Sequence-Specific Targeting of Drosophila rox Genes by the MSL Dosage Compensation Complex

Transcription

1 Molecular Cell, Vol. 11, , April, 2003, Copyright 2003 by Cell Press Sequence-Specific Targeting of Drosophila rox Genes by the MSL Dosage Compensation Complex Yongkyu Park, 1 Gabrielle Mengus, 4,6 Xiaoying Bai, 2 Yuji Kageyama, 1,7 Victoria H. Meller, 5 Peter B. Becker, 4 and Mitzi I. Kuroda 1,2,3, * 1 Howard Hughes Medical Institute 2 Program in Developmental Biology 3 Department of Molecular and Cellular Biology Baylor College of Medicine One Baylor Plaza Houston, Texas Ludwig Maximilian Universitat Munchen Adolf Butenandt Institut Molekularbiologie Schillerstrasse 44 D80336 Munchen Germany 5 Department of Biology Tufts University Medford, Massachusetts Summary MSL complexes bind the single male X chromosome in Drosophila to increase transcription approximately 2-fold. Complexes contain at least five proteins and two noncoding RNAs, rox1 and rox2. The mechanism of X chromosome binding is not known. Here, we identify a 110 bp sequence in rox2 characterized by high-affinity MSL binding, male-specific DNase I hypersensitivity, a shared consensus with the otherwise dissimilar rox1 gene, and conservation across species. Mutagenesis of evolutionarily conserved sequences diminishes MSL binding in vivo. MSL binding to these sites is rox RNA dependent, suggesting that complexes become competent for binding only after incorporation of rox RNAs. However, the rox RNA segments homologous to the DNA binding sites are not required, ruling out simple RNA-DNA complementarity as the primary tar- geting mechanism. Our results are consistent with a model in which nascent rox RNA assembly with MSL proteins is an early step in the initiation of dosage compensation. Introduction The organization of genes into chromatin is thought to be fundamental to proper function of the genome (Far- kas et al., 2000). Covalent modifications of chromatin components play key roles in both stable and dynamic aspects of genome organization (Kingston and Narlikar, 1999). One idea is that chromatin domains are estab- lished and maintained by chromatin remodeling com- plexes that can spread from nucleation sites into flank- ing sequences (Bannister et al., 2001; Ho et al., 2002; Lachner et al., 2001; Lee and Jaenisch, 1997; Nakayama et al., 2001). The Drosophila dosage compensation complex is thought to associate with the male X chromosome through recognition of two rox genes and perhaps other sites that subsequently allow the complex to spread in cis along the chromosome (Kelley et al., 1999). The complex consists of at least five MSL proteins, including the MOF histone acetyltransferase, and two untranslated rox RNAs (Park and Kuroda, 2001). Each MSL protein is required for male viability and the wild-type pattern of X chromosome association of the MSL complex (Pannuti and Lucchesi, 2000). In contrast, males mutant for either rox1 or rox2 survive, but double mutant males die with a disrupted pattern of X localization of MSL complexes (Franke and Baker, 1999; Meller and Rattner, 2002). Although wild-type MSL complexes bind to a precise, reproducible pattern of sites in vivo on male polytene chromosomes, there is no known DNA binding subunit within the complex, and DNA sequence requirements for MSL binding sites have not been identified. Special attention has focused on 35 sites that are bound by partial MSL complexes in msl mutants, which have been proposed to act as chromatin entry sites (Kelley et al., 1999). So far, two such sites have been identified and tested for their recruitment of MSL complexes to other chromosomes (Kelley et al., 1999; Meller et al., 2000). These sites map to the rox1 and rox2 loci, which code for the untranslated rox RNAs that are components of MSL complexes (Amrein and Axel, 1997; Meller et al., 1997). Transgenes containing either rox1 or rox2 are sufficient to attract the MSL complex to ectopic sites on autosomes, from which complexes can spread into flanking chromatin (Kelley et al., 1999; Meller et al., 2000; Park et al., 2002). This result is the basis for a model in which MSL complexes initially assemble at chromatin entry sites, and then spread in cis. The spreading step is not limited to X chromosomal DNA, since autosomal chromatin that is not normally a target of MSL com- plexes can be bound when located in cis to a rox transgene. The rox1 and rox2 RNAs have redundant genetic function and both corresponding genes attract MSL complex, suggesting that they would share common sequences. However, initial sequence comparisons between the complete genes revealed only a small 25/30 nt identity (Franke and Baker, 1999). This sequence is neither necessary nor sufficient for MSL binding at rox1. Instead, a distinct 217 bp fragment of the rox1 gene attracts MSL complexes (Kageyama et al., 2001). That a small subclone from the 3.7 kb rox1 RNA transcription unit is sufficient for this attraction demonstrated that local synthesis of a wild-type rox RNA in cis is not an absolute requirement for MSL binding. A male-specific DNase I hypersensitive site forms at the MSL binding site within the endogenous rox1 gene, and chromatin IP experiments show that the MSL complex can be crosslinked to this region in SL2 male tissue culture *Correspondence: mkuroda@bcm.tmc.edu 6 Present address: IGBMC, BP10142, Illkirch, France. 7 Present address: Graduate School of Biological Sciences, Nara Institute of Science and Technology, Takayama, Ikoma, , Japan. cells (Kageyama et al., 2001). Therefore, we searched for

