Regulation of Histone H4 Lys16 Acetylation by Predicted Alternative ...

Download PDF
3 downloads 116 Views 839KB Size Report
Comment
Feb 10, 2008 - Mailing address: Department of Cell Biol- ogy and Molecular Medicine, UMDNJ-New Jersey Medical School,. 185 South Orange Ave., Newark, ...
MOLECULAR AND CELLULAR BIOLOGY, Aug. 2008, p. 4952–4962 0270-7306/08/$08.00⫹0 doi:10.1128/MCB.00219-08 Copyright © 2008, American Society for Microbiology. All Rights Reserved.

Vol. 28, No. 16

Regulation of Histone H4 Lys16 Acetylation by Predicted Alternative Secondary Structures in roX Noncoding RNAs䌤† Seung-Won Park,1‡ Mitzi I. Kuroda,2,3 and Yongkyu Park1* Department of Cell Biology and Molecular Medicine, UMDNJ-New Jersey Medical School, Newark, New Jersey 07103,1 and Harvard-Partners Center for Genetics & Genomics, Brigham & Women’s Hospital,2 and Department of Genetics, Harvard Medical School,3 Boston, Massachusetts 02115 Received 10 February 2008/Returned for modification 14 April 2008/Accepted 29 May 2008

Despite differences in size and sequence, the two noncoding roX1 and roX2 RNAs are functionally redundant for dosage compensation of the Drosophila melanogaster male X chromosome. Consistent with functional conservation, we found that roX RNAs of distant Drosophila species could complement D. melanogaster roX mutants despite low homology. Deletion of a conserved predicted stem-loop structure in roX2, containing a short GUb (GUUNUACG box) in its 3ⴕ stem, resulted in a defect in histone H4K16 acetylation on the X chromosome in spite of apparently normal localization of the MSL complex. Two copies of the GUb sequence, newly termed the “roX box,” were functionally redundant in roX2, as mutants in a single roX box had no phenotype, but double mutants showed reduced H4K16 acetylation. Interestingly, mutation of two of three roX boxes in the 3ⴕ end of roX1 RNA also reduced H4K16 acetylation. Finally, fusion of roX1 sequences containing a roX box restored function to a roX2 deletion RNA lacking its cognate roX box. These results support a model in which the functional redundancy between roX1 and roX2 RNAs is based, at least in part, on short GUUNUACG sequences that regulate the activity of the MSL complex. sidering that there was no additional sequence homology apparent between the two RNAs, it seemed likely that roX RNAs instead have common functional domains, such as degenerate primary sequences and/or secondary structures. Using a comparative evolutionary approach with several Drosophila species (14), conserved short primary sequences and putative stemloop structures were found within the roX RNAs, in spite of the low homology between their full-length sequences. Specifically, multiple GUb (GUUNUACG box) sequences, including an exact match within the original 25/30-nt region of identity between D. melanogaster roX1 and roX2 (3), were found in the 3⬘ ends of both roX RNAs in all species. Interestingly, a key predicted stem-loop structure in roX2 RNA includes a GUb sequence in its 3⬘ stem, and expression of an RNA consisting of six tandem repeats of this putative stem-loop (six repeats of 72 nt) is sufficient for MSL targeting and H4K16ac on the X chromosome (14). These results suggest that the GUb sequence is likely to play an important functional role in both roX RNAs but that its importance was missed previously due to its presence in multiple copies. In this study, we explored this hypothesis by testing roX RNAs of distantly related Drosophila species for function and by deletion or mutation analysis of conserved sequences in D. melanogaster roX genes. Our results provide strong evidence that a common sequence motif in roX RNAs, previously designated the GUb, is responsible, at least in part, for the functional redundancy seen between and within roX RNAs. We propose renaming the GUb the “roX box” in recognition of its key function in roX RNAs.

Dosage compensation of the X chromosome in mammals and Drosophila occurs by global regulation of a specific chromosome, mediated by noncoding RNAs (15). In mammals, Xist RNA is essential for X inactivation, while in fruit flies, two noncoding RNAs (roX1 [3,700 nucleotides {nt}] and roX2 [500 nt]) are functionally redundant despite their differences in size and sequence (3, 11). roX (RNA on X) RNAs are essential components in the male-specific lethal (MSL) complex with MSL1, MSL2, MSL3, MOF (histone acetyltransferase), and MLE (RNA helicase) proteins (10). The MSL complex induces twofold hypertranscription of hundreds of genes on the male X chromosome (4, 7, 10, 20) through specific histone H4 Lys16 acetylation (H4K16ac) (5, 19, 23). However, it is not known how roX RNAs function within the MSL complex and how they are functionally redundant. Previously, the following two common motifs were found from comparison of the two Drosophila melanogaster roX genes: (i) a 25/30-nt region of identity near the 3⬘ end of each transcription unit (3) and (ii) an ⬃110-bp region (internal in roX1 and located downstream of the major 3⬘ end of roX2) capable of attracting the MSL proteins to roX transgenes and forming a male-specific DNase I-hypersensitive site (DHS) (6, 17). Surprisingly, deletion analysis revealed that neither sequence is essential for roX function as an RNA (17, 21). Con-

* Corresponding author. Mailing address: Department of Cell Biology and Molecular Medicine, UMDNJ-New Jersey Medical School, 185 South Orange Ave., Newark, NJ 07103. Phone: (973) 972-2969. Fax: (973) 972-7489. E-mail: [email protected]. † Supplemental material for this article may be found at http://mcb .asm.org/. ‡ Present address: Department of Microbiology, College of Medicine, The Catholic University of Korea, Seoul 137-701, Republic of Korea. 䌤 Published ahead of print on 9 June 2008.

MATERIALS AND METHODS Fly genotypes and cross. The D. melanogaster full genotypes were y w (wild type [WT]) and y w roX1ex6 Df(1)roX252 P{w⫹ 4⌬4.3} (roX⫺ double mutant) (16). The P{w⫹ 4⌬4.3} element contains the essential genes lost in Df(1)roX252

4952

VOL. 28, 2008

REGULATION OF H4K16 ACETYLATION BY roX

4953

FIG. 1. Functional conservation of roX RNAs between Drosophila species. (A) RNA in situ hybridization of D. willistoni polytene chromosomes (blue), using a D. willistoni roX2 probe (red), which shows a metacentric X chromosome with two arms (XL and XR). C, chromocenter. (B to D) Polytene chromosomes of transgenic D. melanogaster male larvae expressing D. ananassae roX1 (B) or roX2 (C) or D. willistoni roX2 (D) in a roX⫺ mutant background. The chromosomes were immunostained with anti-MSL1 (red) and counterstained with DAPI (4⬘,6⬘-diamidino-2-phenylindole) (blue). Arrows, roX transgenes of different species; arrowhead, roX2 DHS of D. melanogaster. (E) Male rescue frequency (male/female ratio) by H83MeroX, H83AnroX, or H83WiroX in the D. melanogaster roX⫺ mutant. Averages of male viability (%) are represented, with standard deviations, for H83MeroX1 (three independent transgenics tested), H83MeroX2 (three transgenics tested), H83AnroX1 (four transgenics tested), H83AnroX2 (four transgenics tested), and H83WiroX2 (three transgenics tested). No, no transgene. (F) Northern analysis showing male-specific expression of H83WiroX2 RNA in a D. melanogaster roX⫺ mutant. Twenty micrograms of total RNA from transgenic adult flies was loaded per lane, and a D. willistoni roX2 probe was used. Sizes of upper (⬃800 nt) and lower (⬃600 nt) bands were estimated from an RNA ladder (Invitrogen). rp49, loading control; M, male; F, female. (G) Alternative splicing of H83AnroX2 and H83WiroX2 in D. melanogaster. Species-specific primers are indicated in the gene structure of roX2, which is composed of exon 1, exon 2, exon 3, and the DHS (17). Thick and dotted lines represent the major and 3⬘-extended minor transcripts, respectively. To check the major (lanes 3 and 6) and minor (lanes 8 and 10) splice forms of roX2 RNA, several primer sets were used for RT-PCR, using total RNAs isolated from adult male transgenic flies. In lane 2, H83AnroX2 plasmid (p) was used as a template to mark the size of the nonspliced form. Lanes 1 and 5, 1-kb Plus DNA ladder (Invitrogen); *, 400 bp, indicated on DNA ladders in the following figures.

