Abstract
Many heterogametic organisms adjust sex chromosome expression to accommodate differences in gene dosage. This requires selective recruitment of regulatory factors to the modulated chromosome. How these factors are localized to a chromosome with requisite accuracy is poorly understood. Drosophila melanogaster males increase expression from their single X chromosome. Identification of this chromosome involves cooperation between different classes of X-identity elements. The Chromatin Entry Sites (CES) recruit a chromatin-modifying complex that spreads into nearby genes and increases expression. In addition, a family of satellite repeats that is enriched on the X chromosome, the 1.688X repeats, promotes recruitment of the complex to nearby genes. The 1.688X repeats and CES are dissimilar, and appear to operate through different mechanisms. Interestingly, the siRNA pathway and siRNA from a 1.688X repeat also promote X recognition. We postulate that siRNA-dependent modification of 1.688X chromatin contributes to recognition of nearby genes. In accord with this, we found enrichment of the siRNA effector Argonaute2 (Ago2) at some 1.688X repeats.Mutations in several proteins that physically interact with Ago2, including the histone methyltransferase Su(var)3-9, enhance the lethality of males with defective X recognition. Su(var)3-9 deposits H3K9me2 on some 1.688X repeats, and this mark is disrupted upon ectopic expression of 1.688X siRNA. Furthermore, integration of 1.688X DNA on an autosome induces local H3K9me2 deposition, but enhances expression of nearby genes in a siRNA-dependent manner. Our findings are consistent with a model in which siRNA-directed modification of 1.688X chromatin contributes to recognition of the fly X chromosome by the MSL complex.
Introduction
Males of many species carry one X chromosome and a gene-poor Y chromosome. Hemizygosity of the male X chromosome produces a potentially lethal imbalance in the ratio of X to autosomal gene products. This imbalance is corrected by a process known as dosage compensation, a specialized type of gene regulation that modulates expression of an entire chromosome. Different strategies to achieve dosage compensation have evolved independently. In Drosophila melanogaster, males increase X-linked gene expression by approximately two-fold (LUCCHESIet al. 2005).This involves the activity of the Male Specific Lethal (MSL) complex. The MSL complex is recruited to active genes on the X chromosome, where it modifies chromatin to increase expression (LUCCHESI AND KURODA 2015). The MSL complex contains five proteins, Male-Specific Lethal 1, -2, and -3 (MSL1, -2, -3), Maleless (MLE), and Males Absent on the First (MOF) (reviewed in (KOYA AND MELLER 2011)). Enhanced transcription by the MSL complex is associated with H4K16 acetylation by MOF (AKHTAR AND BECKER 2000; SMITHet al. 2000). H4K16 acetylation decompacts chromatin, and this may enhance transcriptional elongation of X-linked genes (SHOGREN-KNAAKet al. 2006; LARSCHANet al. 2011).
The MSL complex also contains one of two non-coding RNA on the X (roX1, -2) transcripts (QUINNet al. 2014). While elimination of any one of the MSL proteins is lethal to males, roX1 and roX2 are redundant for compensation. Mutation of both roX genes leads to mislocalization of the MSL proteins to ectopic autosomal sites in male larvae (MELLER AND RATTNER 2002; DENG AND MELLER 2006). X-linked gene expression is reduced in these males, as is survival to adulthood. Both roX genes are located on the X chromosome, and both overlap Chromatin Entry Sites (CES), specialized sites with increased affinity for the MSL complex (KELLEYet al. 1999; ALEKSEYENKOet al. 2008; STRAUBet al. 2008).
Although much is known about the role of MSL complex in dosage compensation, how this complex selectively targets the X chromosome is poorly understood. Recognition and binding to X chromatin is believed to be a two-step process. Initial recruitment of the MSL complex to CES is followed by spreading into nearby transcribed genes (GELBART AND KURODA 2009). Contained within the CES are motifs called MSL Recognition Elements (MREs) (ALEKSEYENKOet al. 2008; STRAUBet al. 2008). MREs are 21 bp GA-rich motifs that bind Chromatin-Linked Adaptor for MSL Protein (CLAMP), a zinc finger protein that is essential for MSL recruitment (SORUCOet al. 2013).Spreading into nearby active genes is supported by interaction of MSL3 with the cotranscriptional H3K36me3 mark (KIND AND AKHTAR 2007; LARSCHANet al. 2007; SURALet al. 2008). These mechanisms describe local recruitment of the MSL complex, but fail to explain how the MSL complex specifically targets the X-chromosome. H3K36me3 is found on active genes throughout the genome, and MREs are only modestly enriched on the X chromosome. Furthermore, CLAMP binds MREs throughout the genome, but only recruits the MSL complex to X-linked CES (ALEKSEYENKOet al. 2008; SORUCOet al. 2013). We conclude that additional mechanisms must distinguish X and autosomal chromatin.
