We used the modular GAL4-UAS mis-expression system [10] to identify endogenous genes that could modify the severity of bw^D silencing when over-expressed in the eye (Figure 1A). We mobilized the mis-expression transposons P[EP] and P[LA], both of which contain a GAL4-dependent promoter at one end of the element that transcribes into flanking DNA sequences [10], [17]. New insertions were combined with the eye-specific GAL4 source GMRGAL and bw^D to test for effects on heterochromatic silencing, and adults with increased or decreased eye color were retained. We categorized pigmentation of the eye on a scale of 1 through 6, where silencing from the bw^D allele with no mis-expression insertion was assigned a score of 2, and full pigmentation in bw⁺ adults was a score of 6 (Figure 1B). Insertions with enhanced silencing were assigned a score of 1, and insertions with de-repressed silencing were ranked 3–6 depending on the extent of de-repression.

We recovered 28 P[EP] modifying insertion lines and 23 P[LA] insertion lines from ∼1100 fertile individual crosses (Table 1). 45 lines showed de-repression of silencing, and 9 lines showed enhanced silencing. 7 of these lines had effects on eye morphology, but changes in silencing were clear even in these cases where the eye was rough. We used inverse PCR to identify the location of the transposon in each line (Table S1). A single gene could not be identified in most P[EP] lines as multiple insertions were present, in part because these lines still carried the donor insertion. The 4 lines with single insertions that could be identified were retained (Table 1; Table S1), and other lines discarded. Inverse PCR successfully identified 22 of the P[LA] insertions. We also tested candidate over-expression lines from public stock centers that we selected based on molecular pathways known to be involved in heterochromatic gene-silencing, or implicated by hits in our screens (Table 1; Table S2). Each line was verified as requiring GAL4 induction of the flanking genomic sequence for effects on bw^D silencing.

Altogether, we identified 36 genes that act as over-expression modifiers of bw^D. These include 7 genes that are known to be required for heterochromatic gene-silencing from previous studies with null alleles (psq [18], Ubp64E [19], ASF1 [20], Acf1 [21], xnp [22], [23], Rm62 [24], and vig2 [25]). Some of these factors have also been implicated in Polycomb-dependent silencing, suggesting that this screen may identify factors that can affect multiple levels of chromatin structure and gene expression. As these factors affect silencing when over-expressed, caution is necessary in inferring their normal functions. Indeed, we noted that in many cases over-expression of a factor had similar effects on silencing as null alleles for that factor, suggesting that their effect is not simply due to increased dosage of the factor. We group these factors according to their annotated biological function below.

We tested our insertions with a series of additional assays. We first determined if over-expression modifiers have general effects on heterochromatic silencing, or are limited to bw^D-mediated silencing. As an independent test of silencing, we used the inversion In(1)w^m4 (w^m4) where the w⁺ gene is juxtaposed to pericentric heterochromatin. Many mis-expression constructs carry a w⁺ marker and cannot be assayed with w^m4. However, the P[LA] element we used is marked with y⁺; thus we could induce our lines using GMRGAL in combination with w^m4 and then assess effects on white silencing. We found that all 8 enhancers of bw^D silencing are also enhancers of w^m4 silencing (Table 2). This implies that these factors do indeed have general effects on heterochromatin. The effects of bw^D de-repressors are more variable. Only 1 line de-represses both bw^D and w^m4, and 9 lines have no effect on w^m4. Surprisingly, 2 lines de-repress bw^D but enhance w^m4. Previous studies have also found that the bw^D and w^m4 rearrangements are not equivalently affected by all modifiers of heterochromatic silencing [35]. These differences suggest that each chromosome rearrangement has a unique combination of gene regulatory elements and heterochromatic sequences that determine the extent of silencing. The testing of bw^D modifiers for effects on silencing of w^m4 is informative, as it reveals that enhancers are general, yet de-repressors are not. The silencing of bw⁺ by bw^D provides a sensitive assay for multiple levels of chromosomal organization. Over-expression lines that perturb bw^D silencing yet have no effect on w^m4 may affect bw⁺ regulation, chromosome pairing, or heterochromatic aggregation, all of which are required for silencing in trans.

