We employed the DamID technique 41 to study the redistribution of HP1 on w^m chromosomes. DamID was previously used to identify the natural binding sites of Drosophila HP1 in Kc cells 1718 and in whole adult flies 42. These studies found a strong enrichment of HP1 in pericentric regions and on chromosome 4, in agreement with immunofluorescence microscopy data 18. DamID maps of HP1 showed a strong overlap with several other heterochromatin proteins but not with euchromatin proteins 171843. DamID also confirmed earlier microscopy observations 30, indicating that the binding of HP1 to chromosome 4 is less dependent on the presence of Su(var)3-9 than binding of HP1 to all other chromosomes 18. These earlier studies thus validated the use of DamID to map HP1 binding.

We expressed the Dam-HP1 fusion or unfused Dam from the hsp70 core promoter 42, which at 25°C drives only extremely low expression levels, as confirmed by the lack of a detectable band on a Western blot (Figure 1A). This trace amount of Dam-HP1 expression relative to endogenous HP1 is not only a requirement for DamID 41 but it also ensures that the Dam-HP1 fusion protein does not significantly alter heterochromatin by itself. In order to map HP1 binding on different variants of w^m inverted X chromosomes we downscaled DamID for use in adult fly heads. We used whole heads because chromosomal conformations correlating with an eye phenotype can be readily observed in the central nervous system 3839. We examined two different chromosomal inversions, w^m4e and w^m51b, which display a different eye phenotype but have a very similar chromosomal structure. w^m4e males have almost completely white eyes with only a few red patches, whereas the eyes of w^m51b males are red and resemble wild-type eyes (Figure 1B). The euchromatic breakpoints of the w^m4e and w^m51b inversions are located ~29 and ~25 kb downstream of the transcriptional start of the white gene (see Methods and 34), which is normally located on the distal tip of the chromosome. The heterochromatic breaks are proximal and distal to the rDNA, in heterochromatin block h28 and h30, respectively 4445 (Figure 1B). As a control we mapped HP1 on the wild-type X chromosome in the Oregon-R-S strain.

To identify the DNA from the chromatin that was bound by HP1 in vivo, we designed a high-density 44 K oligonucleotide microarray with 60-mer probes corresponding to unique sequences from the first 3.2 Mb of the X chromosome (chr). As positive control regions we included parts of the genome where HP1 levels are expected to be high, i.e., the centromere-proximal 0.75 Mb of chr 2R, and the complete mostly heterochromatic chr 4 1842. As a negative control region we included a euchromatic 0.5 Mb segment of chr 2R (position 11.4–11.9 Mb), where HP1 binding is restricted to a small number of genes 1842. We normalized the log₂ binding ratios to the average of this euchromatic chr 2R segment (see Methods). Therefore, positive log₂ binding ratios (i.e. ratios > 1) can be interpreted as more HP1 binding than found on average in euchromatin.

To visualize the binding of HP1 in adult male heads we made chromosomal maps (Figure 1C–H). In agreement with previous observations 18 we detect high HP1 levels on the centromere-proximal 0.75 Mb part of 2R (data not shown) and along most parts of chr 4 (Figure 1D, and data not shown). Next, we focused on the HP1 binding pattern in a 400 kb region surrounding the white gene (Figure 1E–H). Compared with Oregon-R-S (Figure 1E), HP1 levels on w^m4e (Figure 1F) and w^m51b (Figure 1G) are clearly elevated in the regions next to the inversion breakpoint. Effects of the inversion are seen on both sides of the euchromatic inversion breakpoint (vertical dotted lines in Figure 1F–H). HP1 levels are elevated over a ~175 kb region stretching from the break towards CG3603, the gene upstream of roughest (rst), and over a ~30 kb region downstream from the break including Syx4, but clearly not crm (Figure 1H). We will refer to this ~200 kb region of the chr X surrounding the euchromatic inversion breakpoint as X_Syx4-CG3603.