(11). WT Drosophila ananassae and Drosophila virilis flies were obtained from the Tucson Stock Center at the University of Arizona. For male rescue experiments, male viability was scored in the presence of each roX transgene in a roX⫺ double mutant. y w; [roX transgene] males were crossed to y w roX1ex6 Df(1)roX252 P{w⫹ 4⌬4.3} females, and then the ratio of the male (y w roX1ex6 Df(1)roX252 P{w⫹ 4⌬4.3}/Y; [roX transgene]/⫹) to female (y w roX1ex6 Df(1)roX252 P{w⫹ 4⌬4.3}/y w; [roX transgene]/⫹) progeny was calculated using flies collected during the first 10 days of eclosion. The average male rescue frequency (%) was calculated, with standard deviations, from at least three different transgenic locations, except for ⌬A ⌬B ⌬C ⌬D ⌬E transgenics (two lines), in which two locations showed similarly high male rescue frequencies (see Fig. S2 in the supplemental material). For immunostaining of salivary gland polytene chromosomes, y w roX1ex6 Df(1)roX252 P{w⫹ 4⌬4.3}/Y; [roX transgene]/⫹ male larvae obtained from the rescue cross were used. Immunostaining of MSL proteins and RNA in situ hybridization of Drosophila willistoni roX2 RNA were performed as previously described (7, 10). Transgene construction and transformation. According to the predicted sizes of roX genes in other Drosophila species (14), roX1 (H83AnroX1; 3,413 bp) and roX2 (H83AnroX2; 1,487 bp) of D. ananassae and roX2 of D. willistoni (H83WiroX2; 1,485 bp), shown in Fig. 1, were amplified from genomic DNA by use of species-specific primers (see the list of primers in the supplemental

material) and then cloned into the pCRII-TOPO vector (Invitrogen). After being sequenced, these fragments were subcloned into pCaSpeR Hsp83-act [containing an act5C gene fragment to provide a 3⬘ poly(A) site], using EcoRI/NotI (H83AnroX1 and H83AnroX2) or NheI/XhoI (H83WiroX2). To create ⌬A, ⌬B, ⌬C, ⌬D, and ⌬E constructs of roX2 (Fig. 2), inverse PCR was performed with Pfu Turbo DNA polymerase (Stratagene), using primers to delete specific regions from the whole roX2 gene (1,380 bp) within the pCRIITOPO vector (pYP47) (18). For the mX1 construct (Fig. 3), inverse PCR was performed using pYP47 as a template, with primers containing the mutated sequences within the roX2-box1 region (pYP153). For the construction of mX1mX2 (Fig. 3), pYP153 was amplified using inverse PCR with primers containing the mutated sequences within the roX2-box2 region. After self-ligation and sequencing, these deleted or mutated roX2 fragments were subcloned into pCaSpeR Hsp83-act by using XhoI/NotI sites. The WT roX2 fragment (1,380 bp) was also inserted into pCaSpeR Hsp83-act to create H83MeroX2 (18). For the roX2 ⌬X2 construct (Fig. 3), inverse PCR was performed using pYP47 as a template, with primers to delete the 3⬘ minor transcript region (332 bp), including the roX2-box2 and the DHS (pYP8). To create deletions of ⌬F, ⌬5, and ⌬X1 from the ⌬X2 construct (Fig. 2 and 3), pYP8 (roX2 ⌬X2; 1,048 bp) was amplified with inverse PCR to delete specific regions. To create H-X2 from the ⌬X2 construct (see Fig. 6), pYP8 was amplified with inverse primers that incor-

4954

MOL. CELL. BIOL.

PARK ET AL.

FIG. 2. Exon 3 and roX2-box2 sequences of D. melanogaster roX2 gene (mel) and the consensus roX2 sequences (Con) between nine Drosophila species, which were aligned using the ClustalW program (http://www.ebi.ac.uk/clustalw) supplemented by manual inspection (see Fig. S1 in the supplemental material) (14). The previously described “GUb-2” of roX2 (14) was renamed “roX2-box2.” In the consensus sequences (Con), bold red characters represent perfect matches (no mismatch or mismatch in only one of nine species) and bold black characters represent less-thanperfect matches (mismatch in two of nine species) (see Fig. S1 in the supplemental material). In mel sequences numbered from the whole roX2 gene (1,380 bp), the conserved sequences are exhibited by bold black characters. Double arrows represent the deleted regions within the roX2 gene in each construct. Vertical arrows, poly(A) sites mapped in Fig. 3A; gray boxes, 5⬘-stem, roX2-box1, and roX2-box2 regions; blue characters, mutated nucleotides (Fig. 3A).

porated the roX1-box2 sequence into the 5⬘ end of the primer. After self-ligation and sequencing, these roX2 fragments were subcloned into pCaSpeR Hsp83-act. For the mX3 construct of roX1 (see Fig. 5), inverse PCR was performed with primers including the mutated sequences within the roX1-box3 region, using pKSc20roX1inv plasmid as a template, which contains the c20 cDNA (3,390 bp) of roX1 in pBlueScript (21). To create 1⌬X2-mX3 (see Fig. 5), the same primers were used for inverse PCR using pKSc20roX1⌬13 (1⌬X2) as a template, in which 34 bp, including the roX1-box2 sequence, was already deleted from the pKSc20roX1inv plasmid (21). After self-ligation and sequencing, these mutated roX1 fragments were subcloned into the pCaSpeR Hsp83-tra2 plasmid [containing a 450-bp tra2 gene fragment to provide a 3⬘ poly(A) site] by use of the EcoRI site. The WT roX1 (3,390 bp), 1⌬X2 (3,356 bp), and ⌬SL (3,321 bp) fragments were inserted into the same pCaSpeR Hsp83-tra2 plasmid to obtain H83MeroX1, 1⌬X2, and ⌬SL, respectively (21). Transgenic flies were made by P-element-mediated transformation at Model System Genomics of Duke University. Immunoprecipitation, RT-PCR, and 3ⴕ rapid amplification of cDNA ends (3ⴕ RACE) analysis. To detect alternatively spliced forms of roX2 RNAs of other Drosophila species expressed in D. melanogaster (Fig. 1G), oligo(dT)-primed cDNAs were made from 5 ␮g total RNA of transgenic D. melanogaster male adults containing H83AnroX2 and H83WiroX2. Species-specific primers (see the list of primers in the supplemental material) were used for reverse transcriptionPCR (RT-PCR). To check the size of ⌬F RNA, which has a 55-nt deletion (see Fig. S2 in the supplemental material) and the major splice form of roX2 RNA from roX2 transgenes (see Fig. 3D, 4C, and 6B), oligo(dT)-primed cDNAs were produced from 5 ␮g total RNA of male adults containing each transgene in the roX⫺ mutant background. For 3⬘ RACE experiments with roX2 (see Fig. 3D, 4C, and 6B) and roX1 (see Fig. 5E) RNAs, primer TPII (see the list of primers in the supplemental material) was used to make cDNA from 5 ␮g total RNA of male adults containing each transgene in the roX⫺ mutant background. In the primary PCR, primer 2 for roX2 (Fig. 3A) and primer 1 for roX1 (see Fig. 5A) were used with primer SPII (3⬘ primer). Using the primary PCR product as a template, primer 3 for roX2 (Fig. 3A) and primer 2 for roX1 (see Fig. 5A) were used with primer SP (3⬘ primer) in a secondary PCR. Polyadenylation sites of roX RNAs were mapped after sequencing of the purified 3⬘ RACE product from each reaction. These experiments were performed by a method modified from protocol PT3269-1 for Smart RACE cDNA amplification (Clontech). For immunoprecipitation (IP) of MSL complexes (Fig. 3F), total protein extracts from SL2 cells, which were cotransfected with the same amount (2 ␮g) of pCaSpeR Hsp83-act plasmid (C) and ⌬X1-⌬X2 plasmid (⌬X) (Qiagen), were prepared in extraction buffer (9). One milligram of extract was incubated with 3 ␮l MSL protein-specific antiserum or preimmune serum as previously described (10). After IP, total RNAs were isolated and reverse transcribed using an oligo(dT) primer, and then the cDNAs were analyzed by PCR.