X-localization is disrupted in roX1 roX2 males, making them a sensitized genetic background that can be used to identify additional factors contributing to X recognition.Using this strategy, our laboratory demonstrated a role for the siRNA pathway in recognition of the X-chromosome (MENON AND MELLER 2012; MENONet al. 2014). A likely source of siRNA is a family of repeats that is near exclusive to the X chromosome. These are the AT rich, 359 bp 1.688X satellite repeats, a clade of which is found in short, tandem arrays throughout X euchromatin (HSIEH AND BRUTLAG 1979; WARING AND POLLACK 1987; DIBARTOLOMEISet al. 1992; GALLACH 2014). Specific clusters are denoted by a superscript indicating cytological position. In support of the idea that 1.688X repeats assist X recognition, ectopic production of siRNA from one repeat partially rescues roX1 roX2 males (MENONet al. 2014). 1.688X repeats are often close to or within genes, leading to the idea that they function as “tuning knobs” for gene regulation (KUHNet al. 2012). In accord with these ideas, autosomal insertions of 1.688X DNA enable recruitment of functional dosage compensation to nearby autosomal genes (JOSHI AND MELLER 2017).
The 1.688X repeats share no sequence identity with the CES, and appear to act in a genetically distinct manner (JOSHI AND MELLER 2017). The question of how 1.688X DNA promotes compensation of nearby genes is thus of great interest. We pursued the idea that siRNA-directed modifications of chromatin at 1.688X repeats link the repeats and the siRNA pathway to X recognition. Reduction of the siRNA-binding effector protein Argonaute 2 (Ago2) enhances the lethality of partial loss of function roX1 roX2 mutations, and further reduces X-localization of MSL proteins (MENON AND MELLER 2012). We hypothesized that an Ago2-containing complex might localize to and modify 1.688X chromatin in otherwise wild type flies. In accord with this idea, we find that Ago2 is enriched at 1.688X repeats. Proteins interacting with Ago2 may also play a role in dosage compensation. To address this, we tested high confidence Ago2-binding proteins for genetic interactions with roX1 roX2, and found that mutations in several of these genes further reduced the survival of roX1 roX2 males. Of particular interest is the H3K9 methyltransferase Su(var)3-9, which is responsible for enrichment of H3K9me2 at a subset of 1.688X repeats. H3K9me2 enrichment is disrupted upon ectopic expression of 1.688X siRNA. Chromatin flanking an autosomal insertion of 1.688X DNA is enriched for H3K9me2, and enrichment is enhanced by ectopic expression of 1.688X siRNA. In contrast to the repressive nature of H3K9me2, we find that expression of autosomal genes close to the 1.688X transgene is increased in male larvae, and further elevated by ectopic production of 1.688X siRNA. These findings support the idea that X recognition and transcriptional upregulation by dosage compensation are distinct processes, and propose that siRNA-dependent modification of chromatin in or near 1.688X repeats contributes to X recognition in wild type flies. We propose that epigenetic modifications link the siRNA pathway, 1.688X repeats on the X chromosome and X recognition.
Materials and Methods
Fly culture and Genetics
Mutations Dcr1Q1147X (BDSC #32066), Rm6201086 (BDSC #11520), Fmr1Δ113m (BDSC #67403), Su(var)3-91 (BDSC #6209), Su(var)3-92 (BDSC #6210), smg1 (BDSC #5930), Taf111 (BDSC #65410), Taf115 (BDSC #65409), p535A-1-4 (BDSC #6815), p5311-1B-1 (BDSC #6816), foxoΔ94 (BDSC #42220), PIG-Se00272 (BDSC #17833), belL4740 (BDSC #10222), bel6 (BDSC #4024), barrL305 (BDSC #4402), SmD1EY01516 (BDSC #15514), vigC274 (BDSC #16323), Ago1k08121 (BDSC #10772), aubQC42 (BDSC #4968), piwi06843 (BDSC #12225), Su(var)2-102 (BDSC #6235), eggMB00702 (BDSC #22876), G9aMB11975 (BDSC #29933), P{EPgy2}09821 (BDSC #16954), P{EPgy2}15840 (BDSC #21163), and FLAG.HA.Ago2 (BDSC #33242)were obtained from the Bloomington Drosophila Stock Center. Ago2414 (Kyoto #109027) was obtained from the Kyoto Stock Center. Su(var)3-714 was a gift from Dr. P. Spierer (SEUMet al. 2002). ocm166 was a gift from Dr. R. Kelley. ΔDsRedΔupSET (upSET in Figure 2) was a gift from Dr. M. Kuroda (MCELROYet al. 2017). To minimize genetic background effects all mutations were out crossed for five generations using a nearby marked P-element (unmarked mutations) or the laboratory reference yw strain (mutations marked with w+ or y+). Stocks were constructed with outcrossed, rebalanced chromosomes and a reference Y-chromosome (MENON AND MELLER 2009). All mutations were confirmed by phenotype or PCR. Mating schemes to determine the effect of Ago2-interactors on dosage compensation are presented in Figure S1. Each test scored about 1000 flies and was performed in triplicate. To express 1.6883F siRNA in a Su(var)3-9-/- mutant background, we generated [hp1.6883F] [Sqh-Gal4]/In(2LR)Gla wgGla-1; Su(var)3-91/ ™3TbSb flies and selected non-Tb third instar males for ChIP. The [Sqh-Gal4] insertion was a gift of Dr.S. Todi. The [hp1.6883F] transgene contains part of the 1.6883F repeat cluster coned in inverted orientation in pWIZ (SIKLEE AND CARTHEW 2003). Although siRNA accumulates to abnormally high levels in larvae expressing [hp1.6883F], the siRNAs produced appear similar to those isolated from wild type embryos (MENONet al. 2014)