On the wild-type chr X the region corresponding to X_Syx4-CG3603 shows no prominent HP1 binding (comparable to the average of the entire X chr), indicating that this region does not possess an intrinsic bias for HP1. In contrast, on w^m4e and w^m51b almost all microarray probes in this region report elevated levels of HP1. Thus, the inversion induces HP1 association with the entire X_Syx4-CG3603 region, although gaps smaller than the estimated resolution of DamID (about 1–2 kb 41) cannot be ruled out. Despite this apparent contiguous HP1 binding, reproducible local variations in the HP1 DamID signal were observed, with the highest levels of HP1 close to the inversion breakpoint and on the white gene. Overall, the HP1 binding ratios in X_Syx4-CG3603 are somewhat lower than those in the pericentric heterochromatic regions of chr 4 and chr 2R (cf. Figure 1D and data not shown). Taken together, these data suggest that w^m chromosomal inversions induce invasion of HP1 from heterochromatin into the neighboring euchromatic regions. HP1 binds to these regions in a contiguous fashion, but the binding levels show substantial local variation.

These HP1 data are consistent with a previous ChIP analysis of a w^m chromosomal inversion, which detected H3K9me2 in the region from white to rst, with the highest levels on white 46. Direct comparison reveals that, at the six loci in this region for which H3K9me2 levels are known, an excellent correlation (R² = 0.96) exists between our HP1 DamID signals and the H3K9me2 ChIP data (Figure 2). From this we conclude that our HP1 maps are of high quality.

The HP1 pattern on the X chromosome outside X_Syx4-CG3603 looks highly similar in all genotypes examined (Figure 1D–F, data not shown), suggesting that the w^m chromosomal inversions affect HP1 binding patterns only locally. To examine this in more detail we calculated the average change in HP1 binding ratios for each of the probed chromosomal segments (Figure 3A). HP1 binding levels in X_Syx4-CG3603 are 1.6–1.7-fold higher on w^m4e and w^m51b compared with the wild-type X chr, whereas the remainder of the chr X and pericentric heterochromatin on chr 2R were unaltered (< 1.07-fold change). We also noticed a smaller increase (~1.2-fold) in HP1 binding ratios on chr 4, which might be a secondary effect of the X chromosomal inversion, or due to other differences in the genetic background. In any case, the most pronounced effects of the w^m inversions are found locally in X_Syx4-CG3603.

In some rearrangements that give rise to PEV, genes up to ~2 Mb from the breakpoint have been shown to variegate 47. To systematically identify the genes on chr X where HP1 binding was affected most, we calculated the average change in HP1 binding log-ratio (ΔHP1) for each gene on w^m4e and w^m51b chromosomes relative to the wild-type chr X (Figure 3B–C). We then selected the top 5% of genes (n = 16) for which ΔHP1 is largest in w^m4e (Figure 3B) or w^m51b (Figure 3C). Twelve out of 16 affected genes overlap between w^m4e and w^m51b (Figure 3D), and these genes are all located in the X_Syx4-CG3603 region. This result reinforces the notion that effects of the chromosomal inversion on HP1 patterns are mainly found locally.

The white gene is the gene that shows the strongest increase in HP1 levels, both on the w^m4e and on the w^m51b chr. Nevertheless, the eye color phenotype of these two lines is remarkably different: w^m4e males have almost white eyes, whereas the eye color of w^m51b males is virtually wild type (Figure 1B). From the eye color phenotypes we would predict that the white gene is down regulated in the w^m4e line and normally expressed in the w^m51b line. To investigate the transcriptional status of white and other genes, we generated microarray expression profiles of male heads from w^m4e and w^m51b and control Oregon-R-S flies. The MA-plot is a graphical way to visualize expression levels (fluorescence intensity) and change in expression levels (log-ratios) at the same time, that can be used to identify differentially expressed genes. The MA-plot of a set of Oregon-R-S self-self hybridizations (Figure 4A) was used to estimate the biological and technical noise. MA-plots of w^m4e against Oregon-R-S and w^m51b against Oregon-R-S are shown in Figure 4B and 4C, respectively. First, we focused on the 20 genes in X_Syx4-CG3603 (Table 1). More than half of these genes had an expression level that was too low to allow detection of differential expression on our microarray platform (Table 1, genes with A < 7, stringent and arbitrary cutoff). These genes are in the left part of the A-axis in Figure 4B and 4C. Eight out of 20 genes were expressed at sufficient levels (A > 7) to detect their possible differential expression. As expected, white was down-regulated in w^m4e (log₂ ratio = -2.11, P < 10^-45, Figure 4B and Table 1) but not in w^m51b (log₂ ratio = -0.21, P = 0.18, Figure 4C). Other than white only one out of the eight genes (CG14419) showed a modest down-regulation (log₂ ratio = -1.10) in w^m4e compared with Oregon-R-S, while none of the other genes displayed a detectable reduction in their expression. In w^m51b none of the eight genes was more than about 30% up- or down-regulated compared with Oregon-R-S (Table 1).

