fig. 6b and Supplementary Table 3). Notably, mRNA arising from the truncations in this exon was detected at approximately the same frequency as the WT allele in DLBCL patients and lymphoma cell lines^16,50 (Extended Data Fig. 6d). Moreover, we observed a band migrating at the expected molecular size in two lymphoma cell lines, indicating that such premature stop codons are not causing nonsense-mediated decay (Extended Data Fig. 6d).

We thus cloned truncations spanning the central part of RFX7. Overexpression of WT human RFX7 in HEK293T cells increased the transcription of known target genes such as MXD4 and PIK3R3³ (Extended Data Fig. 6e). In contrast, the four tested RFX7 mutants failed to do so despite being expressed at high levels (Extended Data Fig. 6e,f). All RFX7 mutants appeared to lower MXD4 and PIK3R3 mRNA levels in comparison to mock-transfected cells, suggesting a dominant-negative effect. We therefore selected the 1–688 mutant to test its effects on WT RFX7. Confirming our hypothesis, the 1–688 mutant reduced the capacity of WT RFX7 to transactivate target genes, even when the latter was expressed more abundantly (Fig. 6c).

To approach the question in a more physiological way, we overexpressed WT, 1–688 or 1–782 RFX7 in DLBCL-derived U-2932 cells (Extended Data Fig. 6g). As a negative control, we also included a nuclear localization signal (NLS) mutant (NLSm) version of RFX7³. In agreement with the expected effects of RFX7 on cellular metabolism³, U-2932 cells overexpressing WT RFX7 exhibited lower SSC compared to the control condition overexpressing NLSm or truncated RFX7 forms (Fig. 6d). Quantitative PCR with reverse transcription (RT–qPCR) and RNA-seq data confirmed the upregulation of known target genes by WT RFX7 (Fig. 6e and Extended Data Fig. 6h). In line with primary B cell results, GSEA corroborated the downregulation of MYC-dependent pathways, whereas we observed the ‘P53 pathway’ among the upregulated signatures^4,5 (Figs. 2c,d and 6e and Supplementary Table 4). Coexpression of the truncated RFX7 forms (co-expressed at levels comparable to WT RFX7; Extended Data Fig. 6g) counteracted the effects of WT RFX7 on SSC, on MYC targets’ signatures and on the transactivation of a large fraction of target genes, including DDIT4, MXD4, PIK3R3 and REC8 to a smaller extent^3,5 (Fig. 6d–f and Extended Data Fig. 6i). The truncated RFX7 forms either did not induce target genes or reduced their transcripts below the level of RFX7 NLSm (Fig. 6e); only a negligible number of genes hinted to potential neomorphic features, which were largely lost in the context of coexpression with WT RFX7 (Extended Data Fig. 6j).

In silico analyses predicted that the 1–688 and 1–782 truncations lack protein domains positioning in proximity of the DNA, indicating that the C-terminal part of the protein might regulate chromatin interaction at target gene promoters (Extended Data Fig. 6k). We thus went back to HEK293T cells to test the ability of RFX7 1–688 and 1–782 to bind to target gene promoters^3,5 by chromatin immunoprecipitation (ChIP) assay. The mutants not only retained the capacity to occupy the promoter of DDIT4, MXD4 and PIK3R3 but also exhibited an increased ability to do so as compared to WT RFX7 (Fig. 6g). In agreement with a lesser effect of the truncations on REC8 mRNA abundance, promoter occupancy was similar to the one of WT RFX7, suggesting that REC8 transactivation has different requirements compared to that of other targets (Fig. 6f,g). Taken together, these results indicate that RFX7 mutations and transcript abundance can impact on the overall activity of RFX7 and on disease outcomes.

Patient cohort prognosis and RFX7 mRNA analysis

For the previously published cohorts^49,50 and normal B cells⁶⁶, we reprocessed the RNA-seq data from raw FASTQ files. Reads were aligned to the GRCh38 reference genome using STAR, and gene-level quantification was performed with featureCounts. For the lymphoma analyses, normalized RFX7 expression values were incorporated as a continuous covariate in Cox proportional hazards models to evaluate associations with OS and PFS using the survival R package. Furthermore, we employed the survminer R package (https://cran.r-project.org/web/packages/survminer/index.html) to determine the optimal cutoff point for dichotomizing patients into high and low RFX7 expression groups in terms of OS. Specifically, we used the surv_cutpoint function, which tests all possible expression thresholds iteratively and selects the value that best separates survival outcomes based on the maximally selected rank statistic. This cutoff was then used to define binary groups for OS and PFS outcomes in Kaplan–Meier analysis and Cox modeling. Cell of origin, as described in the relative original publications^49,50, was used in a multivariate Cox analysis with the stratification induced by RFX7 expression.

