X chromosome dosage compensation via enhanced transcriptional elongation in Drosophila.

The evolution of sex chromosomes has resulted in numerous species in which females inherit two X chromosomes but males have a single X, thus requiring dosage compensation. MSL (Male-specific lethal) complex increases transcription on the single X chromosome of Drosophila males to equalize expression of X-linked genes between the sexes. The biochemical mechanisms used for dosage compensation must function over a wide dynamic range of transcription levels and differential expression patterns. It has been proposed that the MSL complex regulates transcriptional elongation to control dosage compensation, a model subsequently supported by mapping of the MSL complex and MSL-dependent histone 4 lysine 16 acetylation to the bodies of X-linked genes in males, with a bias towards 3' ends. However, experimental analysis of MSL function at the mechanistic level has been challenging owing to the small magnitude of the chromosome-wide effect and the lack of an in vitro system for biochemical analysis. Here we use global run-on sequencing (GRO-seq) to examine the specific effect of the MSL complex on RNA Polymerase II (RNAP II) on a genome-wide level. Results indicate that the MSL complex enhances transcription by facilitating the progression of RNAP II across the bodies of active X-linked genes. Improving transcriptional output downstream of typical gene-specific controls may explain how dosage compensation can be imposed on the diverse set of genes along an entire chromosome.

To investigate how MSL complex specifically elevates transcription of X-linked genes, we performed GRO-seq in SL2 cells, a male Drosophila cell line that has been extensively characterized for MSL function4,9. To display the average enrichment across genes, a 3 kb ‘metagene’ profile was plotted in which the internal regions were rescaled so that all genes appear to have the same length (Fig. 1). Analysis was restricted to expressed genes that were sufficiently large (> 2.5 kb) so that gene-body effects could be clearly assessed (822 X-linked genes, 3420 autosomal genes), and all gene profiles were normalized by their copy-number as determined by analysis of SL2 DNA content10. High correlation coefficients were observed between replicate libraries (Pearson correlation coefficient: ≥ 0.98; Fig. S1). The metagene profiles revealed a prominent 5′ peak of paused RNAP II consistent with previous ChIP and RNA-seq analysis of short 5′ RNAs11,12. In addition, a peak of RNAP II density downstream of the metagene 3′ processing site is evident, possibly due to slow release in regions of transcription termination8. The 3′ peak is present even when the influence of neighboring gene transcription is eliminated (Fig. S2).

Figure 1

The male X chromosome has higher levels of engaged RNAP II over gene bodies relative to autosomes

(a) Average GRO-seq profiles of expressed genes are shown for X (red) and autosomes (blue). Read counts on all chromosomes were normalized to genomic read coverage to control for copy number variation, mappability and other potential biases. To construct a metagene profile, genes are scaled as follows: 1) the 5′ end (1 kb upstream of the transcription start site (TSS) to 500 bp downstream) and the 3′ end (500 bp upstream of the transcript termination site (TTS) to 1 kb downstream) were unscaled; 2) The remainder of the gene is scaled to 2 kb (see Supplementary Methods). (b) Pausing indices (PI) do not differ between X (red bar) and autosomal genes (blue bar). Elongation density indices (EdI) are significantly different between X (red bar) and autosomal genes (blue bar). Error bars represent a 95% confidence interval for the mean PI or EdI (1.96*SE: n = 1344 [X-genes]; n = 6090 [A-genes]. The definitions of PI and EdI are shown in the schematic. The PI and EdI are calculated with unscaled GRO-seq tag counts.

The central question with regard to dosage compensation is how genes on the X chromosome differ on average from genes on autosomes. Overall, we found that RNAP II density on active X-linked genes was higher than on autosomal genes, specifically over gene bodies (Fig. 1a). The increase in tag density over the bodies of X-linked genes compared to autosomal genes was approximately 1.4-fold, consistent with previous estimates of MSL-dependent dosage compensation9,10,13. We also performed RNAP II ChIP in SL2 cells, confirming higher occupancy on X-linked genes compared to autosomes but with lower resolution and reduced sensitivity (Fig. S3). Therefore, we proceeded with GRO-seq to analyze X and autosomal differences.

