Recruitment of the male-specific lethal (MSL) dosage compensation complex to an autosomally integrated roX chromatin entry site correlates with an increased expression of an adjacent reporter gene in male Drosophila.

Drosophila dosage compensate (equalize X-linked gene products) by doubling the transcription of most X-linked genes in males. The MSL (male-specific lethal) ribonucleoprotein complex consisting of at least five proteins and two non-coding RNAs (roX1 and roX2) is essential for this transcription response. Recently it has been shown that the X-linked roX1 and roX2 genes each contain at least one chromatin entry site for the MSL complex. In this study we show that insertion of either roX1 or roX2 DNA sequences, upstream of an insulated lacZ reporter gene controlled with the constitutive armadillo promoter (arm-lacZ), results in a significant elevation of expression of lacZ in males. However, full compensation, that is a precise doubling of lacZ expression in males relative to females, was only observed in some lines carrying autosomal insertions of either roX1-arm-lacZ or roX2-arm-lacZ transgenes. Furthermore, we found that a 419-base pair fragment of roX1 that contains an MSL binding site was sufficient to cause a modest elevation of expression of lacZ in males, but this response was significantly less than obtained with a full-length roX1 cDNA. This is the first direct demonstration that insertion of an MSL chromatin entry site on an autosome results in elevated expression in males of genes near the entry site.

In the vinegar fly Drosophila melanogaster, males have one X chromosome and females have two. Males dosage compensate by doubling the transcription of most X-linked genes (1). The MSL 1 complex is required for this male-specific hypertranscription (2,3). The complex is comprised of at least five proteins, MSL1, MSL2, MSL3, MLE, and MOF and two non-coding RNAs, roX1 and roX2. All components of the complex co-localize to hundreds of sites along the male X chromosome (4 -6). Two of the proteins have enzymatic activity; MLE is an RNA helicase (7) and MOF is a histone acetylase (8). In addition, a kinase, JIL-1, preferentially associates with the male X chromosome (9), but it is not known if this protein is essential for dosage compensation. The complex only assembles in males as one of the components, MSL2, is not present in females (10 -12).
The complex is initially targeted to the male X chromosome through binding to 30 -40 "high affinity" or "chromatin entry" sites (13). The complex is then thought to spread from these sites to other sites on the X chromosome (2). Two of these sites are the X-linked roX1 and roX2 genes (14). That is, the same genes that encode the RNA components of the complex also appear to contain DNA sequences that are recognized by the complex.
It has recently been shown that a 217-bp DNA fragment of roX1 is sufficient to produce an ectopic chromatin entry site when inserted on an autosome (15).
Previously, we developed an insulated reporter gene system to search for cis-acting X-linked DNA sequences that are required for dosage compensation (16). The system consists of the constitutive armadillo promoter driving expression of the lacZ reporter gene and flanked by SCS and SCSЈ insulator elements. Seven X-linked DNA fragments totaling 63 kb were tested with the system, but none were found to contain DNA sequences that caused elevated expression of the reporter in males. Here we report that insertion of either roX1 or roX2 DNA sequences upstream of the armadillo promoter results in elevated expression of the lacZ reporter gene in male Drosophila.

