We determined the positioning of nucleosomes across exons 7, 8, 9, 10 or 14 of the MLH1 gene in LCLs derived from six Lynch syndrome patients. Splice site mutations within or near these exons affected splice donor or acceptor sites (Table 1). NOMe-seq and single molecule sequencing allowed us to distinguish mutant from wild-type alleles using the splice site sequence mutation. At all sites analysed the majority of molecules were inaccessible at GpC sites and methylated at CpG sites (Figure 1). NOMe-seq analysis of the control gene HSPA5, a constitutively active gene that maintains a nucleosome-depleted region at the transcription start site [18], showed that the CpG island was unmethylated (data not shown) and confirmed that GpC inaccessibility was due to nucleosome occupancy rather than incomplete M.Cvi PI treatment (Figure 2). This shows that the majority of molecules at the sites analysed are occupied by nucleosomes in these cells.

Separation of alleles by presence or absence of the splice site mutation showed no overall difference in either nucleosome positioning or occupancy between wild-type and mutant alleles (Figure 1 and Table 1). In the six splice site mutations analysed, three were located within donor splice sites and three were located within acceptor splice sites (Table 1). Though some difference in precise positioning between mutant and wild-type alleles was observed across some exons (for example, exons 7 and 10, Figure 1B) this was confined to a small subset of molecules, rather than the majority of molecules as would be expected if the mutations affected nucleosome positioning. Our study therefore shows that nucleosome occupancy is unaltered at exon-intron boundaries containing these splice site mutations.

In this study, we show that nucleosome positioning is unaltered at exon-intron boundaries containing splice site mutations. This finding suggests that the positioning of nucleosomes at exon-intron boundaries containing splice site mutations does not play a role in the splicing defect.

NOMe-seq has been used to measure nucleosome occupancy at gene promoters, CTCF binding sites [19] and regulatory elements [20] but this is the first study that has utilised the technique to measure nucleosome occupancy at splice sites. NOMe-seq previously showed that the active CpG island (CGI) promoter of HSPA5 has a nucleosome-depleted region (NDR) upstream of the transcriptional start site [19]. This region was included as a control in each assay, and in all cases the NDR was present. This indicated that the absence of GpC methylation observed in MLH1 was caused by nucleosome occlusion, not an artefact caused by a defective M.Cvi PI enzyme. In addition, CpG methylation was absent from the HSPA5 NDR which is a feature of active CGI promoters [21]. Finally, gene body methylation is a feature of human cells [22],[23] and the majority of CpG sites within the MLH1 exons and introns were methylated. Together, these controls indicated that the NOMe-seq assays captured the true state of nucleosome positioning at MLH1 splice sites.

Alternative splicing of MLH1 occurs in a range of normal tissues [24], but several pieces of evidence indicate that the transcripts observed in our study are caused by the mutation, rather than being normal splicing events. First, the splicing aberrations were predicted with bioinformatic splicing software [13], and second, although one of the aberrant transcripts (Δ10) has been reported among naturally occurring MLH1 splice transcripts, the clinical data associated with these variants are also indicative of pathogenicity as demonstrated by the results of previous multifactorial likelihood analyses [13].

One strength of our study was the ability to determine the allele-specific position of nucleosomes by incorporating a heterozygous single nucleotide variant into each NOMe-seq assay. Small changes in positioning were observed but these molecules constituted approximately the same small proportion of total molecules on both alleles. Subtle changes in nucleosome positioning have been considered important in regulating the expression of cell cycle-dependent genes [18] and enhancer accessibility [25], but in those studies the subtle changes were observed on a significant number of molecules that resulted in an overall change in nucleosome occupancy. There was no consistent alteration in nucleosome occupancy between the wild-type and mutant alleles at a variety of donor and acceptor sites, which would be expected if nucleosome positioning at these sites played a role in mediating the effects of splice site mutations. Together our data shows that alleles containing a splice site mutation show no differences in nucleosome occupancy to wild-type alleles.

A limitation of the approach was that in most of the assays (assays 2 to 6) it was only possible to measure nucleosome positions at either the donor site or the acceptor site. Bisulfite conversion causes fragmentation of genomic DNA and it is technically challenging to amplify fragments greater than 500 bp in length [26]. The donor and acceptor sites were separated by more than 2 kb of intronic sequence, making it impossible to concurrently determine the allele-specific position of nucleosomes at the donor and acceptor site on the same molecule. Although an independent NOMe-seq assay could be performed at the other donor or acceptor site, it would provide no information on allele-specificity. The data from one assay (c.588 + 1 G > T), however, in which both splice sites were present in the one amplicon, showed no significant change in nucleosome occupancy at the donor or acceptor site in intron 7.

Previous studies have utilised genome-wide datasets of micrococcal nuclease (MNase) digested chromatin to investigate nucleosome occupancy [3],[4]. Here we used NOMe-seq to map the position of nucleosomes relative to exons and splice sites in a single gene at single molecule resolution. This approach provides the most accurate possible measurement of nucleosome positioning and enables allele-specific mapping of nucleosomes. In agreement with previous reports in humans [3],[4], Caenorhabditis elegans[3],[4] and Drosophila melanogaster[3], we observed high nucleosome occupancy at exons. Thus, the findings from genome-wide studies were supported by our independent, single-molecule approach at the MLH1 gene.

A key finding of our study was that nucleosome positioning was not significantly affected by mutations at the MLH1 splice sites investigated. Recent work with an in vitro IKBKAP mini-gene system found that alternative splicing changed chromatin organisation, with splice site strength and factors needed for splicing, such as U1 snSNP, playing a role in the regulation of nucleosome occupancy in exons [27]. This and an earlier study [3] indicate that splice site strength is an important determinant of nucleosome occupancy in exons. Exons with stronger polypyrimidine tracts (PPT; one of the conserved DNA elements located at 3′ splice sites) have increased nucleosome occupancy compared to immediately adjacent introns, suggesting that nucleosomes act as a barrier that help to define the physical location of the 3′ splice site [3]. Although we investigated an equal number of donor and acceptor splice site mutations, nucleosome positioning was not significantly different between exons and the immediate 5′ and 3′ intronic sequence on the wild-type or mutant allele. The discrepancy between earlier studies and ours may in part relate to the strength of the MLH1 splice sites investigated. Alternatively, changes to chromatin organisation induced by splice mutations may be different to those caused by normal splicing events. Co-transcriptional splicing involves a complex interplay between RNAPII, the spliceosome [8] and histone modifications [3],[9] and alterations in these factors, in addition to DNA sequence changes may be needed to alter nucleosome occupancy at exons.

In summary this study used NOMe-seq to determine the allele-specific position of nucleosomes in relation to MLH1 splice site mutations. We conclude that splice site mutations that cause aberrant splicing of MLH1 do not alone significantly affect local nucleosome positioning in LCLs from Lynch syndrome patients.