Abnormal Dosage Compensation of Reporter Genes Driven by the Drosophila Glass Multiple Reporter (GMR) Enhancer-Promoter

In Drosophila melanogaster the male specific lethal (MSL) complex is required for upregulation of expression of most X-linked genes in males, thereby achieving X chromosome dosage compensation. The MSL complex is highly enriched across most active X-linked genes with a bias towards the 3′ end. Previous studies have shown that gene transcription facilitates MSL complex binding but the type of promoter did not appear to be important. We have made the surprising observation that genes driven by the glass multiple reporter (GMR) enhancer-promoter are not dosage compensated at X-linked sites. The GMR promoter is active in all cells in, and posterior to, the morphogenetic furrow of the developing eye disc. Using phiC31 integrase-mediated targeted integration, we measured expression of lacZ reporter genes driven by either the GMR or armadillo (arm) promoters at each of three X-linked sites. At all sites, the arm-lacZ reporter gene was dosage compensated but GMR-lacZ was not. We have investigated why GMR-driven genes are not dosage compensated. Earlier or constitutive expression of GMR-lacZ did not affect the level of compensation. Neither did proximity to a strong MSL binding site. However, replacement of the hsp70 minimal promoter with a minimal promoter from the X-linked 6-Phosphogluconate dehydrogenase gene did restore partial dosage compensation. Similarly, insertion of binding sites for the GAGA and DREF factors upstream of the GMR promoter led to significantly higher lacZ expression in males than females. GAGA and DREF have been implicated to play a role in dosage compensation. We conclude that the gene promoter can affect MSL complex-mediated upregulation and dosage compensation. Further, it appears that the nature of the basal promoter and the presence of binding sites for specific factors influence the ability of a gene promoter to respond to the MSL complex.


Introduction
The Drosophila eye is well suited to genetic investigations, as it is a dispensable tissue with mutant phenotypes that are relatively easy to identify [1]. Glass is a Zinc-finger transcription factor involved in photoreceptor development, present in the eyes, ocelli, small areas of the brain, and the embryonic Bolwig organ [2,3]. The glass multiple reporter (GMR) is a pentamer of a 27 bp Glass response element from the Rh1 (ninaE) promoter, upstream of a minimal promoter from the hsp70 gene [4]. The GMR enhancerpromoter drives expression in all cells in, and posterior to, the morphogenetic furrow in the developing eye imaginal disc [5,6]. Expression of the pro-apoptotic hid gene from GMR kills eye cells proportional to the GMR-hid expression level; a smaller eye with higher levels [7]. This reporter system facilitated a mutational screen for apoptosis-related factors, by identifying modifiers of GMR-hid-derived eye size [8]. We sought to use the variable eye size of GMR-hid to screen for modifiers of Drosophila dosage compensation, but observed that GMR-hid did not report on dosage compensation. We have thus investigated why GMR-hid failed to respond, and report that the minimal hsp70 promoter appears refractory to the compensation machinery.
Drosophila males (XY) lack a dose of X-linked genes in comparison to (XX) females, but compensate for this by doubling transcription from the single X chromosome [9,10,11]. Dosage compensation requires the action of the RNA-containing Male-Specific Lethal (MSL) complex, which cannot form in females as translation of msl2 RNA is repressed by the female-specific SXL protein [12]. The MSL complex specifically binds to many sites on the male X chromosome and modifies the chromatin, most notably by acetylating histone H4 at lysine16 (H4K16) [13]. The enrichment for H4K16ac across active genes could lead to a less compact chromatin structure as incorporation of H4K16ac into nucleosomal arrays abolished a salt-dependent compaction into 30 nm-like fibres in an in vitro assay [14]. However, it has been reported that H4K16ac is more strongly associated with DNA replication timing than transcription in Drosophila cells [15]. Thus the exact mechanism of transcription enhancement by the MSL complex is unknown. It has been suggested that the MSL complex may enhance transcription elongation as the MSL complex is enriched on active genes with a bias towards the 39 end [16,17,18,19]. Indeed, using a global run-on sequencing (GROseq) approach, Kuroda and colleagues concluded that the MSL complex enhances transcription by facilitating the progression of RNA polymerase II along X-linked gene [20].
The dosage compensation machinery can also affect genes that are not endogenous to the X chromosome. Translocations of genomic fragments between the X chromosome and autosomes exhibit ''spreading'' of the MSL complex from the X-linked sequences into the adjacent autosomal regions [21]. This ability of the MSL complex to bind autosomal genes is likely the reason why autosomal genes translocated to the X chromosome usually gain increased transcription in males [22,23,24]. Short sequences with strong affinity for MSL complex can recruit the complex to autosomes, and increase the transcription of neighbouring genes [25,26,27,28]. These MSL-binding High Affinity Sites (HAS) are thought to serve as primary attractors of the MSL complex to the X chromosome, from which it can then ''spread'' to affect the transcription of nearby active genes [29]. Indeed, chromosomewide studies of MSL binding find that most MSL-bound genes are active [17]. Thus, we expected the (active) GMR-hid to respond to dosage compensation, but found that it was unable to fully do so. We argue that an active gene is not sufficient for recognition by the dosage compensation machinery, even if near a HAS, and specifically that the hsp70 minimal promoter appears to hinder dosage compensation.