ChIP and real-time PCR analyses. For chromatin IP (ChIP) experiments (see Fig. 5C), 200 pairs of salivary glands were isolated from the third instar male larvae of each roX transgenic carrying the roX⫺ mutant background and then cross-linked with 1.0% formaldehyde in PBST (1⫻ phosphate-buffered saline supplemented with 0.2% Triton X-100) for 5 min at room temperature. Three microliters of rabbit anti-H4K16ac (Serotec) or no antibody was used for ChIP, and the subsequent procedures were performed according to the manufacturer’s instructions (Upstate). Ten nanograms of immunoprecipitated DNA was used as a template for quantitative real-time PCR, which was performed with SYBR green, using an ABI7300 real-time PCR instrument (Applied Biosystems). The changes (ratio of anti-H4K16ac to no antibody) in H4K16ac in the CG3016 and CG1793 genes (see Fig. 5C) were determined by the comparative cycle threshold method (ABI Prism 7700 sequence detection system user bulletin 2; Applied Biosystems). Northern analysis. For Northern analysis (see Fig. 1F, 3E, 4B, 5F, and 6E), total RNA was purified from transgenic adult flies with the roX⫺ mutant background, using TRIzol reagent (Gibco-BRL), and 20 ␮g of total RNA was loaded in each lane. The probes hybridized to the roX RNAs were prepared by random priming (Invitrogen) using PCR products of roX genes amplified from genomic DNA. After overnight hybridization at 42°C in hybridization solution (30% formamide, 1 M NaCl, 100 mM NaPO4, pH 7.0, 7% sodium dodecyl sulfate, 10⫻ Denhardt’s solution, 100 ␮g/ml single-stranded salmon sperm DNA), the membrane was washed twice in 2⫻ SSC (1⫻ SSC is 0.15 M NaCl plus 0.015 M sodium citrate)–0.1% sodium dodecyl sulfate at 42°C. The same membranes were stripped and reprobed with rp49 probes to normalize loading.

RESULTS roX RNAs from diverse Drosophila species function in D. melanogaster in spite of low sequence homology. We previously found that endogenous roX1 and roX2 RNAs from Drosophila species ranging from D. melanogaster to D. virilis (40 million years apart) showed male-specific expression and X-chromosome-specific binding (Fig. 1A) (14, 17, 18), indicating that the function of roX RNAs in dosage compensation of the X chromosome is evolutionarily conserved. In distantly related Drosophila species ranging from D. ananassae to D. virilis, however, the homology to full-length D. melanogaster roX1 and roX2 genes was only 33 to 30% and 45 to 34%, respectively, almost as low as that of unrelated control sequences (28 to 27%) (14). This suggested that roX RNAs retain common

VOL. 28, 2008

REGULATION OF H4K16 ACETYLATION BY roX

4955

FIG. 3. Defective H4K16ac caused by deletion of a stem-loop region of roX2. (A) Gene structure of roX2 RNA, with a predicted stem-loop structure (14). Vertical arrows, poly(A) sites mapped by 3⬘ RACE; horizontal arrows, primers used for RT-PCR; hatched box in DHS, 25/30-nt identity region found by initial sequence comparison between roX1 and roX2 (3); black box in DHS, 110-bp segment containing conserved sequences for MSL binding located in both roX1 and roX2 (17); Con, conserved roX2 sequence from nine Drosophila species (see Fig. S1 in the supplemental material); N, random; p, purine base; y, pyrimidine base; ⌬X2, deletion of the 3⬘ extension (332 bp) found in the minor transcript, including roX2-box2 and DHS regions; ⌬5, 37-bp deletion removing the 5⬘ stem (see nucleotides detailed in Fig. 2); ⌬X1, 60-bp deletion removing roX2-box1; mX1, point mutations within the roX2-box1 sequence; mX2, point mutations within roX2-box2; blue characters, mutated nucleotides. (B) Rescue frequencies (%) for male viability from transgenic lines. Four independent lines were tested for each construct, except for the WT and ⌬5-⌬X2 constructs (three independent lines tested). (C) Polytene chromosomes of ⌬X2, ⌬5-⌬X2, and ⌬X1-⌬X2 male transgenic larvae immunostained with antibodies against MSL proteins and H4K16ac. Colocalization of red and green immunofluorescence is visualized as yellow. Red, MSL1 and H4K16ac; green, MSL3. (D) RT-PCR with primers in panel A, using total RNAs purified from male transgenic adults carrying a roX⫺ mutant background (no-RT control not shown). Arrows in the 3⬘ RACE analysis represent 3⬘ poly(A) sites (Fig. 2 and panel A), which were sequenced and mapped. ⌬X2 and ⌬5-⌬X2 transcripts (lanes 9 and 10) produced the same poly(A) site (arrow 2). However, the 3⬘ RACE product from ⌬5-⌬X2 RNA was smaller due to the 37-bp deletion in the 5⬘ stem region. E, endogenous roX2 RNA from yw male adults. (E) Northern analysis using the same total RNAs used for panel D. The exposure time was 2 days. ⌬5-⌬X2 RNA (lane 4) showed a slightly smaller size than did ⌬X2 RNA (lane 2) due to the 37-bp deletion in the 5⬘ stem region. rp49, loading control. (F) Interaction of MSL proteins with ⌬X1-⌬X2 roX2 RNA in SL2 cells. To detect RT-PCR products of vector (C; 217 bp) and ⌬X1-⌬X2 (⌬X; 362 bp) RNAs, the same 3⬘ primer (primer 3) and different 5⬘ primers (primers 1 and 2) were used. PCR was performed for 30 (before IP [TP]) and 35 (after IP [MSL proteins]) cycles. E, endogenous roX2 (343 bp); TP, total protein extracts as an input control; MLE pre, MLE preimmune serum used as a negative control in the IP; X1, roX2-box1 primer shown in panel A.

functional domains, such as short primary sequences and/or secondary structures, in spite of considerable sequence divergence. To test this hypothesis, we asked whether roX RNAs of distantly related Drosophila species could function in D. melanogaster. We expressed the D. ananassae roX1 (H83AnroX1), D. ananassae roX2 (H83AnroX2), and D. willistoni roX2 (H83WiroX2) genes under the control of a constitutive promoter (hsp83) in D. melanogaster roX1⫺ roX2⫺ double mutants. The transgenes were named by the promoter utilized (H83), the species (“An” for D. ananassae and “Wi” for D.

willistoni), and the roX gene name. In each case, the transgenic RNAs could paint the X chromosome (X) in conjunction with the endogenous MSL proteins of D. melanogaster (Fig. 1B to D) and could rescue the male-specific lethality of roX⫺ double mutants (Fig. 1E). In addition, these hybrid complexes had the ability to spread in cis (two or three bands) from the autosomal insertion site of the transgene (arrows in Fig. 1B and C) and could bind to the roX2 DHS of D. melanogaster (Fig. 1D, arrowhead in small rectangle). These results demonstrate that in spite of considerable sequence divergence, roX RNAs from

4956

PARK ET AL.