Mice

Rfx7-floxed mice³ were crossed to VaviCre and R26-stop-EYFP mice (The Jackson Laboratory), Cd19Cre⁶⁷ and to Cγ1Cre mice³⁵ to generate VavRfx7^fl/fl, Cd19Rfx7^fl/fl and Cγ1CreR26-stop-EYFPRfx7^fl/fl mice. Further, Cd19Rfx7^fl/fl and control Cd19Rfx7^wt/wt mice were crossed onto mice expressing the Iμ-Bcl6 transgene²³ or with floxed Trp53 gene (The Jackson Laboratory) mice and additionally onto Myc-floxed mice³³ or AID full body knockout mice²². All animal experimental protocols were approved by the Veterinary Office regulations of the Cantone Ticino, Switzerland, and all methods were performed in accordance with the Swiss guidelines and regulations. Both male and female mice were used. The mice were housed in individually ventilated cages under standardized conditions (20 ± 2 °C, 55 ± 8% relative humidity, 12 h light/dark cycle) with cage enrichment, and access to food and water ad libitum.

Survival

Cd19p53^fl/fl, Cd19p53^fl/flRfx7^fl/fl, Cd19p53^fl/flRfx7^fl/flMyc^fl/fl, Cd19p53^fl/flRfx7^fl/flMyc^fl/wt and Cd19p53^fl/flRfx7^fl/flAicda^−/− mice were followed by visual inspection and by measuring weight and temperature once per week for 300 days (±7 days). Humane endpoints were based on cumulative scores requiring daily monitoring and consideration of euthanasia if not improved after 1 day when reflecting isolated symptoms or combinations of mild symptoms (including reduced grooming or activity, limited weight or temperature loss) and immediate euthanasia with scores reflecting severe symptoms, such as moribund state, visibly enlarged organs, ≥12% bodyweight loss and temperature lower than 34.5 °C. Iμ-Bcl6Cd19Rfx7^fl/fl and Iμ-Bcl6Rfx7^fl/fl mice were inspected visually, and weight was measured every second week from 6 months onwards. Humane endpoints were based on cumulative scores requiring daily monitoring and consideration of euthanasia if not improved after 1 day when reflecting isolated symptoms or combinations of mild symptoms (including reduced grooming or activity, limited weight loss) and immediate euthanasia with cumulative scores reflecting concomitant moderate symptoms or any severe symptom, such as moribund state, abnormal masses and ≥15% bodyweight loss.

Human lymphoma cell lines

NU-DUL-1 (ACC 579 (ref. ⁶⁸)) B cell lymphoma cell line and CA-46 (ACC 73 (ref. ⁶⁹)) Burkitt lymphoma cell line were purchased from DSM GmbH and maintained in culture according to manufacturer’s instructions using complete medium (RPMI, 10% FBS, 1% penicillin-streptomycin, 1% MEM NEAA, 1% GlutaMAX, 1% sodium pyruvate, 0.1% 2-mercaptoethanol, all from Gibco) as well as U-2932 (DLBCL, provided by F. Bertoni, Bellinzona). Whole-cell lysates were prepared using E1A buffer (250 mM NaCl, 1% NP-40, 50 mM Hepes pH 8) and equal amount of material was used in western blots as described in ‘Immunoblot analysis.’

Lymphoma cell growth in vitro and transplantation

Spleen from healthy controls (age-matched Cre-negative littermates) or tumor-bearing mice (increased percentage of blastoid cells) were processed to single-cell suspensions and treated with hypotonic buffer (155 mM NH₄Cl, 10 mM KHCO3, 0.1 mM Na₂EDTA) to lyse erythrocytes. Cells were plated at a concentration of 1–2.5 × 10⁶ cells ml⁻¹ (in different experiments) in complete medium (RPMI, 10% FBS, 1% penicillin-streptomycin, 1% MEM NEAA, 1% GlutaMAX, 1% sodium pyruvate, 0.1% 2-mercaptoethanol, all from Gibco). Cells were split according to density and grown for a period of at least 2 weeks. Survival of cells was checked by flow cytometry staining using propridium iodide (ImmunoChemistry Technologies) and cells expanding for at least 2 weeks were considered ‘lines.’ For testing transplantability, 1 × 10⁶ lymphoma cells from the established cell lines were injected subcutaneously into the flank of C57BL/6 recipient mice. Tumor size was measured regularly and the mice were euthanized 10–13 days post-transplantation, when the tumor was between 100 mm³ and 600 mm³ (maximal tumor size permitted was 1,000 mm³). Humane endpoints were based on cumulative scores requiring daily monitoring and consideration of euthanasia when mice presented moderate symptoms (including reduced activity, slightly worsened body condition) and immediate euthanasia with symptoms such as moribund state, tumor size more than 1,000 mm³ and worsened body condition.