We also report two further observations of oddities in expression from GMR that will be of interest to those working with this enhancer-promoter. Activity of beta-galactosidase produced by a GMR-lacZ transgene was not linear with respect to gene dose, at least in the environment of the female X chromosome. The activity in females with two copies of the transgene was more than twice the activity in single-copy females, suggesting some synergy in homozygosity. Further, we detected some repression of expression from GMR in the larval imaginal tissue, consistent with that first reported for an earlier Glass-based eye reporter that contained a longer glass response element [30]. Both these observations have implications for dosage compensation, but in general they also call for careful use of the GMR enhancerpromoter.

GMR-hid is a poor reporter of dosage compensation
Our understanding of the mechanism of X chromosome dosage compensation may need to better appreciate the balance between the MSL complex and the wider network of chromatin regulators. Identification of all factors that play a role in dosage compensation is not trivial. The five MSL proteins (MSL1, MSL2, MSL3, MLE, and MOF) were identified based on mutant phenotypes of male-specific lethality [31,32]. Loss of function mutations in genes for more generally acting epigenetic modifiers would be expected to be homozygous lethal for both sexes. However, some factors have been implicated to have a role in dosage compensation based on hypomorphic phenotypes of reduced male viability or altered male X chromosome morphology [10]. Such factors include the DNA supercoiling factor [33], JIL-1 histone kinase [34] and the heterochromatin-associated protein SU(VAR)3-7 [35]. The initial aim of this study was to search for novel dosage compensation factors, not by screening for male-lethality, but by using the reduced eye size of GMR-hid to screen for mutations that affected the end product of dosage compensation: the equalization of phenotype.
We followed two strategies to adapt the GMR-hid system to report on Drosophila dosage compensation, with the aim of then conducting a mutagenesis screen for modifiers of dosage compensation. To obtain a sex-specific difference in eye size, we created P-element transgenic lines carrying GMR-hid, and screened for those genetically linked to the X chromosome ( Figure 1A). Transgenic insertions on the X chromosome usually acquire a degree of compensation [22,23,24], meaning that males with one transgene copy express more than single-copy females, and perhaps as much as two-copy females. In all cases, flies homozygous for the GMR-hid transgene had ablated eyes ( Figure 1A) representing high levels of transgene expression, but little information as to what level of expression. All lines homozygous for any GMR-hid transgene were weak, presumably due to leaky expression of hid from the basal hsp70 TATA, or expression in other neuronal tissues. Flies heterozygous for all autosomal GMR-hid insertions had similar eye sizes in each sex ( Figure 1B). Surprisingly, lines with X-linked insertions of GMR-hid also showed very little difference in eye size between the sexes. In most cases, (hemizygous) male eyes were only slightly smaller than heterozygous female eyes; both sexes having a small range of sizes that just overlapped. The line with greatest difference and least overlap between the sexes (line C60) was selected for further analysis.
To validate the usefulness of the line in a mutagenesis screen, we crossed line C60 to flies deficient in msl1, msl2, and mle. Flies heterozygous for both GMR-hid and each msl deficiency (GMR-hid/ +; msl1 msl2 mle/+) had similar sized eyes to those without the msl deficiency ( Figure S1), indicating an inability to report on mutations in known MSL components. As an alternative genetic test to determine if GMR-hid could respond to the MSL complex, we crossed line C60 to a line that constitutively expresses MSL2. Females with hsp83-msl2 can assemble MSL complex, erroneously up-regulate X-linked gene expression, and thus have poor viability [36]. However, GMR-hid/+; hsp83-msl2/+ females were identical to females that lacked the hsp83-msl2 transgene ( Figure S1), indicating that the sex-specific eye size may not have reflected action by the MSL complex. The strategy seemed unsuitable for the proposed mutagenic screen.
In a complementary approach, we sought to directly attract MSL complex to autosomal insertions of GMR-hid by coupling the transgene to the strongest known MSL binding sites. The MSL complex binds with high affinity to a DNase I hyper-sensitive site (DHS) in the X-linked roX1 gene [26]. The long roX1 noncoding RNA is a component of the MSL complex. roX1 expression is tightly regulated through a combination of constitutive repression and male-specific hyper-activation by the MSL complex that antagonises the repression [37]. Ectopic autosomal insertions of the roX1 DHS attract significant amounts of MSL complex to the insertion site, clearly visible by immuno-fluorescent labelling of polytene chromosomes, and a nine-copy multimer of the DHS is particularly effective [26]. The roX1 DHS is also capable of mediating increased activity of a coupled lacZ reporter gene at autosomal insertions [25].
We created roX1 DHS-GMR-hid, and (roX1 DHS)9-GMR-hid, and observed the eye phenotype in flies carrying autosomal insertions of each ( Figure 1C,E). Lines carrying single insertions of roX1-DHS-GMR-hid had eyes of equal size in each sex. Those with the nine-copy multimer of the binding site displayed a range of eye sizes, mostly larger than for roX1 DHS-GMR-hid, with male eyes in most cases slightly smaller than female eyes. Thus, eye size did not respond well to any MSL complex recruited by the roX1 DHS.