distant species have conserved the ability to interact with MSL proteins and with the roX2 DHS regulatory site in D. melanogaster. In Northern analysis, D. melanogaster carrying an H83WiroX2 transgene as its only functional roX gene expressed a malespecific roX2 RNA (Fig. 1F), while the RNA was not detected in females. This result indicates that D. willistoni roX2 transcripts expressed from the constitutive hsp83 promoter are stable only in males, presumably because males produce MSL proteins, while females do not. Previously, we found that alternative splicing of D. melanogaster roX2 is important for localization of the MSL complex to the X chromosome (18). A major splice form of roX2 RNA is composed of exon 1 joined to exon 3 (Fig. 1G), and the exon 2 region exhibits numerous combinations of alternative miniexons to create multiple minor splice forms, which are evolutionarily conserved in other Drosophila species (14, 18). These alternative splicing patterns still occurred when H83AnroX2 and H83WiroX2 transgenes were expressed in D. melanogaster (Fig. 1G), consistent with previous evidence that alternative splicing of roX2 RNA is relevant to its function (18). Genetic complementation in D. melanogaster indicates that the functional domains of roX RNAs are evolutionarily conserved even in distantly related Drosophila species, despite low overall sequence homology. Functional redundancy within roX2. Considering the fact that roX2 RNA (500 nt) is much smaller than roX1 RNA (3,700 nt) but is still functionally equivalent, we focused initially on identification of functional domains in roX2. The exon 3 (431 nt) region encompasses evolutionarily conserved primary sequences and a predicted secondary structure in all nine Drosophila species (Fig. 2; see Fig. S1 in the supplemental material) (14). We started by making transgenic flies with constructs that deleted 17-bp segments from each of four specific regions containing evolutionarily conserved sequences (⌬A, ⌬〉, ⌬C, and ⌬⌭) and from a nonspecific region (⌬D) of the complete roX2 gene (1,380 bp) under the control of the constitutive hsp83 promoter (Fig. 2). To analyze any functional defects in these deleted roX2 RNAs, we expressed these transgenes in roX⫺ double mutants and scored the survival of male flies. Unexpectedly, all of them functioned normally, although the rescue frequency of ⌬C (64%) was slightly lower than that of a WT roX2 transgene (80%) (see Fig. S2 in the supplemental material). One explanation for the function of these deletion constructs is that redundancy exists within the roX2 RNA, similar to the internal redundancy in roX1 revealed by deletion analysis (21). Another possibility is that roX2 is a noncoding RNA which might be tolerant of small deletions or variations of nucleotides. Therefore, we made a larger deletion (⌬F; 55 bp), including the conserved “〈” segment (17 bp) and another conserved region 5⬘ of the “〈” segment (Fig. 2). At this point, to exclude the possibility of redundant function from the 3⬘ region of roX2, the ⌬F construct was also deleted for the 3⬘ region downstream of the major polyadenylation site (⌬X2; 332 bp, including roX2-box2 and DHS regions) (Fig. 2 and 3A), which normally contributes to a minor transcript (17). The 55-nt deletion of exon 3 in the ⌬F-⌬X2 transcript was confirmed compared to a ⌬X2 transcript (see Fig. S2 in the supplemental material). However, this 13% deletion from exon 3 also maintained normal function of roX2 RNA, including

MOL. CELL. BIOL. TABLE 1. Efficiencies of H4K16ac in different roX transgenic flies % of MOF immunostaininga

% of H4K16aca

H83roX2 transgenics WT ⌬X2 ⌬5-⌬X2 ⌬X1-⌬X2 mX1 mX1-mX2 H-X2 W-SL-6 M-SL-6

99 (526) 98 (479) 99 (394) 65 (455) 98 (522) 96 (389) 92 (494) 95 (383) 0b

80 (323) 97 (354) 11 (256) 20 (335) 61 (607) 32 (296) 63 (430) 57 (463) 0b

H83roX1 transgenics WT 1⌬X2 mX3 1⌬X2-mX3 ⌬SL

100 (535) 99 (559) 100 (499) 97 (411) 93 (401)

83 (553) 57 (546) 63 (394) 22 (315) 78 (264)

Transgenic background

a Percentages (%) of immunostaining of MOF or H4K16ac (red) on the X chromosomes colocalized with MSL3 (green) on the polytene chromosomes of male transgenic larvae carrying the roX⫺ mutant background. First, the numbers of nuclei showing MSL3 staining were counted (in parentheses), and next, double staining for MSL3 plus MOF or MSL3 plus H4K16ac was counted. The percentage of double staining was calculated as follows: 关(MSL3 ⫹ MOF)/MSL3兴 or 关(MSL3 ⫹ H4K16ac)/MSL3兴 ⫻ 100 (%). b No staining of MSL3, MOF, or H4K16ac was detected from M-SL-6 transgenic flies which carried point mutations (Fig. 3A) within the roX2-box1 sequence of W-SL-6 (14).

MSL binding to the X chromosome, H4K16ac, and high rescue of male-specific lethality in a roX⫺ double mutant (see Fig. S2 in the supplemental material). This indicates that although they are evolutionarily conserved, these sequences in roX2 RNA might still be internally redundant under the conditions assayed. Deletion of a stem-loop region in roX2 RNA results in defective H4K16ac. Previously, a stem-loop structure was predicted to form in the 3⬘ end of exon 3 of roX2 RNA (14). This predicted secondary structure, with conserved primary sequences in the 5⬘ and 3⬘ segments, was also found in nine different Drosophila species (Fig. 2; see Fig. S1 in the supplemental material) (14). We previously discovered that expression of six tandem repeats of this stem-loop region was sufficient for targeting the MSL complex and inducing H4K16ac on the X chromosome, even in the absence of other regions of roX2 RNA (Table 1), although male rescue was relatively low (17%) (14). To determine whether this region is essential for roX2 function, we created transgenic flies with constructs that deleted 37 bp from the 5⬘ stem region (⌬5) or 60 bp including the 3⬘ stem region (⌬X1) (Fig. 2 and 3A). These deletion mutants were produced using a ⌬X2 construct, which lacks sequences downstream of the major 3⬘ end of roX2 RNA (Fig. 2 and 3A) (17). Similar to those from the WT and ⌬X2 constructs (Fig. 3D, lanes 2 and 3), transcripts from ⌬X1-⌬X2 and ⌬5-⌬X2 were still capable of producing the major spliced roX2 transcript (Fig. 3D, lanes 4 and 5). In 3⬘ RACE analysis, polyadenylation of ⌬X2 RNA occurred at a single site in the 3⬘ end of exon 3 (Fig. 3A, vertical arrow 2, and D, lane 9, arrow 2), in contrast to two sites of poly(A) in the endogenous roX2 gene and the WT roX2 transgene (Fig. 3A, vertical arrows 2 and 4, and D, lanes 7 and 8, arrows 2 and 4). The ⌬5-⌬X2 and