Functional RNAi systems are required for heterochromatic silencing in eukaryotes. Nuclear complexes containing small RNAs are thought to target histone modifications to homologous repetitive sequences, and to promote the retention and subsequent degradation of nascent transcripts from those repeats [36]. In Drosophila, the RNA endonuclease Dcr2 is required for both production of post-transcriptional silencing small RNAs and for small RNAs derived from transposable elements in the genome [37]. Genetic evidence in Drosophila indicates a link between RNAi and heterochromatic silencing as well [38]–[40], although the mechanism of how RNAi is converted into chromatin structure has not been detailed. We recovered 10 lines that are implicated in RNA metabolism in our screen (Table 1), and we tested whether over-expression lines might affect bw^D silencing by inhibiting RNAi. We used a hairpin construct that eliminates w⁺ message (GMR-wIR) to assay the effectiveness of RNAi [41]. GMR-wIR eliminates almost all pigmentation in the eye, and this effect requires Dcr2 activity. As the hairpin construct and the over-expressed modifiers are both induced from GMR promoters, the double-strand RNA for knockdown and the modifier are expressed at the same time starting late in the development of the eye.

There were no significant effects of over-expression lines on the extent of w⁺ knockdown by GMR-wIR, although 3 lines showed a slight increase in eye pigmentation, and 4 lines had a slight decrease (Table 1, Figure S1). In these 7 lines the mis-expression P elements carry a w⁺ marker, and the marginal effects on knockdown may be due to different levels of w⁺ expression. Regardless, the severe effect of these insertions on heterochromatic silencing with little effect on knockdown implies that these factors do not affect silencing by altering RNAi.

The GMRGAL driver we used in our screen produces high levels of GAL4 late in eye development, immediately before the last S phase and cell division of pigment cells in the 3^rd instar imaginal disc [42]. Previous experiments have indicated that heterochromatic silencing varies during the development of the eye [43]. We used the modular nature of the GAL4-UAS system to test if continual production of factors in the eye would also alter silencing. The eyGAL driver produces moderate levels of GAL4 in the eye primordium starting in embryogenesis, and shuts off just before the last cell division in the developing eye [44]. We anticipated that factors may only be effective when expressed with GMRGAL if they are required at high levels to modify silencing. Indeed, we found that the majority (29/34) of lines have no effect on silencing when induced by eyGAL (Table 1).

Only five factors affected bw^D silencing when induced by eyGAL4 (Table 1). Induction of the chromatin remodelers xnp and Ino80 have quantitatively similar effects on silencing whether they are induced by GMRGAL or by eyGAL. This implies that moderate expression of these factors is sufficient for their effect. GMRGAL induction of ACF1 has a dramatic de-repression of silencing, but de-repression is more moderate with eyGAL, suggesting that amounts of ACF1 are limiting for de-repression. Late induction of vig2 enhances silencing, but early induction de-represses silencing. Vig2 is normally produced early in development and may promote the formation of heterochromatin, while the related Vig protein may take over its functions in later development [25]. Perhaps early over-expression of Vig2 interferes with function in early development, while later expression interferes with Vig function. Finally, crol is an exceptional case, because over-expression with eyGAL gives more dramatic de-repression than induction by GMRGAL. This line is examined in more detail below.

The effects of early over-expression on silencing could not be determined for 5 lines that are lethal or severely distort the eye in combination with eyGAL (Table 1). These appear to be cases where continuous expression is toxic to cells. It is notable that toxic effects are infrequent in this collection of modifiers. Indeed, when we ubiquitously induced modifier lines with the A5CGAL driver, 19 had no effect on viability (Table 1). This includes seven lines with strong de-repressive effects on silencing when induced by GMRGAL. This is consistent with the observation that some modifiers of heterochromatin silencing are largely dispensable for viability in Drosophila [45]. Lines that are lethal when constitutively expressed are likely to have more general effects on chromatin regulation.

The crol transcription factor is one of the few factors tested that de-represses silencing when continually expressed in the eye (Table 1). Strikingly, continual expression of crol changes the pattern of silencing in bw^D/bw⁺ animals (Figure 2A). The bw^D allele normally causes speckled variegation of bw⁺ that is thought to result from the disruption and re-establishment of inter-chromosome interactions every cell cycle as the eye grows [15], [16]. However, bw^D/bw⁺ animals with continual expression of crol frequently have de-repressed sectors in the eye. Most animals with crol expression show one or more sectors, implying that de-repression is frequent in this genotype (Figure 2B). Sectors appear in a speckled background, implying that bw^D silencing remains severe for some cells.

We verified that de-repressed sectors were due to crol expression using an independent over-expression insertion line (d03228, Exelexis Stock Center). Furthermore, increased crol expression with two eyGAL4 drivers also increases the frequency of de-repressed sectors. Continual expression of crol also de-represses w^m4 (Figure 3), indicating that this factor can generally modify heterochromatic silencing. Late induction of crol moderately de-represses silencing with bw^D and has little or no effect with w^m4, implying that the timing of crol expression during development is important for de-repression.

