RBFOX2 and alternative splicing in B-cell lymphoma

RBFOX2 is a master regulator of alternative splicing¹. This RNA-binding protein (RBP) is expressed in the brain², muscle³, and embryonic stem cells⁴. RBFOX2 is required not only for the proper splicing of target RNAs, but also for cerebellar development², myogenesis³, and for survival of human embryonic stem cells⁴.

Hitherto, not much is known about the expression and function of RBFOX2 in hematopoetic tissues. In an early report, RBFOX2 had been shown to be capable of promoting inclusion of exon16 in protein 4.1R⁵. This splicing event is important for erythropoiesis because it increases the affinity of 4.1R for target genes⁵. In a recent study, expression of RBFOX2 was detected in the human T-cell line JURKAT and a functional antagonism of the RBPs RBFOX2 and CELF2 was demonstrated⁶. We wanted to describe the expression patterns of RBFOX2 in hematopoetic malignancies, to discover target genes and to unravel the consequence of RBFOX2 repression for target gene splicing and isoform expression.

Expression array and Western blot analysis showed that human B non-Hodgkin lymphoma (B-NHL) cell lines are RBFOX2 negative or positive (Supplement 1A). To find the potential targets of the splice factor RBFOX2, we compared the expression of individual exons in RBFOX2-negative and RBFOX2-positive cell lines. This approach relied on the assumption that the differential expression of this RBP would provoke changes in the expression of individual exons and would thereby allow identification of target genes. Supplement 1B shortlists these genes ordered by statistical significance. Expression of the individual exons and joining sequences of MALT1 is shown as heatmap in Fig. 1a. The full-length MALT1 variant was associated with RBFOX2 expression (Fig. 1a).

Fig. 1. RBFOX2 and RBFOX2 targets in B-NHL cell lines.

a Heatmap of RNAseq data showing expression of individual MALT1 exons and corresponding joining sequences (red boxes: exon 7 and exon 7 joining sequences). Expression of full-length MALT1 correlates with expression of RBFOX2. b Expression array analysis (upper) and RT-PCR analysis (lower) revealed that increasing expression of RBFOX2 was paralleled by the full-length isoforms of MALT1, CLSTN1, FMNL3, and MYO9B. Exon numbering refers to the following sequences: MALT1 (NM_006785.3), CLSTN1 (NM_001009566), FMNL3 (ENST00000550488.5), MYO9B (NM_001130065). c Transfection with siRNA oligonucleotides efficiently downregulated expression of RBFOX2 mRNA (upper) and protein (medium). Repression of RBFOX2 resulted in an increase of the short isoforms of MALT1, FMNL3 and MYO9B (lower)

Results of splice variant analysis with a larger panel of cell lines revealed a striking association between expression of RBFOX2 and expression of the full-length forms of all four candidate target genes, MALT1, CLSTN1, FMNL3, and MYO9B (Fig. 1b). The short variants were expressed in RBFOX2-negative cell lines only (Fig. 1b). Two of these potential RBFOX2 target genes (CLSTN1 and FMNL3) had already been described in the context of RBFOX2-mediated splicing⁷. The RBFOX2 target sequence “UGCAUG” was present in all introns following the retained exons, indicating that high RBFOX2 levels might be the cause of the full-length forms in the RBFOX2 positive cell lines. Supporting the notion that RBFOX2 was important for splicing of these genes was also the finding that RBFOX2 was the sole gene that was significantly overexpressed in cell lines expressing full-length MALT1 when compared to cell lines expressing MALT1 without exon 7 (Supplement 1C).

We performed knockdown experiments to test whether RBFOX2 was responsible for retaining MALT1 exon 7, FMNL3 exon 26, and MYO9B exon 37. siRNAs reduced expression of RBFOX2 in RBFOX2-positive cell lines BL-2, SU-DHL-5, and HT by more than 50% (Fig. 1c, Supplement 1D). Repression of RBFOX2 induced the shorter isoforms of MALT1 (w/o exon 7), MYO9B (w/o exon 37), and FMNL3 (w/o exon 26) (Fig. 1c, Supplement 1D). The long form of CLSTN1, the fourth gene tested here, was not or only marginally expressed in BL-2, SU-DHL-5, and HT cells, explaining why we could not observe an increase of the short isoform of this gene after RBFOX2 knockdown (data not shown). In sum, our data showed that isoforms of MALT1, MYO9B, and FMNL3 in B-NHL cell lines are controlled by RBFOX2: the full-length RNAs were expressed when RBFOX2 was high, the short variants prevailed when RBFOX2 was repressed.