VOL. 28, 2008

⌬X1-⌬X2 transcripts were also capable of 3⬘-end processing (Fig. 3A, vertical arrows 1 and 2, and D, lanes 10 and 11, arrows 2 and 1, respectively). The synthesis sites of ⌬5-⌬X2 and ⌬X1-⌬X2 transcripts (sites of transgene insertion are shown by arrows in Fig. 3C) could still attract the MSL proteins, including the MOF acetyltransferase, and the MSL complex was detected along the length of the X chromosome by immunostaining of polytene chromosomes (Fig. 3C). However, a key difference between deletions of the stem-loop and other deletions in roX2 was that the intensity of H4K16ac on the X chromosome was greatly diminished (Fig. 3C; see Fig. S2 in the supplemental material). To calculate the efficiencies of MSL complex binding and H4K16ac on the X chromosome, we performed double staining with anti-MSL3 and anti-MOF or anti-H4K16ac antibodies on the polytene chromosomes of transgenic larvae (Table 1). We identified nuclei positive for MSL3 staining and then scored these for MOF or H4K16ac double staining. The ⌬5⌬X2 and ⌬X1-⌬X2 RNAs showed similar or slightly lower percentages of MOF binding (99% and 65%, respectively) than those of the WT and ⌬X2 RNAs (99% and 98%, respectively). However, efficiencies of H4K16ac on the X chromosomes of ⌬5-⌬X2 and ⌬X1-⌬X2 transgenics were significantly decreased (11% and 20%, respectively) compared to those for WT and ⌬X2 transgenics (80% and 97%, respectively). This indicates that although all five MSL proteins, including MOF, were localized to the X chromosome in the transgenics, this MSL complex could not acetylate histone H4K16 at WT levels. Consistent with the defect in histone modification on the X chromosome, ⌬5-⌬X2 and ⌬X1-⌬X2 transgenics exhibited poor rescue of roX⫺ mutants (20% and 6%, respectively), in contrast to WT (80%) and ⌬X2 (79%) transgenics (Fig. 3B). Although the ⌬X1-⌬X2 transcript could be detected by RTPCR (Fig. 3D), this transcript was not detected by Northern analysis (Fig. 3E, lane 3), and only low levels of ⌬5-⌬X2 transcripts were detected (Fig. 3E, lane 4). Binding efficiencies of MLE protein were 74% and 40% in ⌬5-⌬X2 and ⌬X1-⌬X2 transgenics, respectively, which are comparable to those for WT and ⌬X2 transgenics (91% and 100%, respectively) (see Table S1 in the supplemental material). At this point, we do not know yet if the ⌬5-⌬X2 and ⌬X1-⌬X2 RNAs are unstable due to the reduced interaction with MLE protein or due to the deletion itself. It is possible that the instability of deletion RNAs might result in the defect of H4K16ac on the X chromosome. However, a transcript composed of six tandem repeats of the roX2 stem-loop region (W-SL-6), which was similarly unstable (14), showed a high efficiency (57%) of H4K16ac on the X chromosome, in contrast to the ⌬5-⌬X2 RNA (11%) (Table 1). In addition, genomic roX2 constructs (M-GMroX2 and m-GMroX2) which were incapable of MSL localization to the X chromosome due to the absence of alternative splicing of roX2 RNA could still induce histone H4K16ac on their transgenic sites of synthesis, in spite of transcript instability (18; data not shown). These data suggest that deletion of the stem-loop region within roX2 RNA might directly affect the enzyme activity of the MSL complex rather than being an indirect effect of inefficient targeting. To confirm the interaction of ⌬X1-⌬X2 RNA with MSL proteins, including the MOF acetyltransferase, transgenic RNAs were tested for interaction with MSL proteins in the

REGULATION OF H4K16 ACETYLATION BY roX

4957

SL2 cell culture system. After cotransfection of two plasmids, including a construct encoding ⌬X1-⌬X2 RNA, total protein extracts were subjected to IP with anti-MSL antibodies. After IP, RNAs were recovered and subjected to RT-PCR analysis (Fig. 3F). Both control (C) and experimental plasmid (⌬X) transcripts were detected in total protein extracts before IP. In contrast, after IP, only the ⌬X1-⌬X2 RNA and the endogenous roX2 RNA were detected, indicating a specific interaction between ⌬X1-⌬X2 RNA and MSL proteins (Fig. 3F). Point mutations in “roX boxes” of roX2 result in defective H4K16ac. Upon detection of multiple GUb sequences in roX RNAs of all species, we renamed the core conserved sequence the “roX box.” To determine whether deletion of the evolutionarily conserved roX2-box1 sequence within the stem-loop region was responsible for the defect in H4K16ac, we made transgenic flies with a construct (mX1) that contained several base substitutions within the roX2-box1 sequence (blue nucleotides in Fig. 2 and 3A) but retained other parts of the roX2 gene intact. These point mutations were designed to interrupt the stem-loop structure in roX2 RNA. Previously, we introduced the same point mutations within the roX2-box1 sequence into a transcript composed of six tandem repeats of the stem-loop region (M-SL-6), and this mutation interfered with MSL targeting, and thus H4K16ac, on the X chromosome (Table 1) (14). Unexpectedly, this mX1 RNA exhibited normal roX2 function, including high rescue of male-specific lethality (Fig. 3B), MSL binding to the X chromosome, and H4K16ac (Fig. 4A), similar to other site-specific deletions within the exon 3 region (Fig. 2; see Fig. S2 in the supplemental material). This suggested that in the context of a full-length roX2 gene, the roX2-box1 sequence was not essential. Interestingly, the mX1 transcript was longer (⬃570 nt) (Fig. 4B, lane 2) than the mature WT roX2 RNA (⬃500 nt) (Fig. 4B; lane 1). One possibility is that the mX1 transcript contains additional mini exons from the exon 2 region instead of the exon 1-exon 3 major splice form normally predominant in the WT (Fig. 4C, lane 2). However, RT-PCR demonstrated the clear presence of the same major form in the mX1 RNA (Fig. 4C, lane 3). Another possibility is that the larger size of mX1 RNA might be contributed by extension of the 3⬘ end into the region usually found only in minor transcripts. In 3⬘ RACE analysis, WT roX2 transcripts are processed at two sites, namely, a major polyadenylation site at the 3⬘ end of exon 3 and a minor site at the 3⬘ end of the roX2-box2 sequence (Fig. 3A, vertical arrows 2 and 4, and 4C, lane 6, arrows 2 and 4) (17). By 3⬘ RACE analysis, we determined that mX1 RNA switched most 3⬘-end processing to the second site (Fig. 3A, vertical arrow 4, and 4C, lane 7, arrow 4). The 3⬘ RACE data were consistent with the larger size of mX1 RNA (⬃570 nt) detected in Northern analysis. ⌬A, ⌬〉, ⌬C, ⌬D, and ⌬⌭ transgenics (Fig. 2), which also showed normal function, like mX1, did not exhibit the longer transcript in Northern analysis found for the mX1 RNA (data not shown). These results suggest that point mutations within the roX2-box1 sequence of mX1 RNA can be compensated by a longer transcript containing the roX2-box2 region. To test if the roX2-box1 and roX2-box2 sequences are functionally redundant, we created transgenic flies with a combined construct (mX1-mX2) that contained additional point mutations within the roX2-box2 sequence (blue nucleotides in Fig.

4958

PARK ET AL.

MOL. CELL. BIOL.

FIG. 4. Defective H4K16ac caused by mutations in roX2 box sequences. (A) Polytene chromosomes from mX1 male transgenic larvae immunostained with anti-MSL1 (red) and anti-MSL3 (green) plus anti-H4K16ac (red). The arrow represents the insertion site of the transgene. (B) Northern analysis using total RNAs purified from male adults of each transgenic line in the roX⫺ mutant background. The WT RNA in lane 1 was from the same total amount (20 ␮g) as that in lane 1 of Fig. 3E. However, to see the difference in size between WT and mX1 roX2 RNAs, the exposure time was kept shorter (6 h). The mX1-mX2 RNA was not detected even with a longer exposure (data not shown). RNA sizes of WT (⬃500 nt) and mX1 (⬃570 nt) bands were estimated from an RNA ladder (Invitrogen). (C) RT-PCR analysis using the same total RNAs used for Northern analysis in panel B, with the primers shown in Fig. 3A (no-RT control not shown). Arrows in the 3⬘ RACE analysis represent 3⬘ poly(A) sites from each roX2 RNA (Fig. 3A). (D) Polytene chromosomes of male transgenic larvae of mX1-mX2, immunostained with antibodies against MSL proteins and H4K16ac. All antibodies except for anti-MSL3 (green) were detected with a red signal. (E) Predicted roX2 stem-loop structure (roX2-SL) with 5⬘ stem and roX2-box1 (top) or roX2-box2 (bottom). Bold red and black characters represent the consensus sequences shown in Fig. 2 and 3A. Blue characters are nucleotides mutated in the mX1 and mX2 constructs (Fig. 3A).