2 Molecular Cell 978 pendent transgenic lines (Figure 2A). Hereafter, we will use DHS and MSL binding site synonymously. Bind- ing of MSL complexes to the ectopic rox2 DHS in polytene chromosomes was robust, but unlike the rox1 multimer, we did not observe detectable MSL spreading from rox2 DHS multimer constructs inserted at six differ- ent autosomal locations (Figure 2B and data not shown). MSL binding at the rox2 DHS was msl3 independent (Figure 2C), which is a hallmark of rox genes and 33 other proposed chromatin entry sites on the X chromosome (Kelley et al., 1999). Northern analysis using a probe from the DHS in rox2 confirmed that the major rox2 transcript does not overlap with the MSL binding sequences (Figure 1C). Therefore, unlike in the rox1 gene, the MSL binding site in rox2 is separable from the segment that encodes the majority of rox2 transcripts. This organization has facili- tated the separation of rox2 RNA production from MSL binding at rox2 in subsequent analyses. features common to the rox1 and rox2 MSL binding sites. DNase I hypersensitivity assays, analysis of in vivo binding to rox2 transgenes, and sequence comparison with rox2 genes from related Drosophila species have revealed a 110 bp segment with islands of conserved sequence between the MSL binding sites of rox1 and rox2. We show that these sequences, and rox RNAs, are important for MSL binding to rox genes in vivo. Results An MSL Binding Site in rox2 Is Coincident with a Male- Specific DNase I Hypersensitive Site The size of rox2 transcripts in adult flies was initially reported to be 1.1 kb (Amrein and Axel, 1997), while in Smith et al. (2000) the major species was ca. 600 nt. To analyze the gene structure of rox2, we mapped the 5 end of rox2 transcripts using 5 RACE and RT-PCR with successively more 5 primers (Figures 1A and 1B). These results suggest most transcription initiates within a small interval. In contrast, 3 RACE analyses showed one major 3 end and two distinct minor 3 ends (Figure 1A), in agreement with several existing ESTs and rox2 cdnas. Splicing of exon 1 to exon 2 in combination with use of the major 3 end would be consistent with 600 nt transcripts observed in Smith et al. (2000). Use of a major 3 site coupled with inclusion of the intron between exons 1 and 2 could be consistent with a 1.1 kb transcript reported by Amrein and Axel (1997). Use of the distal, minor 3 end would produce a 1.4 kb transcript. A combination of cdna sequencing and RT-PCR analysis showed that rox2 transcripts are alternatively spliced in a surprisingly complex pattern (Y.P., unpublished data). The most commonly recovered form in RT- PCR and 3 RACE analyses contained the two common exons and ended at the shortest putative 3 end (Figure 1A). To confirm that the major rox2 transcript has this structure, probes from different parts of rox2 were used in Northern analyses of total RNA from adult flies (Figure 1C) as well as larvae and tissue culture cells (data not shown). The blots demonstrated that the major rox2 species are indeed around 600 nt long in each of these stages. A minor signal most likely representing unspliced transcripts of approximately 1.4 kb size was also detected. The intron probe hybridized in a diffuse smear indicating the presence of several minor splice forms (data not shown). A strong MSL binding site within the rox1 gene was previously mapped by a combination of male-specific DNase I hypersensitivity (DHS) assays and transgenic deletion analyses (Kageyama et al., 2001). The binding site was narrowed down to 200 bp, centrally located within the rox1 transcription unit. Transposons carrying the 200 bp rox1 DHS segment could attract MSL complexes to ectopic sites on autosomes, and were sufficient, at least as a nine-copy multimer, to occasionally nucleate limited spreading of MSL complexes into flanking chromatin (Kageyama et al., 2001). Based on this information, we checked the rox2 gene for male-specific DNase I hypersensitivity and found evidence for a DHS region at the 3 end of the gene (Figure 1D). Although this site was not as prominent as its rox1 counterpart, we found that this 270 bp segment was also sufficient to attract MSL complexes to autosomes in multiple inde- The DHS Sequence Is the Principal MSL Binding Site in the rox2 Gene, and Has No Essential Activity as RNA One role for MSL binding to rox DNA might be to capture rox RNA cotranscriptionally. If so, MSL binding within or adjacent to the site of rox RNA synthesis might be critical for rox RNA assembly into functional MSL complexes. To test this, we assayed Hsp83roX2 transgenes lacking the 3 DHS sequences for functional complementation of rox1 rox2 mutants. Surprisingly, all four transgenic lines tested produced rox2 RNA that could fully rescue mutant males, resulting in a wild-type pat- tern of MSL complexes and rox RNA on the X chromosome (Figures 3A 3D). This occurred in spite of the fact that these transgenes showed weak or undetectable MSL binding at their insertion sites (Figures 3B 3D). We draw several conclusions from these results. First, the 3 DHS sequence is the primary MSL binding site within the rox2 gene. Second, MSL binding at the adjacent 3 DHS is not a prerequisite for rox2 RNA assembly into MSL complexes. Third, soluble MSL subunits are able to capture rox RNAs in the nucleoplasm and assemble them into functional MSL complexes, without a strong MSL signal at the site of the transgene. Whether faint MSL staining at the truncated rox2 transgene detected in some lines is due to a low-affinity secondary DNA binding site for MSL complexes or reflects assembly of MSL proteins on nascent RNA has not been determined. Fourth, minor rox2 transcripts that extend beyond the major 3 end are not required for rox2 function. The rox2 DHS sequence has no essential activity as RNA. A similar finding has been made for the rox1 DHS se- quence (C. Stuckenholz, personal communication). Another way to test the ability of rox RNAs to assem- ble into functional complexes is to use a spreading competition assay (Park et al., 2002). Previously, we showed that spreading from genomic GMroX transgenes inserted on autosomes was eliminated by competition from ex- pressed rox1 cdnas at other locations. To test whether the presence of a DHS sequence influences the ability of rox RNAs to compete for MSL complex assembly, we tested transgenes containing (1) a rox1 or rox2 DHS site or (2) an expressed rox1 or rox2 cdna without a DHS site (transgenes such as P{w GMroX2}97F will be