Tissue collection and chromatin preparation
Embryo collection and chromatin preparation was as previously described (KOYA AND MELLER 2015). Briefly, 0.5 g of 0 - 12 hr embryos were collected on molasses plates with yeast. Embryos were dechorionated for 2.5 min in bleach, crosslinked in 50 mM HEPES, 1 mM EDTA, 0.5 mM EGTA, 100 mM NACl, 1 % formaldehyde with heptane for 20 min. Crosslinking was quenched with 125 mM glycine, 0.01 % Triton X-100, 1 X PBS for 30 min. Embryos were washed with 10 mM HEPES, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA and 0.01 % Triton X-100 and suspended in 2.5 ml of 10 mM HEPES, 1 mM EDTA, 0.5 mM EGTA, 0.1 % Na-deoxycholate and 0.02 % Na-azide for sonication. Sonication, performed on ice at 35 % amplitude, 30 sec on, 59 sec off for a total time 15 min using a Fischer Scientific Model FB505 sonicator, produced 300-600 bp fragments. Chromatin was clarified by centrifuging at 13,000 rpm for 15 min, diluted 1:1 with 2 X RIPA buffer (2 % Triton X-100, 0.2 % Na-deoxycholate, 0.2 % SDS, 280 mM NaCl, 20 mM Tris-HCl pH 8.0, 2 mM EDTA, 0.02 % Na-azide, 2 mM DMSF with complete protease inhibitor (Roche)). Chromatin solution (5.5 ml) was preabsorbed by incubation at 4° for 30 min with 55 μl of blocked Pierce™ Protein A agarose beads (Catalog #20333) and aliquots stored at -80°.
For larval chromatin, a modified protocol from (KUZUet al. 2016) was used. 150 larvae were frozen in liquid N2 and ground in a chilled mortar. The powder was transferred to a cooled 15 ml Dounce and homogenized with a loose pestle (10 strokes) and a tight pestle (15 strokes) in 10 ml PBS with protease inhibitor. Homogenate was made to 40 ml with PBS, crosslinked with 1 % formaldehyde for 20 min and quenched with 125 mM glycine for 30 mins. Crosslinked material was pelleted, washed once with wash buffer A (10 mM Hepes pH 7.6, 10 mM EDTA, 0.5 mM EGTA, 0.25 % Triton X-100, protease inhibitor and 0.2 mM PMSF), once with wash buffer B (10 mM Hepes pH 7.6, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1 % Triton X-100, protease inhibitor and 0.2 mM PMSF), and 3 times with TE wash buffer (10 mM Tris pH 8.0, 1 mM EDTA, 0.01 % SDS, protease inhibitor and 0.2 mM PMSF). The pellet was resuspended in 2 ml pre-RIPA buffer (0.1 % SDS, 10 mM Tris-HCl, 1 mM EDTA, protease inhibitor and 0.2 mM PMSF). Sonication was performed at settings described above for 2 min. Sonicated samples were diluted with 1 % Triton X-100, 0.1 % Na-deoxycholate, and 140 mM NaCl, centrifuged at 1500 g to clarify, aliquoted and stored at -80°.
Chromatin Immunoprecipitation
Seventy five micrograms of chromatin was incubated overnight at 4° with 4 μl anti-H3K9me2 (Abcam, ab1220) or 8 μl anti-H3K9me3 (Abcam, ab8898), clarified by centrifugation at 14,000 rpm for 5 min and supernatants transferred to tubes containing 40 μl blocked Pierce™ Protein A agarose beads (Catalog #20333) and incubated 4 h at 4°. Following washing, reverse crosslinking, organic extraction and precipitation, DNA was suspended in 50 μl distilled water.
ChIP-qPCR
Duplicate 20 μl reactions consisting of 2 μl DNA, 10 μl BioRad iTaq (#172-5101), and primers were amplified using an Mx3000P Real-Time PCR system (Stratagene).Standard error was derived from the mean Ct values of biological replicates. qPCR analysis was previously described (KOYA AND MELLER 2015). Each ChIP pull down was validated. For H3K9me2, primers in an H3K9me2-enriched region of the 3rd chromosome and dmn served as positive and negative controls. ChIP primers are presented in Table S1. Primer specificity for 1.688X repeats was ensured by anchoring one primer in flanking unique sequence (1.6881A, 1.6883C, 1.6884A and 1.6887E) or by designing primers to unique sequences within repeats and testing with genomic DNA from a strain deleted for the repeat cluster (1.6883F and 1.6887F; see Menon et al., 2014). Primer efficiencies were determined using MxPro qPCR software. Repeat copy number is normalized by expressing enrichment as percent input.