The MSL complex binds the X chromosome most strongly at hundreds of MSL Recognition Elements (MREs), GA-rich motifs that are found predominantly in intergenic or intronic sequences [38,39]. The MRE at cytological position 18D11 has strong affinity for MSL complex, and can attract the complex to an ectopic autosomal insertion site, especially when multimerized [40]. We placed both the MRE from 18D11 and the tetramer of the site in the 39 UTR of GMR-hid, to mimic the normally high concentrations of MSLs towards the 39 end of a transcript. However, flies transgenic for either construct still displayed eyes of similar size in males and females (Figure 1 D, F). It appeared that directly attracting MSL complex to GMR-hid did not improve the ability of the reporter to respond to dosage compensation.
The failure of GMR-hid to report might have been due to an inability to report two-fold transcriptional changes as an altered eye size. However, we observed a wide range of possible eye sizes (that likely reflected position effects on transgene expression), and the system has displayed great sensitivity previously [8]. Further, the ablated eye phenotype of two copy females showed that the system was dose-responsive. Instead, we theorized that GMR-hid was not amenable to dosage compensation. An alternative, although unlikely explanation, was the MSL complex was not recruited to the transgenes containing a roX or 18D11 high affinity site. Yet, in (roX1 DHS)9-GMR-hid lines (Figure 1 E) female eye color was generally lighter than males, and in some cases indistinguishable from white. Additionally, there was variegation in pigmentation in some lines (e.g. C20, C21). This is consistent with general repression of the mini-white transformation marker gene, responsible for the red pigment, with male-specific relief by the MSL complex, as previously observed at some autosomal insertions of mini-white when coupled to a roX1 transgene [41]. This suggests the MSL complex was recruited to the multimer of the roX1 DHS, sufficient to relieve repression of mini-white, but with little effect on GMR-hid expression.

The promoter can affect compensation
To more quantifiably measure compensation of a GMR-driven transgene, we used the enhancer-promoter to drive the betagalactosidase gene lacZ, and compared beta-galactosidase activity from several related constructs ( Figure 2). To measure the degree of up-regulation and compensation supported by GMR-mediated expression, we compared the response of GMR-lacZ at X-linked sites to that of lacZ driven by the constitutive promoter from the Xlinked armadillo (arm) gene. We have previously shown that Xlinked insertions of arm-lacZ generally acquire dosage compensation [24,42]. To remove position effects of integration, we used the phiC31 recombinase system to target transgenic constructs to defined attP landing sites [43]. We tested five X-linked landing sites  [44] for efficiency of transformation, but could not generate insertions at attP-3B, and found that insertions at attP-3Aa were either male-lethal or ectopically integrated. Of the three remaining sites, attP-2A and attP-6E were within 20 kb of MSL binding sites noted in MSL immuno-precipitates [19], and the peri-centromeric attP-20C was devoid of bound MSL. Flies carrying GMR-hid at each of these locations had eyes similar to those with P-element insertions of the transgene ( Figure 1A, lines C70, C72, C74).
We targeted both arm-lacZ and GMR-lacZ to each of the three X-linked attP sites, and measured beta-galactosidase activity ( Table 1). We compared activities in (hemizygous) males to those in females heterozygous for the insert, as a measure of male hyperactivity ( Figure 3, top panels). Separately, we compared males to homozygous females, to measure dosage compensation ( Figure 3, bottom panels). We took the log 2 of the ratios, to enable a more accurate comparison of ratios above and below zero. A two-fold increase would give a log 2 of 1.0, whereas equal expression would give 0. As expected, arm-lacZ males had nearly twice the activity of single-copy females (top panels), nearly sufficient to compensate for deficiency against two-copy females (bottom panels), although the response at 6E was slightly lower. However, male activity in lines with GMR-lacZ was only slightly above single-copy female activity, providing effectively no dosage compensation against two-copy females. The apparent discrepancy between the two ratios (slight hyper-activity but no dosage compensation) is explored below. As the lacZ sequences were identical between arm-lacZ and GMR-lacZ, these results imply that the GMR-hsp70 regulatory sequences were responsible for poor dosage compensation.
Why did GMR-lacZ fail to compensate? Perhaps the construct actively blocked the compensatory machinery, although no such elements have yet been reported. Alternatively, the construct may have lacked some feature necessary for transcription to be compensated. We selected the landing site at 2A, which supported strong expression and rates of transformation, and tested the effect of a variety of modifications to GMR-lacZ on the male hyperactivity and dosage compensation. Since the MSL complex binds to the body of active X-linked genes with a bias towards the 39 end, we first considered replacing the 39 hsp70 sequences with those from SV40 that were used for arm-lacZ. The GMR-lacZ-SV40 line produced much higher levels of beta-galactosidase than GMR-lacZ, possibly because the lacZ RNA is less stable with hsp70 39 UTR sequences [45]. There was also a small increase in male hyper-activity and dosage compensation, in comparison to GMR-lacZ but the increase was not statistically significant. To enhance binding of the MSL complex to the 39 end of the transgene, we inserted of an MRE (from 18D11) in the SV40 39 UTR of GMR-lacZ-SV40. However, we found no measureable effect on male hyper-transcription (Table 1).