EXPERIMENTAL PROCEDURES
Recombinant DNA-All recombinant DNA manipulations were carried out using standard procedures (17) unless otherwise specified. The insulated reporter P transformation vector pHF11 contains a unique EcoRI site between the SCSЈ element and the armadillo promoter (16). pHF11 also contains a unique NotI site immediately upstream of SCSЈ. A derivative of pHF11, pRH07, which contains a unique NotI site between SCSЈ and the armadillo promoter, was constructed by first deleting the NotI site of pHF11 then inserting a linker that contains a NotI site into the EcoRI site. roX DNA fragments were inserted into either the EcoRI or NotI sites of pHF11 or pRH07. The fragments were a 4.9-kb EcoRI genomic roX1 (14), a 3.7-kb NotI fragment containing roX1 cDNA (18), a 1.1-kb NotI/PspOMI fragment containing roX2 cDNA (18), a 2.2-kb NotI/PspOMI fragment containing the hsp83 promoter (19) and roX2 cDNA, a 4.0-kb NotI genomic fragment containing the roX2 gene (6), and 419-and 246-bp fragments of roX1 generated by polymerase chain reaction. The primers used to obtain the 419-bp fragment called roX1 BS were 5Ј-GTCGAATTCGAACGAAAGAGACAAA-TGAACCC-3Ј and 5Ј-GTCGAATTCTTATGGCGATTCTACGCTCCTG-3Ј. The primers used to generate the 246-bp fragment called roX1 SIM were 5Ј-GTCGAATTCGAAAAACACATTTACTAACAAATAA-3Ј and 5Ј-GTCGAATTCCCCAAAGAAATCCACATAACAT-3Ј. The reaction mixture (50 l) contained the following: 10 pmol of each primer, 0.1 ng of plasmid DNA containing the 4.9-kb roX1 genomic DNA fragment as template, 0.2 mM dNTP, 1ϫ ELONGASE buffer, and 1 l of ELONGASE enzyme mix (Life Technologies). The reactions were subject to the following thermal profile: 5 cycles of 94°C for 30 s, 40°C for 30 s, and 68°C for 60 s, followed by 30 cycles of 94°C for 30 s, 62°C for 30 s, and 68°C for 60 s. Following amplification the fragments were digested with EcoRI then ligated with EcoRI-cut pHF11. Based on the numbering of the roX1 gene (GenBank accession number U97114), roX1 BS is from nucleotides 1152 to 1570, and roX1 SIM is from 3556 to 3801. All roX fragments were inserted in the same 5Ј33Ј direction as lacZ except roX1 4.9 genomic, roX2 4.0 genomic, and roX2 cDNA, which are in the 3Ј-5Ј orientation. The construct containing both roX1 and roX2 cDNAs was made by insertion of the 3.7-kb roX1 cDNA into the NotI site of pRH09 (pRH07 containing the roX2 cDNA).
Drosophila Stocks-Flies were raised on standard cornmeal-yeastsugar-agar medium with methyl paraben. Crosses were performed at 25°C unless otherwise indicated. All stocks not specifically mentioned are described in Lindsley and Zimm (20).
Germ-line Transformation-Manually dechorionated y w embryos were injected with a mixture of P transformation vector and helper plasmid DNA using standard procedures (21). Microinjections were performed using an Eppendorf transjector and Femtotips. Single G 0 adults were mated with y w, and offspring of these crosses were examined for non-white eye color. Single G 1 transformants were backcrossed with y w. Homozygotes were selected in subsequent generations on the basis of a darker eye color. Linkage of P[w ϩ ] was determined by following w ϩ segregation in the appropriate crosses. For some lines the sequence flanking the integrated transgene was determined by inverse polymerase chain reaction according to the method of J. Rehm (www-.fruit fly.org/methods). The chromosomal site of integration could then be inferred by comparison of the flanking sequence with the sequence of the Drosophila genome.
␤-Galactosidase Assays-␤-Galactosidase assays were performed as described previously (16). For each transgenic line ␤-galactosidase activity was standardized by both wet weight measurement and total protein microassays (Bio-Rad). Initial statistical analysis (standard error, 95% confidence limits) was determined by using Microsoft Excel 98. To make comparisons of lines carrying different constructs we performed an analysis of variance, using the line-means as data. The means were weighted by using 1/S.E. 2 ϭ Number of observations contributing to Mean/Variance of observations contributing to mean to allow for the fact that some means are more precise estimates because they are based on more observations, or on observations which are less variable, than others. The analysis was performed using SAS Proc Mixed. In most cases only two Treatments (e.g. two lines of the same dose) are being compared, and so the significance level of the Treatment effect is the significance level of the difference. When comparing the four combinations of Dose and line, Tukey's honestly significant difference was used to compare the means for the different combinations. When analyzing ratios, there are good theoretical reasons for analyzing the log of the ratio, rather than the ratio itself; the mean of the ratio of A:B is not the ratio of the mean of A:mean of B, whereas the mean of log(A:B) is approximately log(mean of A:mean of B). However, in our analyses both the raw ratios and the log(ratios) were analyzed and produced similar results.