To measure how X and autosomes differed on average in distribution of elongating RNAP II, we segmented genes into their 5′ 500 bp and the remainder of the coding region. We further subdivided the remainder of the coding region into 5′ and 3′ segments (25% and 75% respectively). Using this segmentation, we quantified RNAP II pausing and elongation separately based on unscaled GRO-seq signal (Fig. 1b). The pausing index (PI) was previously defined as the ratio of GRO-seq signal at the 5′ peak to the average signal over gene bodies8. Here, we calculated the PI for X and autosomal genes as the ratio of the 5′ peak (segment A) to the first 25% of the remaining gene body (segment B), and found no statistically significant difference (Fig. 1b).

To separately examine transcription elongation across gene bodies, we defined the Elongation density Index (EdI) as the ratio of tag density in the 3′ region of each gene (segment C) compared to its 5′ region after the first 500 bp (segment B). In contrast to our analysis of 5′ pausing, we found statistically significant differences in EdI (P-value < 0.0162) between X and autosomes (Fig. 1b), regardless of how 5′ and 3′ regions of genes were divided (Table S1). As defined, the average PI (log scale) is a positive number because RNAP II generally is enriched at 5′ ends compared to gene bodies; the average EdI (log scale) is a negative number, as the relative density of RNAP II typically decreases from the beginning to the end of gene bodies. We conclude that X–linked genes, on average, exhibit a significantly smaller decrease in RNAP II density along their gene bodies when compared to autosomal genes.

To measure the specific contribution of MSL complex to the increase in RNAP II within X-linked gene bodies, we used MSL2 RNAi to reduce complex levels in male SL2 cells as described previously9. Excellent correlations between replicate data sets were observed (Fig. S1). To confirm the X-specific effect of MSL2 RNAi, we computed the distributions of GRO-seq signal (averaged over the bodies of genes excluding the 5′ peak) for all genes before and after RNAi. When comparing X vs. autosomes, we found a preferential decrease on the X-chromosome, with an average control:MSL RNAi ratio of 1.4 (Fig. 2a). MSL-dependent changes in average GRO-seq density showed weak but statistically significant correlation with changes in steady-state mRNA levels assayed by expression array9 (Pearson correlation = 0.22, P-value < 1 × 10−15) or mRNA-Seq10 (Pearson correlation = 0.30, P-value < 1 × 10−15). These results confirm that MSL-dependent changes in steady-state RNA levels reflect differences in active transcription on the X chromosome.

Figure 2

MSL complex increases engaged RNAP II density on the male X chromosome

(a) The log ratio of sense-strand reads in the MSL2 RNAi sample to the control RNAi sample was computed within the body of each gene. Here, the distributions of these ratios are plotted for all genes on X and autosomes. (b) GRO-seq sense-strand read densities within the roX2 gene for the untreated, control RNAi and MSL2 RNAi samples. Schematic below GRO-seq profiles indicates the location of the DHS (DNase I Hypersensitive Site), which contains sequences that can recruit MSL complex to the X chromosome.

In addition to assessing the average decrease of X-linked RNAP II density after MSL2 RNAi, we asked whether any genes showed strong MSL-dependence, a hallmark of the roX genes that encode RNA components of the complex14,15. We found that roX2 showed a strong loss in GRO-seq density after MSL2 RNAi as predicted (9-fold) (Figs. 2b, S4). Interestingly, in the untreated or control RNAi samples, there is a prominent GRO-seq peak downstream of the major roX2 3′ end, coincident with an MSL recruitment site (see discussion below). roX1 expression is low in this isolate of SL2 cells, and no other expressed genes on X or autosomes displayed strong MSL-dependence in our assays (> 6-fold). Examples of additional individual gene profiles are shown in Figs. S5, S6.