We next considered that the GMR-hsp70 enhancer-promoter lacked elements that are present in the arm promoter and are required to recruit/respond to the MSL complex. The arm promoter fragment contains short runs of GA, similar to the binding sites for the GAGA factor [46] (Figure 2). The GAGA factor has itself been linked to dosage compensation, as it genetically interacts with msl mutations to alter male lethality [47]. We also identified a cluster of four DRE-like elements in the armadillo promoter. When analysed on a cDNA array, DRE is over-represented in fragments bound by MSL1 immuno-precipitates, compared to those not bound (28 and 16%, respectively), and is slightly enriched on the X chromosome [18], implying a possible role in dosage compensation. The arm promoter does not contain an MRE. However, an MRE is found within the gene, but this is downstream of the promoter fragment used in arm-lacZ (A. Alekseyenko, personal communication). We isolated a short fragment of the armadillo promoter (arm*) spanning the cluster of DRE-like elements and five GAGA-like elements, and tested it upstream of GMR-lacZ. The fragment did not include the transcription start site. The activity levels and sex ratio from arm*-GMR-lacZ were similar to those in flies without the arm fragment (Table 1). We next tested an oligonucleotide containing four consecutive GAGA sites and three DRE sites, designed to mimic sequences known to bind GAGA factor and DREF in vitro [48,49]. The activity of DRE/GAGA-GMR-lacZ was lower in both sexes than GMR-lacZ alone, suggesting that the DRE and/or GAGA elements have inhibited GMR-mediated transcription enhancement. However, the male hyper-activity was significantly higher than for GMR-lacZ (P,0.05, T-test).
We expanded on these observations with DRE/GAGA-GMR-lacZ. We targeted the construct to the attP landing sites and 6E and 20C, and found similar responses of the transgene at all locations. We checked that the observed increase was specific to the X chromosome, by targeting the construct to an autosomal attP (at 86F). Indeed, males and females with DRE/GAGA-GMR-lacZ at 86F had similar activity levels (Table 1). We crossed the DRE/ GAGA-GMR-lacZ 2A line to an hsp83-msl2/Sb line, and found that female offspring with hsp83-msl2 had a significantly higher betagalactosidase activity (P,0.05, T-test) than those without (Table 2); a response to induced MSL complex equal in magnitude to that of arm-lacZ. This argues that the male-specific increases observed above were likely due to the MSL complex. Unlabelled elements are identical to the construct immediately above. A, F, G) Sequences identical to, or one mis-match from, the DRE consensus sequence WATCGATW [82], and GAGA (or TCTC), are indicated with points and + symbols, respectively. F) The fragment from the armadillo promoter included in arm*-GMR-lacZ is underlined in panel A. doi:10.1371/journal.pone.0020455.g002 Lastly, we considered that the GMR-hsp70 enhancer/promoter does not respond to the MSL complex because of the hsp70 minimal promoter. To separate the GMR elements from the hsp70 sequences, we replaced the hsp70 TATA with a basal promoter (242 to +23) from the X-linked dosage compensated Pgd gene [50]. The activity from GMR-pgd-lacZ was much reduced compared to GMR-lacZ in both sexes (Table 1), which was not unexpected given that the hsp70 promoter has a relatively high basal activity at 25uC [51]. However, male hyper-activity was significantly higher than for GMR-lacZ (P,0.05, T-test) and dosage compensation appeared to be partially restored.

Hyper-activity of GMR-lacZ in homozygous females
The two methods for comparing male and female activity of all GMR-based constructs appeared to give slightly different results. Constructs with very low hyper-transcription (e.g. GMR-lacZ, arm*-GMR-lacZ) had no dosage compensation, and those with high male hyper-transcription (GMR-pgd-lacZ, DRE/GAGA-GMR-lacZ) had only modest dosage compensation. Because (hemizygous) males in each measurement were genetically identical, this anomaly can only be explained by differences in female expression. Due to experimental variation, the raw male or female activities cannot be directly compared without first controlling for variation by normalization, as with a male to female ratio. But if we instead calculate the inverse (female to male) ratios, then we have comparable activity estimates of females, normalized to single-copy males ( Table 3). As expected, females carrying two copies of arm-lacZ had about twice the activity of single-copy females, which is why a male hyper-activity of two-fold (log 2 of 1.0) is sufficient to dosage compensate. But females with two copies of all GMR-based constructs have more than twice the activity of single-copy females. This explains the discrepancy: a slight male hyper-activity was only sufficient to equal about half the activity of paired females, and full compensation would require male hyper-activity of around 2.5 to 3.0-fold (log 2 of 1.3-1.6). The significance of this homozygous effect is discussed below.
Early or constitutive expression of GMR-lacZ is insufficient to lead to compensation It was also possible that GMR-lacZ failed to compensate because of problems with expression time or specificity. Perhaps dosage compensation was less efficient in the developing eye tissue. The set of MSL-bound genes in larval salivary glands appears to be a subset of the total MSL-bound genes in early embryos, prompting the suggestion that MSL complex may bind early, and maintain largely stable binding through development [18]. The same conclusion was reached after observing that MSL complex can colocalize with RNA polymerase II on salivary gland chromosomes, but remain bound after polymerase dis-associates [52]. We next tested the relative importance of early and constitutive expression on the late, tissue-specific compensation of GMR-lacZ.
