Trithorax regulates systemic signaling during Drosophila imaginal disc regeneration

Regeneration is a complex process through which an organism replaces damaged or lost tissue. Although planarian flatworms and freshwater Hydra are capable of replacing a complete organism (Elliott and Sánchez Alvarado, 2013; Galliot, 2012) and urodele amphibians and teleost fish can replace whole appendages (Gemberling et al., 2013; McCusker and Gardiner, 2011), the regenerative capacity of mammals is restricted and decreases significantly with maturity and age. Therefore, understanding the regulatory mechanisms that permit and promote regeneration in model organisms is of great importance to the field of regenerative medicine, which seeks to enhance the regenerative capacity of human tissues.

Recent work in different model systems has begun to identify the genes and signal transduction pathways that control regeneration. However, it is not clear how tissue damage leads to the activation of these signals and expression of regeneration genes. Interestingly, specific chromatin modifiers are important for regeneration in several organisms, suggesting that chromatin modification regulates the expression of at least some regeneration genes. For example, pharmacological inhibition of histone deacetylases blocks Xenopus tail regeneration (Tseng et al., 2011), the H3K27me3 demethylase Kdm6b.1 is required for zebrafish fin regeneration (Stewart et al., 2009), the PRC1 component Bmi1 is required for a regenerative response to pancreatitis in mice (Fukuda et al., 2012), several members of the Set1/MLL family of histone methyltransferases are required for the stem cell-based regeneration that occurs in planaria (Hubert et al., 2014), the SWI/SNF component Brg1 (Smarca4 – Mouse Genome Informatics) is essential for mouse epidermal wound repair and hair regeneration (Xiong et al., 2013) and its Drosophila homolog, Brahma, is important for midgut regeneration (Jin et al., 2013). In most of these cases, however, the extent to which tissue damage induces chromatin modification and the genes regulated by these chromatin modifiers during regeneration remain unknown.

Drosophila melanogaster imaginal discs undergo wound repair and regenerative growth, replacing lost tissue and patterning (reviewed by Worley et al., 2012). These tissues are an excellent system for studying regeneration because they are a simple columnar epithelium, but they have complex patterning and fate determination that have been well characterized. Furthermore, the genetic tractability of Drosophila and plethora of available reagents are an advantage over many vertebrate models of regeneration. The recent development of genetic tools that induce tissue ablation and allow regeneration to occur in situ has enabled analysis of the complex signaling and patterning events that occur during imaginal disc repair (Bergantiños et al., 2010; Smith-Bolton et al., 2009). Here, we used these tools to identify the chromatin-modification gene trithorax (trx) (Breen and Harte, 1991; Kuzin et al., 1994) in an unbiased forward genetic screen for genes important for imaginal disc regeneration.

By analyzing regeneration in imaginal discs with reduced levels of Trx, we demonstrate that mutant damaged tissue failed to express sufficient levels of the insulin-like peptide dILP8, which delays entry into metamorphosis to allow time for regeneration to occur (Colombani et al., 2012; Garelli et al., 2012). We propose a model in which Trx is important for ensuring proper levels of Jun N-terminal kinase (JNK) signaling after tissue damage, which in turn induces expression of dilp8 (Ilp8 – FlyBase). This identification of a mechanism through which one chromatin modifier regulates the local and systemic response to wounding confirms a role for epigenetic regulation of regeneration and opens the door for further identification of regeneration genes.

To enable large-scale regeneration experiments and forward genetic screens, we use genetic tools to induce tissue damage and regeneration in the developing Drosophila wing imaginal disc (Smith-Bolton et al., 2009). By using rotund (rn)-GAL4 and tubulin-GAL80^ts to regulate expression of the pro-apoptotic gene UAS-reaper, most of the wing pouch (>94%) can be ablated in a spatially and temporally defined manner in hundreds of third instar larvae simultaneously (Smith-Bolton et al., 2009). Briefly, the animals are maintained at 18°C until 7 days after egg laying, which is early third instar. The vials are then placed in a 30°C circulating water bath for 24 h before rapidly cooling to 18°C in an ice-water bath to halt expression of UAS-reaper and returning to an 18°C incubator. The 24 h induction of UAS-reaper is sufficient for almost complete ablation of the rn-expressing cells. After ablation, the wing pouch regrows and re-patterns, producing an adult wing upon metamorphosis. Measuring the adult wing size provides a quantification of the extent of regenerative growth (Fig. 1A-F). Measuring a large population of wings determines the average amount of regenerative growth for a specific genetic background.