Translocation PCR-based analyses

Total genomic DNA from mouse lymphoma B cell lines was extracted using the QIAamp Blood Mini Kit (Qiagen). For each cell line, 500 ng of genomic DNA, corresponding to 10⁵ cells, was used as a template for PCR amplification. Nested PCR was performed using the Expand Long Template PCR System (Roche) with primers described by Ramiro and colleagues⁷⁰. Briefly, for derivative chromosome (chr.) 12 translocations, the first PCR used the following primers: 5′-TGAGGACCAGAGAGGGATAAAAGAGAA-3′ and 5′-GGGGAGGGGGTGTCAAATAATAAGA-3′; the second round of PCR used the following primers: 5′-CACCCTGCTATTTCCTTGTTGCTAC-3′ and 5′-GACACCTCCCTTCTACACTCTAAACCG-3′; for derivative chr. 15 translocation the first round of PCR used the following primers: 5′-GTGAAAACCGACTGTGGCCCTGGAA-3′ and 5′-ACTATGCTATGGACTACTGGGGTCAAG-3′; the second round of PCR used the following primers: 5′-GTGGAGGTGTATGGGGTGTAGAC-3′ and 5′-CCTCAGTCACCGTCTCCTCAGGTA-3′. All primers were used at 300 nM final concentrations. PCR conditions were as follow: 92 °C for 2 min followed by ten cycles of (92 °C 10 s; 60 °C 30 s; 68 °C 7 min) and 25 cycles of (92 °C 10 s; 60 °C 30 s; 68 °C 7 min (with an additional extension time of 20 s per cycle)) for the first round. A 1 μl aliquot of the first round PCR product was used in the second round with the following cycling conditions: 92 °C for 2 min followed by ten cycles of (92 °C 10 s; 60 °C 30 s; 68 °C 7 min) and 19 cycles of (92 °C 10 s; 60 °C 30 s; 68 °C 7 min, with an additional extension time of 20 s per cycle). PCR products were analyzed on a 0.8% agarose gel. DNA corresponding to the band of interest was extracted from the agarose using the QIAquick Gel Extraction KIT (Qiagen) and sent for sequencing (ONT sequencing Microsynth). Sequences were analyzed by mapping the results to the mouse genome (GRCm39).

Immunization

Mice were immunized with 100 µg (4-hydroxy-3-nitrophenylacetyl–keyhole limpet hemocyanin (NP–KLH) (Biosearch Technologies), precipitated with 0.6 mg Imject-alum (Thermo Scientific) in PBS. Mice were injected intraperitoneally and analyzed 10 days later.

B cell activation

B cells were isolated from the spleen using EasySep Mouse B cell Isolation Kit (STEMCELL Technologies). For flow cytometry analysis, FO B cells were further FACS-sorted (SORP FACSymphony S6, BD Biosciences) for B220⁺/CD19⁺/CD93⁻/CD23⁺ and seeded at 100,000 cells per well in 96-flat bottom well plate in complete medium (RPMI, 10% FBS, 1% penicillin-streptomycin, 1% MEM NEAA, 1% GlutaMAX, 1% sodium pyruvate, 0.1% 2-mercaptoethanol, all from Gibco). Cells were activated by CpG (Microsynth) added to a final concentration of 2 µg ml⁻¹ or by anti-IgM (Jackson ImmunoResearch), coated overnight at a concentration of 5 μg ml⁻¹, together with IL-4 (PeproTech) added to a concentration of 2.5 ng ml⁻¹ and cultured at 37 °C for 18 h. For proliferation analysis, sorted FO B cells were labeled with Cell Trace Violet (Thermo Fisher) according to manufacturer’s instructions and seeded and stimulated as above, for 40 h (CpG) or 48 h (IL-4 or anti-IgM + IL-4). For immunoblot analysis, B cells were seeded at 1.3 × 10⁶ cells per well in 24-flat bottom well plate. Cells were then activated by CpG (2 µg ml⁻¹) or plate-bound anti-IgM (5 μg ml⁻¹) + IL-4 (2.5 ng ml⁻¹) and cultured at 37 °C in complete medium for 0 h, 6 h and 18 h.