The sectored pattern of variegation suggested that continual crol expression causes somatically heritable de-repression. To test this idea, we reduced the strength of crol over-expression by raising animals at 18°C, where GAL4 is less effective as an activator [46]. Indeed, raising animals at 18° completely blocks the appearance of de-repressed sectors (Figure 4A–4C). This allows us to use temperature shift experiments to determine the developmental timing when crol causes de-repression. We found that animals raised at 18° for early development and then shifted to 25° showed reduced de-repression (Figure 4A and 4B). Strikingly, some animals raised in this regimen showed numerous small sectors (Figure 4D), consistent with the idea that de-repression does not occur early in this regime but often occurs in later development. This idea is supported by our observation that animals shifted to 25° after 1–2 days at 18° show more de-repression than animals shifted to 25° after 3–4 days (Figure 4B). We conclude that crol-stimulated de-repression can occur sporadically throughout development.

To determine if crol over-expression is required for the establishment of de-repression, for its stable inheritance, or for both, we transiently expressed crol early in development. We raised animals at 25° for embryonic and early larval stages, and then shifted them to 18°. De-repression persisted in this temperature regimen and often appeared as a single sector in the eye (Figure 4E), demonstrating that de-repression can be maintained in the absence of crol expression. Thus, the strong effect of continual crol expression appears to result from multiple de-repression events throughout development (Figure 4F). We conclude that crol is required for de-repression, but de-repression can be maintained through cell divisions without over-expression of the factor.

How does crol de-repress silencing? We noted that GMRGAL-induced crol expression resulted in a slight roughening of the eye, suggesting that there may be proliferation defects. Indeed, over-expression of crol promotes cell division in developing wing discs [47]. We confirmed that crol over-expression also promotes cell cycle progression in eye discs. In late third instar larvae, the eye disc contains both mitotically active cells and differentiating cells, and the last two waves of cell divisions in the eye occur on either side of the morphogenetic furrow (MF; [48]). Over-expression of crol causes a substantial increase in the number of mitotic cells on both sides of the MF (Figure 5A and 5B). This is accompanied by increased cell death in these zones (Figure 5C and 5D). Previous studies have shown that increased cell proliferation induces compensatory cell death in developing imaginal discs [49]. Thus, crol induces both accelerated cell cycles and stable de-repression of silencing in the developing eye.

Acceleration of the cell cycle by crol over-expression is suppressed by mutations the mitotic regulator cyclin B (cycB; [47]). We used this to test if cell cycle acceleration by crol causes de-repression. We found that cycB² dominantly reduces de-repression by crol over-expression (Figure 5E). We conclude that de-repressed clones result from an acceleration of the cell cycle. Notably, cycB mutations have no dominant effect on bw^D silencing, demonstrating that silencing and clonal de-repression are genetically distinct processes.

If accelerated cell cycles induced by crol over-expression cause de-repression, then crol over-expression in post-mitotic cells should have no effect on silencing. The GMRGAL driver induces transgenes immediately before the last cell division in the eye, and induction of crol with this driver moderately de-represses bw^D silencing. We used the cyclin inhibitor p21 to eliminate the last division in the eye disc [50]; in this background GMRGAL induces crol after the last cell division. We found that eliminating the last cell cycle blocks the de-repressive effect of crol, confirming that crol over-expression is only effective in cycling cells (Figure 6).

A second factor we identified also implicated cell cycle progression in de-repression. The bantam microRNA promotes cell growth, and indeed, late over-expression of this factor with GMRGAL leads to both de-repression of silencing and expansion of the eye (Figure 6). To determine whether the de-repressive effects of bantam are also limited to cycling cells, we tested if de-repression could occur when p21 was also expressed. We found that eliminating the last cell cycle greatly reduces de-repression caused by bantam over-expression (Figure 6).

Finally, we tested if other over-expression modifiers are also only effective in mitotically active cells. We focused on the 9 lines that show dramatic de-repression (Rank 6 in Table 1). All 9 of these factors show dramatic de-repression when expressed either before or after the last division in the eye (Figure 6). We conclude that the silenced chromatin state is plastic, and over-expression can overcome silencing even in differentiated cells. The bantam and crol factors are exceptional, in that they are only effective in cycling cells.

Over-expression of crol generates bw⁺ clones, implying that some cells establish de-repressed early in development and their daughter cells maintain de-repression. We used inducible modifiers we recovered to test if enhancers could inhibit establishment or the maintenance of crol-mediated de-repression. Jim, CG8036, CG10630, and Ubp64E are enhancers of silencing when induced late in development, but have no effect when expressed early (Table 1). To test if these factors affect clonal de-repression, we induced each of these factors and crol early in development. While crol induction results in extensive clonal de-repression, contemporaneous expression of Jim, CG8036, or CG10630 strongly reduced clones (Figure 7A and 7B), suggesting that establishment of de-repression is more sensitive than bw⁺ expression later in development. In contrast, contemporaneous expression of crol and Ubp64E dramatically increases clonal de-repression in the eye (Figure 7B), implying that more cells sporadically switch to a de-repressed state. Thus, while the establishment of de-repression is sensitive to some modifiers of heterochromatic silencing, these appear to be genetically distinct.