Using this system, we designed an unbiased forward genetic screen to identify novel and unpredicted regeneration genes that regulate all steps of regeneration, including wound closure, blastema formation, growth, repatterning and the whole-animal developmental arrest that occurs during imaginal disc regeneration. To identify loci that regulate regeneration, screening conditions were established such that discs in most lines in an isogenic collection of mutations regenerated a moderate amount, producing wings that were smaller than undamaged wings (Fig. 1G; Smith-Bolton et al., 2009). Mutations that led to a consistent increase or decrease in average regenerated wing size could then be isolated. Two pilot dominant-modifier genetic screens (Smith-Bolton et al., 2009) that used isogenic deficiencies (Ryder et al., 2007) and a collection of growth-control mutants in an isogenic background (Tapon et al., 2002) successfully identified mutations that impaired or promoted tissue regeneration.

These initial screens isolated a deficiency, Df(3R)ED5644, that impaired regeneration as assessed by adult wing size when compared with other isogenic deficiencies (Smith-Bolton et al., 2009; Fig. 1G). In screening conditions, most adult wings from damaged and regenerated Df(3R)ED5644/+ wing discs were 25% the size of a normal wing, whereas isogenic deficiencies that did not impact regeneration resulted in adult wings that were most often ≥50% the size of a normal wing (Fig. 1G; Smith-Bolton et al., 2009).

To identify the gene responsible for the Df(3R)ED5644/+ phenotype, we tested smaller deficiencies and mutations in candidate genes within the region, including trx and suppressor of Hairy wing [su(Hw)]. As an additional control in all experiments, we compared regeneration in these mutant lines with regeneration in the commonly used control line w¹¹¹⁸ (Hazelrigg et al., 1984), because it was comparable to the isogenic deficiencies that did not impact regeneration in our pilot screen (data not shown). A strong hypomorphic allele of trx, trx^E2 (Kennison and Tamkun, 1988), consistently regenerated worse than w¹¹¹⁸ (Fig. 1H; Fig. S1A).

To confirm our semi-quantitative screen findings, we imaged all wings and calculated the area per wing to quantify the difference in size between trx^E2/+ and control regenerated wings (Fig. 1I). To determine whether the difference in size was the result of fewer cells or smaller cells, we calculated cell density by counting the actin-rich trichomes that protrude from one vertex of each cell. Cell density was not different between the control and the mutant (Fig. 1J). To confirm that the poor regeneration phenotype observed in the trx^E2/+ mutants was indeed the result of a reduction in Trx levels, we used RNAi to knock down trx in the rn-expressing cells that survived ablation. We used two independently generated RNAi transgenes targeting trx: JF01557 from the Transgenic RNAi Project (TRiP) collection (Ni et al., 2009), which we used in conjunction with UASdicer2 as recommended, and KK108122 from the Vienna Drosophila Resource Center (VDRC) (Dietzl et al., 2007), which has been confirmed to knock down trx in imaginal wing discs (Mohan et al., 2011). Expression of either RNAi in the subset of blastema cells that expressed rn yet survived ablation impaired regeneration (Fig. S1B,C).

Trx acts in the TAC1 complex to methylate Lysine 4 of Histone 3 (H3K4) and is required for the expression of homeotic genes during development (reviewed by Grimaud et al., 2006). Although Trx has been linked to H3K4 trimethylation, most H3K4 trimethylation is carried out by set1, with some contribution by trithorax related (trr) (Ardehali et al., 2011; Hallson et al., 2012). Loss of trx does not lead to a visible decrease in H3K4me3 levels in imaginal discs (Kanda et al., 2013). Indeed, recent work has suggested that Trx is a monomethyltransferase (Tie et al., 2014).