Immunoblot analysis

B cells enriched and activated as described above, were washed with ice-cold PBS and lysed in Laemmli buffer (187.5 mM Tris-HCl pH 6.8, 6% SDS, 0.03% Bromophenol Blue, 30% glycerol). Cell lysates were boiled for 5 min at 95 °C and separated on a precast (4–20%) gradient SDS–PAGE gel (GenScript) run in MOPS buffer. Proteins were then transferred onto nitrocellulose membranes. Membranes were blocked with 5% milk in PBS 0.1% TWEEN 20 (Sigma) for 15 min at 20–25 °C followed by overnight incubation with the primary antibodies against LDHA (polyclonal), RPL11 (D1P5N), MYC (D84C12), CcnaA2 (E9Q5G), CDC20 (D6C2Q), P53 (D2H90), γH2AX (polyclonal), AID (L7E7), β-ACTIN (E4D9Z), GAPDH (D4C6R), β-TUBULIN (D3U1W) and Histone H3 (D1H2) from Cell Signaling Technology, RFX7 (NBP1-71819) from Novus (1:1,000 dilution), RFX7 (A303-061A) from Bethyl Laboratories (1:1,000 dilution) and VINCULIN (7F9) from Santa Cruz Biotechnology (1:3,000 dilution) in 5% milk in PBS–0.1% TWEEN 20 at +4 °C. Blots were washed with PBS 0.1% TWEEN 20 and incubated with respective HRP-conjugated IgG secondary antibodies from Cell Signaling Technology (1:2,500 dilution) for 30 min at 20–25 °C followed by six rounds of washing and developed using chemiluminescent substrate solution (WesternBright ECL HRP Substrate, Advansta) and imaged by image analyzer machine (Fusion FX, Vilber). Immunoblot images were analyzed using ImageJ software⁷¹.

Confocal microscopy

Spleens were fixed in 4% paraformaldehyde overnight at 4 °C and sections (150 μm) were cut with a Leica VT1200S Vibratome. Sections were stained for IgD (11–26c.2a conjugated to Alexa Fluor 594, Biolegend) and GL-7 (GL-7, conjugated to Alexa Fluor 647, Biolegend). Mosaic Z stack images were taken with a Leica Stellaris 8 laser scan confocal microscope, with a HC PL APO CS2 ×10/0.40 numerical aperture DRY objective. Excitation was performed with a White Light Laser (excitation lines 590 nm and 653 nm) and collection windows were set to 595–652 nm and 662–829 nm, respectively. Acquisition was set to pinhole 2 a.u. in bidirectional mode, with a pixel size of 570 nm, pixel dwell time of 0.6 µs and line average of 2. Mosaic Z stacks of a total size of about 50 µm each with a Z step of 5 µm were acquired and then later reconstructed with LAS X software. Images were analyzed using ImageJ software⁷¹: from Z-stacks, a Maximum Intensity projection was created, minimum, maximum and contrast were set linearly according to control images and then images were overlaid to a final composite and converted to RGB images. Regions of interest were drawn manually.

Histology and immunohistochemistry

Tissues were fixed in 4% formaldehyde for up to 48 h. The tissues were embedded in paraffin, cut to a thickness of 2 µm and then stained with hematoxylin and eosin (H&E) (Medite cat. no. TST 15). The slides were analyzed at the Istituto Cantonale di Patologia di Locarno, EOC.

For immunohistochemistry (IHC), following paraffin embedding, serial sections (4-μm thick) were obtained using a microtome and incubated at 60 °C for 1 h in a dry oven. Deparaffinization was performed with two consecutive 5-min incubations in OTTIX Plus (Diapath, cat. no. X0076), followed by a 5-min incubation in OTTIX Shaper (Diapath, cat. no. X0096). Heat-induced epitope retrieval was carried out in either citrate buffer (pH 6.0) or Tris-EDTA buffer (pH 9.0), depending on the antibody, at 98 °C for 20 min. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide (VWR Chemicals, cat. no. 23615.248) for 10 min and nonspecific binding was minimized using Protein Block solution (DAKO, Agilent Technologies, cat. no. X0909) for 10 min. Sections were incubated for 1 h at 20–25 °C with primary antibodies against PNA (VectorLabs cat. no. B-1075-5 1:400 pH 6), CD45R/B220 (BioLegend cat. no. RA3-6B2 1:1,000 pH 9) and IRF-4 (CST cat. no. E8H3S 1:600 pH 6) followed by appropriate secondary detection and chromogenic visualization. Slides were counterstained with H&E, dehydrated and coverslipped. Slides were scanned using the Aperio AT2 Slide Scanner (Leica Biosystems) and digital images were analyzed using ImageScope software.