3 Sequence-Specific MSL Binding Sites in rox Genes 979 Figure 1. The DNase I Hypersensitive Region of rox2 Is Located Downstream of the Major 3 Processing Site of Most rox2 Transcripts (A) The gene structure of rox2 includes Promoter (P), Exon 1 (110 bp), Intron (545 bp), Exon 2 (430 bp), and DHS (270 bp) regions. The major and minor 3 ends are indicated as three arrowheads within the line representing the gene. Thick line, major transcript; dotted line, 3 region included in minor transcripts. Hatch box, 30 nt similarity of unknown function between rox1 and rox2 (Franke and Baker, 1999); black box, 110 bp segment containing islands of conserved sequences (Figure 5A). (B) RT-PCR to map the 5 end of rox2. RT-PCR products using primers indicated in Figure 1A show the major spliced form. 1, 100 bp ladder; 2 11, RT-PCR; 12, PCR using rox2 genomic fragment as template. (C) Northern analysis of rox2 RNA from adult flies. The membranes were exposed for 36 hr (for rox2 probes of different sizes but approximately the same specific activity) or 24 hr (for rp49 as loading control). The weaker signal from the exon1 probe could be due to its smaller size. Left numbers, RNA ladder; M, male; F, female. (D) DNase I hypersensitivity within the rox2 locus. DNase I cleavages were mapped within a BglII fragment containing the rox2 gene, schematized to the left (open box, 1 indicates the approximate 5 end of the gene). Nuclei from female (left) or male (right) adult flies were treated with increasing concentrations of DNase I. DNA was isolated and digested to completion with BglII. DNase I cleavages were revealed by Southern blot using a probe adjacent to a BglII site located 2466 bp downstream of the rox2 transcription start site. The double-headed arrow highlights male-specific DNase I hypersensitivity in rox2 chromatin. The numbers to the right correspond to a DNA size marker and indicated sizes in bp. referred to as [transgene name-location], i.e., [GMroX2-97F]). Spreading competition was measured as the ability to influence spreading from a competing GMroX transgene. Neither rox1 nor rox2 DHS sequences alone were able to influence spreading from [GMroX2-97F] (arrows, Figures 3E and 3F), consistent with previous results that rox RNA transcription is required for competition to occur. However, expression of rox1 or rox2 cdnas lacking the DHS sequences provided strong competition for MSL complex spreading from either [GMroX2-97F] (arrows, Figures 3G and 3H) or [GMroX1-67B] (Table 1). This result provides additional evidence that the assembly mechanism by which nascent rox RNAs attract MSL proteins does not require a linked MSL binding site. Identification of Consensus Sequences in the MSL Binding Sites of rox1 and rox2 through Analyses of rox2 from Related Drosophila Species A direct comparison of the MSL binding sites in rox1 and rox2 did not yield an obvious consensus sequence for MSL recognition. To search for conserved se-

4 Molecular Cell 980 Figure 2. The DHS Region of rox2 Is an MSL Binding Site In Vivo Polytene chromosome squashes from transgenic larvae immuno- stained with rabbit anti-msl1 (red) and counterstained with DAPI (blue). (A) Male nucleus containing a monomer of the DHS fragment [rox2dhs-59d]. (B) Male nucleus containing a multimer [rox2dhs-91a]. (C) Monomer [rox2dhs-47b] in a female nucleus carrying [Hsp83MSL2] and mutant for msl3. (D) Male nucleus containing a 110 bp rox2 monomer [rox2cr-78e]. quences that might help us identify important functional elements in the rox genes, especially within the MSL binding sites, we cloned rox genes from related Dro- sophila species using low-stringency PCR (Figure 4A, see Experimental Procedures). By low-stringency Northern analysis, we knew that D. simulans and D. erecta had male-specific RNAs of the appropriate size that hybridized to a D. melanogaster rox2 probe (Figure 4B). After cloning and sequencing, we found that the D. simulans rox2 gene displayed more than 90% identity in all regions, but the D. erecta rox2 gene showed less homology within the intron (82%) and the MSL binding site (84%) (Figure 4C). In addition, although exon 2 of D. erecta rox2 was 93% identical to D. melanogaster rox2, it had a deletion of 35 bp, resulting in a smaller transcript (Figure 4B). The rox1 gene from D. simulans was also cloned and sequenced (90% identity to D. melanogaster rox1) to assist in the comparison between the regions of MSL binding in the two genes. Once the five sequences were aligned (GCG program), islands of consensus sequences were found over a 110 bp region which had previously escaped notice (Figure 5A). The consensus sequences lie completely within the previously defined 200 bp DNase I hypersensitive region (Figure 1A). The spacing of sequence elements is nearly identical in each DHS, and the consensus includes several GAGA sequences, including a palindrome in regions 1 and 2 indicated in Figure 5A. Finding conserved elements shared between the MSL binding sites within the rox1 and rox2 genes made this consensus sequence a prime candidate for the actual MSL target. To test this idea, we chose five short blocks of sequence within the rox2 DHS for mutagenesis. We replaced five base pairs within each block, with the substitutions listed below the consensus sequence in Figure 5A. Blocks 1, 2, 4, and 5 were absolutely con- served in all five rox sequences, while block 3 was dissimilar between rox1 and rox2. The mutations were Figure 3. rox RNA Is Assembled and Functional without a Linked MSL Binding Site All experiments were performed in a rox1 rox2 genetic background. Red, rabbit anti-msl1; blue, DAPI. (A) Rescue frequency (male:female ratio) of rox1 rox2 mutants by four different [H83roX2 DHS] transgenic lines. (B D) MSL complexes containing rox2 DHS RNA paint the X chromosome although the [H83roX2 DHS] insertion sites show no or weak binding of MSL proteins because of the absence of the DHS region. (E H) Competition for cis-spreading from [GMroX2-97F] (arrow) with rox cdna constructs (arrowheads). (E) [rox1dhs-31f]; (F) [rox2dhs- 59D]; (G) [H83roX1 DHS-92C]; (H) [H83roX2 DHS-83C]. [roxdhs] transgenes alone cannot compete for cis-spreading from [GMroX2-97F]. However, [H83roX DHS] transgenes show strong competition.