Next, we compared the average RNAP II density along X and autosomal metagene profiles after control and MSL2 RNAi. Unlike our initial analysis of X and autosomes, where different gene populations were compared (Fig. 1), here we could examine the same genes in the presence and absence of MSL complex (Fig. 3). We found that after MSL2 RNAi, the density of elongating RNAP II over the bodies of X-linked genes decreased, approaching the level on autosomes (Figs. 3, S7). The presence of MSL complex affected RNAP II density starting just downstream of the 5′ peak and continuing through the bodies of X-linked genes (Figs. 3, S7). Thus, GRO-seq functional data correlate with physical association of MSL complex which is biased towards 3′ ends of active genes on the male X chromosome4,5.

Figure 3

MSL complex facilitates the progression of engaged RNAP II across transcription units

(a) Metagene profiles of expressed X chromosome genes and autosomal genes in control RNAi and MSL2 RNAi samples. Higher RNAP II density can be seen within the bodies of genes on the X (solid-red) compared to those on autosomes (solid-blue) in the control RNAi sample. After MSL2 RNAi, average RNAP II density on X decreases over gene bodies (dashed-red) becoming similar to autosomal gene bodies (dashed-blue). (b) Ratios of pausing indices (PI) between control and MSL2 RNAi treated cells are not significantly different for genes on the X (red bar) compared to those on autosomes (blue bar). In contrast, ratios of elongation density indices (EdI) between the control and MSL2 RNAi sample decreased significantly for genes on the X (red bar) compared to those on the autosomes (blue bar). Pausing indices (PI) and elongation density indices (EdI) were calculated as described for Figure 1. Error bars represent a 95% confidence interval for the mean PI or EdI (1.96*SE: n = 1358 [X-genes]; n = 6135 [A-genes].

To quantify the differences in density of engaged RNAP II over X-linked genes in the presence and absence of MSL complex, we calculated the pausing (PI) and elongation density indices (EdI), expressing them as ratios comparing MSL2 and control RNAi treatment. We found that both X and autosomes increased PI and decreased EdI after MSL2 RNAi treatment (Fig. S8). However, in each case the change was larger on X than on autosomes, and the most profound difference was an MSL-dependent change in EdI on the X compared with autosomes (P < 1 × 10−15; Fig. 3b). EdI was computed, as before, by defining the 5′/3′ regions as 25%/75% of the gene body after removing the 5′ peak, but the difference was statistically significant for all other values until the 3′ end was reached (Table S1). When these analyses were performed separately for two independently prepared sets of GRO-seq libraries (Fig. S9), the results were also statistically significant (P-value < 7.6 × 10−14, P-value < 1.1 × 10−4 for each of two replicates). We conclude that MSL complex causes the transcriptional elongation profiles of X-linked genes to differ from those of autosomal genes.

To visualize the location along gene bodies at which MSL complex functions, we calculated control:MSL2 RNAi GRO-seq ratios and generated a metagene profile (Fig. 4a). Here, values above zero represent higher relative amounts of engaged RNAP II in the presence of MSL complex compared to after RNAi treatment. In contrast, values below zero represent a relative increase in engaged RNAP II after MSL2 RNAi. In the absence of MSL complex, there is a relative increase in the amount of RNAP II localized to 5′ ends of both autosomal (blue) and X-linked genes (red), perhaps due to relocalization of RNAP II from the bodies of X-linked genes (Fig. 4a). A limitation of the GRO-seq assay is that we cannot currently distinguish between initiating and 5′ paused polymerase, so we cannot assign a definitive role for this 5′ increase in RNAP II after MSL2 RNAi treatment. However, relative RNAP II levels over autosomal gene bodies do not increase, suggesting that any relocalized enzyme in this experiment is likely to remain paused rather than progressing across transcription units. This is consistent with a model in which the functional outcome of MSL2 RNAi is to shift RNAP II density away from productive transcription through X-linked gene bodies.