RBFOX2 is part of a mesenchymal splicing network⁷. The gene is also essential for the viability of human embryonic stem cells and for normal cerebellar development of mice^2,4. Thus, RBFOX2 appears to have different functions in cells of different origin. To check whether RBFOX2 induced splicing of MALT1, CLSTN1, FMNL3, and MYO9B in hematopoetic cells other than B-cells, we tested RBFOX2 expression and expression of the splice variants of the putative RBFOX2 targets in T- and myeloid cell lines.

Neither T-cell lines nor myeloid cell lines reached the RBFOX2 mRNA expression level of the B-cell line HT. Nevertheless, the T- and myeloid cell lines with highest RBFOX2 mRNA levels also expressed the protein (Supplement 1A). T-cell lines with high RBFOX2 levels expressed the large isoform of MALT1. However, an apparent universal dependence of RBFOX2 and the large isoforms of MALT1, CLSTN1, FMNL3, and MYO9B comparable to that in B-cell lines was neither found in T- nor in myeloid cell lines. This does not necessarily mean that RBFOX2 is functionless in cells of the T-lymphoid and myeloid lineages. RBFOX2 is expressed in cell lines of both entities, and the full-length MALT1 isoform is expressed in RBFOX2-positive T-cells. The other three genes (CLSTN1, FMNL3, and MYO9B) appear to be targets of RBFOX2 in the B-lineage only. These results suggest that tissue-specific factors might contribute to the splicing process mediated by RBFOX2.

We limited our further studies to B-NHL, because we had identified the RBFOX2 target genes in B-NHL cell lines. As shown in cell lines, also primary tumor cells of patients with diffuse large B-cell lymphoma (DLBCL) show differential RBFOX2 gene expression (Fig. 1b upper, Supplement 1E). We analyzed RNAseq data from patients with different forms of B-NHL (ICGC MMML-Seq consortium) to find out whether primary tumor cells exhibited the same correlation between RBFOX2 expression and the RBFOX2 target gene isoforms as detected in B-NHL cell lines. We checked samples from patients with DLBCL (n = 78), Burkitt lymphoma (BL) (n = 21), follicular lymphoma (FL) (n = 87), and FL-DLBCL (n = 15). Germinal center (GC) B-cells (n = 5) and naive B-cells (n = 5) were included as controls.

RBFOX2 expression and MALT1 exon 7 inclusion were positively correlated in BL, FL, activated B-cell (ABC), and GC DLBCL (p < 0.05) (Table 1). In contrast, no such correlation was found for healthy controls, DLBCL (type III) and FL-DLBCL (Table 1). Supporting the notion that RBFOX2 regulates splicing in all four proposed RBFOX2 target genes (MALT1 exon 7, CLSTN1 exon 11, FMNL3 exon 26, and MYO9B exon 37), we found a statistically significant positive correlation between expression of RBFOX2 and inclusion of target exons in FL, BL, and in at least one subtype of DLBCL (Table 1). The data had been normalized against target gene expression levels to avoid a potential bias through target gene expression levels.

Table 1.

Correlation between expression of RBFOX2 and inclusion of exons in RBFOX2 target genes

	MALT1 exon 7		CLSTN1 exon 11		FMNL3 exon 26		MYO9B exon 37
	Correlation	p-value	Correlation	p-value	Correlation	p-value	Correlation	p-value
DLBCL ABC (n = 26)	0.496	0.009	0.450	0.023	0.139	0.496	0.338	0.09
DLBCL GCB (n = 37)	0.591	0.0001	0.147	0.383	0.558	0.0004	0.561	0.0003
DLBCL type III (n = 15)	0.125	0.675	0.385	0.156	1	0	0.307	0.265
All DLBCL (n = 78)	0.522	1.2 x 10−6	0.303	0.007	0.300	0.008	0.566	1.1 x 10−7
BL (solid ped BL) (n = 21)	0.703	0.0005	0.669	0.001	0.484	0.027	0.672	0.001
FL (n = 87)	0.321	0.003	0.437	2.3 x 10−5	0.227	0.034	0.397	0.0001
FL-DLBCL (n = 15)	0.464	0.083	0.029	0.923	0.503	0.058	0.435	0.106
GC B cells, control (n = 5)	−0.9	0.083	0.2	0.783	0.3	0.683	−0.2	0.783
Naive B cells, control (n = 5)	0.6	0.083	-0.354	0.783	-0.8	0.683	−0.2	0.783