5 Sequence-Specific MSL Binding Sites in rox Genes 981 Table 1. Competition by rox DHS RNAs Inhibits cis-spreading from Autosomal GMroX Transgenes MSL Binding at Transgene A Transgene A Transgene B a N b Single (%) Spreading (%) GMroX2-97F c GMroX2-97F H83roX1 DHS-23B GMroX2-97F H83roX1 DHS-48E GMroX2-97F H83roX1 DHS-92C GMroX2-97F H83roX2 DHS-79F GMroX2-97F H83roX2 DHS-83C GMroX2-97F rox1dhs-31f GMroX2-97F rox2dhs-59d GMroX1-67B c GMroX1-67B H83roX1 DHS-92C GMroX1-67B H83roX2 DHS-83C All experiments were performed in rox1 rox2 double mutants. a [H83roX1 DHS] showed variable MSL binding to the site of the transgene. [H83roX2 DHS] showed no or weak MSL binding to a single band. [roxdhs] showed a strong single band of MSL binding. b Total number of nuclei counted. c From Park et al. (2002). assayed for in vivo MSL binding within the context of a only that specific segment of rox2. We found that MSL 270 bp rox2 P element insert. Each animal assayed binding to the 110 bp segment of rox2 was robust (Figcarried two similar rox2 DHS transgenes, the mutant ure 2D), comparable to the full 270 bp fragment (Figure sequence (M) and a wild-type 270 bp element (W) serving 2A). Thus, evolutionarily conserved consensus se- as an internal positive control. The results are sum- quences within the rox2 gene are both necessary and marized below the altered sequences in Figure 5A, and sufficient to attract the MSL complex. chromosome binding data are displayed in Figure 5B. We found that mutagenesis of consensus sequences The Search for Additional MSL Binding Sites within the rox2 MSL binding site resulted in decreased or rox Genes binding of MSL complexes in vivo. Changes in region 2 Our original model for binding and spreading of the MSL nearly abolished MSL binding, while mutations in region complex on the X chromosome proposed that there 4 resulted in a milder decrease. DHS transgenes con- were approximately 35 chromatin entry sites on the X taining mutations in regions 1 and 5 were affected less (Kelley et al., 1999). This was based on the observation than 2 and 4 but still displayed weak binding relative to that partial complexes bound to 35 cytological locations, the reference wild-type transgene. In contrast, changes including the rox1 and rox2 genes, in the absence in region 3, which was not conserved between rox1 of complete or functional complexes (Kelley et al., 1999; and rox2, resulted in retention of robust MSL staining Meller et al., 2000). Therefore, we used the Pattern comparable to the wild-type control. Thus, in contrast Search program ( to search the to most simple protein-dna interactions, binding determinants Drosophila genomic sequence for regions of similarity for the large MSL complex (approximately 2 to our newly derived consensus sequences, hoping to MDa) are spread over a broad 110 bp region. locate additional putative binding sites. We searched To test whether the 110 bp segment was sufficient for the GAGAGN 4-5 TC[T/C]CTCTC palindromic sequence for MSL binding, we constructed transgenes carrying containing region 1 and 2 (Figure 5C) and found five Figure 4. rox2 Genes from Other Drosophila Species (A) The divergence distance of the three species. Mya, million years. (B) Low-stringency Northern analysis using rox2 and rp49 probes from D. melanogaster. (C) Sequence similarity of rox2 from D. melanogaster, D. simulans, and D. erecta analyzed using the GCG program.

6 Molecular Cell 982 Figure 5. Consensus Sequences within the DHS Are Important for Binding of MSL Complex (A) Consensus sequences (gray box) common to the MSL binding sites of rox1 and rox2. Conserved (1, 2, 4, and 5) and nonconserved (3) sequences were mutagenized by five base substitutions. The degree of MSL binding was determined by comparing binding to wild-type (wt) and mutant transgenes of [rox2dhs] in the same nucleus. The scores represent comparisons of multiple combinations of wt and mutant transgenes analyzed independently by four different people. (B) Immunostaining with anti-msl1 antibodies (red) reveals binding to a wild-type [rox2dhs] (W, arrow) at 21E (panels 1 5), or to mutant [rox2dhs] (M, arrowhead) at 60E (mutant 1), 29B (mutant 2), 30C (mutant 3), 92F (mutant 4), or 100D (mutant 5). (C) Two blocks of consensus sequences are not enough to attract MSL complex. Blue blocks (1 and 5) are regions that show a mild effect when mutagenized, and red blocks (2 and 4) are regions that show a strong effect. 1B7, 11D1, and 12E8 were candidate chromatin entry sites that were unable to attract MSL proteins to autosomal transgenes. candidates on the X chromosome. Two of these were to be missed in our limited analysis, carry a different rox1 and rox2, and the other three candidates were consensus sequence, or somehow attract MSL com- located near cytologically mapped chromatin entry sites plexes through another unknown mechanism. (1B7, 11D1, and 12E8). We analyzed whether the three candidates were MSL binding sites by constructing Binding of MSL Complexes to rox DHS Transgenes transgenic flies containing bp inserts encompassing Requires rox RNAs the short sequence similarity, but none of the MSL proteins fail to bind rox genes without the aid of candidates created new MSL binding sites in the polytene the MLE RNA helicase, suggesting that RNA might play chromosome assay (Figure 5C and data not some role (Kageyama et al., 2001; Meller et al., 2000). shown). These results demonstrate that similarity to regions Coupled with the observation that the MSL proteins 1 and 2 of the rox DHS sites is not sufficient for cannot bind the X normally in the absence of rox RNA, MSL binding in vivo. we tested whether rox RNA contributes to the binding We also used all four conserved blocks of sequence specificity of the MSL complex. To test this, we crossed (Figure 5C) in the rox DHS sites to query the flyenhancer transgenes containing the 300 bp MSL binding sites web tool (Markstein et al., 2002). The search found both of either rox1 or rox2 into rox1 rox2 double mutants. rox1 and rox2, but all other matches were very dissimilar Nearly all the resulting males die before adulthood, but to the rox DHS sequence in that the individual sequence their larval polytene chromosomes can be recovered elements appeared in different spacing or order. The for analysis. Although rox1 rox2 double mutants show recovered sequences were randomly distributed on weak variable MSL binding to the X chromosome as well both autosomes and the X, with the X hits failing to map as ectopic binding to autosomes and heterochromatin close to any predicted entry site. Since there are several (Meller and Rattner, 2002; Figure 6A), we found that MSL apparent differences in the behavior of the rox genes proteins no longer recognized the 300 bp MSL binding and other putative chromatin entry sites (see Discus- sites in the absence of rox RNA (Figures 6B and 6G). sion), it remains possible that they are dissimilar enough Just as rox RNAs are redundant for function in dosage