Protein Isolation from embryos
Fifty mg of 0 - 12 hr embryos were homogenized in 250 μl RIPA buffer on ice. Homogenate was passed through a 26 gauge needle 10 - 12 times to shear DNA. Particulate matter was removed by centrifugation, and supernatant was mixed with an equal volume of 2 X SDS Sample buffer and boiled for 5 min before separation on a 15 % SDS polyacrylamide gel.
Protein blotting
Polyacrylamide gels were equilibrated in transfer buffer (48 mM Tris, 39 mM glycine, 1.3 mM SDS, 20 % methanol) for 20 min. A PVDF membrane was activated in 100 % methanol for 1 min. Filter paper and activated PVDF membranes were saturated in transfer buffer and proteins transferred using a Trans-Blot SD Semi-Dry Transfer Cell (BIO-RAD). The membrane was washed in TBST (10 mM Tris-Cl, 200 mM NaCl, 0.1 % Tween 20, pH 7.5), blocked in 5 % BSA and probed overnight at 4° using 1:2000 mouse anti-H3K9me2 diluted in blocking solution (Abcam, ab1220) or 1:4000 goat anti-tubulin (Developmental Studies Hydrinoma Bank, E7). After washing with TBST, the membrane was incubated with alkaline phosphatase conjugated secondary antibodies (goat anti-mouse, Sigma, A3562 or rabbit anti-goat, Sigma, A4062), washed and developed in 100 mM diethanolamine, 100 mM NaCl, 5 mM MgCl2, pH 9.5 containing 33 μg/ml Nitroblue Tetrazolium (NBT) and 165 μg/ml 5-Bromo-4-chloro-3-indolyl phosphate (BCIP). Signals were quantified by ImageJ.
Quantitative RT-PCR
Total RNA was isolated from 50 third instar male larvae or 100 mg dechorionated embryos using Trizol reagent (Invitrogen) as previously described (KOYA AND MELLER 2015). One microgram of RNA was reverse-transcribed using random hexamers and ImProm-II reverse transcriptase (Promega). Duplicate reactions were amplified using iTaq Universal SYBR Green Supermix (Bio-Rad) with an Mx3000P Real-Time PCR system (Stratagene). Primers are in Table S1. For determining relative transcript abundance, values were normalized to dmn (DCTN2-p50). To calculate fold change values were normalized to dmn and to a reference strain. Expression was calculated using the efficiency corrected comparative quantification method (PFAFFL 2001).
Strains and materials used in this study are available upon request. The authors affirm that all data necessary to confirm the conclusions of this study are presented within the article, figures and tables.
Results
Ago2 localizes at 1.688X repeats
We took advantage of the resolution of ChIP and a FLAG-tagged Ago2 transgene to determine if Ago2 localizes to 1.688X repeats. FLAG-Ago2 was first tested for rescue of the dosage compensation function of Ago2. Males with the partial loss of function roX1ex40roX2Δ chromosome have high survival, as do Ago2-/- flies, but synthetic lethality is observed in roX1ex40roX2Δ/Y; Ago2-/- males (MENON AND MELLER 2012). One copy of a FLAG-Ago2 transgene rescues these males, demonstrating that the FLAG tag does not disrupt the dosage compensation function of Ago2 (Figure 1A). Chromatin from FLAG-Ago2; Ago-/- embryos, and from a reference strain lacking the FLAG-Ago2 transgene, was immunoprecipitated with anti-FLAG antibodies and enrichment determined by quantitative PCR (qPCR). FLAG-Ago2 was enriched at the Hsp70 promoter, a site known to bind Ago2 (CERNILOGARet al. 2011) (Figure 1B). In contrast, a control region in the dmn gene displayed no enrichment. We then examined FLAG-Ago2 enrichment at a panel of six representative 1.688X repeat clusters that differ in location and environment (within, near or far from protein coding genes), transcription level and sequence (Table 1). Interestingly, five of these show enrichment of FLAG-Ago2 over the repeats, but little or no enrichment in flanking regions (Figure 1C). We conclude that Ago2 localizes at many 1.688X repeats, a finding that is consistent with involvement of Ago2 in siRNA-directed recruitment of chromatin modification at or around these regions.
Genetic interactions identify an Ago2-interaction network that participates in dosage compensation
Argonaute proteins in the RNA Induced Transcriptional Silencing (RITS) complexes of S. pombe and plants recruit chromatin modifiers to nascent transcripts (reviewed in (MELLERet al. 2015)). To explore the possibility of Ago2-interacting proteins participating in X chromosome recognition, we screened genes in an Ago2-interaction network for genetic interaction with roX1 roX2. A map of high probability Ago2-interactors was created using BioGRID (STARKet al. 2006), and esyN (BEANet al. 2014) (Figure 2A; see File S2 for inclusion criteria). Members of this network were examined for genetic interactions with the partial loss of function roX1ex33roX2Δ X chromosome. roX1ex33roX2Δ males display partial mislocalization of MSL proteins and eclose at 20 % of normal levels (DENGet al. 2005b). Reduction of proteins that participate in X recognition further disrupts X localization and enhances roX1ex33roX2Δ male lethality (MENON AND MELLER 2012). Females are fully viable and fertile when the roX genes are mutated. roX1ex33roX2Δ females were mated to males that were heterozygous for a mutation in the gene being tested (Figure S1A). All sons are roX1ex33roX2Δ/Y, and heterozygous (experimental) or wild type (control) for the gene of interest. A reduction in normalized survival (experimental /control) reveals enhancement of roX1 roX2 male lethality (Figure 2B, C). Daughters, which do not dosage compensate and are heterozygous for roX1ex33roX2Δ, do not display altered survival upon mutation of Ago2-interacting genes. As G9a is located on the X chromosome, a modified strategy to test this gene is presented in Figure S1B.