We placed the tetO tetracycline operator upstream of GMR-lacZ to allow additional regulation through the tetracycline system [53]. DNA binding of the tetracycline-controlled transcriptional activator (tTA) is inhibited if tetracycline is added to the culture media (i.e. a tet-OFF system). Early expression of tTA was driven by promoters from the nullo and serendipity alpha genes [54]. nullo and serendipity alpha encode proteins that are required for cellularization and are predominately expressed in the early stages of embryogenesis [55]. Expression from each activator (n1-tTA and s1-tTA, respectively) is sufficient to lead to cell death through expression of a tetO-hid ala5 transgene, but only in the absence of tetracycline [54]. arm-tTA [56] was used as a constitutive driver of tTA. We crossed flies carrying tetO-GMR-lacZ to lines with arm-tTA, n1-tTA, or s1-tTA activators, and raised the offspring in the absence of tetracycline. We measured the male hyper-activity of betagalactosidase in adult heads (Table 4), and compared the responses to those of arm-lacZ and GMR-lacZ ( Figure 4A). Males and (singlecopy) females of all crosses had very similar activities. The response to additional constitutive expression provided by arm-tTA was evident in the very high activities in offspring of this cross. Further,   the increase in lacZ expression was dependent on tTA as there was no response when flies were raised with tetracycline (mean male activity of 2.1560.06, female of 1.6460.07 OD/min/mg). A separate cross to y w flies (parental strain), which lack tTA, showed that the tetO sequences alone had little effect on activity from GMR-lacZ (Table 4). To confirm that the tTA drivers were active in the early embryo, we crossed the tTA driver lines to a tetO-YFP responder line [57] and examined the offspring for yellow fluorescence ( Figure 4B). We found that all tTA drivers were effective at inducing YFP expression well above background levels.
To confirm that regulation via the tetracycline system was capable of shifting the expression pattern of tetO-GMR-lacZ, we crossed the reporter line to arm-tTA flies and observed the altered expression pattern in climbing third instar larvae ( Figure 5). The GMR enhancer/promoter is active in all cells in, and posterior to, the morphogenetic furrow in the developing eye-antennal disc in third instar larvae [6]. In a control cross to y w flies, betagalactosidase staining could be detected in the developing eye, posterior to the morphogenetic furrow, and to a limited extent in the brain, consistent with Glass expression [30]. When crossed to arm-tTA, beta-galactosidase could also be detected in the brain and most other larval tissues ( Figure 5). This response was specific to arm-tTA as it was inhibited with the addition of tetracycline to the culture medium. Given that the tetO-GMR-lacZ reporter could respond to additional regulation through the tetracycline system, and that the activators were all active in the early embryo, we concluded that the above test of effect on dosage compensation was valid. Therefore, additional early or constitutive expression of GMR-lacZ was insufficient to affect dosage compensation. The time and specificity of expression may have been relatively less important than the nature of the promoter/enhancer. However, the staining for beta-galactosidase ( Figure 5) revealed a surprising observation. The pattern of staining seen with arm-tTA activation of tetO-GMR-lacZ did not equal the pattern from the arm-lacZ positive control line. The tetO-GMR-lacZ reporter was widely expressed with the notable exception of the developing imaginal discs, beyond the Glass-expressing areas of the eyeantennal disc. As the fragment of armadillo was the same in arm-tTA and arm-lacZ [56], it appears that tetO-GMR-lacZ was being repressed in these tissues. These results suggest that the 29 bp GMR enhancer contains the binding site for a repressor that is being expressed in imaginal cells.

Discussion
Our attempt to screen for modifiers of dosage compensation was thwarted by the failure of GMR-hid to be well compensated. The transgene did not acquire compensation in an X chromosome environment, nor when coupled to strong MSL attractors. Quantitative measurements with the lacZ gene confirmed that the GMR-hsp70 enhancer-promoter was at fault. We believe that this is not due to lack of local MSL complex, but rather an inability of the complex to regulate GMR-mediated transcription.