There are two possible explanations for the poor regeneration in trx^E2/+ animals. Regeneration might require a global relaxation of chromatin-mediated gene silencing, as is thought to underlie the aberrant cell fate changes called transdetermination that can follow tissue damage (Lee et al., 2005). Alternatively, chromatin changes might occur only at specific genes that must be induced or silenced to facilitate regeneration, as can occur during developmental patterning and growth control (Classen et al., 2009; Oktaba et al., 2008). Indeed, chromatin remodeling regulates specific regeneration genes after tissue damage, including notch1 and bmp2 in Xenopus tails, Shh in mouse skin and dlx4a in zebrafish fins (Stewart et al., 2009; Tseng et al., 2011; Xiong et al., 2013).

To distinguish between these possibilities, we examined expression of ultrabithorax (ubx), which is normally expressed in the haltere and third thoracic leg discs but is silenced in the wing disc, except when chromatin-mediated silencing is disrupted (Fig. S2A,B; Glicksman and Brower, 1988; Wang et al., 2010). ubx was not expressed in damaged wing discs, indicating that tissue damage did not alleviate silencing at this locus (Fig. S2C-F). In addition, we examined overall levels of H3K4me3 and Histone 3 Lysine 27 (H3K27) trimethylation, which can indicate chromatin state. Changes in overall methylation levels can be detected in imaginal discs by immunostaining (Hallson et al., 2012). Immunostaining did not show overall increases or decreases in H3K4me3 or H3K27me3 in regenerating wing discs 24 h after tissue damage (recovery time 24 or R24; Fig. S2G-N). A previous report has noted increased H3K4me3 by immunostaining after induction of sporadic cell death in wing imaginal discs (Herrera and Morata, 2014) but did not specify when the increases were observed or whether they were observed in dying or surviving cells. Our results suggest that tissue damage did not induce global changes in levels of these histone methylation marks or the deregulation of an epigenetically controlled homeotic gene. Thus, Trx likely regulates chromatin at specific loci to regulate discrete processes that occur upon wing disc injury.

To quantify gene expression during regeneration, we identified two genes, gapdh2 (Tso et al., 1985) and CG12703, used as controls in previous studies of imaginal disc growth (Classen et al., 2009), that did not change relative expression after tissue damage (Fig. S3A). Using these reference controls and mRNA from whole wing imaginal discs, we detected elevated relative expression of genes that are upregulated after tissue damage, such as puckered and cabut (Blanco et al., 2010), or expressed only outside the ablation zone, such as teashirt (Wu and Cohen, 2002; Fig. S3B-D). Interestingly, trx itself was not upregulated in the damaged and regenerating wing discs (Fig. S3E).

Our analysis of Trx in regeneration has been carried out in trx^E2/+ tissue, because homozygous trx mutant animals are embryonic lethal (Kennison and Tamkun, 1988) and mitotic clones of trx homozygous tissue in imaginal discs fail to grow and are eliminated through programmed cell death (Kanda et al., 2013). The trx^E2 allele has been used to assess the role of Trx in a variety of processes (Kanda et al., 2013; Klymenko and Müller, 2004; Maurange and Paro, 2002). Our genetic analysis, comparing this allele with a weaker trx allele (trx¹) and a chromosomal deficiency, suggests that it is a very strong hypomorph (Fig. S4). Thus, we sought to identify how regeneration was impaired in the trx^E2/+ mutant tissue.

We examined trx^E2/+ regenerating wing discs for a regeneration blastema, which is the zone of proliferating cells that forms by R24 (Kiehle and Schubiger, 1985; Smith-Bolton et al., 2009). The signaling molecule Wingless (Wg) is expressed in the regeneration blastema when damage is caused by a cut or by tissue ablation (Gibson and Schubiger, 1999; Schubiger et al., 2010; Smith-Bolton et al., 2009). Wg was expressed in the trx^E2/+ regenerating tissue (Fig. 2A-C). Furthermore, marking cells in S phase using EdU incorporation demonstrated that a blastema formed in the trx^E2/+ damaged wing discs (Fig. 2D-F). The intensity of 5-ethynyl-2′-deoxyuridine (EdU) immunostaining was not different between control and trx^E2/+ regenerating discs at R24 (Fig. 2G). We also used phospho-histone H3 to mark mitotic cells (Hendzel et al., 1997). At R24, we found no difference in the number of mitotic cells per area between control and trx^E2/+ regenerating discs (Fig. 2H-K). Although increases were detected in EdU incorporation in R48 trx^E2/+ regenerating discs both within and outside of the blastema, following a peak of cyclinE (cycE) expression (Richardson et al., 1993) at R24 (Fig. S5A-G), no increases were detected in the frequencies of cells in mitosis or in the expression of cyclinA (Lehner and O'Farrell, 1989) and cyclinB (Lehner and O'Farrell, 1990; Fig. S5H-M). Thus, the mutant tissue formed a blastema that appeared to proliferate appropriately.