2 and 3A). The mX1-mX2 transcript was still spliced (Fig. 4C, lane 4) and polyadenylated (Fig. 3A, vertical arrows 3 and 5, and 4C, lane 8, arrows 3 and 5). However, mX1-mX2 transgenic flies also displayed an H4K16ac defect on the X chromosome (Fig. 4D), similar to the ⌬X1-⌬X2 and ⌬5-⌬X2 transgenics (Fig. 3C). Although binding of MOF protein to the X chromosome was evident (96%) (Table 1), the detection of H4K16ac on the X chromosome was significantly lower (32%), as was male rescue (8%) (Fig. 3B). Considering that ⌬X2 and mX1 RNAs still contain normal levels of H4K16ac (Fig. 3C and 4A), these results suggest that the roX2-box1 and roX2box2 sequences (GUUNUACG) of roX2 RNA are functionally redundant and are important for the H4K16ac activity of the MSL complex. The fact that the steady-state RNA level in mX1-mX2 transgenic flies was undetectable (Fig. 4B, lane 3), while its transcript could be detected in RT-PCR analysis (Fig. 4C, lanes 4 and 8), suggests that the roX2-box1 and roX2-box2 sequences also function in the stability of roX2 RNA. Figure

4E shows a possible interaction of the 5⬘ stem with the roX2box1 and roX2-box2 sequences. roX box mutations in roX1 also result in defective H4K16ac. Considering that roX1 (3,700 nt) and roX2 (500 nt) RNAs are functionally redundant in the MSL complex, despite their different sizes and primary sequences, we searched roX1 RNA for a functional domain similar to the roX2 stem-loop region. Previously, the 3⬘-terminal 600-nt segment of roX1 (Fig. 5A, double arrow) was identified as the most sensitive to mutation (21). This region contains a predicted stem-loop region (59 nt) in which we noticed a GUUNUCCG sequence in the 3⬘ stem region similar to the roX box sequence (GUUNUACG) (Fig. 5A) (14, 21). To determine if this roX1 stem-loop has a similar function to the roX2 stem-loop, we tested a mutant deleting the 69-nt roX1 stem-loop region (⌬SL) from the whole roX1 transcript (H83MeroX1; 3,390 bp) (21). However, in the stem-loop deletion mutant of roX1, the MSL complex was localized to the X chromosome and displayed efficient H4K16ac (Fig. 5B and

VOL. 28, 2008

REGULATION OF H4K16 ACETYLATION BY roX

4959

FIG. 5. Redundant function of roX1 boxes within roX1 RNA. (A) Gene structure of roX1 RNA showing a predicted stem-loop region (21) and roX box sequences (14). Dashed line with arrowheads at both ends, an essential domain (⬃600 bp) found by nested deletion (⬃300 bp) analysis of roX1 (21); vertical arrows, poly(A) sites mapped by 3⬘ RACE (E); horizontal arrows, primers used in panel E; tra-2, vector region (450 bp) used for polyadenylation; Con, conserved roX1 sequence from eight Drosophila species (14); WT, H83MeroX1 containing roX1 c20 cDNA in pCaSpeR-hsp83T3; 1⌬X2, 34-bp deletion removing the roX1-box2 sequence (21); mX3, point mutations within the roX1-box3 sequence; blue characters, mutated nucleotides. (B) Polytene chromosomes of male transgenic larvae of 1⌬X2, mX3, 1⌬X2-mX3, and ⌬SL transgenics immunostained with antibodies against MSL3 protein (green) and MOF protein (red) or H4K16ac (red). ⌬SL, 69-bp deletion removing the predicted stem-loop region shown in panel A (21). (C) H4K16ac ChIP analysis of an X-chromosomal target gene (CG3016) (8) in roX2 or roX1 transgenics. ChIP was performed on salivary gland chromatin, using anti-H4K16ac or no antibody, followed by real-time PCR analysis. Changes in H4K16ac (ratio of anti-H4K16ac to no antibody) in three independent experiments were averaged and are represented with standard deviations. The level of H4K16ac in the CG3016 gene (red) was normalized to a fourth chromosomal nontarget gene (CG1793; blue) as an internal control (onefold change) in each transgenic. (D) Male rescue frequency (male/female ratio) for roX1 transgenic lines. For WT, 1⌬X2, and ⌬SL transgenics, three independent lines were tested, and for mX3 and 1⌬X2-mX3 transgenics, four independent lines were tested. Averages of male viability (%) are represented, with standard deviations. (E) 3⬘ RACE analysis of total RNAs from male transgenic adults (roX⫺ mutant background), using the primers shown in panel A (no-RT control not shown). Arrows represent 3⬘ poly(A) sites (A) from the roX1 RNAs. Because of the 34-bp deletion in the roX1-box2 region, lane 4 shows a smaller product than that of the WT (lane 3). E, endogenous roX1 RNA from yw adult males. (F) Northern analysis using the same total RNAs used for panel E. The size of the roX1 RNA was estimated from an RNA ladder (Invitrogen).

Table 1). These results suggest that although the 3⬘ stem region of the roX1 stem-loop has a sequence similar to that of the roX box, this roX1 stem-loop might function differently from the roX2 stem-loop region. In contrast to the initial report (21), we measured high rescue of the roX1 ⌬SL mutant construct (53% average rescue [Fig. 5D] versus 22% highest rescue in the initial report), consistent with its ability to support histone acetylation on the X chromosome. There are three conserved roX box sequences in the 3⬘ end of roX1 RNA, at locations similar to those within roX2 (Fig. 5A) (14). However, previously, a deletion (34 nt) of the roX1box2 sequence from H83MeroX1 (1⌬X2) did not affect the function of roX1 RNA (Fig. 5A to D) (21). It remained a possibility that a malfunction induced by the roX1-box2 deletion of roX1 could be compensated by the presence of other roX box sequences in roX1, as in roX2 (Fig. 3 and 4). Therefore, we made transgenics including point mutations in the roX1-box3 sequence (blue characters in Fig. 5A) from H83MeroX1 (mX3) and a combination mutant (1⌬X2-mX3). First, we checked 3⬘-end processing of these roX1 mutant

RNAs because the roX box sequences are located in the 3⬘ end of roX1 RNA. In 3⬘ RACE analysis of endogenous roX1 RNA, we found two 3⬘-end processing sites (Fig. 5E, lane 2, arrows 1 and 2) located in the 3⬘ ends of the roX1-box2 and roX1-box3 regions (Fig. 5A, vertical arrows 1 and 2). WT roX1 RNA constructed from the c20 cDNA (H83MeroX1) also exhibited two 3⬘ ends in the same positions as those for the endogenous roX1 RNA (Fig. 5E, lane 3). However, the 1⌬X2 or mX3 RNA showed only one processing site, 3⬘ of the roX1-box3 or roX1box2 sequence, respectively (Fig. 5E, lanes 4 and 6). This indicates that the 1⌬X2 and mX3 RNAs preferentially retain an intact roX1 box sequence at their 3⬘ ends. Double mutant 1⌬X2-mX3 RNA failed to process its transcript until reading through ⬃100 nt of tra-2 sequence (Fig. 5A, vertical arrow 3, and E, lane 5, arrow 3). This alteration of 3⬘ processing was previously seen in roX1 deletion mutants, in which the additional tra-2 sequences were observed not to interfere with localization of the MSL complex on the X chromosome (21). The mX3 RNA still had normal function of roX1 RNA, including high rescue of male-specific lethality (Fig. 5D), MSL

4960

PARK ET AL.

MOL. CELL. BIOL.