This was most evident in the phenotype of the GMR-hid lines with the 9-copy multimer of the roX1 DHS. In most of these lines,  females had a lighter eye color than males, and there was obvious variegation in pigmentation patterns in several lines, suggesting repression of the mini-white marker gene. It had been previously shown that recruitment of the MSL complex to a roX1 transgene led to relief of repression of mini-white [41]. Similarly, MSL complex recruited to the roX1 DHS in the (roX1 DHS)9-GMR-hid lines may have been sufficient to relieve repression on mini-white. However, there was little difference in eye size, suggesting that GMR-hid was not responding to the bound MSL complex. Lack of response to MSL complex is also consistent with our observations that strong MSL binding sites had no effect on GMR-lacZ betagalactosidase activity. While formally possible that local MSL complex was limiting at each site, the large number of lines screened makes this unlikely. Certainly local MSL complex at the 2A, 6E, and 20C attP sites was sufficient to up-regulate arm-lacZ, yet had no effect on GMR-lacZ. The most likely explanation is that expression from the GMR-hsp70 enhancer-promoter was not responsive to MSL complex. Our observation that the promoter can affect dosage compensation was somewhat surprising. The MSL complex binds the male X chromosome most strongly at 150 to 300 mostly noncoding GA-rich MRE sequences [38,39] but is also highly enriched across most active X-linked genes, with a bias towards the 39 end [17,18,19]. It has been proposed that the complex initially binds to the MRE sites and then spreads to active genes [38]. In support of this model, the MSL complex is attracted to ectopically-transcribed portions of the X chromosome, created when random insertions of a UAS-hsp70 TATA transgenic promoter are induced with GAL4 [58]. Further, the complex is recruited to transgenic autosomal insertions of the X-linked mof gene, but only when mof is transcribed [59]. As recruitment still occurs when mof is transcribed from a tubulin promoter, GAL4 UAS sequences, or even in an antisense direction, Kind and Akhtar conclude that ''the type of promoter and direction of transcription are not [important]''. It appears that some feature of transcription itself, or an active chromatin state, may be responsible for recruitment of MSL complex to the many genes on the X chromosome. This may involve the chromo domain of MSL3, which recognizes histone H4 monomethylated at lysine 20 [60,61], a marker for active genes in human cells [62]. Accordingly, autosomal genes transposed to the X chromosome acquire MSL binding, H4K16 acetylation, and dosage compensation, but only when the transposed genes are transcribed [63]. However, transcription was disabled by deleting promoter regions. Figure 5. tetO-GMR-lacZ responded to tTA in brain and other tissues but not imaginal discs. Third instar larvae of the indicated strains, raised with (+) or without (2) tetracycline (tet) in the diet, were dissected and stained with X-gal for lacZ expression. Additional induction of tetO-GMR-lacZ occurs with arm-tTA in the absence of tetracycline. y w and arm-lacZ provided as negative and positive controls for lacZ expression, respectively. All staining and photographic conditions were equal across all sets. Bar = 0.2 mm. doi:10.1371/journal.pone.0020455.g005 Our observation that GMR-lacZ failed to acquire compensation, at the same X-linked sites where arm-lacZ did, argues that transcription alone is not sufficient for a gene to be compensated, and that the importance of a gene's promoter in dosage compensation cannot be ignored. Indeed, studies of the X-linked white gene concluded that the gene contains multiple elements that influence dosage compensation, some near the promoter and some in the coding region [64,65]. Further, it has been suggested that DNA supercoiling factor bound at promoter regions may facilitate loading of the MSL complex onto active genes [33]. DNA supercoiling factor contributes to Drosophila dosage compensation [10].
Our efforts to determine why genes driven by the GMR enhancer-promoter do not respond to the MSL complex were complicated by the surprising observation that activity in homozygous females was more than twice that of hemizygous females. We believe this may be because of a complex interplay between the Glass transcription factor and an unknown repressor. As a consequence of the hyperactivity in homozygous females, the comparison of the activity of one dose males with one dose females provides a more accurate view of male hyperactivation. Of the various gene constructs evaluated, two showed a significant increase in activity in males. In one construct, the hsp70 basal promoter was replaced with the weaker basal promoter from the X-linked dosage compensated Pgd gene. This suggests that the Pgd basal promoter contains elements that are important for recruitment of the MSL complex to the reporter gene and that the hsp70 basal promoter lacks these elements. The white gene promoter was concluded to be important for dosage compensation because autosomal white transgenes driven by the full heatinducible hsp70 promoter were expressed equally in males and females [64]. Autosomal white transgenes are typically expressed at higher levels in males than females [66]. An alternative interpretation of these results is that the hsp70-white transgene is not compensated because the hsp70 promoter does not respond to the MSL complex. The Pgd promoter did not contain any obvious binding sites for known transcription factors. Widely expressed genes such as Pgd and arm generally contain similar levels of RNA polymerase II at 59 and 39 regions of the transcription units [67]. It has been suggested that RNA polymerase II stalling at 59 ends depends upon core promoter elements that are absent in constitutively expressed genes [67]. Thus it may be of interest to determine if the MSL complex can only regulate those X-linked genes that do not have stalled promoters. The other construct that showed significant increased activity in males contained binding sites for the GAGA and DREF DNA binding proteins upstream of the GMR-hsp70 enhancer-promoter. The DREF and/or GAGA factors could create a local chromatin environment that facilitates recruitment of the MSL complex or could interact directly with the complex. It may now be beneficial to separate the effects of DRE and GAGA elements, and determine such characteristics as motif strength, placement, and repeat spacing. Such studies may explain why addition of a fragment from the arm promoter that contained predicted GAGA and DREF sites, did not lead to a significant increase in GMR-lacZ expression in males. We found no support for the notion that late or tissue-specific expression was a factor in poor compensation, although it is possible that these factors affect compensation of other genes. Thus we conclude that genes driven by the GMR-hsp70 enhancer-promoter do not respond to the MSL complex because of the nature of the hsp70 basal promoter and because it lacks elements such as GAGA and DREF binding sites.