CDR3-sequencing

Genomic DNA from formalin-fixed and paraffin embedded samples was extracted using DNeasy Blood and Tissue Kit (Qiagen), following the manufacturer’s instructions. Optimal cutting temperature compound-frozen samples were processed and embedded subsequently in paraffin before DNA extraction. For sequencing, libraries were constructed using Ion AmpliSeq Mouse BCR IGH SR Assay (Thermo Fisher Scientific) according to the manufacturer’s instructions, using a DNA input ranging from 50 ng to 100 ng. Template preparation and enrichment were carried out on Ion Chef instrument (Thermo Fisher Scientific) and pooled libraries were sequenced on Ion GeneStudio S5 Prime (Thermo Fisher Scientific) using Ion 550 Chips, yielding a total of 1.3–4.6 million reads per sample.

Sequencing data were analyzed using Ion AmpliSeq Mouse BCR IGH SR (w.1.4) workflow on Ion Reporter software v.5.20.8.0 (Thermo Fisher Scientific). Resulting sequences were analyzed with the online version of IgBLAST using the default alignment options (https://www.ncbi.nlm.nih.gov/igblast/igblast.cgi) to analyze mutations in the CDR3 sequence.

Flow cytometry

Spleen, BM and iLNs were processed to single-cell suspensions and erythrocyte lysis was performed for spleen and BM with hypotonic buffer (155 mM NH₄Cl, 10 mM KHCO₃, 0.1 mM Na₂EDTA). For flow cytometry analysis, cells were preincubated with anti-CD16/32 (clone 2.4G2) from invivoMab, to block Fc receptors and then surface stained with antibodies against CD19 (6D5), CD21 (7E9), CD23 (B3B4), CD38 (90), CD86 (GL-1), CD93 (AA4.1), CD138 (281-2), B220 (RA3-6B2), IgM (11/41), IgD (11-26.c2a), IgK (187.1), IgL (RML-42) and CXCR4 (L276F12). Antibodies were purchased from BioLegend, Invitrogen and eBioscience, anti-CD95 (Jo2) from BD Biosciences. Dead cells were stained using propidium iodide (ImmunoChemistry Technologies), or Zombie yellow Fixable Viability kit (Biolegend). Staining was performed with appropriate combinations of fluorophores. Intracellular stainings were performed with antibodies against Ki67 (SolA15), Bcl6 (7D1), Irf-4 (IRF-4.3E4), Pax5 (1H9) and Blimp-1 (5E7) purchased from BioLegend and Invitrogen and with c-Myc/n-Myc (cat. no. D3N8F) from Cell Signaling Technology using the Thermo Fisher Scientific Foxp3/Transcription Factor Staining Buffer Set (cat. no. 00-5523-00). After surface staining, cells were fixed with fixation/permeabilization buffer and then further permeabilized according to the manufacturer’s instructions. The intracellular staining antibody was added in permeabilization buffer for 60 min at 20–25 °C. Data were acquired with a Becton Dickinson LSRFortessa flow cytometer and analyzed in FlowJo software (Tree Star).

Cloning and transfection

The pLenti-ORF clone of human WT RFX7 with a Myc-DDK tag (C-terminal) construct was purchased from OriGene (RC211086L1). The different truncations of RFX7 were generated by PCR using the Q5 High Fidelity polymerase and subcloned (BamHI–MluI) into pLenti-Myc-DDK vector. The following primers were used: common forward primer (5′-GACTGGATCCGGTACCGAGGA-3′) and distinct reverse primers for each truncation: Q355rev (5′-AAAACGCGTTTGAGGAGAAAGGATTGAAGGATTTCC-3′) Q688rev (5′-AAAACGCGTCTGCCCTTCTATGGTAGCTGCTG-3’) W782rev (5’-AAAACGCGTCCATCCATTTGGATTAAAGCTGCCAAC-3′) Q870rev (5′-AAAACGCGTCTGGGAAACAGAAGGCTCAAACTC-3′). pcDNA3.1(+)-N-term. eGFP-RFX7 WT and mutants and pcDNA3.1(+)-N-term. mCherry-RFX7 WT were outsourced (GenScript cloning service). These constructs contain the full CDS of human RFX7 bearing a STOP codon immediately after the amino acid Q688 or W782. The NLS mutant contains the following single amino acid mutation: K673A–K674A.