The lines referred to as ‘GMRGAL’ and ‘A5CGAL’ are previously described white-deficient versions of drivers for late eye-specific and constitutive expression of GAL4, respectively [65]. For constitutive eye-specific expression, we used the P[eyGAL, w⁺]3-8 line, referred to as eyGAL [66]. For experiments with In(1)w^m4, the eyGAL driver was destabilized using TMS, P[Delta2–3] to generate a white-deficient insertion that retained eye-specific expression of GAL4.

We used st or v^36f to eliminate all ommochrome pigments from the eye. In these backgrounds, bw⁺ cells appear red, while cells with bw⁺ silencing appear white. A preliminary screen was performed using a w⁺-marked P[EP]2339 (inserted at 59E) as a donor for mutagenesis. We crossed P[EP]2339/CyO; st virgins to Dr/TMS, P[Delta2–3] males, and then crossed individual Cy⁺ Sb male progeny to GMRGAL bw^D/CyO; st females. Cy⁺ Sb⁺ progeny with enhanced or de-repressed bw⁺ silencing were recovered and mated to a w¹¹¹⁸ stock for extraction of new P[EP] insertions.

A second screen used the y⁺-marked P[LA] construct for mutagenesis. We first transposed P[LA]4 [17] onto FM7i using P transposase, and then used the resultant chromosome FM7i, P[LA]4.2 as a donor chromosome. We crossed y/FM7i, P[LA]4.2; st females to Dr/TMS, P[Delta2–3] males, and then crossed individual progeny B Sb males to y; GMR bw^D/CyO; st females. Cy⁺ Sb⁺ y⁺ males with enhanced or de-repressed bw⁺ silencing were recovered and mated to a y stock for extraction of new P[LA] insertions.

Candidate gene insertion lines were obtained from the Bloomington Drosophila Stock Center (Bloomington IN) and the Exelexis Stock Center (Boston MA) and tested for effects on bw^D silencing in v^36f; GMRGAL bw^D/+ males. To assess if effects were dependent on expression of a gene adjacent to the P element, each insertion from screens and candidate tests were re-assessed with GAL4 (GMRGAL bw^D/insertion) and without GAL4 (bw^D/insertion). Progeny from crosses were scored and photographed 3–4 days after eclosion as previously described [67].

New insertion sites were mapped using inverse PCR according to published protocols [68]. Genomic DNA from balanced lines was purified and digested using MspI or RsaI restriction enzymes, ligated, and used for PCR amplification using the following primers: P[LA] 5′ ends, LA(f).1/LA(r).1; P[LA] 3′ ends, Pry4/Pry1, or Sp6/Pry4; P[EP] 5′ ends, Pwht1/Plac1; P[EP] 3′ ends, Pry4/Pry1. Products from all 3′ ends were sequenced using the nested primer Spep1, for P[LA] 5′ends using LA(f)seq1, and for P[EP] 5′ ends using Plac1. The gene responsible for effects on heterochromatic gene-silencing was inferred to be the nearest gene down-stream of the inducible promoter.

We tested whether P[LA] modifiers of bw^D silencing altered w^m4 silencing by crossing In(1)w^m4h; GMRGAL females to each insertion line and scoring silencing in male progeny. Insertion-bearing progeny were divided into 5 ranks based on the extent of w^m4 silencing and compared to silencing in siblings carrying a dominant marker (CyO or Sco for chromosome 2 inserts, and Sb for chromosome 3 inserts). At least 40 flies were scored for each genotype, and assessed for statistical significance using Mann-Whitney U tests. To test if insertions affected RNAi-mediated gene-silencing, we crossed each insertion line to GMRGAL; P[GMR-wIR] [69] and scored w⁺ expression. To determine if modifying effects of insertions required expression in dividing cells, we used the cyclin inhibitor p21 to block cell cycle progression in the GMR expression domain of the eye [50]. We crossed each insertion line to v^36f; GMRGAL bw^D; P[GMR-p21,w⁺] and scored silencing in male progeny.

Imaginal eye-antennal discs were dissected from late 3^rd instar larvae in PBS. For detection of apoptosis, discs were incubated in 5 µg/mL acridine orange/PBS for 5 minutes, and then imaged using FITC excitation and emission filters. For detection of mitotic cells, discs were fixed with 2% formaldehyde, and stained with antisera to the mitosis marker H3-S10-phosphorylation (Millipore).