A curious possibility is that any one of the promoter modifications could have caused or increased repression of the transgene, perhaps via the putative repressor detected in larval staining for beta-galactosidase. This may be reflected in the generally lower beta-galactosidase activity in lines with pgd-GMR-lacZ and DRE/GAGA-GMR-lacZ transgenes. Knowing that MSL complex can act to relieve repression in males, such changes would also register in our tests as an apparent increase in male hyperactivity. However, we cannot distinguish between general repression with male-specific relief, and more simple male enhancement, as our assays were designed to measure the final result of dosage compensation.
Can our results shed any insight into why other X-linked genes are not dosage compensated? The non-compensated X-linked Lsp1alpha gene has been particularly well studied [42,68,69]. Lsp1alpha is not enriched for H4K16ac in the tissue in which it is active [42]. Similarly, the runt gene, which is dosage compensated by an MSL-independent mechanism [70], is not enriched for H4K16ac in Drosophila embryos [16]. Thus it appears that these genes are not hypertranscribed by the MSL complex due to a failure to recruit the complex rather than because their promoters lack certain elements. Indeed, X-linked lacZ transgenes driven by the Pgd promoter are enriched for H4K16ac [42]. Whether or not this is the general explanation for why some X-linked genes are not compensated would require a more global approach to compile a list of non-compensated genes. To our knowledge no such list exists, however it may be possible to do this using the extensive existing data sets of male vs female expression, MSL2 RNAi knockdown in cell lines and MSL complex binding and H4K16ac enrichment profiles [13,38,71,72]. However, in general almost all active genes on the male X chromosome are enriched for H4K16ac, which has a broader distribution profile than the MSL complex [13]. Thus since there are very few active X-linked genes that are devoid of H4K16ac, either there are very few noncompensated genes or other factors such as the nature of the gene promoter influence male hypertranscription.
Beyond the promoter, our experiments also revealed that another factor can affect the degree of compensation. Two copies of all GMR cassettes had more than twice the beta-galactosidase activity of one copy, instead increasing three-to four-fold. As we only measured this effect in females with X-chromosomal insertions, it would be interesting to extend the experiments to autosomes and males. The non-linear effect did not occur with arm-lacZ, and is not a property of the beta-galactosidase assay, which is linear across a 100-fold range of protein extract concentration [51]. As a similar effect occurred in all cassettes with GMR, regardless of the surrounding sequences, the effect appears to be due to paired GMR enhancer elements. Homologous chromosomes in Diptera pair through all mitotic stages, not just during meiosis [73]. Paired homologues can share transcription factors or chromatin regulators, which can affect transcription in a variety of ways known as 'transvection' [74]. The pairing effect of GMR may similarly reflect synergy in the binding or activity of positive factors such as Glass. Conversely, negative regulators such as the putative repressor could have less effect on the paired GMR elements than would be expected for two unpaired elements. We attempted to determine if the non-linear response was pairing-dependent by measuring two copies of GMR-lacZ cassettes at different locations, but found that the differing activity levels at each location precluded conclusions based on mixed genotypes. Whatever the cause, the non-linear effect has consequences for dosage compensation of GMR-lacZ, as male expression needs to be more than doubled to equal the paired female levels.
More generally, a description of X chromosome dosage compensation as a transcriptional doubling by the MSL complex may be too simple. When quantified by microarray, X chromosomal transcripts reduced a variable amount after knockdown of msl-2, with median decreases of about 1.3 to 1.7-fold [72,75]. Recent estimates based on high-throughput sequencing of RNAs in S2 cells put the effect of MSL-mediated up-regulation at about 1.4-fold [71]. Zhang et al. believe that the remaining upregulation required to equalize gene dose is likely provided by a separate mechanism that also compensates for aneuploid regions genome-wide. Gene expression in aneuploid regions of S2 cells, or in chromosome aberrations, can be ''buffered'' to varying extents, with levels closer to wild-type than would be expected from gene copy number [71,76,77]. If Drosophila males have indeed evolved to co-opt gene buffering as a part solution to X chromosomal haploidy, then dosage compensation should be viewed as a combinatorial process. At the level of an individual X-linked gene, male expression may be affected by any or all of several mechanisms, including aneuploidy-type buffering, dose-dependent regulatory feedback, and MSL-mediated up-regulation. Given that the paired female X-linked genes might also be affected by transvection, dosage compensation could be seen as merely the result of a combination of mechanisms. Care must be taken in the design of experiments to ensure the correct biological process is measured.
Finally, it should be noted that the tissue-specificity of the GMR enhancer appears to reflect more than simple activation by the Glass transcription factor. We observed a general lack of GMR-lacZ expression in larval imaginal tissues beyond those that express Glass. This repression pattern is similar to that observed for another Glass reporter, ''38-1'', a pentamer of a larger (38 bp) fragment from the Rh1 proximal enhancer spanning the Glass binding site [30]. 38-1 cannot respond to glass overexpression (from a hsp70-glass transgene) in imaginal tissue with the exception of a small number of cells in the leg disc. In contrast, in response to glass overexpression, 38-1 directs lacZ expression in many cells in the brain and central nerve cord. A shorter 29 bp enhancer (GMR) responded more strongly to glass overexpression, directing lacZ expression in many cells in the brain and in all cells of the eye/antennal and leg discs. It was concluded that shortening the enhancer to 29 bp had deleted the binding site for a putative repressor that inhibits Glass activity in non-photoreceptor cells [30]. DNA sequencing of GMR-lacZ used in this study confirmed that we were using the pentamer of the 29 bp element. One explanation for our results is that the repressor binding site was not completely removed when shortening the fragment from 38 to 29 bp, merely weakened to allow dominance of high levels of Glass over the repressor. Alternatively, the 38 bp enhancer contains binding sites for two repressors, one of which is retained in the 29 bp GMR enhancer. Additional studies would be required to distinguish between these two possibilities. In either case, it appears that the tissue-specific expression directed by the 29 bp GMR enhancer is due to combined action of the Glass activator and an unknown repressor.