HEK293T cells (from ATCC) were seeded at 2.5 × 10⁶ per 100-mm dish and transfected 24 h later using JetPrime reagent (Polyplus) according to the manufacturer’s instructions always keeping the final amount of DNA to 5 µg. Transfected cells were harvested 24 h post-transfection and FACS-sorted according to eGFP to select transfected cells. Protein expression was verified by immunoblot.

U-2932 cells were transfected using the NEON system (Invitrogen) and the 100-µl gold-coated tips. Briefly, 3 million cells were resuspended in 110 µl of R buffer (resuspension buffer as provided by the Neon Transfection System 100 µl kit) and plasmids encoding the protein of interest were added keeping the amount of final DNA constant (21 µg). Electroporation setting: 1,350 V, 20 ms, two pulses. Immediately after electroporation, cells were kept overnight in a 6-well plates in RPMI antibiotic free medium. Cells were analyzed 24 h post-transfection.

RT–qPCR

Cell pellets were frozen in TRIzol Reagent (Life technologies) and RNA was extracted according to manufacturer’s instructions. Genomic DNA was removed by treating the RNA with Turbo DNase (Thermo Fisher) before RT, according to the manufacturer’s instructions. In some experiments RNA was extracted using RNeasy Plus Mini Kit (QIAGEN) according to the manufacturer’s instructions. A total of 2 μg of RNA was retrotranscribed using either random primers (Life technologies) or oligo (dT)20 primer (Microsynth). Annealing was performed at 70 °C for 5 min, followed by incubation with M-MLV Retro transcriptase (Promega), RNasin (Promega) and deoxyribonucleotides (Roche) at 40 °C for 10 min, 45 °C for 50 min and 70 °C for 15 min. Primers used for quantitative PCR (qPCR) were reported previously³ or described in Supplementary Table 5. qPCR was performed using SYBR GREEN based master mix (PerfeCTa SYBR Green SuperMixes and FastMixes; Quantabio).

Chromatin immunoprecipitation

HEK293T cells were seeded at 2.5 × 10⁶ per 100-mm dish 24 h before transfection with empty plasmid (mock) and eGFP-tagged WT RFX7 or indicated mutants using JetPrime reagent according to manufacturer’s instructions, always keeping the final amount of DNA to 5 µg. Cells were harvested and sorted for eGFP-positivity 24 h later. Chromatin was prepared using the truChIP Chromatin Shearing Kit (Covaris). Briefly, after sorting cells were washed twice in ice-cold PBS (200rfc, 5 min), resuspended in 1.5 ml of fixing Buffer A and 11.1% paraformaldehyde added to reach 1% final concentration. The cell suspension was incubated for 10 min at 20–25 °C and quenched by adding quenching buffer E for 5 min at 20–25 °C. Cells were collected by centrifuging at 500rcf for 5 min at 20–25 °C. The cell pellet was washed twice with ice-cold PBS. Nuclei were prepared following manufacturer’s instructions for high cell numbers. Shearing of the chromatin was performed using 1-ml AFA tubes (Covaris) using Covaris sonicator and following the manufacturer’s instructions. Chromatin was reverse cross-linked and purified using QIAGEN DNA purification columns and quality checked on a 1% agarose gel (fragment range between 200 bp and 1,000 bp). DNA LoBind Eppendorf tubes or Safe Lock Eppendorf tubes were used throughout the process. Sheared chromatin (10 µg) was precleared using 60 µl of Sepharose 6B beads supplemented with Protease Inhibitor (Cocktail 2 SIGMA) and Salmon Sperm DNA (10 mg ml⁻¹; Invitrogen) (1:100 dilution) in a final volume of 600 µl for 1 h at 4 °C with rocking. After centrifugation at 5,200g, 1% of material was transferred to a clean 1.5-ml tube and used as input. The remaining precleared chromatin was incubated overnight at 4 °C with either 10 µg of anti-GFP antibody (abcam ab290) or Isotype Control antibody (Cell Signaling). The next day, 30 µl per reaction of Protein A beads was added to the mix of chromatin and antibody followed by 90 min of incubation at 20–25 °C with rotation. After this time, beads went through a series of washes (seven to ten washes). Tubes were changed at each buffer change. After the final wash, 110 µl of room temperature (20–25 °C) Elution Buffer (EB: 10 mM Tris-HCl pH 8.0, 5 mM EDTA, 300 mM NaCl, 0.5% SDS—prepared and stored at 20–25 °C after filtering) were added to the beads and to the input material. For each sample a pre-mix of: 6 µl of EB and 2 µl of RNase A (10 mg ml⁻¹) was prepared and 8 µl added to each tube. Tubes were incubated at 37 °C for 30 min with shaking. Meanwhile, for each sample a pre-mix of: 3 µl of EB and 2.5 µl of Proteinase K (20 mg ml⁻¹) was prepared and 8.5 µl added to the tubes after the 30 min incubation with RNaseA. Tubes were incubated for additional 2 h at 37 °C shaking, followed by overnight incubation at 65 °C. The next day, beads were centrifuged at 5,200g and the supernatant transferred to a clean tube. DNA was extracted using the QIAGEN DNA purification columns. qPCR was performed using the SYBR GREEN based master mix (PerfeCTa SYBR Green SuperMixes and FastMixes; Quantabio) and the primers described in Supplementary Table 5.