roX1 and roX2 DNA Sequences Are Sufficient to Cause Elevated Expression of Autosomally Integrated arm-lacZ Reporter
in Males-We have previously developed an insulated reporter gene system to search for X chromosome-linked DNA sequences that are required for dosage compensation (16). The gene system is shown schematically in Fig. 1. The constitutive armadillo promoter controls expression of the lacZ reporter gene, which encodes ␤-galactosidase. The arm-lacZ construct is bracketed by the SCS and SCSЈ insulator elements to protect against possible repressive effects of an autosomal environment. The insulated reporter is expressed equally in males and females when inserted on an autosome and fully dosage compensated when on the X chromosome (16). X-linked DNA fragments are generally inserted between the SCSЈ element and the arm promoter (position B in Fig. 1). If the fragment contains a DNA sequence necessary for dosage compensation, then transgenic males carrying an autosomal insert of the construct should produce twice the ␤-galactosidase activity of females.
We made a series of constructs where either roX1 or roX2 genomic or cDNA fragments were inserted into site B of the insulated reporter vector. Because both roX1 and roX2 contain binding sites for the MSL complex, we anticipated that DNA sequences from either gene should cause elevated expression of arm-lacZ in males if recruitment of the MSLs is sufficient to cause hypertranscription. We assayed three lines carrying autosomal inserts of a 4.9-kb roX1 genomic-arm-lacZ construct (Table I). This construct would be expected to express roX1 RNA. We assayed males and females carrying either one or two copies of the transgene. For all lines, the male to female (M/F) ratio of ␤-galactosidase activity was significantly greater than one. However, the ratios were also significantly less than the 2-fold increase in activity in males expected if the arm-lacZ reporter was fully compensated. We assayed four lines carrying autosomal inserts of a 3.7-kb roX1 cDNA-arm-lacZ construct only two of which could be made homozygous (Table I). This construct would not be expected to make roX1 RNA unless by chance the transgene has become inserted adjacent to a promoter. As for the roX1 genomic construct, all lines showed significantly elevated expression of arm-lacZ in males. Furthermore, both lines 1 and 2 showed full male-specific hyperactivation, that is an M/F ratio of close to 2, when homozygous for the transgene. Inspection of the data suggested that lines, which were homozygous for the roX1 cDNA construct, gave higher male/female ratios than homozygous lines carrying the roX1 genomic construct. To determine if this difference was statistically significant, we performed an analysis of variance (ANOVA) using weighted means of the lines as data. We found that the homozygous lines carrying the roX1 cDNA did give a significantly higher M/F ratio than the homozygous roX1 genomic lines (protein ratio, p ϭ 0.0009; log(protein ratio), p ϭ 0.0006; weight ratio, p ϭ 0.016; log(weight ratio), p ϭ 0.024). Further analysis showed that the one dose roX1 cDNA lines also gave significantly higher M/F ratios than the two dose roX1 genomic lines (p Ͻ 0.05 for both protein and weight ratios).
Lines carrying autosomal insertions of either a 4.0-kb roX2 genomic fragment-arm-lacZ or roX2 1.1-kb cDNA-arm-lacZ construct were assayed for ␤-galactosidase activity (Table II). All lines gave M/F ratios of ␤-galactosidase activity that were significantly greater than one. However, there was significant variation between the roX2 1.1-kb cDNA-arm-lacZ lines. Line 1

FIG. 1. Schematic representation of the insulated reporter gene construct.