To confirm that the wing primordium was regrowing at the same rate in control and trx^E2/+ regenerating discs, tissue size was compared by measuring the area of the wing disc that expressed the wing primordium marker nubbin (Ng et al., 1995). Average wing primordium size was not different between control and trx^E2/+ regenerating discs at R24 (Fig. 2L-N) or R48 (Fig. 2O-Q).

Given that the wing primordia in trx^E2/+ and control regenerating discs at R48 were the same size, the deficit in regeneration in the trx^E2/+ tissue presumably occurred after this time. Immunostaining for cleaved Caspase 3 at R48 stained only pockets of debris, marked by absence of Nubbin and whole nuclei, and did not show any regenerated tissue undergoing apoptosis in trx/+ discs (Fig. 3A-H). Therefore, the trx^E2/+ regenerating tissue was not lost via apoptosis.

Tissue damage in imaginal discs activates a developmental checkpoint, which induces a delay in pupariation. This developmental delay requires retinoid synthesis and the insulin-like peptide dILP8 (Colombani et al., 2012; Garelli et al., 2012; Halme et al., 2010). To determine whether trx^E2/+ regenerating animals activated the developmental checkpoint, we quantified pupariation rates in undamaged and regenerating animals. Although normally developing trx^E2/+ and control animals pupariated at the same rate, trx^E2/+ animals with damaged wing discs appeared to pupariate approximately 1 day before control animals with damaged wing discs when new pupae were counted once a day (Fig. 3I). We confirmed this premature pupariation by quantifying the number of animals that had pupariated every 12 h (Fig. 3J). Therefore, although these two genotypes had regenerated to a similar extent at R48, the trx^E2/+ animals entered pupariation, while the control animals continued to regenerate for approximately 1 day more. Furthermore, use of RNAi to reduce Trx levels also led to premature pupariation (Fig. S6A,B).

These differences in timing of pupariation were due to differences in timing of Ecdysone signaling, as measured by expression of the Ecdysone-responsive gene E74 (Fig. 3K; Burtis et al., 1990). Levels of E74 expression remained low at R24 in control and trx^E2/+ larvae (Fig. 3K). However, E74 expression had peaked at R48 in trx^E2/+ larvae, while E74 expression in control larvae was only beginning to increase (Fig. 3K). Expression was high at R60 in control larvae, after all trx^E2/+ animals had pupariated (Fig. 3K).

We determined whether this early entry into pupariation halted regenerative proliferation by dissecting and staining regenerating wing discs at R56, when most trx^E2/+ animals had pupariated but control animals had not. The trx^E2/+ everted discs were significantly smaller than normal discs and folded abnormally, rendering them difficult to identify and image. However, PH3 immunostaining in the trx^E2/+ wing discs at R56 showed very few mitotic cells, demonstrating that regenerative proliferation had halted (Fig. S6C-E).

To determine whether this premature entry into pupariation accounted for the diminished regenerative capacity of the trx^E2/+ tissue, we artificially extended the third larval instar. Larvae fed on erg2 mutant (erg2Δ) yeast fail to make Ecdysone, leading to a delay or absence of pupariation (Bos et al., 1976; Katsuyama and Paro, 2013; Parkin and Burnet, 1986). Transferring larvae to food made of erg2Δ yeast, agar and water 2 days after egg deposition delayed pupariation. The consistency of the erg2Δ yeast food affected the length of the pupariation delay. When transferred to softer erg2Δ yeast food, made with less agar, most larvae failed to pupariate. Those that did pupariate delayed by approximately 7 days (data not shown) and had fully regenerated wings (Fig. 3L). When transferred to firmer erg2Δ yeast food, made with more agar, larger numbers of larvae pupariated after a delay of 3-7 days. The trx^E2/+ animals grown on this firmer erg2Δ yeast food had wings comparable in size to those of control animals grown on standard food (Fig. 3M), showing rescue of the regeneration defect. However, control animals grown on the firmer erg2Δ yeast food delayed pupariation longer and had fully regenerated wings (Fig. 3M).