Recombinant DNA
Standard molecular cloning techniques [78] were used, with clones confirmed by restriction analysis, PCR, or DNA sequencing.
hid gene constructs. An EcoRI to EcoRI fragment of pBS SK hid cDNA [7], encoding hid cDNA, was inserted into the EcoRI site of pGMR-1 [4] to create pGMRhid, which contains the GMR-hid cassette used for most GMR-hid lines. A XhoI to NotI fragment of pCaSpeR3 roX1 DHS short or pCaSpeR3 roX1 DHS short multimer [26], containing the roX1 DHS or its multimer, was inserted into the XhoI site of pGMR-1, with the aid of oligonucleotides (59-TCGACGTTTAAACGGTTGGCC-39 and 59-AATTGGCCAACCGTTTAAACG-39) annealed to link EcoRI and XhoI sites, to create pCL02 or pCL03, respectively. A EcoRI to EcoRI fragment of pBS SK hid cDNA containing hid was inserted into the EcoRI site of pCL02 or pCL03, to create pRGH and pR9GH, which contained the roX1 DHS-GMR-hid or (roX1 DHS)9-GMR-hid cassettes, respectively. pCaSpeR3 18D10-L monomer or pCaSpeR3 18D10-L4mer [40] was digested with NotI, treated with Klenow fragment, then digested with BamHI, and the fragment encoding the 18D site, or its multimer, was inserted into the BglII and StuI sites of pGMRhid to created p18DGH, and p18D4GH, which contain the GMR-hid-18D and GMR-hid-(18D)4 cassettes, respectively. An EcoRI to EcoRI fragment of pTAattB [43], encoding attB, was treated with Klenow fragment and inserted into the EcoRV and SmaI sites of pBluescript II KS+ (Invitrogen) to create pBSattB. pBSattB was partially digested with XhoI and re-ligated to remove the XhoI site, then a HindIII to HindIII fragment of pGMR-1, encoding mini-white and GMR, was inserted into the HindIII site to create pBSw+GMRattB. An EcoRI to EcoRI fragment of pBS SK hid cDNA, encoding hid cDNA, was inserted into the EcoRI site of pBSw+GMRattB to create pGHattB, which contains the GMR-hid cassette used for phiC31-mediated transformations (lines C70, C72, C74).
Right-hand-side adult eyes that best represented the modal average eye size of each line were illuminated with reflected halogen light. Images of the lines carrying (roX1 DHS)9-GMR-hid were captured with a Magnafire camera and software (Optronics), and the RGB colour balance manually altered to best match other images. All other images were captured with an Olympus DP-70 camera, with the provided software or with Soft Imaging System analySIS FiVE 5.0, and balanced RGB on white card to control for exact lighting conditions. The exposure time was automatically determined by the software.
P-element mediated transformation of Drosophila was performed as previously described [24]. phiC31 recombinase mediated transformation was performed using strains that contained both an attP landing site and expressed recombinase in the germ-line [44]. Transgenic flies were back-crossed to the y w line, established as homozygous or balanced lines, and confirmed by PCR both within the insert and across the insert/landing site boundary.
Beta-galactosidase assays of hemisected adults or heads were conducted based on the procedure of Simon and Lis [51]. If necessary, tetracycline was added at 10 ug/mL to the media, and flies raised in the dark. New emerged flies were collected and then aged 3 to 5 days at before the assay. For each replicate, 9 females or 12 males were hemisected to remove wings, legs, and abdomens, or 15 female and 20 male heads used. The mean activities of males, females, or ratios between the sexes, are displayed with standard errors (one standard deviation / square root of sample size (3)), with a log 2 transformation then often applied to allow better comparisons of different ratios. Two sample, two-tailed Student's t-tests were conducted in the R software environment, using (the default) Satterthwaite's approximation. Tissue dissected from climbing third instar larvae was stained with X-gal essentially as described by Glaser et al. [81]. The stained tissue was mounted in 90% glycerol, and photographed under transmitted white light at 1/90 second exposure, ISO 200, with an Olympus DP-70 camera. Figure S1 GMR-hid did not respond to the MSL complex. Eyes of flies that carried a single copy of the GMRhid transgene (line C60), and were either wild type for the msl genes, heterozygous for msl1, msl2 and mle or constitutively expressed msl2 (hsp83-msl2). If GMR-hid on the X chromosome responded to levels of the MSL complex then the eye size in msl1 msl2 mle heterozygous males should have been larger than control. Similarly, eyes of females that expressed msl2 should have been smaller than control. (TIF)