Expression of lacZ is controlled by the constitutive promoter from the armadillo gene and protected from position effects by the SCS and SCSЈ insulator elements. For most constructs, roX sequences were inserted at position B, between the SCSЈ insulator and the armadillo promoter. In one construct, roX2 cDNA linked with the hsp83 promoter was inserted at position A, upstream of the SCSЈ element.
showed only a small increase in expression in males whereas both one dose and two dose line 3 flies had M/F ratios close to 2, indicating near full compensation.
Several arm-lacZ lines carrying either roX1 or roX2 DNA sequences do not appear to be fully compensated, that is the M/F ratios are less than two. Although there are several possible explanations for these results (see "Discussion"), one possibility we considered was that a single roX gene integrated in an autosomal environment may not recruit sufficient copies of the MSL complex to achieve full hyperactivation of the reporter. Thus we made a construct that carried both roX1 and roX2 cDNAs (in that order) upstream of arm-lacZ. Two lines carrying autosomal insertions of this construct were assayed for ␤-galactosidase activity (Table III). Both lines showed a significant elevation of arm-lacZ expression in males, but the M/F ratios were also less than 2. Thus, insertion of both roX1 and roX2 did not lead to a greater male-specific hyperactivation of arm-lacZ than either alone. In addition to the autosomal lines, we obtained one line (number 3) with the construct integrated onto the X chromosome. In this line lacZ was fully compensated with one dose males having twice the ␤-galactosidase activity of one dose females (Table III). This was the expected result, because we had previously found that the insulated arm-lacZ reporter was fully compensated when in-serted onto the X chromosome (15).
A Fragment of roX1 Is Sufficient to Cause a Small Increase in arm-lacZ Expression in Males-Two regions of roX1 have been indicated as potentially being important for function. A 217-bp fragment near the 5Ј-end has been shown to contain a binding site for the MSL complex (15) (Fig. 2). At the 3Ј-end of roX1 there is a region of 30 bp that shows high similarity to a sequence in roX2 (5). We tested whether either of these regions are sufficient to cause elevated expression of lacZ in males. A 419-bp fragment containing the MSL binding site (roX1 BS) (Fig. 2) and a 246-bp fragment that contains the region of similarity (roX1 SIM) (Fig. 2) were inserted into site B of the insulated arm-lacZ reporter (Fig. 1). Three lines carrying autosomal insertions of each construct were assayed for ␤-galactosidase activity. We found that all lines carrying the roX1 BS-arm-lacZ construct had a small but generally significant elevation of arm-lacZ expression in males (Table IV). All lines had M/F ratios that were significantly greater than one with the exception of line 2 standardized by protein where the 95% confidence interval (1.04 -1.28) is only slightly higher than one. In contrast all lines carrying the roX1 SIM-arm-lacZ construct had M/F ratios that were not significantly higher than 1.0. Thus the fragment that contains the MSL binding site but not the fragment that has the region of similarity with roX2 con-  tains a sequence that is sufficient for at least partial compensation of the reporter. However, it appears that the level of elevation of the reporter in males carrying the roX1 BS construct is less than was obtained with the full-length cDNA (Table I). To determine if this difference was significant we again performed an ANOVA on weighted line means. We found that the lines carrying the full-length roX1 cDNA had significantly higher M/F ratios of ␤-galactosidase activity than lines carrying the roX1 BS construct (protein ratio, p ϭ 0.0002; weight ratio, p ϭ 0.0018). We considered several explanations for the difference in M/F ratios of the roX1 BS and roX1 3.7 cDNA lines. One possibility is that the MSL binding site within the roX1 BS fragment is not as effective in recruiting complex as full-length cDNA. If this was the case it may mean that the complex would only be bound to the roX1 BS site in a subpopulation of cells in a tissue whereas in most or all cells the complex would bind to the site containing the full-length cDNA. To examine this possibility we prepared polytene chromosomes from male larvae from several lines (roX1 4.9 genomic line 2, roX1 3.7 cDNA line 2, roX2 1.1 cDNA line 1, roX1 SIM line 2, roX1 BS lines 1 and 3) and examined them for MSL binding using standard immunostaining techniques. As anticipated we detected MSL binding to a single autosomal site for all of the lines carrying a roXarm-lacZ construct with the exception of roX1 SIM (Fig. 3). We did not observe spreading of the MSL complex from the transgene integration site in any of the lines, but this only occurs frequently in some lines carrying autosomal roX genes (14). We counted the number of nuclei from salivary glands that showed MSL binding to an autosomal site for one roX1 3.7 cDNA and two roX1 BS lines. The proportion of nuclei that showed binding to an ectopic site was similar for all three lines (roX1 BS line 1 69/75 (92%), roX1 BS line 3 179/246 (73%), roX1 3.7 cDNA line 2 256/316 (81%)). Thus at least in the salivary gland there is no difference in the proportion of cells that show ectopic binding of the MSL complex in the roX1 BS and roX1 3.7 cDNA lines. Similar observations have been made by Kageyama et al. (15).