FIG. 6. Restoration of H4K16ac on the X chromosome by a hybrid roX2-roX1 fusion RNA. (A) Sequence alignment between the stem-loop region of roX2 RNA and the roX1-box2 region of roX1 RNA. Bold red and black characters represent consensus sequences in Fig. 3A and 5A (14). Each sequence is numbered from the whole roX2 or roX1 gene (14). The sequence of H-X2 shows replacement of the roX2-box1 region of roX2 RNA (52 nt) with the roX1-box2 region of roX1 RNA (47 nt). Vertical arrows represent poly(A) sites mapped by 3⬘ RACE analysis using H-X2 RNA (lane 8 in panel B). (B) RT-PCR analysis using total RNAs purified from male transgenic adults, using the primers shown in Fig. 3A (no-RT control not shown). ⌬X2 of roX2 deletes the roX2-box2 region but keeps the roX2-box1 sequence in its 3⬘ end (Fig. 3A). Primer 3 in Fig. 3A was used as a 5⬘ primer in lanes 2 to 5. As a 3⬘ primer, the 2-X1 (roX2-box1) or 1-X2 (roX1-box2) primer was used (see the list of primers in the supplemental material). (C) Polytene chromosomes of male larvae with the H-X2 transgene in the roX⫺ mutant background, immunostained with antibodies against MSL3 protein (green) and MOF protein (red) or H4K16ac (red). (D) Male rescue frequencies (male/female ratio) for H-X2 transgenic lines (four independent lines tested). Averages of male viability (%) were graphed, with standard deviations. Rescue frequencies for ⌬X2 and ⌬X1-⌬X2 transgenics from Fig. 3B are shown again here for comparison. (E) Northern analysis using the same total RNAs used for Fig. 3E and panel B. The same amount (20 ␮g) of total RNA was loaded per lane. (F) Possible stem-loop structure between the 5⬘ stem of roX2 and the roX1-box2 of roX1 within the H-X2 RNA.

localization, and histone H4K16ac on the X chromosome (Fig. 5B and Table 1). However, the double mutant of two roX boxes (1⌬X2-mX3) revealed a defect in H4K16ac, despite apparently normal localization of the MSL complex to the X chromosome (Fig. 5B and Table 1). Using ChIP analysis, we specifically measured H4K16ac on an X-chromosomal target gene (CG3016) (8) in the roX2 and roX1 transgenics (Fig. 5C). Compared to the ⌬X2 mutant in roX2 (77-fold) and the 1⌬X2 mutant in roX1 (54-fold), the ⌬X1-⌬X2 mutant in roX2 (9fold) and the 1⌬X2-mX3 mutant in roX1 (11-fold) exhibited significantly reduced levels of H4K16ac (Fig. 5C). Consistent with the defective histone acetylation, the 1⌬X2-mX3 roX1 transgenic failed to rescue male-specific lethality (9%), in contrast to the single 1⌬X2 (64%) and mX3 (87%) mutants (Fig. 5D). In Northern analysis, the steady-state RNA level of the 1⌬X2-mX3 transcript was detectable (Fig. 5F, lane 4), although it was unstable compared to the 1⌬X2 and mX3 RNAs (Fig. 5F, lanes 2 and 3) and exhibited readthrough to a 3⬘ poly(A) site in flanking sequences. A hybrid roX2-roX1 fusion RNA can restore histone H4K16ac to the X chromosome. The similar locations and functions of multiple roX box sequences might explain the functional redundancy between the two roX RNAs. To inves-

tigate this, we made transgenic flies expressing a hybrid RNA (H-X2), fusing the 5⬘ stem of roX2 RNA and the roX1-box2 sequence of roX1 RNA (Fig. 6A). Because the ⌬X1-⌬X2 roX2 transgene was nonfunctional without the roX2-box1 region (Fig. 3), we asked whether the addition of the roX1-box2 sequence of roX1 RNA would recover function (Fig. 6A). Therefore, the H-X2 and ⌬X2 transcripts contain one roX box (from roX1 or roX2) in the 3⬘ end. Using oligo(dT)-primed cDNAs from transgenics, we confirmed that the ⌬X2 and H-X2 transcripts contained the roX2-box1 sequence of roX2 and the roX1-box2 sequence of roX1, respectively (Fig. 6B, lanes 2 to 5). We also verified that the H-X2 RNA had no defect in alternative splicing of roX2 RNA (Fig. 6B, lanes 9 and 10). However, we found two unexpected 3⬘ processing sites (Fig. 6B, lane 8) in the middle of the roX1-box2 sequence (Fig. 6A, arrows 6 and 7). This suggests that the H-X2 RNA must retain the roX1-box2 sequence by utilizing a downstream polyadenylation site. Interestingly, the H-X2 RNA was able to complement the defect in H4K16ac found in the ⌬X1-⌬X2 transgenic (Fig. 6C and Table 1). In addition, male rescue by the ⌬X1-⌬X2 transgenic (6%) was increased to 36% in the H-X2 strain (Fig. 6D). Among four H-X2 transgenic lines tested, one line showed

VOL. 28, 2008

REGULATION OF H4K16 ACETYLATION BY roX

male rescue of up to 48%. This indicates that the roX1-box2 sequence can replace the roX2-box1 sequence within roX2 RNA. However, the H-X2 transgenic could not fully rescue male-specific lethality. When the steady-state level of H-X2 RNA was checked by Northern analysis, we found that the H-X2 RNA was undetectable (Fig. 6E, lane 3), like the ⌬X1⌬X2 roX2 RNA (Fig. 3E, lane 3). Nevertheless, the H-X2 transgenic showed a higher rescue level (36%) (Fig. 6D) than that of the ⌬5-⌬X2 roX2 transgenic (20%) (Fig. 3B) that produced more stable RNA (Fig. 6E, lanes 1 and 3). These data suggest that formation of a secondary structure is important for roX box function. Figure 6F shows a possible interaction between the 5⬘ stem of roX2 and the roX1-box2 sequence of roX1 within the H-X2 RNA, which could account for its partial restoration of function. DISCUSSION We found that roX RNAs of distantly related Drosophila species had sufficient function in D. melanogaster to target the MSL complex and H4K16ac to the X chromosome (Fig. 1). This indicates that roX RNAs in diverse species contain common functional domains for dosage compensation of the X chromosome, in spite of a considerable lack of conservation in overall sequence. By mutating the conserved stem-loop region of roX2 RNA and “roX box” sequences in both roX RNAs, we found that these regions are essential for normal levels of enzymatic activity of the MSL complex (Fig. 3 to 5). However, other short, conserved sequences in roX2 were not essential upon deletion (Fig. 2; see Fig. S2 in the supplemental material), suggesting internal redundancy, as shown in other noncoding RNAs (roX1 and Xist) (21, 22). One possibility is that these conserved regions are functionally redundant with each other, as shown for the roX box sequences (Fig. 4 and 5). Another possibility is that overexpression of these constructs from the constitutive hsp83 promoter can overcome defects in these RNAs. Expression of WT roX2 RNA from the hsp83 promoter results in approximately threefold higher amounts of stable roX2 RNA than of the endogenously expressed RNA (18). However, the stem-loop deletion and mutant roX box constructs resulted in a defective phenotype even when they were expressed from the hsp83 promoter. Our results strongly suggest that the roX box sequences in roX2 RNA are functionally redundant in forming a stem-loop structure at the 3⬘ end of the RNA. Mutation of the roX2-box1 sequence (in mX1 RNA) resulted in more-than-normal accumulation of a longer RNA including the roX2-box2 sequence (Fig. 4B and C). We propose that the roX2-box2 sequence could form a stem-loop structure with the 5⬘ stem sequence when the roX2-box1 sequence is mutated (Fig. 4E). Alternatively, it is possible that the roX2-box2 sequence can make its own stem-loop structure within the 3⬘ extension (67 nt). There is an additional potential 5⬘ stem sequence upstream of the roX2-box2 sequence that shows evolutionary conservation (Fig. 2; see Fig. S1 in the supplemental material). In 3⬘ RACE analysis of endogenous and WT transgenic (H83MeroX1) roX1 RNAs, we found two 3⬘ processing sites, located 3⬘ of roX1-box2 and roX1-box3 (Fig. 5A and E). These were consistent with reported 3⬘ ends of roX1 cDNAs (U85980 and c3) (2, 12), with mapping on flybase (http://www.flybase