Normalized survival of roX1ex33roX2Δ males with mutations in the Ago2-interaction network is presented in Figure 2B. Genes displaying significant interactions are noted by pink symbols, and those showing no interaction are blue in Figure 2A. We confirmed a previously identified siRNA-processing sub-network containing Dcr2, Elp1, and loqs (Figure 2A, dotted line; (MENON AND MELLER2012)). The present study identified several additional Ago2-interactors, including a potential chromatin-modifying sub-network containing Dcr1, Fmr1, Rm62, and the histone methyltransferase Su(var)3-9 (green, Figure 2A). Su(var)3-9 deposits H3K9me2 and acts with Rm62 to re-silence active chromatin (BOEKEet al. 2011).
Additional chromatin modifiers and genes in other small RNA pathways were also tested (Figure 2C). A previous study found no interaction between roX1ex33roX2Δ and the piRNA pathway genes aub and piwi, or the miRNA pathway gene Ago1, a finding replicated here (MENON AND MELLER 2012). Since our findings point towards involvement of chromatin modifiers, we tested the chromatin regulatory factor Su(var)2-10 and two additional H3K9 methyltransferases, eggless (egg) and G9a (Figure 2C).None of these modified roX1ex33roX2Δ survival Mutations in Su(var)3-7, important for heterochromatin formation, and upSET, which maintains heterochromatin and H3K9me2 levels, enhance roX1ex33roX2Δ male lethality (SPIERERet al. 2008; MCELROYet al. 2017). Over compensating males (ocm) has an unusual dosage compensation phenotype as mutations in ocm rescue males with insufficient MSL activity, suggesting that it might act as a governor of activation (LIM AND KELLEY 2013). Interestingly, mutation of ocm significantly increased the survival of roX1ex33roX2Δ males, supporting the idea that ocm normally restrains activation. The P{EPgy2}09821 and P{EPgy2}15840 strains, used to outcross Su(var)3-9 and barr mutants, display no interaction and serve as controls for genetic background. Taken together, these findings suggest that several genes that deposit H3K9me2, maintain this mark or participate in heterochromatin formation also contribute to X chromosome dosage compensation. At first glance these observations appear to be at odds with X chromosome hypertranscription, the ultimate consequence of X chromosome recognition.
Ectopically expressed 1.6883F siRNA disrupts H3K9me2 patterns
Previous studies found that ectopically produced 1.6883F siRNA partially rescues roX1 roX2 males and increases X localization of the MSL complex (MENONet al. 2014). The mechanism by which siRNA promotes X recognition is unknown. The discovery that insertion of 1.688X DNA on an autosome enables functional compensation of nearby genes, and the enhancement of this effect by ectopic 1.6883F siRNA, suggests siRNA action through cognate genomic regions (JOSHI AND MELLER 2017). In accord with this idea, an autosomal roX1 transgene also enables compensation of nearby genes, but is unaffected by 1.6883F siRNA. To test the idea that 1.6883F siRNA directs epigenetic modification of 1.688X chromatin, we used ChIP to analyze chromatin around 1.688X repeats on the X chromosome. ChIP-qPCR detected H3K9me2 enrichment in 4 out of 6 repeats (white bars, Figure 3). As H3K9me2 enrichment was not uniform, we considered additional factors that might determine this mark, and noted that only repeats showing evidence of transcription were enriched for H3K9me2, consistent with the idea of Ago2-dependent recruitment to nascent transcripts (Figure S4) (VERDELet al. 2004). Upon ectopic expression of 1.6883F siRNA a dramatic disruption of H3K9me2 was observed in and around 1.688X repeats (black bars, Figure 3). For example, 1.6883F and 1.6884A display peaks of H3K9me2 in wild type flies, but this mark was reduced over the repeats and increased in surrounding regions by elevated 1.6883F siRNA. The reduction in H3K9me2 over repeats themselves was unexpected and could represent repositioning of nucleosomes or blocking of the methylated site by another protein. Untranscribed repeat clusters at 1.6881A and 1.6887E show no H3K9me2 enrichment in wild type flies, but gained H3K9me2 upon expression of 1.6883F siRNA. In contrast, no enrichment of H3K9me3 in or near 1.688X repeats was detected in wild type or 1.6883F siRNA-expressing embryos (Figure S2). We conclude that some 1.688X repeats are enriched for H3K9me2, and that cognate siRNA broadly disrupts this mark within and several kb adjacent to 1.688X DNA.