To verify our microarray expression data we used quantitative RT-PCR (qRT-PCR) to determine the relative expression of several genes (Figure 4D). This confirmed that white is down-regulated in w^m4e (P = 2*10^-4, Student's t-test) but not in w^m51b (P = 0.48). None of three other tested genes in X_Syx4-CG3603, including CG14419, showed significant down-regulation in either w^m4e or w^m51b. Taken together, these results identify white as the gene with the most prominent response to heterochromatin, while other genes in the X_Syx4-CG3603 region are not (or only marginally) affected, despite the fact that they show elevated HP1 levels.

Heterochromatin formation is generally thought to lead to increased compaction of chromatin. Previously it was shown that upon heterochromatinization various genes in yeast and flies become less accessible to methylation by (unfused) Dam 48495051. To study the effects of the w^m rearrangements on chromatin accessibility, we analyzed the methylation levels as reported by the Dam-only channel of our DamID microarray data. Indeed, a majority of probes in the X_Syx4-CG3603 region show a reduction in methylation by Dam-only in w^m4e and w^m51b flies compared with wild-type flies (Figure 5A and data not shown). This change is not found outside of the X_Syx4-CG3603 region, which is consistent with the unaltered binding of HP1. In contrast to Dam-only, the Dam-HP1 methylation channel shows a specific enrichment in the X_Syx4-CG3603 region (Figure 5B), due to targeting of the fusion protein to heterochromatin. These data show that the invasion of HP1 due to w^m rearrangements is accompanied by a general decrease in chromatin accessibility.

Because the expression levels of all genes in X_Syx4-CG3603 except white are not detectably altered in w^m flies (Figure 4), we wondered whether the compaction of chromatin may be restricted to intergenic regions. Subsequent analysis revealed that promoters of the genes in X_Syx4-CG3603 show a significant reduction in accessibility (Figure 5C). At the same time, the accessibility of the coding regions of these genes is not reduced. These results suggest that heterochromatinization differentially affects the accessibility of promoters and coding regions. The reduction in promoter accessibility at most genes in X_Syx4-CG3603 is apparently not sufficient to cause detectable changes in gene expression, except in the case of white.

Su(var)3-9 is one of the strongest modifiers of PEV known 52. It was previously shown that the loss of a single allele of Su(var)3-9 dramatically decreases the silencing of reporter genes, including white, in a number of PEV reporter assays 53. Indeed, heterozygous loss of Su(var)3-9 changes the eye color of w^m4 males from nearly white to almost completely red (inset Figure 6A). This suggested that heterozygous loss of Su(var)3-9 may lead to destabilization of heterochromatin in the X_Syx4-CG3603 region. To test this, we constructed HP1 binding maps in heads of mutant w^m4; Su(var)3-9⁰¹/+ males, and of sibling w^m4;+/TM3, Sb Ser control males. Flies carrying the TM3 balancer chromosome were used as a control because this balancer does not affect PEV 545556. The w^m4 and w^m4e X chromosomes originate from the same fly stock 34 and we confirmed that they have the same euchromatic breakpoint (see Methods). The binding maps of the control flies showed elevated HP1 levels in X_Syx4-CG3603 on w^m4, very similar to w^m4e (Figure 6A and 6B). In contrast, heterozygous loss of Su(var)3-9 leads to a modest (log₂ ratio -0.27, corresponding to ~0.8-fold) but statistically significant reduction of HP1 binding to X_Syx4-CG3603 (Figure 6B and 6E), and especially to white (Figure 6B). Unlike in the X_Syx4-CG3603 region, the HP1 levels on the probed heterochromatic segments of chr 2R and 4 are not affected by heterozygous loss of Su(var)3-9 (Figure 6C–E).

To systematically identify the genes on chr X that show the strongest reduction of HP1 after removal of one allele of Su(var)3-9, we calculated the change in HP1 binding per gene (Figure 6F) and selected the bottom 5% of genes (n = 16). Ten out of these 16 genes are located in the X_Syx4-CG3603 region. Interestingly, these 10 genes are the same genes that gain HP1 on w^m4e and w^m51b. Thus, HP1 binding to the X_Syx4-CG3603 region of w^m4 is exceptionally sensitive to the levels of Su(var)3-9, whereas HP1 binding to other heterochromatic regions is more robust.

