We used PCA to help identify groups of histone marks () that displayed distinct profiles within active genes (Supplementary Fig. 1
). Among these, we identified a group of 10 marks biased for 5′ introns and roughly reciprocal with H3K36me3 near exons ( and Supplementary Fig. 2
). Sequence-related nucleosome occupancy alone was not sufficient to explain transitions between intronic and exonic histone modification regions (Supplementary Fig. 5
). The genome-wide phenomenon of intronic and exonic marks suggests that a general mechanism associated with splicing influences histone marking, as speculated previously15
. Thus, we tested if levels of RNA splicing influenced the profiles of these histone modifications. Neither inducible changes in exon inclusion in YPEL5 after several hours (), nor changes in inclusion of multiple exons in CD45 over numerous cell divisions () resulted in significant differences in H3K79me2 or H3K36me3. This contrasts starkly with transcriptional induction, in which H3K79me2 and H3K36me3 levels can increase within one hour34
Our results rule out a simple correlation between the extent of RNA splicing and histone modification levels. Intriguingly, regulated skipping of CD45 exon 4 is achieved by stalling spliceosome assembly at an early step following exon definition35
; therefore, spliceosome assembly may be sufficient to pattern histone modifications at CD45
exon 4, regardless of splicing outcomes. As it is now recognized that promoter marks are relatively stable36,37
and may reflect transcription initiation38
rather than productive transcription, we suggest by analogy that intronic and exonic marks are relatively stable and may reflect an aspect of the splicing process, such as exon definition, rather than productive splicing per se
If a feature related to splicing were required to influence histone modifications, we would expect a decrease in 5′ intron marks and an increase in H3K36me3 specifically at bona fide
exons, but not at ECRs, which have similar sequence composition, but are not spliced. Indeed, ECRs do not direct increased levels of H3K36me3 as at exons or decreased levels of certain 5′ intronic marks (i.e. H3K4me1, H3K4me2, H3K9me1, H3K23ac) (Supplementary Fig. 5
). In contrast, some 5′ intronic marks (e.g. H3 lysine 79 methylations) have nearly equivalent decreases near exons and ECRs, suggesting that the local sequence composition of exons (and ECRs) may be sufficient to influence these histone modification profiles. Thus, cis
-acting sequences, functioning as DNA or RNA, could signal to change histone modifications. For example, SR protein binding sites at exons could affect chromatin marking through “reverse” coupling8
. Additionally, elevated nucleosome occupancy at exons13,16-19
may play a role, independently or in conjunction with cis
-acting sequences. Extensive mutagenesis will be essential to distinguish between these possibilities and identify necessary sequences.
How distinct histone modification patterns are determined remains largely unknown. In agreement with Spies et al.18
, we find that histone modification patterns do not merely reflect nucleosome enrichment at exons, indicating the need for additional steps of regulation beyond nucleosome occupancy such as chromatin modifying enzyme action. Spatial similarities among the intronic marks could reflect cross-regulation between histone modifications39
. For example, the extensive overlap between H2Bub and H3K79me2/3 (Supplementary Fig. 2
) is consistent with direct stimulation by H2Bub of H3 lysine 79 methyltransferase activity by DOT1L and a DOT1L-containing protein complex40,41
. Also, H2Bub can directly stimulate human SET1-mediated methylation of H3 lysine 4 in vitro
, so H2Bub within 5′ introns may help to increase H3K4me1 and H3K4me2 (refs. 42,43
We propose that promoter, intronic, and exonic intragenic histone modification regions constitute three distinct chromatin zones at active human genes4,23
, analogous to those found in yeast10,44
: i) the “backbone” or promoter zone contains H3K4me3, ii) a 5′ zone includes H3K4me2 (and 9 other marks in humans, including H3K79me2), and iii) a 3′ zone contains H3K36me3 (). However, most yeast genes are short and intronless, in contrast to human genes, which frequently contain large first introns. Also, we found occasionally that these different zones may interdigitate, e.g.
there was switching back and forth between H3K79me2 and H3K36me3 within STIM1
(Supplementary Fig. 3
). This implies that these zones are not strictly 5′ and 3′ with respect to genic positioning, but rather that introns and exons act to organize intragenic histone modifications.
Intronic and exonic histone modifications can, in principle, affect many aspects of gene expression7
. For example, H3K36me3-modified nucleosomes could directly influence splicing12
. Furthermore, H3K36me3 levels on average are slightly lower at alternatively spliced exons compared to nearby constitutive exons14,15,17
. An open question for future studies remains: how are these marks established and maintained at particular introns and exons within the same transcribed unit? A working knowledge of such localization mechanisms will be paramount to understanding temporal and spatial regulation of chromatin marks in humans.