4961

.org/), and with 3⬘ ends determined by RNase protection assay (21). Similar to roX2, mature roX1 RNAs display the roX box at their 3⬘ ends. Our results strongly suggest that roX1 and roX2 RNAs utilize roX box sequences at the 3⬘ end for normal function of the MSL complex. The alternative 3⬘-end processing and the locations and functions of the two roX box sequences are very similar in roX1 and roX2 (Fig. 3 to 5). Most importantly, when it was expressed as a hybrid RNA (H-X2), the roX box region from roX1 was able to compensate for the loss of a roX box segment from roX2 (Fig. 6). Therefore, the presence of roX boxes in both roX1 and roX2 explains at least part of the functional redundancy between roX RNAs. At this point, we do not know if the roX1-box1 sequence in roX1 has the same function as the roX1-box2 and roX1-box3 sequences. When we carefully analyzed the alignment of roX1 genes from eight Drosophila species (14), we found an evolutionarily conserved sequence upstream (135 nt) of the roX1box1 sequence, which might function as a 5⬘ stem (data not shown). However, we currently cannot exclude the possibility that the roX1-box2 and roX1-box3 sequences in roX1 RNA function as primary sequences rather than in secondary structures. Our results for the first time uncouple X chromosome targeting of the MSL complex from histone acetylation activity. The stem-loop region in the 3⬘ end of roX2 RNA and the roX box sequences in the 3⬘ ends of roX1 and roX2 RNAs are critical for full histone acetylation by the MSL complex. It was reported that the histone acetyltransferase activity of the MOF protein is stimulated by the interaction of MOF with MSL1 and MSL3 in vitro (13). However, in vivo histone acetyltransferase activity of MOF is low in a roX⫺ mutant (11), suggesting that roX RNA is required for the optimum activity of MOF in vivo. MOF and MSL3 proteins have RNA binding activity in vitro and are dissociated from the MSL complex by RNase treatment in vivo (1). In future analyses, it will be very interesting to determine whether roX RNAs directly or indirectly stimulate the acetylation activity of the MSL complex. ACKNOWLEDGMENTS We thank H. Oh, Y. Kang, J. Sypula, and T. Chan for critical readings of the manuscript and for technical support. This work was supported by grants from the American Heart Association (0535548T) and the New Jersey State Commission on Cancer Research (08-1082-CCR-E0) to Y. Park and by NIH grant GM45744 to M. Kuroda. REFERENCES 1. Akhtar, A., D. Zink, and P. B. Becker. 2000. Chromodomains are proteinRNA interaction modules. Nature 407:405–409. 2. Amrein, H., and R. Axel. 1997. Genes expressed in neurons of adult male Drosophila. Cell 88:459–469. 3. Franke, A., and B. S. Baker. 1999. The roX1 and roX2 RNAs are essential components of the compensasome, which mediates dosage compensation in Drosophila. Mol. Cell 4:117–122. 4. Hamada, F. N., P. J. Park, P. R. Gordadze, and M. I. Kuroda. 2005. Global regulation of X chromosomal genes by the MSL complex in Drosophila melanogaster. Genes Dev. 19:2289–2294. 5. Hilfiker, A., D. Hilfiker-Kleiner, A. Pannuti, and J. C. Lucchesi. 1997. mof, a putative acetyl transferase gene related to the Tip60 and MOZ human genes and to the SAS genes of yeast, is required for dosage compensation in Drosophila. EMBO J. 16:2054–2060. 6. Kageyama, Y., G. Mengus, G. Gilfillan, H. G. Kennedy, C. Stuckenholz, R. L. Kelley, P. B. Becker, and M. I. Kuroda. 2001. Association and spreading of the Drosophila dosage compensation complex from a discrete roX1 chromatin entry site. EMBO J. 20:2236–2245. 7. Kelley, R. L., V. H. Meller, P. R. Gordadze, G. Roman, R. L. Davis, and M. I.

4962

8. 9.

10.

11.

12.

13.

14.

15. 16.

PARK ET AL.

Kuroda. 1999. Epigenetic spreading of the Drosophila dosage compensation complex from roX RNA genes into flanking chromatin. Cell 98:513–522. Kind, J., and A. Akhtar. 2007. Cotranscriptional recruitment of the dosage compensation complex to X-linked target genes. Genes Dev. 21:2030–2040. Lanz, R. B., N. J. McKenna, S. A. Onate, U. Albrecht, J. Wong, S. Y. Tsai, M. J. Tsai, and B. W. O’Malley. 1999. A steroid receptor coactivator, SRA, functions as an RNA and is present in an SRC-1 complex. Cell 97:17–27. Meller, V. H., P. R. Gordadze, Y. Park, X. Chu, C. Stuckenholz, R. L. Kelley, and M. I. Kuroda. 2000. Ordered assembly of roX RNAs into MSL complexes on the dosage compensated X chromosome in Drosophila. Curr. Biol. 10:136–143. Meller, V. H., and B. P. Rattner. 2002. The roX RNAs encode redundant male-specific lethal transcripts required for targeting of the MSL complex. EMBO J. 21:1084–1091. Meller, V. H., K. H. Wu, G. Roman, M. I. Kuroda, and R. L. Davis. 1997. roX1 RNA paints the X chromosome of male Drosophila and is regulated by the dosage compensation system. Cell 88:445–457. Morales, V., T. Straub, M. F. Neumann, G. Mengus, A. Akhtar, and P. B. Becker. 2004. Functional integration of the histone acetyltransferase MOF into the dosage compensation complex. EMBO J. 23:2258–2268. Park, S.-W., Y. I. Kang, J. G. Sypula, J. Choi, H. Oh, and Y. Park. 2007. An evolutionarily conserved domain of roX2 is sufficient for induction of H4Lys16 acetylation on the Drosophila X chromosome. Genetics 177:1429– 1437. Park, Y., and M. I. Kuroda. 2001. Epigenetic aspects of X-chromosome dosage compensation. Science 293:1083–1085. Park, Y., R. L. Kelley, H. Oh, M. I. Kuroda, and V. H. Meller. 2002. Extent

MOL. CELL. BIOL.

17.

18.

19.

20.

21.

22.

23.

of chromatin spreading determined by roX RNA recruitment of MSL proteins. Science 298:1620–1623. Park, Y., G. Megnus, X. Bai, Y. Kageyama, V. H. Meller, P. B. Becker, and M. I. Kuroda. 2003. Sequence-specific targeting of Drosophila roX genes by the MSL dosage compensation complex. Mol. Cell 11:977–986. Park, Y., H. Oh, V. H. Meller, and M. I. Kuroda. 2005. Variable splicing of non-coding roX2 RNAs influences targeting of MSL dosage compensation complexes in Drosophila. RNA Biol. 2:157–164. Smith, E. R., A. Pannuti, W. Gu, A. Steurnagel, R. G. Cook, C. D. Allis, and J. C. Lucchesi. 2000. The Drosophila MSL complex acetylates histone H4 at lysine 16, a chromatin modification linked to dosage compensation. Mol. Cell. Biol. 20:312–318. Straub, T., G. D. Gilfillan, V. K. Maier, and P. B. Becker. 2005. The Drosophila MSL complex activates the transcription of target genes. Genes Dev. 19:2284–2288. Stuckenholz, C., V. H. Meller, and M. I. Kuroda. 2003. Functional redundancy within roX1, a non-coding RNA involved in dosage compensation in Drosophila melanogaster. Genetics 164:1003–1014. Wutz, A., T. P. Rasmussen, and R. Jaenisch. 2002. Chromosomal silencing and localization are mediated by different domains of Xist RNA. Nat. Genet. 30:167–174. Yokoyama, R., A. Pannuti, H. Ling, E. R. Smith, and J. C. Lucchesi. 2007. A plasmid model system shows that Drosophila dosage compensation depends on the global acetylation of histone H4 at lysine 16 and is not affected by depletion of common transcription elongation chromatin marks. Mol. Cell. Biol. 27:7865–7870.