To examine the effect of heterozygous Su(var)3-9 loss on the expression of white and other genes, we made expression profiles in male heads from mutant w^m4; Su(var)3-9⁰¹/+ and control w^m4;+/TM3, Sb Ser flies (Figure 7A and 7B). The MA-plot of the mutant w^m4; Su(var)3-9⁰¹/+ shows that only two of the 16 genes at which HP1 levels are most decreased (i.e. genes indicated with black dots in Figure 6F) also have an altered expression level (Figure 7B). These genes are white, which is expressed ~3-fold higher (log₂ ratio = 1.56, P = 4.2*10^-41), and CG14419, which is expressed slightly lower (log₂ ratio = -0.38, P = 2.1*10^-8). To confirm this result we repeated the microarray experiment using w¹¹¹⁸ flies, instead of Oregon-R-S as source of the wild-type Y chromosome and autosomes. The correlation between these two experiments was high (Spearman's rho = 0.54, P < 2.2*10^-16, data not shown). We confirmed the up-regulation of white (log₂ ratio = 1.28, P < 10^-45), and found that CG14419 was slightly, but not significantly, down-regulated (log₂ ratio = -0.16, P = 0.02).

Su(var)3-9 loss possibly affects HP1 binding and gene expression levels outside the X_Syx4-CG3603 region. However, a bivariate scatterplot (Figure 7C) did not reveal any correlation between the changes in HP1 binding and the changes in expression level (Spearmans's rho = -0.08). This suggests that there is no association between HP1 loss and change in gene expression outside X_Syx4-CG3603. Thus, in a w^m4 background, loss of one allele of Su(var)3-9 mainly affects the expression level of white. None of the other genes in the X_Syx4-CG3603 region are upregulated as a result of the reduced Su(var)3-9 dosage.

It has been suggested that binding of transcriptional activators to the promoter can counteract heterochromatin-mediated silencing 57. This protection was shown to depend on the level and timing of transcriptional activator expression. In particular, promoter activation in the embryonic stage could prevent silencing of a reporter later in development 57. To investigate whether such a phenomenon may explain the preferential silencing of white by heterochromatin, we analyzed the embryonic expression levels 58 of all 20 genes in the X_Syx4-CG3603 region (Figure 8). Several of these genes are expressed at even lower levels than white, suggesting that the absence of embryonic activity of white is not the sole determinant of its preferential silencing by heterochromatin.

Oregon-R-S (#4269) and w¹¹¹⁸ (#3605) were obtained from Bloomington Drosophila Stock center. The w^m4;Su(var)3-9⁰¹/TM3, Sb Ser stock, In(1)w^m51,w^m51bct and In(1)w^m4,w^m4e (denoted as w^m51b and w^m4e, collectively referred to as w^m) were kindly provided by P Talbert, Fred Hutchinson Cancer Research Center, Seattle, WA, USA. The inverted X chromosomes in these stocks are described in 34. We examined the positions of the euchromatic inversion breakpoints of the w^m4, w^m51b and w^m4e stocks by PCR (data not shown). The w^m51b break is located between position 2,665,094 and 2,666,052 (R5), and the w^m4 and w^m4e breaks are both between position 2,661,012 and 2,662,278, which is ~1,000 bp more upstream than was previously thought 44. This corresponds to ~(24.5–25.5) and ~(28.2–29.5) kb upstream of the beginning of the white gene (R5.4).

The Dam-HP1 transgenic line (w¹¹¹⁸;P{Dam-Myc-HP1,w^+mC} HP23/TM6B) was made in parallel with the Dam-HP1 transgenic line used in 42, but has not been published before. Dam-only transgenic lines (w¹¹¹⁸; P{Dam, w^+mC} 1-1M/TM6B and w¹¹¹⁸; P{Dam, w^+mC} 1-4M/CyO) were constructed for this study. The SacII/EcoRI fragment from pNDamMyc 68 encoding Myc tagged EcoDam was cloned into pUAST 69. Germline transformations into w¹¹¹⁸ were done by BestGene Inc, Chino Hills, CA, USA. Transformants were identified by their eye color and balanced. Dam and Dam-HP1 expression are driven from the un-induced, truncated heat-shock promoter in pUAST, yielding extremely low expression levels below the detection limit of Western blotting or immunofluorescence microscopy 68. Thus, it is highly unlikely that the Dam-HP1 protein itself will alter heterochromatin structure.