Ex vivo and CpG-stimulated B cell and U-2932 cell library preparation and sequencing

FO B cells were FACS-sorted as B220⁺/CD19⁺/CD23⁺ and either processed directly or activated with CpG (2 μg ml⁻¹) for 18 h in vitro. Total RNA was extracted from sorted cells with Monarch Total RNA Miniprep Kit (New England Biolabs) according to the manufacturer’s instructions. For U-2932, cells were transfected using the NEON system (Invitrogen) and 100-µl gold-coated tips (three independent transfections). Briefly, three million cells were resuspended in 110 µl of R buffer and plasmids encoding the protein of interest were added keeping the amount of final DNA constant (21 µg). Electroporation setting: 1,350 V, 20 ms, 2 pulses. Immediately after electroporation, cells were seeded in a 6 wells plate in RPMI antibiotic free medium and analyzed 24 h later. Total RNA was extracted using RNeasy Plus Mini KIT (QIAGEN) from double- or single-positive eGFP/mCherry sorted cells according to manufacturers instruction. Quality controls were performed on Bioanalyzer 2100 using RNA Integrity Number (RIN) > 8 (Agilent Technologies) and Qubit V4 (Invitrogen Thermo Fisher Scientific). NEBNext Ultra Directional RNA Library Prep Kit for Illumina (New England BioLabs Inc.) was employed with the NEBNext Multiplex Oligos for Illumina (New England BioLabs Inc.) and NEBNext Poly(A) mRNA Magnetic Isolation Module for cDNA synthesis with the addition of barcode sequences. The sequencing of the pre-pools was performed using the NextSeq2000 (Illumina) with the P2 reagents kit V3 (100 cycles; Illumina). Samples were processed starting from stranded, single-ended 120-bp-long sequencing reads.

Ex vivo, CpG-stimulated B cell and transfected U-2932 bioinformatic data analysis

Read quality was verified initially using FastQC (https://www.bioinformatics.babraham.ac.uk/index.html) and MultiQC⁷². Reads were trimmed subsequently to remove sequencing adapters using cutadapt v.2.8⁷³ or Trimmomatic v.0.39. Estimates of transcript abundance for all transcripts was performed using HISAT2 v.2.1.0⁷⁴ using either the mouse reference genome version GRCm39 or GENCODE (v.GRCh38). HTSeq-count v.2.0.2⁷⁵ was used to generate the table of counts containing the number of reads mapping to each feature in each sequencing sample. Expression levels were calculated as counts per million (CPM). Statistical testing of differential expression of genes was performed using DESeq2 with the Wald test⁷⁶. For mouse data, transcripts were considered differentially expressed when false discovery rate (FDR) < 0.05, whereas for U-2932 transcripts were considered differentially expressed when FDR = 0.05 and log₂ fold change (FC) = 1. GSEA was performed using the Hallmark gene sets from MSigDB³¹ using fast GSEA (fGSEA)⁷⁷, a geneset was considered significant if Q < 0.05. Statistical computation and visualization (heatmap, volcano plot, activation maps) were performed using the Omics Playground v.4.0.4 (ref. ⁷⁸).

