Cis-regulatory somatic mutations and gene-expression alteration in B-cell lymphomas

Background With the rapid increase of whole-genome sequencing of human cancers, an important opportunity to analyze and characterize somatic mutations lying within cis-regulatory regions has emerged. A focus on protein-coding regions to identify nonsense or missense mutations disruptive to protein structure and/or function has led to important insights; however, the impact on gene expression of mutations lying within cis-regulatory regions remains under-explored. We analyzed somatic mutations from 84 matched tumor-normal whole genomes from B-cell lymphomas with accompanying gene expression measurements to elucidate the extent to which these cancers are disrupted by cis-regulatory mutations. Results We characterize mutations overlapping a high quality set of well-annotated transcription factor binding sites (TFBSs), covering a similar portion of the genome as protein-coding exons. Our results indicate that cis-regulatory mutations overlapping predicted TFBSs are enriched in promoter regions of genes involved in apoptosis or growth/proliferation. By integrating gene expression data with mutation data, our computational approach culminates with identification of cis-regulatory mutations most likely to participate in dysregulation of the gene expression program. The impact can be measured along with protein-coding mutations to highlight key mutations disrupting gene expression and pathways in cancer. Conclusions Our study yields specific genes with disrupted expression triggered by genomic mutations in either the coding or the regulatory space. It implies that mutated regulatory components of the genome contribute substantially to cancer pathways. Our analyses demonstrate that identifying genomically altered cis-regulatory elements coupled with analysis of gene expression data will augment biological interpretation of mutational landscapes of cancers. Electronic supplementary material The online version of this article (doi:10.1186/s13059-015-0648-7) contains supplementary material, which is available to authorized users.


GNA13
The G-protein subunit α − 13 (GNA13) has been previously linked to tumour progression and has been described as an important mediator of prostate cancer cell invasion [2]. It has also been previously identified as a recurrent target of mutation with inactivation impact in DLBCLs [3]. In our xseq analysis of cohort2, GNA13 was predicted to be mutated with altered expression in two Burkitt lymphoma samples. One sample (SA321004) was associated to a SNV in the protein-coding space whereas another one (SA32848) was associated to a SNV lying within three overlapping TFBSs (for the TFs GATA1, GATA3, and JUND). We predicted the SNV as disrupting the TFBS associated to GATA3. Given the GATA3 TF binding profile, we observed that the mutation was one of the most severe mutation that can occur to the TFBS (see Figure 8FGHIJ from the manuscript). In the corresponding sample, we observed an up-regulation of GNA13 transcription consistent with an important role of GNA13 for cancer cell invasion. Notice that GNA13 was also found up-regulated when associated to a mutation in the sample SA321004 associated to a mutation in an exon (Additional file 1: Figure S6).
The Burkitt lymphoma SA320848 sample highlighted a deregulation of expression of genes known to interact with GNA13 (see Additional file 1: Figure S6). We observed the up-regulation of the known oncogenes VAV1 [4] and ECT2 [5]. Moreover, the RHOB gene was down-regulated and has been shown to associate with cell proliferation and an increase in DNA double strand breaks [6].