SCSЈ Insulator Element Does Not Block Hyperactivation of the Reporter in Males-We have previously found that the insulated arm-lacZ reporter is fully compensated when inserted on the male X chromosome (16). Thus the insulator elements did not appear to be able to block the transcription elevation due to the MSL complex. Because the roX genes contain MSL binding sites, we decided to test this more directly by inserting roX sequences either upstream or downstream of the SCSЈ insulator element (Fig. 1). We used a hsp83.roX2 1.1 cDNA construct, which should express a roX2 RNA from the constitutive hsp83 promoter (19). The construct was inserted either upstream of the SCSЈ element (position A in Fig. 1) or between the SCSЈ element and the arm promoter (position B in Fig. 1). We found that all lines gave a significant elevation of expression of the arm-lacZ reporter in males (Table V). Furthermore, there was no significant difference between the lines that had the hsp83.roX2 construct inserted upstream of SCSЈ (i.e. position A) compared with those lines where the construct was inserted downstream of SCSЈ (position B). Thus the SCSЈ insulator element appears to be unable to block hyperactivation of the arm-lacZ due to insertion of an MSL chromatin entry site. DISCUSSION There are several lines of evidence that have shown that the MSL complex binds to hundreds of sites along the male X chromosome and causes a 2-fold increase in expression of most X-linked genes in male Drosophila (1,2). The most direct evidence for the latter is that males that are homozygous for mutant alleles of msl1, msl2, or mle have significantly reduced levels of X-linked but not autosomal enzymes and mle males have a lower overall rate of X-chromosome transcription (22). In this study we have shown that binding of the MSL complex to either roX1 or roX2 DNA sequences integrated at autosomal sites correlates with an elevated expression of an adjacent lacZ reporter gene. Indeed, in some of the lines that had either roX1 or roX2 cDNA inserted upstream of lacZ we observed full compensation, that is a doubling of expression in males compared with females carrying the same number of copies of the construct. However, in all of the lines carrying roX1 genomic fragments, a line with a roX2 genomic fragment, and some of the lines with roX cDNA upstream of lacZ we observed partial compensation. That is the male/female ratios were significantly greater than one but less than the 2-fold expected if recruitment of the MSL complex leads to a precise doubling of transcription as occurs on the X chromosome. We considered that this partial compensation might be because the MSL complex is only binding to the autosomal roX sequence in a fraction of the cells in a tissue. However, we found that the proportion of nuclei that showed binding to the roX1 BS construct, which shows partial compensation, was the same as compared with a line that showed full compensation (3.7 cDNA line 2). Thus it appears that partial compensation cannot be explained by variability in MSL binding between cells within a tissue. However, we have only analyzed cells from one tissue, third instar larvae salivary glands. We also considered the possibility that one roX sequence might not be sufficient in some autosomal locations to recruit enough MSL complexes to achieve full compensation. If so, a construct that had both roX1 and roX2 cDNAs inserted upstream of lacZ may be more effective at recruiting the complex and would thus show full compensation in most lines. However, the lines tested both showed partial compensation. Thus there must be some other explanation for why some lines show only partial compensation. We think the most likely ex-  3. MSL complex binds to autosomal roX1 transgenes. MSL3 localization in male nuclei determined by immunostaining with anti-MSL3 antibodies (green) and DAPI (blue), which binds to all chromosomes. A, roX1 3.7 cDNA line 1 (chromosome 3, 85B). B, roX1 BS line 3 (chromosome 3, 79C). C, roX1 SIM line 2 (chromosome 3, 93B). In all nuclei strong binding is seen to many sites on the male X chromosome. In A and B, MSL3 binding is also seen at one autosomal site (arrowhead). planation is that the local chromatin environment at the autosomal integration site influences the level of hyperactivation of the lacZ reporter by the MSL complex. It's known that some autosomal sites are much more permissive to spreading of the complex than others, indicating that the local autosomal chromatin environment can affect at least one function of the MSL complex.