JASPAR analysis

Transcription factor binding sites were analyzed with the JASPAR tool (http://jaspar.genereg.net/)²⁹. Promoter sequences were collected in the eukaryotic promoter database (https://epd.expasy.org/epd/), to span −700 to +300 bp over the described transcription start site of each gene. A relative profile score threshold of 85% was used to screen for similarities with the matrix profile MA1554.1 described in the JASPAR CORE database. Screened promoters belong to the control, the Rfx7-target and the E2F upregulated gene groups. The latter corresponds to the top 20 upregulated genes of the E2F signature²⁹ in CpG-activated Rfx7-deficient B cells based on our RNA-seq data (Stmn1, Ccnb2, Cdca3, Birc5, Plk1, Kif22, Tcf19, Hmmr, Cdk1, Cdca8, Cks1b, Tk1, Rrm2, Tacc3, Ccne1, Aurka, Cdc20, Rad51c, Spag5 and Mxd3). The control group corresponds to 34 nonmodulated genes (log₂FC > −0.018 and <0.024; meta.Q > 0.99, chosen randomly) selected from the list of genes from our RNA-seq (Adamdec1, Btnl6, C1qa, C6, Ccr3, Cd14, Cd200r1, Cd276, Cd300a, Cd55b, Clec4n, Clec7a, Clu, Cxcl9, Dysf, Ephx2, Fcgr1, Fpr1, Gad1, Hcar2, Hpgd, Hs3st2, Ifit3b, Kctd12b, Loxhd1, Lsmem1, Mafb, Paqr9, Pf4, Pilrb2, Pparg, Siglec1, Slco2b1 and Tlr8). The 42 Rfx7-targets were selected among RFX7 targets identified by ChIP⁵ and found to be tendentially decreased (log₂FC < −0.2) in the current or previously published mouse RNA-seq data³ (Dip2a, Klf9, Ptms, Tob2, Arl15, Dolpp1, Maf, Top2b, Atrx, Mxd4, Rap2a, Cabin1, Emc9, Rev3l, Tsc22d1, Rfx5, Ccnd1, Fam214a, Pdcd4, Rmnd5a, Tspyl2, Ccng2, Hnrnpul2, Ube2h, Cdk4, Pik3ip1, Sesn3, Cic, Ip6k2, Slc43a2, Ypel2, Junb, Plcxd2, Sox12, Cks2, Kdm4a, Pnrc1, Sox4, Ddit4, Kdm6b, Prkcz and Tob1).

Protein structural prediction of truncated and full-length RFX7

The full-length structure of human RFX7 (UniProt ID Q2KHR2) was predicted using the Boltz-1 server⁷⁹, which implements an AlphaFold3-based algorithm for protein structure prediction. The sequence of RFX7 was retrieved from the UniProt database and used as input. The prediction was performed including DNA as a ligand and using the X-ray structure of hRFX1 DNA-binding domain (PDB 1DP7) as template for the RFX7 DNA-binding domain. The structural models of the RFX7 forms ending at residue Q688 (RFX7 1–688) and at residue W782 (RFX7 1–782) were generated using SwissModel⁸⁰, using the full-length RFX7 model predicted by Boltz-1 as the template.

The structural stability of the predicted RFX7 1–688 and RFX7 1–782 models was evaluated using molecular dynamics simulations performed with GROMACS 2019 (v.2019.6). Each model was solvated in a TIP3P water box, neutralized with counterions, and energy-minimized before simulation. Production molecular dynamics runs were carried out under NPT conditions for 100 ns at 300 K using the AMBER99SB-ILDN force field⁸¹. Trajectory analysis, including root-mean-square deviation, was performed to assess model stability.

Statistical analysis

Statistical analyses were performed using Prism software (GraphPad v.5.0). Statistical tests are described in the figure legends. Outliers present in the dataset of Fig. 3b (and Extended Data Fig. 3c) were analyzed according to the literature⁸². After graphing outliers and checking with normal probability plots, we excluded normality of data. Thus, we used the interquartile range method to create outlier fences. Once outliers have been identified, we checked that the percentage was below a standard limit (such as 5% in trimmed mean) and we excluded those values, being confident of noninformativeness of outliers.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Peer review information

Nature Immunology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: L. A. Dempsey, in collaboration with the Nature Immunology team.

Extended data

is available for this paper at https://doi.org/10.1038/s41590-026-02526-2.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41590-026-02526-2.