To compare control and trx^E2/+ regenerating animals with the same pupariation timing, we collected pupae that formed during 24 h windows and assessed their wings upon eclosion. Although variation occurred in individual experiments, trx^E2/+ animals had either similar or slightly reduced adult wing size compared with control animals (Fig. 3N). Thus, extending the time to pupariation largely rescued the regeneration defect of trx^E2/+ wing discs. Other, unidentified factors might account for the small remaining difference between trx^E2/+ and control regeneration.

Given that dILP8 is expressed in damaged imaginal discs and is required for induction of the delay in pupariation (Colombani et al., 2012; Garelli et al., 2012), we examined dilp8 expression. Interestingly, dilp8 expression in trx^E2/+ regenerating discs was reduced at R24 compared with control regenerating discs (Fig. 4A).

The reduced dilp8 expression could account for the shortened time for regeneration and smaller wings in trx^E2/+ animals. Indeed, animals heterozygous for the hypomorphic allele dilp8^MI00727 (Colombani et al., 2012) showed premature pupariation and reduced adult wings after damage (Fig. 4B,C). Importantly, overexpressing dILP8 using a UAS-dilp8 transgene (Garelli et al., 2012) expressed via rn-Gal4 during the thermal shift in the rn-expressing cells that survived ablation restored proper dilp8 expression levels and rescued the premature pupariation phenotype in the trx^E2/+ regenerating animals, as well as overall wing size (Fig. 4D-F; Fig. S7A,B).

Because dilp8 had significantly reduced expression in trx^E2/+ damaged tissue, it is possible that Trx is required to modify chromatin at the dilp8 locus after wounding to alleviate silencing. However, the region around dilp8 is characterized as open chromatin in S2 cells and lacks H3K4me3 and H3K27me3 in embryos, whole larvae and adults according to data deposited in the modENCODE database (Celniker et al., 2009; Kharchenko et al., 2011). Given that H3K4 becomes methylated at the start site of many transcriptionally active genes, including dilp8 in regenerating tissue (data not shown), it is possible that Trx acts directly on the dilp8 locus after tissue damage. However, it is also possible that Trx controls expression of a regulator of dilp8, either in the disc itself or in another tissue that impacts disc growth. Changing Ecdysone levels, by feeding the larvae either 20-hydroxyecdysone or erg2Δ yeast, did not affect dilp8 expression (Fig. S7C,D). Therefore, we sought to identify a regulator of dilp8 expression in the imaginal disc that is misregulated in the trx^E2/+ regenerating tissue.

Previous reports have suggested that JNK signaling regulates dilp8 expression (Colombani et al., 2012; Katsuyama et al., 2015). Indeed, examination of the dilp8 locus in GenomeSurveyor (Kazemian et al., 2011) shows conservation of a predicted AP-1 binding site (Perkins et al., 1988) about 4 kb upstream of the dilp8 (CG14059) start site. If Trx acts through JNK signaling to regulate dILP8 expression, modification of JNK signaling might replicate or rescue the trx^E2/+ phenotype. Indeed, wing discs heterozygous mutant for the gene encoding JNK, basket (bsk) (Sluss et al., 1996), regenerated poorly as assessed by adult wing size, similar to wing discs heterozygous for trx^E2 (Fig. 5A). In addition, increasing Puc levels via rn-Gal4 and UAS-puc (Bischof et al., 2013) eliminated the normal damage-induced developmental delay and any regenerative response (Fig. 5B,C). Given the central role that JNK signaling plays in wound closure, regeneration and the damage-induced developmental delay, this complete abrogation of regeneration was not surprising.