We found that lines that were homozygous for the roX1 4.9 genomic-arm-lacZ construct gave significantly lower male/female ratios of ␤-galactosidase activity than lines that were either heterozygous or homozygous for the roX1 3.7 cDNA-arm-lacZ construct. Because both constructs contain a binding site for the MSL complex, it's not obvious why the roX1 genomic construct should give lower hyperactivation of lacZ in males. One possibility is that this is simply because by chance in all the roX1 genomic lines the transgene integrated into a negative chromosomal environment that was not permissive to full hyperactivation of lacZ in males. However, this would seem unlikely, because all three lines gave similar male/female ratios of ␤-galactosidase. The genomic construct contains additional DNA sequences from the roX1 gene region compared with the cDNA construct. It's possible that these additional sequences may somehow inhibit hyperactivation of lacZ by the MSL complex. The genomic construct would be expected to produce roX1 RNA, whereas the promoter-less cDNA construct would not. The function of the roX1 RNA in the complex is not known. If the RNA has an inhibitory role, then a localized excess of synthesis of roX1 RNA might result in assembly of an MSL complex that is less effective at elevating expression of the adjacent reporter gene. It has been suggested that in vivo there must be some mechanism for dampening the transcription elevation due to the MOF histone acetylase, because in vitro recombinant MOF is able to increase expression from a nucleosomal template far more than 2-fold (23). Clearly, further experiments with additional constructs are required to determine the biological significance (if any) of the difference in lacZ hyperactivation between roX1 genomic and cDNA constructs.
The level of elevation of ␤-galactosidase activity in males carrying the 419-bp fragment of roX1 that contains the MSL binding site was significantly less than obtained with 3.7-kb roX1 cDNA. It's possible that by chance all three roX1 BS lines are inserted into negative chromatin environments that inhibit the MSL complex. We think this is unlikely, because all three lines gave very similar male/female ratios of ␤-galactosidase activity. Rather it is more likely that binding of the MSL complex to the site in roX1 BS is not sufficient to achieve full compensation. This suggests that other sequences in roX1 3.7 cDNA in addition to the MSL binding site in roX1 BS are required for full roX1 function.
In Drosophila, the SCS and SCSЈ insulator elements are able to protect a gene from position effects and block a transcription enhancer from acting on a promoter (24,25). It has been previously shown that the SCS and SCSЈ elements do not block genes from being dosage compensated when inserted onto the X chromosome (16,24). This suggests that these insulators cannot prevent hypertranscription in males due to the MSL complex. In this study we have tested this directly by placing an SCSЈ insulator between a roX2 sequence, which contains an MSL binding site, and the lacZ reporter. The SCSЈ insulator was unable to block elevation of expression of the reporter in males. Because this insulator can block transcription enhancers, this suggests that the mechanism by which the MSL complex affects transcription may be different from how an enhancer acts on a promoter.