RNASeq data from lymphoma and control, mapped to hg38 with segemehl 2.0; data were normalized against target

gene expression. Bold: statistically significant. Normalization: transformation to target gene expression levels

RBFOX2 is a member of the RBFOX family of RBP, also including RBFOX1 and RBFOX3. All three proteins recognize the same sequence (UGCAUG) in regulated exons or in flanking introns⁸. To analyze whether RBFOX1 and RBFOX3 might also contribute to the splicing of our four target genes, we tested for correlation between expression of these RBFOX family members and inclusion of exons in target genes. We did not find a statistically significant correlation between RBFOX1 or RBFOX3 expression and inclusion of exons in CLSTN1 and FMNL3 (Supplement 1F). MALT1 and MYO9B showed this correlation only in selected tumor variants, but not in BL, ABC DLBCL, or GC DLBCL, when the latter two were analyzed as individual lymphoma entities (Supplement 1F). Thus, RBFOX2 was the only RBFOX family member whose expression was positively correlated with the full-length isoforms of the target genes (MALT1, CLSTN1, FMNL3, and MYO9B) in BL, FL, and DLBCL.

These data suggest that RBFOX2 is a regulator of splicing in B-NHL. This notion is based on (i) the positive correlation between RBFOX2 expression and expression of the full-length variants of the putative RBFOX2 target genes in B-NHL cell lines and in primary B-NHL samples, and (ii) results of knockdown experiments demonstrating that RBFOX2 is responsible for inclusion of exons in MALT1 and other target genes.

MALT1 appears to be of special interest as it encodes a protease that activates the IKK complex⁹. In lymphocytes, MALT1 cleaves RelB, which also leads to the activation of NFkB¹⁰. Both MALT1 isoforms (with and w/o exon 7) are expressed in T-lymphocytes and expression of the individual variants has consequences for T-cell receptor triggered signal transduction¹¹. As part of the CARMA1–BCL10–MALT1 complex, MALT1 is also a central regulator of the B-cell receptor (BCR) / NFkB pathway. ABC-type DLBCL cells rely on the constitutive activation of this pathway to block apoptosis¹². Recurrent mutations in CD79A/B, CARD11, and other BCR/NFkB pathway genes have been described¹³. Like Brutons Tyrosine Kinase, upstream to MALT1 in the BCR/NFkB pathway, also MALT1 is a potential target for precision therapy¹⁴. Future studies shall elucidate whether the two MALT1 isoforms display different capacities to activate NFkB in B-NHL, which might be of importance for the clinical application of MALT1 inhibitors.

In summary, (i) RBFOX2 is expressed in hematopoetic cell lines of different origin; (ii) expression of RBFOX2 correlates with isoforms of potential target genes in B-NHL cell lines and in primary B-NHL cells; and (iii) knockdown experiments suggest that RBFOX2—directly or indirectly—contributes to the splicing of target genes including MALT1, a protease in the BCR/NFkB pathway.

Author contributions

H.Q. Study conception and design, manuscript writing. C.P.: Analysis of RNAseq data, expression array analysis, statistical analysis. S.H.B.: Analysis of RNAseq data, statistical analysis. W.G.D.: Knockdown experiments. V.H.: Acquisition of data. S.H.: RNAseq data analysis. S.N.: Knockdown experiments. R.S.: Coordination of the ICGC-MMML-Seq and the primary lymphoma data generation and provision, interpretation of data. C.C.U.: Analysis and presentation of data. M.Z.: Acquisition of data. H.G.D.: Provision of cell lines and study conception. All authors read and approved the final manuscript.

The authors declare that they have no conflict of interest.

Electronic supplementary material

Supplementary Information accompanies this paper at (10.1038/s41408-018-0114-3).