To determine whether JNK signaling was reduced in trx^E2/+ regenerating wing discs, we assessed levels of phosphorylated JNK, expression of a transgenic reporter of JNK signaling (Chatterjee and Bohmann, 2012), and expression of the JNK signaling target gene puckered (puc) (Martín-Blanco et al., 1998). To assess levels of activated JNK, we immunostained regenerating discs using an anti-phospho-JNK antibody. In two independent experiments, phospho-JNK was significantly reduced in the trx^E2/+ regenerating discs (Fig. 5D-F).

To assess transcription downstream of JNK, we used a transgenic reporter (TRE-red) that consists of four AP-1 binding sites controlling expression of the dsRED gene (Chatterjee and Bohmann, 2012). This reporter measures activity of the JNK-dependent AP-1 transcriptional complex, which consists of Jun and Fos (Perkins et al., 1988). Expression of the TRE-red reporter mirrored that of the puc-lacZ enhancer trap after tissue damage in control imaginal discs (Fig. 5G-J). The dsRed was observed in the regenerating tissue, as well as in the cellular debris that remained adjacent to the epithelium and was occasionally present in the images. To quantify levels of dsRed in the blastema, we calculated the average pixel intensity in the blastema, which was defined by Wg expression. Levels of dsRed remained elevated throughout regeneration, owing to either continued signaling or dsRed perdurance (Fig. 5K).

To compare transcription downstream of JNK signaling in control and trx^E2/+ wounded discs, we imaged and quantified dsRed in the blastema at R24 (Fig. 5L-U). In two independent experiments, dsRed expression was significantly reduced in the trx^E2/+ regenerating tissue compared with control regenerating tissue at R24 (Fig. 5U). In a third experiment, expression was highly variable and not statistically different between the genotypes (data not shown). Given that dilp8 expression, phospho-JNK levels and the TRE-red reporter were all reduced in the mutant regeneration blastema, we concluded that JNK signaling was reduced.

As a third assessment of transcription downstream of JNK signaling, we assessed expression of the target gene puc, which encodes a phosphatase that dephosphorylates and negatively regulates JNK (Martín-Blanco et al., 1998). To our surprise, puc expression levels were significantly elevated in the trx^E2/+ regenerating tissue compared with control regenerating tissue at R24, as measured by both an enhancer trap (Ring and Martinez Arias, 1993) and qRT-PCR of puc mRNA (Fig. 6A-C). This result raised the possibility that Trx negatively regulates puc and that the elevated puc caused the reduction of JNK signaling and dilp8 expression in the trx^E2/+ mutant (Fig. 6D). Importantly, RNAi against trx also caused an increase in puc expression and a reduction in dilp8 expression (Fig. S8).

To confirm that elevated puc expression was responsible for the decrease in dilp8 expression, we quantified dilp8 expression in regenerating discs that were heterozygous mutant for both trx and puc. In these doubly heterozygous mutant animals, puc expression was similar to expression in control regenerating tissue (Fig. 6E). Importantly, dilp8 expression levels were similar to expression levels in control regenerating discs (Fig. 6F). These animals pupariated at the same time as control regenerating animals and regenerated to the same extent as control animals (Fig. 6G,H). Thus, reduction of puc in the trx^E2/+ regenerating tissue restored systemic signaling, developmental timing and regenerative capacity. Although the genetic interaction experiments alone can only suggest a regulatory relationship, the changes in puc expression, phospho-JNK levels and the AP1-dsRED reporter in the trx^E2/+ regenerating tissue, as well as the restoration of proper puc and dilp8 expression levels in the puc^E69, trx^E2/+ regenerating tissue, indicate that puc expression and JNK activity are downstream of trx. As Trx is generally thought to promote rather than repress gene expression, Trx is not likely to act directly on the puc locus in the damaged tissue. Therefore, we propose that Trx regulates expression of an unknown factor, which in turn regulates puc expression.

This work sought to identify the role that the chromatin modifier Trx plays in regulating regeneration. Our results demonstrate that the primary problem in trx heterozygous regenerating animals was insufficient time for regeneration. We propose a model in which reduced Trx levels lead to abnormally high expression of puc, which suppresses JNK signaling, which leads to insufficient dilp8 expression (Bosch et al., 2008, 2005; Colombani et al., 2012). Although JNK signaling was reduced in the trx heterozygous tissue, it was still sufficient to promote wound healing, blastema formation and regenerative growth. Thus, dilp8 expression seems more sensitive to changes in JNK signaling than other transcriptional targets of this pathway or it might require additional input from a Trx-dependent but JNK-independent mechanism (Fig. 6D). It is likely that further reduction in Trx levels or activity would impair other JNK-dependent aspects of regeneration. Thus, we have identified a Trx-dependent mechanism that controls the scope and magnitude of regeneration signaling.