Su(var)3-9 deposits H3K9me2 at 1.688X repeats
The identification of Su(var)3-9 as an indirect binding partner of Ago2, observation of a genetic interaction between roX1 roX2 and Su(var)3-9 and enrichment of H3K9me2 on some 1.688X repeats suggested that Su(var)3-9 could be modifying 1.688X repeats. D. melanogaster has 3 histone H3K9 methyltransferases, Su(var)3-9, eggless, and G9a, but only Su(var)3-9 mutations enhance the male lethality of roX1 roX2 (Figure 2; (SWAMINATHANet al. 2012)). To determine if Su(var)3-9 is responsible for H3K9me2 at 1.688X chromatin, we generated strains carrying Su(var)3-9 over a marked balancer, enabling selection of homozygous Su(var)3-9 mutant larvae. H3K9me2 enrichment is virtually eliminated over 1.688X repeats in Su(var)3-9-/- mutants (Figure 4, gray) and remains low in Su(var)3-9-/- larvae that express 1.6883F siRNA (Figure 4, black). This reveals that Su(var)3-9 deposits H3K9me2 at 1.688X chromatin in wild type flies, and eliminates the possibility that a different methyltransferase is recruited to these regions following ectopic expression of 1.6883F siRNA. Disruption of H3K9me2 upon expression of 1.6883F siRNA thus reflects changes in the localization or activity of Su(var)3-9.
To determine how far from 1.688X repeats the H3K9me2 disruption extends, regions 10 to 26 kb from repeats were examined. In each case, increased H3K9me2 was observed in embryos with ectopic 1.6883F siRNA expression (Figure S3A). This suggested the possibility of a global change in H3K9me2 levels. To address this we probed protein blots from wild type and 1.6883F siRNA-expressing embryos to determine global levels of this modification, but found no evidence of a change in H3K9me2 (Figure S3B). As most H3K9me2 is found in heterochromatic regions that comprise > 30 % of the fly genome, changes in euchromatin may represent a negligible portion of the nuclear pool.
H3K9me2 is generally thought to be repressive, but compensation in flies occurs by increased expression of X-linked genes. To determine if changes in H3K9me2 enrichment correlate with changes in transcription, regions flanking 1.688X repeats were examined in wild type and 1.6883F siRNA-expressing embryos. Consistent with H3K9me2 having a repressive effect, 1.6883F siRNA decreases accumulation of mRNA, as well as non-coding intragenic and intronic regions, with elevated H3K9me2 (Figure S4). The apparent increase in 1.6883F expression presumably originates from the transgene used to produce ectopic 1.6883F siRNA. We detected dramatic reductions in messages immediately adjacent to 1.6881A (tyn, G9a) and 1.6883F (ec, roX1). In spite of a 90 % reduction in ec transcript in embryos expressing 1.6883F siRNA, adults do not display the rough eye ec phenotype. It is possible that ectopic 1.6883F siRNA has a more pronounced effect in embryos, whose lack of differentiation may make them particularly susceptible to epigenetic disruption. Mature patterns of chromatin organization are established later in development, and these may be more resilient. To test this, we examined expression in 1.6883F siRNA-expressing third instar male larvae, and found that tyn, G9a and ec regained wild type transcript levels, and roX1 was largely restored (Figure S4). The precise reason for the differences between embryos and larvae are uncertain, but restoration of normal gene expression by the third larval instar is consistent with the lack of phenotype in otherwise wild type flies that ectopically express 1.6883F siRNA (MENONet al. 2014).
The finding that animal age influenced response to ectopic siRNA prompted us to determine the time point at which H3K9me2 is established at 1.688X repeats. A possible scenario is that this mark is placed before MSL localization, and acts in some way to guide X recognition. X-localization of the MSL complex occurs 3 h after egg laying (AEL) (RASTELLIet al. 1995; MELLER 2003). We measured H3K9me2 enrichment at 1.6883F in embryos before the MSL complex binds to the X (1.5 - 3 h), during initial MSL recruitment (3 - 4 h), and at 4 - 6 h and 6 -12 h. In contrast to our prediction, H3K9me2 is first detected on 1.6883F between 6 and 12 h AEL, after X localization of the MSL complex has occurred (Figure S5). We conclude that H3K9me2 at 1.688X repeats is unlikely to guide initial X recognition, but may serve at a later time point to facilitate spreading of this mark or enforce the stability of X recognition. As lethality due to failure of dosage compensation occurs at the end of the 3rd instar, rescue of male viability could presumably occur during a window of several days.