Flies were raised at 25°C on standard cornmeal/molasses/agar medium. For DamID of HP1 we crossed w^m51b, w^m4e or Oregon-R-S virgin females to Dam (line 1-4M) and Dam-HP1 males. The heads of male progeny (expressing the Dam transgene and containing an inverted X chromosome) were removed using a razor blade, stored at -80°C and used for DamID. To map HP1 in presence of heterozygous Su(var)3-9⁰¹, we crossed w^m4;Su(var)3-9⁰¹/TM3, Sb Ser virgin females to Dam (line 1-1M) and Dam-HP1 males, which both have the transgenes inserted in the 3^rd chr. Heads of male progeny expressing Dam or Dam-HP1 and heterozygous for Su(var)3-9⁰¹, and heads of males expressing Dam or Dam-HP1 and with balancer (TM3, Sb Ser) were collected. For expression profiling we crossed w^m4;Su(var)3-9⁰¹/TM3, Sb Ser virgin females to Oregon-R-S or w¹¹¹⁸ males. Heads of male progeny with and without Su(var)3-9⁰¹ were collected. Heads were collected < 25 hours after eclosion.

Total RNA was extracted from fly heads using Trizol (Invitrogen-Life Technologies). Labeling and hybridizations were done according to standard protocols 70 using printed oligonucleotide arrays 70. Spot fluorescence ratios were normalized using a lowess fit per subarray 72. Each set (e.g. w^m4e vs. Oregon-R-S, or w^m4;Su(var)3-9⁰¹/+ vs. w^m4;+/TM3, Sb Ser) consisted of four hybridizations: two biological replicates were each done in a technical dye-swap fashion. Replicates were combined into weighted average ratio and confidence level (P-value) was calculated per gene using an error model 73, which was fine-tuned by self-self hybridizations.

Total RNA was DNAse treated and reverse transcribed (Invitrogen, ThermoScript RT-PCR System for first-strand cDNA synthesis). qPCRs were run using TaqMan chemistry on a BioRad DNA Engine Peltier thermal cycler. Primers and probe sequences are available upon request. Expression levels of each gene were normalized to Ide, a housekeeping gene located on 3L. For each of the genotypes w^m4e, w^m51b and Oregon-R-S five RNA isolations (20 fly heads per isolation) were done, and normalized expression ratios were calculated for randomly selected w^m/Oregon-R-S pairs. The resulting five ratios were averaged and are plotted in Figure 3D.

DamID was done as described 42 with minor modifications. Detailed protocols are available at 74. For each microarray hybridization the experimental and reference sample consisted of five heads each. Each experiment consisted of four hybridizations using biologically independent samples. 1 μg of amplified methylated DNA was labeled with ULS Cy-dyes (Kreatech, Amsterdam, The Netherlands). Dam-HP1 and Dam-only methylated DNA was co-hybridized in a two-color design on a custom-designed 4 × 44 K Agilent microarray. The Dam-HP1/Dam-only methylation ratio represents the level of HP1-targeted methylation, corrected for local differences in chromatin accessibility 41. Probe sequences are based on the complete Drosophila melanogaster genome sequence, release 5, downloaded from 75 on March 7, 2007. Median probe spacing is 136 bp. Probe sequences do not contain GATC, the recognition sequence of Dam.

All statistical analyses were performed in the R language and environment 76. DamID data were normalized using R-packages limma 77 and vsn 78. Raw data files were loaded in R and the weight of control spots was set to zero to exclude them from the results. We did not apply background correction to the data. Data was initially normalized between different arrays (method = 'vsn') and subsequently to the mean log₂ binding ratio in the euchromatic part of 2R. Hence, a log-ratio of 0 corresponds to the mean HP1 binding level found in euchromatin. Because Dam and HP1-Dam transgenic flies were made using mini-white (w^+mC) as a marker gene, all array probes overlapping with mini-white were excluded from analysis.

DamID and expression data are available from the Gene Expression Omnibus 79, accession GSE12395.