7 Sequence-Specific MSL Binding Sites in rox Genes 983 Figure 6. MSL Binding on rox DHS Sequences Is Dependent on rox RNA Red, rabbit anti-msl1; blue, DAPI staining. (A) rox1 rox2 double mutant male nucleus showing weak MSL binding to the X chromosome and ectopic binding to autosomal sites and heterochromatin (arrowhead). (A F) [rox2dhs-59d] transgenic males. In (B), the transgene (arrow) is not recognized by MSL proteins in the rox1 rox2 double mutant male. (A) and (B) show the same nucleus, with (A) at lower magnification. (C and D) [rox2dhs-59d] is recognized by MSL complexes in rox1 or rox2 single mutant males. (E and F) Recognition of [rox2dhs-59d] is restored in a rox1 rox2 male by expression of [H83roX2 DHS-83C] or [H83roX1 DHS-23B]. (G K) [rox1dhs-31f] transgenic males. In (G), the transgene (arrow) is not recognized by MSL proteins in the rox1 rox2 double mutant male. Stars indicate sites of ectopic autosomal binding of MSL complexes that occurs frequently in double mutant males. (H and I) [rox1dhs-31f] is recognized by MSL complexes in rox1 or rox2 single mutant males. (J and K) Recognition of [rox1dhs-31f] is restored in a rox1 rox2 male by expression of [H83roX2 DHS-83C] or [H83roX1 DHS-23B]. retically possible that MSL complexes might recognize some sequence-independent structural characteristic in chromatin entry sites, here we demonstrate that a broad domain of small islands of consensus sequences is important for MSL binding at rox genes. Computer-based comparisons of the rox1 and rox2 sequences had failed to identify this region. Only after male-specific DNase I hypersensitive sites were identified within each gene and assayed for MSL binding activity in vivo did the consensus target sequence become apparent. We were unable to identify candidates for the additional 33 pos- tulated chromatin entry sites by searching for se- quences similar to the consensus MSL binding sequence in rox genes. This may be due to a failure of our search parameters. Alternatively, this result is consistent with a model in which rox genes are thought to be fundamentally different from other entry sites (Kage- yama et al., 2001; Meller et al., 2000; Park et al., 2002). The most prominent feature of the two rox genes is that they produce noncoding RNA components of MSL complexes. When either is mutant, the other is sufficient for MSL function, but males mutant for both rox RNAs cannot localize their MSL complexes properly. This shows that if any of the other postulated entry sites produce an RNA component of MSL complexes, it is not sufficient to replace these two key components. Several additional lines of evidence now point to the existence of only two rox genes, rather than several dozen. First, SAGE analysis for sex-specific transcripts in adult heads easily found rox1 and rox2 but no other candidate male-specific noncoding RNAs (Fujii and Amrein, 2002). Second, the rox genes differ from the other entry sites in being highly MLE dependent. Finally, if the conserved DHS sequence is a signature for rox genes, it occurs only twice in the genome. compensation, they are also redundant for recognition of MSL binding sites in rox genes, as single rox1 or rox2 mutants produced MSL complexes that retained the ability to recognize either rox1 or rox2 DHS trans- genes (Figures 6C, 6D, 6H, and 6I). The RNA-depen- dence of MSL binding at rox genes provides strong evidence for an RNA-first model for MSL complex assembly. MSL proteins may only acquire high-affinity chromatin binding activity following addition of rox RNA. Of the many mechanisms which might explain how rox RNAs promote recognition of rox genes by the MSL proteins, perhaps the simplest would be some type of RNA-DNA heteroduplex between rox RNA from the MSL complex and the chromosomal DHS DNA sequence. This might be catalyzed by the MLE helicase. We showed above that rox2 RNA lacking any 3 DHS sequence had full activity in the male viability assay and X localization of the MSL complex. However, to test this RNA sequence for a more subtle activity, we asked whether MSL complexes containing rox2 DHS RNA could bind an isolated rox2 DHS transgene. Surpris- ingly, truncated rox2 DHS RNA was sufficient for robust recognition of both rox1 and rox2 DHS (Figures 6E and 6J), excluding the simplest model for RNA-DNA comple- mentarity for binding site recognition. Likewise, MSL complexes containing rox1 RNA deleted for its DHS sequence were also competent for rox1 and rox2 DHS binding (Figures 6F and 6K). Discussion Previous searches for elements necessary to target dosage compensation to the X chromosome failed to yield cis-acting DNA sequences. Although it has been theo-