BCL6
BCL6 was predicted with altered expression associated to mutations in four samples (one corresponding to a Burkitt lymphoma and three to DLBCLs) from cohort2. Three of the altered expression were associated to mutations in TFBSs (two of which were predicted to be disruptive) and one to a missense mutation in a protein-coding exon. Interestingly, we observed an up-regulation of BCL6 in samples SA320848 and SA320968 which were associated to mutations disrupting TFBSs. On the contrary, a down-regulation was observed in SA320962 associated to mutations in TFBSs (not predicted to be disruptive) and no expression change was observed in SA320872 associated to a mutation in an exon. All the mutations but one overlapping TFBSs and associated to BCL6 in the corresponding samples were found in the promoter region of BCL6 (see Additional file 1: Figure S7). In sample SA320962, a mutation was found to lie within two predicted overlapping STAT3 TFBSs. The mutation did not seem to disrupt the TFBSs.
The up-regulation of BCL6 in the two samples was associated to several mutations in TFBSs (7 SNVs in SA320848 lied within 10 TFBSs, 3 of which predicted to be deleterious; 5 SNVs in SA320968 lied within 8 TFBSs, 1 of which predicted to be deleterious). Three SNVs in SA320848 were predicted to disrupt TFBSs of the TFs BRCA1, E2F1, USF1, and BHLHE40 (a mutation overlaps two TFBSs). In SA320968, a SNV was predicted to disrupt a GATA3 TFBS. The presence of multiple SNVs lying within TFBSs in the core promoter of the BCL6 gene is an indication of a potential deregulation of BCL6 expression through the alteration of transcriptional regulation. The absence of mutation within the exonic regions in the same samples is supporting this hypothesis.
We next focused on the two disruptive TFBS samples to determine the potential cascading effect arising from the up-regulation of BCL6 in SA320848 and SA320968. Over expression of BCL6, an oncogene, enables for accelerated proliferation and makes the cell more lenient to DNA damage [7]. Both SA320848 and SA320968 samples showed a down-regulation of SMAD3 or SMAD4 and TXNIP which are all known to have biological interactions with BCL6 (see Additional file 1: Figure S8). The DLBCL sample SA320968 presented a down-regulation of SMAD3 and SMAD4 whereas the Burkitt lymphoma sample SA320848 displayed a down-regulation of SMAD3 (see Additional file 1: Figure S8). Over expression of BCL6 also inhibited the tumour suppressors SMAD3 and SMAD4. TXNIP was down-regulated in both samples. The down-regulation of TXNIP has already been observed in many cancer types including DLBCLs, and has shown to be a tumour suppressor in thyroid cancer [7].

ROBO1
Our approach allows for extracting genes for which little is known regarding their involvement in cancer. For instance, ROBO1, a transmembrane protein, was predicted with altered expression associated with mutations in six samples from cohort1 (see Figure 6 from the manuscript and Additional file 1: Figure S9). Five of these samples highlighted mutations in TFBSs and one sample was associated to a mutation in the proteincoding space. ROBO1 has been previously characterized as a potential tumour suppressor in different types of cancer [8,9,10]. Another study highlighted a tumour specific hypermethylation of the promoter region of the gene in breast tumours, repressing the transcription of the ROBO1 gene [9]. Here, we found that ROBO1 was down-regulated in the five samples associated to mutations in TFBSs whereas it was slightly up-regulated in the sample harbouring a mutation in the protein-coding region. Even though we did not predict any of the mutations to disrupt the overlapped TFBSs, they might impact the binding of the TFs or even create new TFBSs for competing TFs.
To identify the potential impact of the mutated predicted tumour suppressor ROBO1, we inspected the expression deregulation of known interacting genes. We focused on genes repeatedly shown to be deregulated when ROBO1 was predicted to be deregulated by mutations. Four out of the five DLBCL samples where ROBO1 was down-regulated and harbouring TFBS mutations exhibited either SOS1 or SOS2 as downregulated (see Additional file 1: Figure S10), which are Son of Sevenless (SOS) genes known to be involved in the Raf/Ras/MEK/ERK/MAPK pathway where the interaction of Ras with GDP/GTP-exchange factors is mediated by SOS [11]. Another ROBO1 interacting gene commonly deregulated in association with ROBO1 was RAC1. RAC1 has a high probability of being up-regulated in four out of the five samples where ROBO1 was associated to TFBS mutations; one other sample harboured a missense protein coding mutation in RAC1 (see Additional file 1: Figure S10). [8] showed that RAC1 was activated when ROBO1 was lost, inducing cell migration in lymphoma cells. This observation is in agreement with the up-regulation of RAC1 combined with the down-regulation of ROBO1 in all but one sample where ROBO1 was predicted to be associated to mutations within TFBSs. All these results taken together shed light into the supposed tumour suppressor role of the ROBO1 gene. We highlighted here that ROBO1 might be down-regulated at the transcriptional level by mutations in cis-regulatory elements while no mutations were found in the protein-coding space.