It is possible that Trx regulates JNK signaling in a similar manner in other contexts. Indeed, puc expression levels appeared to be slightly increased in undamaged trx^E2/+ discs compared with controls (Fig. 6A), and phospho-JNK levels appeared to be reduced outside the regeneration blastema in trx^E2/+ discs (Fig. 5D,E). Interestingly, when a trx RNAi construct was expressed in the developing notum, adult flies had a weak notum malformation phenotype, while a bsk RNAi construct induced a moderate notum malformation phenotype, suggesting that Trx can affect the JNK-dependent process of notum fusion (Mummery-Widmer et al., 2009). Furthermore, animals homozygous for one mutant allele, trx⁰⁰³⁴⁷, are reported to have a groove in the adult notum suggestive of incomplete notum fusion (‘Insertion alleles’, communication to Flybase by Berkeley Drosophila Genome Project, 1993). To our knowledge, however, no report has implicated Trx in regulation of the JNK-dependent process of embryonic dorsal closure or described a dorsal closure phenotype for any trx allele.

We do not yet know how Trx regulates puc expression after tissue damage. The puc region contains histone modifications and binds chromatin modifiers, including members of the NURF complex and Polycomb, according to the modENCODE database (Celniker et al., 2009). Although these modENCODE data are not from isolated imaginal discs, these findings suggest that regulatory mechanisms other than AP-1 are likely to contribute to the control of puc expression throughout development. Our results caution against using puc expression as the sole indicator of JNK signaling activity. As Trx is generally thought to promote rather than repress gene expression, Trx might regulate a repressor of puc expression. Indeed, GenomeSurveyor (Kazemian et al., 2011) predicts binding sites for many transcription factors in the cis-regulatory modules defined by the AP-1 binding sites in the puc locus. Several of these predicted binding factors can act as repressors of transcription and are in turn located in genomic regions that, according to modENCODE data, contain highly modified chromatin, and thus are candidates for mediating the influence of Trx on puc. We tested mutants of the five genes that we identified as fitting these criteria: sloppy-paired 1, tramtrack, caupolican, traffic jam and earmuff (Gómez-Skarmeta et al., 1996; Grossniklaus et al., 1992; Harrison and Travers, 1990; Kawashima et al., 2003; Weng et al., 2010). None of the mutants had impaired regeneration; therefore, none is likely to be the factor that regulates puc downstream of trx (Fig. S9A-E), which remains to be identified.

We had initially predicted broad changes in gene expression in the trx mutant regenerating tissue, resulting in multiple defects throughout the regeneration process. However, expression of many of the regeneration genes we tested was not impaired (Fig. S9F,G), possibly because the trx heterozygote reduces the gene dosage only by half. Strong changes in expression were detected in only a few specific genes, such as cycE, puc and dilp8. However, it remains possible and even likely that additional genes are regulated by Trx after tissue damage that were not detectable in the heterozygous mutant. Indeed, chromatin modification could rapidly and efficiently alter the developmental program in damaged tissue to enable regeneration. Importantly, RNAi knockdown of the planarian homolog of trx, Smed-mll1/2, did not prevent formation of a regeneration blastema but did impair regeneration of particular cells (Hubert et al., 2014), suggesting that Trx and its homologs play specific roles in regeneration across species.

In summary, our unbiased genetic screen identified a chromatin-modification gene, trx, as a key regulator of regeneration. Reducing levels of Trx in damaged tissue led to reduced JNK signaling and limited dILP8 expression, resulting in a failure to complete regenerative growth before the onset of pupariation and metamorphosis. Thus, we have proposed a model in which Trx is important for regulating the expression of the phosphatase Puckered, which modulates JNK activity. We have not ruled out the possibility that Trx also regulates dilp8 directly or that it regulates additional regulators of dilp8 expression. This work has demonstrated that chromatin modification after wounding can regulate specific signaling events and expression of key genes. A genome-wide examination of changes in histone modification and chromatin state in regenerating tissue will identify more genes that are regulated epigenetically upon wounding, including novel regeneration genes that will contribute to our understanding of wound repair. Our findings also illustrate the importance of fine-tuned regulation of regeneration signaling, because changes in the magnitude or duration of feedback inhibition can significantly alter the regenerative outcome.