H3K9me2 is enriched at regions flanking autosomal 1.6883F transgenes
One challenge of studying recruiting elements on the X chromosome is that the redundancy and proximity of these elements complicates interpretation. To overcome this we tested autosomal integrations of 1.6883F DNA or roX1 (Figure 5A) (JOSHI AND MELLER2017). H3K9me2 ChIP was performed on chromatin from third instar male larvae with 1.6883F (Figure 5B) or roX1 (Figure 5C) on 2L (gray bars), and in the same genotypes with ectopic expression of 1.6883F siRNA (black bars). H3K9me2 within 5 kb of the integration site is not strongly enriched in control males, or in males with a roX1 transgene, but is striking elevated when 1.6883F DNA is present. Consistent with our observations in embryos, ectopic 1.6883F siRNA further elevated H3K9me2 near the 1.6883F integration. This contrasts with negligible enrichment flanking the roX1 transgene (Figure 5C). For unknown reasons, enrichment over the integrated 1.6883F DNA was itself undetectable. We conclude that autosomal insertion of 1.6883F DNA makes flanking chromatin subject to siRNA-induced H3K9me2 deposition. Taken together, these studies support the idea that the 1.688X repeats influence patterns of H3K9me2 nearby, but CES-containing roX1, with a different class of recruiting element, has little effect.
To determine the influence of 1.6883F and roX1 on transcription of autosomal genes on 2L, we performed quantitative RT-PCR (qRT PCR) on total RNA from third instar male larvae of the genotypes described above. The 1.6883F and roX1 integration site is in an intron of haf, over 17 kb from the closest exon. We also examined RFeSP, CG33128, Eno (30, 89 and 142 kb from the insertion, respectively), and CG31778 and Rpl37A, 2.1 and 3.5 Mb distant (Figure 5D). The presence of 1.6883F or roX1 integrations alone had no effect on the most distant genes, CG31778 and Rpl37A. A roX1 integration increased expression of haf 2.5 fold, more than expected from full compensation. This may reflect the fact that MSL recruitment to a roX1 transgene can overcome local, chromatin-based silencing (KELLEY AND KURODA 2003). Addition of 1.6883F siRNA increased haf expression slightly, and similarly increased expression of CG33128 and Eno (light gray bars, Figure 5E).
A 1.6883F insertion produced a four-fold increase in haf, and a slight increase in Eno, 141 kb from the integration site. But upon expression of 1.6883F siRNA, haf expression increased to 8 fold wild type levels, and CG33128 and Eno both increased to 2-fold wild type levels, consistent with full compensation. We conclude that an autosomal insertion of 1.688X DNA induces H3K9me2 deposition on flanking chromatin, but also increases expression of genes on 2L in a manner that is consistent with recruitment of the MSL complex. Both H3K9me2 enrichment and increased expression is enhanced by 1.6883F siRNA, suggesting that X identification involves an siRNA-directed mechanism that operates through 1.688X repeats.
Discussion
Molecularly distinct dosage compensation strategies have arisen independently in different organisms, but a shared feature is the ability to selectively recognize and alter an entire chromosome. How a regulatory system is directed to a single chromosome is poorly understood. The discovery that 1.688X satellite DNA promotes recruitment of dosage compensation to nearby genes supports the idea that these repeats are important for selective recognition of X chromatin (JOSHI AND MELLER 2017). How the 1.688X repeats accomplish this is a question of great interest. Involvement of the siRNA pathway, and siRNA from a 1.688X repeat, in X recognition points to the possibility that siRNA-directed modification of chromatin around 1.688X repeats plays a role in dosage compensation in normal males. The findings of the current study support this idea.
Although numerous studies point to small RNA regulation of chromatin in flies, this process is better understood in other organisms (reviewed in (MELLERet al. 2015)). Small RNA directed heterochromatin formation was discovered in S. pombe (reviewed in (MOAZED 2009)). Heterochromatic regions are transcribed during S phase, and transcripts are processed into siRNAs that guide the Ago1-containing RITS complex to complementary, nascent transcripts (VERDELet al. 2004). In addition to several other activities, RITS recruits the H3K9 methyltransferase Clr4 (ZHANGet al. 2008). We propose that a similar process is occurring at 1.688X chromatin in flies. Most 1.688X repeats bind Ago2, and many are transcribed. Several of the 1.688X repeats that we examined are enriched for H3K9me2 deposited by Su(var)3-9, an ortholog of Clr4. Our screen identified genetic interactions between roX1 roX2 and members of a possible RITS-like complex consisting of Ago2, Rm62 and Su(var)3-9. Finally, H3K9me2 enrichment in, and around, 1.688X repeats is responsive to 1.688X siRNA, and enrichment is blocked by loss of Su(var)3-9. Taken together, these findings suggest that a RITS-like complex normally modifies chromatin at 1.688X repeats.