Figure 4

MSL function correlates with the presence of H4K16 acetylation

(a) The MSL2-dependent effect on RNAP II density as shown by metagene profiles of control: MSL2 RNAi GRO-seq sense-strand reads shown on log scale (base 2). The black line (y = 0) indicates no change after MSL2 RNAi treatment. The cumulative effect of MSL2 RNAi treatment peaks toward the 3′ ends of X-linked genes (red) while having less effect on autosomal genes (blue). (b) Similar to the effect of MSL complex on engaged RNAP II, H4K16 acetylation on the male X chromosome localizes to the bodies of active genes with at 3′ bias (red). On autosomes, H4K16 acetylation is present at 5′ ends (blue) as described previously7.

We plotted the local effect of MSL complex in Fig. 4a to compare it to the status of H4K16 acetylation (Fig. 4b), catalyzed by the MOF component of MSL complex3,16. H4K16 acetylation typically is enriched at 5′ ends of most active genes in mammals and flies6,17; in contrast, a 3′ bias of this mark is a distinctive characteristic of the dosage compensated male X chromosome in Drosophila3,6,7. Interestingly, there is an overall coincidence across gene bodies between MSL complex-dependent GRO-seq signal and the presence of H4K16 acetylation (Fig. 4a;7). How might H4K16 acetylation biased toward the 3′ end of genes generate the improved transcriptional elongation indicated by our GRO-seq results? During transcription elongation, nucleosomes are thought to comprise a barrier to the progress of RNAP II18-20 and several well studied elongation factors, including Spt6 and the FACT complex, are proposed to function by removing nucleosomes that block RNAP II progression and replacing them in the wake of transcription18,21. Interestingly, H4K16 acetylation of nucleosomes has been observed to act in opposition to the formation of higher order chromatin structure in vitro22,23. Thus, H4K16 acetylation is likely to further reduce the steric hindrance to RNAP II progression through chromatin. Improving the entry of RNAP II into the bodies of genes may allow 5′, gene-specific events to proceed at an increased but still regulated rate. Furthermore, reduction in the repressive effect of nucleosomes could increase mRNA output by improving the processivity of RNAP II on each template. Available methodologies cannot distinguish between these mechanisms in vivo, and therefore future approaches will be required to assess their relative contributions to dosage compensation.

In addition to increasing the transcription of X-linked genes for dosage compensation, MSL complex also positively regulates the roX noncoding RNA components of the complex, to promote their male-specificity14,15. roX1 expression is low in our SL2 cell line, but our GRO-seq data indicate that active transcription of roX2 is highly dependent on MSL2 as predicted (Fig. 2b; Fig. S4). Interestingly, there is a strong GRO-seq peak at the 3′ roX2 DHS (DNase I Hypersensitive Site) which contains sequences important for targeting MSL complex to the X chromosome. Sites of roX gene transcription are thought to be critical for MSL complex assembly24,25. Therefore, it is possible that paused RNAP II at the roX2 DHS could promote an open chromatin structure that facilitates MSL complex targeting or incorporation of noncoding roX2 RNA into the complex.

In summary, we hypothesize that MSL complex functions on the male X chromosome to promote progression and processivity of RNAP II through the nucleosomal template. Improving transcriptional output downstream of typical gene-specific regulation makes biological sense when compensating the diverse set of genes found along an entire chromosome.

Methods Summary

To measure the density of engaged RNAP II, GRO-Seq experiments were conducted on DRSC SL2 cells grown in Schneider’s medium with 10% FBS8. To determine how MSL complex contributes to dosage compensation, MSL2 and control (GFP) RNAi treatments were conducted using a bathing protocol9. Nuclei were subjected to GRO-seq analysis after RNAi treatment. Two biological replicates were performed for the untreated, control RNAi, and MSL2 RNAi experiments.