8 Molecular Cell 984 We have not precisely mapped the locations of MSL copies of the GAGA sequence separated by conserved proteins bound to the X chromosome in rox1 rox2 mu- distances. Several Drosophila proteins are known to tant males due to the poor morphology of the chromo- bind similar sequences, including the GAGA factor ensomes, but it resembles the pattern of chromatin entry coded by the trithorax-like gene and the pipsqueak prosites. If true, this would indicate that MSL proteins can tein (Farkas et al., 1994; Lehmann et al., 1998). GAGA bind weakly to numerous sites on the X, but have a strict factor is thought to keep the chromatin of regulatory RNA requirement to bind the rox1 and rox2 genes. Such regions, such as promoters and Polycomb response a rox RNA dependence is consistent with an earlier elements, in an accessible DHS configuration (Cavalli, report that rox genes differ from other entry sites in that 2002; Granok et al., 1995), possibly by targeting the they are not bound by any MSL subunit unless the MLE nucleosome remodeling factor NURF to these sites (Xiao helicase is present. In previous studies, we found that et al., 2001). However, the action of GAGA factor is not the cytological locations of chromatin entry sites, visual- limited to male flies and hence cannot explain a maleized as partial MSL complex binding, were very similar specific DHS. Although we cannot exclude the possibilin msl3, mle, ormof mutants (Kageyama et al., 2001; ity that the altered chromatin structure in this region Lyman et al., 1997; Meller et al., 2000). While the vast precedes MSL binding, it seems more likely that MSL majority of sites were common in the three different binding to this sequence in males induces the more genotypes, the rox2 site at 10C was specifically absent exposed structure. Simple protein-dna contacts often in the mle mutant. When a rox1 cdna transgene was cover bp, so finding essential MSL recognition assayed in isolation, it was also found to require mle elements distributed over several turns of the DNA helix for binding (Kageyama et al., 2001). A requirement for suggests a requirement for multiple factors to create a rox RNA in complexes that bind these sites would be context for MSL binding. It is not known whether any consistent with this mle requirement, as MLE was pre- of the five characterized MSL proteins directly contact viously shown to be critical for rox RNA inclusion in DNA, but it is interesting to note that in the absence of partial MSL complexes (Meller et al., 2000). rox RNA, most MSL proteins are lost from the X and are Based on these and previous results (Park et al., 2002), instead ectopically bound to centric heterochromatin. we propose that rox RNAs can assemble into complexes Satellite IV sequences are located in the centric heterolocally at their sites of transcription. rox RNAs are unsta- chromatin, making up over 1% of the diploid genome ble in the absence of MSL proteins (Meller et al., 2000), (Ashburner, 1989). It consists of the sequence (AAGAsuggesting that complex assembly must occur rapidly GAG) n, which resembles conserved elements in the rox for the RNA to escape destruction. Although we had DHS. considered that the MSL proteins might be preposi- Why does MSL binding at rox genes appear to differ tioned at rox genes to facilitate capture of nascent tran- from binding of the many MSL targets on the X chromoscripts, we instead found that MSL proteins become some? Two functions seem possible. First, although our competent to bind rox genes only after rox RNA is incor- isolated rox2 DHS transgenes did not support ectopic porated into the complex. Only a few other chromatin MSL spreading over flanking autosomal chromatin, in binding proteins have been shown to require an RNA the context of complete rox genes, high-affinity MSL component for chromatin interaction. Recently, HP1, a binding might facilitate epigenetic MSL spreading. Secmajor constituent of heterochromatin, was shown to ond, little is known about the transcriptional control of require RNA for chromatin association (Muchardt et al., the rox genes or how dosage compensation causes 2002). Heterochromatin silencing in fission yeast may only a 2-fold upregulation of X-linked genes. Perhaps also require an RNA component (Hall et al., 2002; Volpe bound MSL complex contributes to regulation of rox et al., 2002). Likewise, in plants, dsrna can lead to RNA transcription, to provide a precise level of MSL gene silencing not only by destruction of cognate RNA complexes for hypertranscription of the X chromosome. through a standard RNAi mechanism, but also by methylating the gene producing the offending RNA (Mette et Experimental Procedures al., 2000). In this case, a large multisubunit complex with an RNA helicase subunit is thought to use a short 22 Fly Genotypes and Transgene Construction nt RNA component to search the genome for sequence The full genotypes used in this study were: rox1- y w rox1 ex6, rox2- w Df(1)roX2 52 ;P{w 4 4.3}, and rox1- rox2- y w rox1 homology. The rox DHS sequence is found in the middle ex6 Df(1)roX2 52 P{w 4 4.3} (Park et al., 2002). The P{w 4 4.3} element of all rox1 RNAs and at the 3 end of some rox2 tran- is needed to supply essential genes lost in Df(1)roX2 52 (Meller and scripts. Our initial test to determine if the MSL complex Rattner, 2002). used this segment of rox RNA as a template to search To construct the monomer rox2dhs transgene, an NdeI/SpeI re- the genome for homology ruled this out. However, we striction fragment (270 bp, ) isolated from the rox plasmid (Amrein and Axel, 1997) was filled in and cloned in EcoRIcannot exclude the possibility that other short elements digested and filled-in pcasper3 (Thummel and Pirrotta, 1991). For within rox RNA might play such a role. Alternatively, the the multimer rox2dhs construct, the DHS fragment (213 bp, 1028 rox RNA may play some structural role in positioning 1210 in rox ) was amplified using primers containing an AvaI the MSL proteins so that they can make specific DNA site, digested, multimerized, filled in, cloned in the pbluescript SmaI contacts. site, and then subcloned in EcoRI/BamHI digested pcasper3. For An important clue leading to the identification of the construction of Hsp83roX2 DHS, primer 1 (5 -AGATGTTGCGGCATT MSL binding sequence was the discovery of male-spe- CGCGG-3 ) and7(5 -ACTGGTTAAGGCGCGTAAAAC-3 ) located in GMroX2 (Meller et al., 2000) were used in PCR. This fragment (1047 cific DNase I hypersensitive sites within rox1 and rox2. bp) was cloned in the pcrii-topo vector (Invitrogen) and, after Nuclease sensitivity is often attributed to mobile or ab- sequencing, was subcloned into XhoI/HpaI-digested pcasper sent nucleosomes exposing DNA to nuclear proteins. Hsp83-act containing an act5c gene fragment to provide a 3 poly-a The most conspicuous feature within the DHS is three site.