The idea that repressive H3K9me2 marks participate in a process culminating in a two-fold increase in expression is counterintuitive, but X recognition is complex. This process involves CES sites that directly recruit the MSL complex and the 1.688X repeats, acting indirectly to enhance X recognition (Figure 6). It is possible that X recognition uses epigenetic marks, such as H3K9me2, that are distinct from the activating marks deposited within genes by the MSL complex. We propose that robust X recognition results from cooperation between two distinct pathways that guide this process. Interestingly, numerous studies have discovered links between the compensated X chromosome of male flies and repressive marks. For example, the male X is enriched in HP1, a major constituent of heterochromatin that binds H3K9me2 (DE WIT et al. 2005; LIUet al. 2005). The structure of the polytenized male X chromosome is extraordinarily sensitive to altered levels of genes that participate in heterochromatin formation or silencing, such as HP1, Su(var)3-7 and ISWI. Mutations in these genes produce a general disruption of polytenization that is strikingly specific to the male X (DEURINGet al. 2000; SPIERERet al. 2005; ZHANGet al. 2006). JIL-1, a kinase that enforces boundaries between heterochromatin and euchromatin, is enriched on the X chromosome and thought to participate in compensation (JINet al. 2000; WANGet al. 2001; EBERTet al. 2004; DENGet al. 2005a). Upon loss of JIL-1, polytenized structure is disrupted and H3K9me2 invades euchromatic chromosome arms, but the X chromosome is most severely affected (ZHANGet al. 2006). Finally, the MSL proteins themselves have an affinity for heterochromatin. In roX1 roX2 mutant males the MSL proteins become mislocalized to ectopic autosomal sites (MELLER AND RATTNER 2002).For reasons that are still unclear, the most prominent of these sites are the heterochromatic 4th chromosome and chromocenter (DENG AND MELLER 2006; FIGUEIREDOet al. 2014). Taken together, these observations suggest that X recognition, or spreading of the MSL complex, could be facilitated by repressive marks. One intriguing possibility is that 1.688X repeats guide deposition of H3K9me2 and this mark, directly or indirectly, assists localization of the MSL complex. Although MSL-mediated compensation initiates at 3 h AEL, before H3K9me2 enrichment over 1.688X chromatin, male killing due to loss or mislocalization of the MSL complex occurs several days later at the larval/pupal transition. It is possible that ectopic production of 1.688X siRNA drives enrichment of H3K9me2 across the X, supporting X recognition or MSL complex spreading at later developmental stages. This would explain why roX1 roX2 mutant male larvae, defective for X recognition, display increased X-localization of the MSL proteins and elevated viability upon expression of 1.688X siRNA (MENONet al. 2014).
An intriguing aspect of dosage compensation is the evolutionary convergence of mechanisms. For example, long non-coding RNA also plays a central role in X recognition in mammals, where expression of the X inactive specific transcript (Xist) RNA guides X inactivation (LEE 2009). Furthermore, repetitive LINE-1 elements that are enriched on the mammalian X chromosome are proposed to facilitate X inactivation (LYON 1998; BAILEYet al. 2000). Interestingly, some LINE-1 elements are transcribed during the onset of X inactivation, producing endo-siRNAs that may guide local spreading of heterochromatin into regions that are otherwise prone to escape (CHOWet al. 2010). These parallels are particularly striking as the outcomes, silencing of an X chromosome in mammalian females and activation of the single X in male flies, appear unrelated.We propose that cooperation between distinct chromatin-modifying systems that rely on long and short non-coding RNAs is one strategy to selectively modulate an entire chromosome.
File S1. Estimating tandem repeat copy number in the laboratory reference strain.
To determine the number of tandem repeats at various cytological positions in the laboratory reference (yw) strain, we performed PCR using primers flanking each repeat cluster. The length of the amplicon attributable to repeats was divided by 359 to obtain the copy number. Prior studies have shown that the copy number of 1.6883F repeats in the laboratory reference strain is 3.5 (Menon et al., 2014). The 1.6884A, 1.6883C, 1.6887E, and 1.6887F repeat clusters span 850 bp, 1200 bp, 900 bp, and 500 bp, respectively, bringing their copy numbers to 2.5, 3.5, 2.5, and 1.5 (Figure S6, Table 1). As the 1.6881A repeats have a copy number of ∼100, it was not possible to use this method to determine 1.6881A repeat length or copy number in the laboratory reference strain.
File S2. Selection criteria for Ago2-interacting proteins
Four criteria were used to rank Ago2-interactors in the Biological General Repository for Interaction Datasets (BioGRID) version 3.4.144, a manually curated database containing high-probability protein interactions (Stark et al., 2006), or identified by the Easy Networks (esyN) search engine (Bean et al., 2014). As BioGRID represents a subset of esyN interactions, all BioGRID interactions in Table S2 are also present in esyN. The Curation Score in Table S2 is the sum of four criteria. The method of detection was scored 0 if high throughput, and 1 if low throughput or experimentally validated. Additional criteria are: a known role in RNA interference; chromatin modification; and chromatin association. Proteins received a score of 1 for each criterion satisfied and 0 if unsatisfied. Most proteins with a score of 2 or more were tested for genetic interaction with roX1 roX2. Phosphatidylinositol glycan anchor biosynthesis class S (PIG-S, curation score 1) was also tested. No genes with a score ≤ 2 displayed a genetic interaction with roX1 roX2, but several with scores of 3 or 4 enhanced the lethality of roX1 roX2 males, suggesting a role in dosage compensation (Table S2, Figure 2).
Acknowledgments
The authors wish to thank colleagues who donated fly strains, Drs. S. Todi, M.Kuroda,R. Kelley and P. Spierer. We thank Drs. S. K. Koya and E. Larschan for advice on ChIP. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) and the Kyoto Stock Center were used in this study. N. D. was supported in part by a Summer Dissertation Fellowship and a Graduate Enhancement award. This work was supported by NIH grant GM 093110 to V.H.M.