Ablation was induced as previously described (Smith-Bolton et al., 2009), using expression of UAS-reaper and a 24 h thermal shift to 30°C to induce cell death. The genetic screen was carried out as previously described (Smith-Bolton et al., 2009). For all experiments, ablation was induced in the early third instar, which is 7 days after egg laying at 18°C. Mock-ablated controls experienced the shift to 30°C for 24 h alongside the ablating animals, but lacked the ablation-inducing transgenes.

Flies were maintained on standard molasses-based food. Regeneration experiments were carried out on modified Bloomington standard media containing malt and 0.3% tegosept (Apex). Fly lines were obtained from the Bloomington Drosophila Stock Center, FlyORF, the Vienna Drosophila Resource Center or colleagues (see supplementary materials and methods for details).

Adult wings were mounted in Gary's Magic Mount [Canada balsam (Sigma) dissolved in methyl salicylate (Sigma)]. Images were taken on an Olympus SZX10 microscope using CellSens Dimension software with the Extended Focal Image feature. Wing area was measured in ImageJ. Samples included both females and males. Undamaged control wings were averaged to give a standard area. The area of each experimental wing was then calculated as a fraction of the standard. Trichomes were counted in ImageJ within a 50×50 pixel box.

Immunostaining was carried out as previously described (Smith-Bolton et al., 2009). Wing discs were imaged on a Zeiss LSM 510 or a Zeiss LSM 700 confocal microscope. Images were processed using ZEN lite (Zeiss), ImageJ (NIH) and Photoshop (Adobe). Details for antibodies, labeling and image analysis can be found in supplementary materials and methods.

qRT-PCR was carried out as previously described (Classen et al., 2009). Wing disc tissue was used for all experiments except the E74 qRT-PCR, in which whole larvae were used. For qRT-PCR of whole larvae, mRNA was extracted using the Omega Total RNA Kit II (VWR). Independent samples consisted of 15 wing discs or five whole larvae. Power SYBR Green Master Mix (ABI) was used and reactions were run on an ABI Step One Plus Real-Time PCR System. Analysis was done by the ΔΔC_T method and expression levels were normalized to gapdh2. Fold changes relative to control undamaged discs are shown. For all primer sequences, sources for primer sequences or sources for primers, see Table S1. The variation observed in fold changes for dilp8 at R24 was probably because expression levels in undamaged discs were extremely low.

Pupariation rates were quantified by counting newly formed pupae every 12 or 24 h. Pupariation was delayed by raising larvae on food made of erg2Δ mutant yeast, strain 4020788 (ATCC, Manassas, VA, USA). Food was prepared as previously described (Katsuyama and Paro, 2011). Soft food contained 6.5% agar. Firm food contained 7.5% agar. Eggs were laid on grape plates, from which larvae were picked on day 2 after egg laying, and transferred to erg2Δ yeast food, 50 larvae per vial. To compare regeneration in animals that pupariated at the same time, pupae formed within 24 h windows were transferred to fresh vials. 20-Hydroxyecdysone (Sigma) was fed to the larvae at a concentration of 0.6 mg/ml of food in a microfuge tube as previously described (Halme et al., 2010). Controls that were not fed 20-hydroxyecdysone were in tubes that contained vehicle (ethanol) only mixed with the food.

All statistical analyses were performed using Student's t-test in Excel except where noted that χ² tests were used. Results were considered statistically significant with P<0.02.

The authors would like to thank A. Brock and P. Newmark for critical reading of the manuscript and helpful discussions; D. Bohmann, M. Dominguez and S. Cohen for reagents; the Bloomington Drosophila Stock Center (NIH P40OD018537); the TRiP at Harvard Medical School (NIH/NIGMS R01-GM084947); the VDRC; and the Developmental Studies Hybridoma Bank (NICHD, University of Iowa).