9 Sequence-Specific MSL Binding Sites in rox Genes 985 To construct point mutations in the rox2dhs, pta-rox2-dhsl analyses. We thank X. Chu, H. Kennedy, and R. Richman for excellent (4170 bp) containing rox2 monomer (270 bp) was amplified using technical support. This work was supported by the Howard inverse PCR with mutant primers containing 5 bp changes. After Hughes Medical Institute and Welch Foundation (M.I.K.), the National self-ligation, mutations of rox2dhs were confirmed by sequencing. Institutes of Health (GM45744 to M.I.K. and GM58427 to V.H.M.), the The mutant and wt rox2dhs fragments were subcloned into NotI/ Human Frontier Science Program (M.I.K. andp.b.b.), and atmrfellowship BamHI-digested pgreen H-Pelican (10.2 kb) containing egfp for of the European Union (G.M. and P.B.B.). M.I.K. is an HHMI promoter/enhancer analysis (Barolo et al., 2000). For the rox2cr Investigator. construct, pta-rox2-dhsl was the template for amplification of the 110 bp conserved region, using primer 44 (5 -AAAGGATCCAGATC Received: December 23, 2002 GATTTAGAG-3 ) and primer 162 (5 -ACGCTCGAGGCAGTTAAT Revised: February 20, 2003 TAGTATTG-3 ). The PCR product was subcloned into pcrii-topo Accepted: February 24, 2003 vector (Invitrogen), checked by sequencing, and cloned into BamHI/ Published: April 24, 2003 XhoI-digested pgreen H-Pelican. For 1B7 (287 bp), 11D1 (307 bp), and 12E8 (253 bp) constructs, References primers 1B-1 (5 -CGAAAGGAGAGACAATTCCC-3 ), 1B-2 (5 -GGAT TCCGGGAATCTCTGGC-3 ), 11D-1 (5 -GACCACCCACTTGCCCA Amrein, H., and Axel, R. (1997). Genes expressed in neurons of adult CCC-3 ), 11D-2 (5 -CTGCATCCAGCTCCCTGTGC-3 ), 12E-1 (5 -CCC male Drosophila. Cell 88, CCAGTTCGAATCGAAAC-3 ), and 12E-2 (5 -GGCATCGCGGCACT Ashburner, M. (1989). Satellite DNA. In Drosophila A Laboratory GCGATG-3 ) were used for PCR from genomic DNA. These frag- Handbook (Cold Spring Harbor, NY: Cold Spring Harbor Laboratory ments were cloned in the pcrii-topo vector (Invitrogen) and, after Press), pp sequencing, were subcloned into NotI/BamHI-digested pcasper3. Bannister, A.J., Zegerman, P., Partridge, J.F., Miska, E.A., Thomas, J.O., Allshire, R.C., and Kouzarides, T. (2001). Selective recognition DNase I Hypersensitivity Assay of methylated lysine 9 on histone H3 by the HP1 chromo domain. The DNA digested by DNase I was prepared as previously described Nature 410, (Kageyama et al., 2001). 10 g of DNA from each DNase I digest was purified and digested to completion with BglII and electrophogalactosidase Barolo, S., Carver, L.A., and Posakony, J.W. (2000). GFP and betaresed on a 1% agarose gel. The DNA was blotted and hybridized transformation vectors for promoter/enhancer analy- as described (Church and Gilbert, 1984). The RNA probe was prepared sis in Drosophila. Biotechniques 29, by in vitro transcription of a 265 bp restriction fragment abut- Cavalli, G. (2002). Chromatin as a eukaryotic template of genetic ting a BglII cleavage site 2466 bp downstream of the rox2 transcrip- information. Curr. Opin. Cell Biol. 14, tion start site. Church, G.M., and Gilbert, W. (1984). Genomic sequencing. Proc. Natl. Acad. Sci. USA 81, Northern Analysis Farkas, G., Gausz, J., Galloni, M., Reuter, G., Gyurkovics, H., and Total RNA from adult flies was prepared using the TRIzol Reagent Karch, F. (1994). The trithorax-like gene encodes the Drosophila (GIBCO-BRL), and 20 g of total RNA was used in each lane. For the GAGA factor. Nature 371, Figure 1C Northern, high-stringency hybridization was used (Church and Gilbert, 1984) and probes were prepared by random priming Farkas, G., Leibovitch, B.A., and Elgin, S.C. (2000). Chromatin orga- PCR products from each region (Exon 1, Exon 2, and DHS). In the nization and transcriptional control of gene expression in Drosophcase of the Figure 4B Northern, low-stringency hybridization solution ila. Gene 253, (30% formamide, 1 M NaCl, 100 mm NaPO 4 [ph 7.0], 7% SDS, Franke, A., and Baker, B.S. (1999). The rox1 and rox2 RNAs are 10X Denhardts, and 100 g/ml ssdna) was used. After overnight essential components of the compensasome, which mediates doshybridization at 42 C, the membrane was washed two times in 2X age compensation in Drosophila. Mol. Cell 4, SSC, 0.1% SDS solution at 42 C. Fujii, S., and Amrein, H. (2002). Genes expressed in the Drosophila head reveal a role for fat cells in sex-specific physiology. EMBO J. Transformation and Immunostaining of Polytene Chromosomes 21, Transgenic flies were made by P element-mediated transformation Granok, H., Leibovitch, B.A., Shaffer, C.D., and Elgin, S.C. (1995). (Spradling and Rubin, 1982). Preparation of polytene chromosomes Chromatin. Ga-ga over GAGA factor. Curr. Biol. 5, and immunostaining using anti-msl1 antibody were performed as Hall, I.M., Shankaranarayana, G.D., Noma, K., Ayoub, N., Cohen, previously described (Kelley et al., 1999). A., and Grewal, S.I. (2002). Establishment and maintenance of a heterochromatin domain. Science 297, RT-PCR To find the 5 and 3 ends of rox2, 5 and 3 RACE were performed Ho, Y., Elefant, F., Cooke, N., and Liebhaber, S. (2002). A defined using the SMART RACE cdna amplification kit (Clontech). For 5 locus control region determinant links chromatin domain acetylation RT-PCR of rox2 (Figure 1B), primers 5 (5 -CATGCAGTTCCCACTA with long-range gene activation. Mol. Cell 9, TATTTTATT-3 ), 4 (5 -GGTATTTTCGCAGTCATTA-3 ), 3 (5 -GTGTG Kageyama, Y., Mengus, G., Gilfillan, G., Kennedy, H.G., Stuckenholz, GCCAAAACTCGAAAA-3 ), 2 (5 -TATCAAGGGCTAGAGCAGCT-3 ), C., Kelley, R.L., Becker, P.B., and Kuroda, M.I. (2001). Association 1(5 -AGATGTTGCGGCATTCGCGG-3 ), and 6 (5 -ATTGCGACTTG and spreading of the Drosophila dosage compensation complex TACAATGTTG-CGTT-3 ) were used. To find rox2 sequences from from a discrete rox1 chromatin entry site. EMBO J. 20, other species, several primer sets from exon 2 were used together Kelley, R.L., Meller, V.H., Gordadze, P.R., Roman, G., Davis, R.L., in one PCR reaction using oligo dt-primed cdna as template (an- and Kuroda, M.I. (1999). Epigenetic spreading of the Drosophila nealing temperature 55 C, 40X). After exon 2 from other species dosage compensation complex from rox RNA genes into flanking was found, a primer from CG11695 (the adjacent gene 5 of rox2 chromatin. Cell 98, in D. melanogaster) and a species-specific primer for exon 2 were Kingston, R.E., and Narlikar, G.J. (1999). ATP-dependent remodeling used to find the promoter, exon 1, and intron regions using genomic and acetylation as regulators of chromatin fluidity. Genes Dev. 13, DNA from each species as template. To find the DHS region, a primer from nod (the 3 downstream gene of rox2) was utilized with a species-specific primer for exon 2. Lachner, M., O Carroll, D., Rea, S., Mechtler, K., and Jenuwein, T. (2001). Methylation of histone H3 lysine 9 creates a binding site for HP1 proteins. Nature 410, Acknowledgments Lee, J.T., and Jaenisch, R. (1997). Long-range cis effects of ectopic We are grateful to R.L. Kelley and H. Oh for critical reading of the X-inactivation centres on a mouse autosome. Nature 386, manuscript and many helpful discussions. We thank C. Stuckenholz Lehmann, M., Siegmund, T., Lintermann, K., and Korge, G. (1998). for D. simulans rox1 DHS sequence and for help with computer The pipsqueak protein of Drosophila melanogaster binds to GAGA

10 Molecular Cell 986 sequences through a novel DNA-binding domain. J. Biol. Chem. 273, Lyman, L.M., Copps, K., Rastelli, L., Kelley, R.L., and Kuroda, M.I. (1997). Drosophila male-specific lethal-2 protein: structure/function analysis and dependence on MSL-1 for chromosome association. Genetics 147, Markstein, M., Markstein, P., Markstein, V., and Levine, M.S. (2002). Genome-wide analysis of clustered Dorsal binding sites identifies putative target genes in the Drosophila embryo. Proc. Natl. Acad. Sci. USA 99, Meller, V.H., Wu, K.H., Roman, G., Kuroda, M.I., and Davis, R.L. (1997). rox1 RNA paints the X chromosome of male Drosophila and is regulated by the dosage compensation system. Cell 88, Meller, V.H., Gordadze, P.R., Park, Y., Chu, X., Stuckenholz, C., Kelley, R.L., and Kuroda, M.I. (2000). Ordered assembly of rox RNAs into MSL complexes on the dosage compensated X chromosome in Drosophila. Curr. Biol. 10, Meller, V.H., and Rattner, B.P. (2002). The rox RNAs encode redundant male-specific lethal transcripts required for targeting of the MSL complex. EMBO J. 21, Mette, M.F., Aufsatz, W., van der Winden, J., Matzke, M.A., and Matzke, A.J.M. (2000). Transcriptional silencing and promoter methylation triggered by double-stranded RNA. EMBO J. 19, Muchardt, C., Guilleme, M., Seeler, J.S., Trouche, D., Dejean, A., and Yaniv, M. (2002). Coordinated methyl and RNA binding is required for heterochromatin localization of mammalian HP1alpha. EMBO Rep. 3, Nakayama, J., Rice, J.C., Strahl, B.D., Allis, D.C., and Grewal, S.I. (2001). Role of histone H3 lysine 9 methylation in epigenetic control of heterochromatin assembly. Science 292, Pannuti, A., and Lucchesi, J.C. (2000). Recycling to remodel: evolution of dosage-compensation complexes. Curr. Opin. Genet. Dev. 10, Park, Y., and Kuroda, M.I. (2001). Epigenetic aspects of X-chromosome dosage compensation. Science 293, Park, Y., Kelley, R.L., Oh, H., Kuroda, M.I., and Meller, V.H. (2002). Extent of chromatin spreading determined by rox RNA recruitment of MSL proteins. Science 298, Smith, E.R., Pannuti, A., Gu, W., Steurnagel, A., Cook, R.G., Allis, C.D., and Lucchesi, J.C. (2000). The Drosophila MSL complex acetylates histone H4 at lysine 16, a chromatin modification linked to dosage compensation. Mol. Cell. Biol. 20, Spradling, A.C., and Rubin, G.M. (1982). Transposition of cloned P-elements into Drosophila germ line chromosomes. Science 218, Thummel, C.S., and Pirrotta, V. (1991). Technical notes: new pcasper P-element vectors. D. I. S. 2. Volpe, T.A., Kidner, C., Hall, I.M., Teng, G., Grewal, S.I., and Martienssen, R.A. (2002). Regulation of heterochromatic silencing and histone H3 lysine-9 methylation by RNAi. Science 297, Xiao, H., Sandaltzopoulos, R., Wang, H.M., Hamiche, A., Ranallo, R., Lee, K.M., Fu, D., and Wu, C. (2001). Dual functions of largest NURF subunit NURF301 in nucleosome sliding and transcription factor interactions. Mol. Cell 8,