Polycomb group gene expression in the peripodial epithelium is required for wing disc eversion

To find genes required for the peripodial EMT the Ubx-GAL4 driver was crossed to UAS-RNAi lines and pharate or eclosed adult flies screened for eversion defects. Phenotypes were categorized in increasing level of severity (Figure 1) as:

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Sce expression in peripodial cells is required for wing eversion. (A-F) Ubx > Sce.IR^V10638 pupae showing increasingly severe categories of adult eversion failure. (G) A Sce^KO MARCM pupa showing a thoracic cleft. (H) Quantification of adult eversion phenotypes. Knockdown of Sce using three different UAS RNAi lines, and three different peripodial GAL4 drivers (Ubx-GAL4, puc-GAL4, odd-GAL4) induces eversion failure phenotypes. Weak phenotype = malformed wing; Strong phenotype = thoracic clefts, single, and double eversion failure (see Table 1). (I) Overnight eversion of cultured third instar imaginal wing discs is inhibited by Ubx-GAL4 driven knockdown of Sce (Ctl, n = 80; Sce.IR, n = 77; proportion everted P = 0.0007) (J-M) Categories of partial EMT and BM breakdown in wing discs cultured for 7 hr. (J) A disc stained for ECad showing intact AJs. (K) A wing disc with a substantial hole in the PE (pe) (arrows), with the underlying disc proper epithelium (dp) showing through. (L) A disc showing no signs of BM breakdown. (M) A disc in which the BM has developed a substantial hole and is pulling away. (N-O) Quantification of 7 hr eversion results. (N) Knockdown of Sce significantly inhibits breakdown of AJs and formation of holes in the PE layer (Ctl, n = 66; Sce.IR, n = 50; proportion intact P < 0.0001). (O) Knockdown of Sce significantly inhibits BM breakdown (Ctl, n = 67; Sce.IR, n = 50; proportion intact P = 0.00845). Error bars = 95% confidence interval (Wilson score method).

1. malformed wing; the thorax is normal but one or both wings are affected in some way such as being smaller, mispositioned, or crumpled (Figure 1B).

2. thoracic cleft: both wings everted but a gap remaining in the middle of the thorax (Figure 1C);

3. single-eversion failure: one wing failed to evert, resulting in an adult lacking half a thorax (Figure 1D);

4. double-eversion failure: neither wing everted and thoracic tissue missing (Figure 1E);

5. early pupal lethal: adult structures such as wings, legs and head not discernible (Figure 1F).

As expected, knockdown of genes known to play a role in eversion such as components of the JNK (fos, slpr) and TGFβ pathways (dpp, punt, Mad) generated eversion phenotypes (data not shown) as did NetA and NetB as previously described (Manhire-Heath et al. 2013). Two other genes with highly penetrant, and phenotypically severe, eversion defects were the PcG genes, Sex combs extra (Sce) and Sex combs midleg (Scm). Knockdown of these genes had similarly strong effects. RNAi to Sce using UAS-Sce.IR^B31612 resulted in a high proportion of single and double eversion failure (18.6%, n = 113) and crumpled wings (16.8%) (Table 1; Figure 1H). Similarly, knockdown of Scm with UAS-Scm.IR^B31614 produced high levels of single and double eversion failure (80.5%, n = 41; Table 1; Figure 1). For further analysis we focused our attention on Sce.

To check for off-target effects, two other RNAi lines for Sce were tested: UAS-Sce.IR^V106328 and UAS-Sce.IR^V27465. At 29° these also produced eversion defects, though in one case (UAS-Sce.IR^V106328) the primary phenotype was early lethality (86.1%, n = 79; Table 1; Figure 1). However, subsequent tests using a temperature shift regime to restrict knockdown to a tighter developmental window, also produced a high proportion of double-eversion failures for this RNAi line (see below), suggesting that the early lethality was due to a stronger RNAi effect. Occasional eversion defects could also be generated by creating random Sce^KO mutant clones using the MARCM technique (Lee and Luo 2001) (Figure 1G).

Immunostaining confirmed that Sce was expressed ubiquitously throughout the wing disc, including the peripodial epithelium, was predominantly nuclear, and appeared relatively constant between third instar and white prepupal stages (Fig. S1A, E). As expected there was a marked reduction of Sce levels in Ubx > Sce.IR^V106328 peripodial cells (Fig. S1C’’).

We next wished to see if Sce RNAi knockdown using other peripodial GAL4 drivers could also disrupt eversion. The PE has genetically distinct subdomains and different drivers express in different regions. The Ubx-GAL4 driver has a broad expression domain throughout the central area of the PE but posterior to the anterior/posterior border, while the odd-GAL4 driver expresses in the medial anterior cells, and the puc-GAL4 driver, a reporter for JNK-activation, expresses strongly in peripodial cells nearest the stalk region (Pastor-Pareja et al. 2004; Tripura et al. 2011; Aldaz et al. 2013) (Fig. S2). Knockdown of Sce with both odd-GAL4 and puc-GAL4 produced eversion failures though the penetrance was less than for Ubx-GAL4 (Table 1; Figure 1H).

Note that although Ubx is part of the bithorax complex along with abd-A and Abd-B, and that region is known to be regulated by PcG repression, our TaDa expression profiling showed that the Ubx locus was not affected by loss of Sce (see below) making it unlikely the Ubx-GAL4 driver was itself being affected by loss of PcG repression.

Taken together these results show that Sce is required for eversion and suggest that target genes of PcG repression must remain repressed for successful eversion to occur.

Sce RNAi affects the partial EMT of the wing discs

Since eversion is a complex multi-step process it can be affected at several stages: the initial apposition of the wing disc to the body wall, the degradation of the BM, the pEMT of the PE, the invasion of the epidermis, or the subsequent epithelial migration (Pastor-Pareja et al. 2004). Previously, we and others have found that the first steps of eversion, the pEMT and BM breakdown, can occur when discs are cultured in the presence of ecdysone (Milner 1977; Aldaz et al. 2010; Manhire-Heath et al. 2013). This provides an opportunity to determine if eversion failures are due to those early events, or later stages of the process. At 29°, eversion typically begins after 6-7 hr of culturing and is complete by 9-10 hr. To obtain an overall readout of eversion success we cultured discs for >16 hr, a period long enough to ensure complete eversion. Under these conditions we have found discs fall into three categories (Golenkina et al. 2021):

1. successfully everted. discs that have flattened, wing-like morphologies and the PE forms a disorganised clump;

2. partially everted. discs show evidence of breakdown of the PE but have not flattened out,

3. uneverted. discs show no evidence of PE and BM breakdown although the DP may have undergone some bending.

When Ubx > Sce.IR ^V106328 discs (hereafter Sce.IR discs) were cultured overnight there was a significant change in eversion outcomes. The proportion of discs that were uneverted increased from 28.8% (n = 80) to 45.45% (n = 77) (P = 0.0007), while successful eversion fell by half, from 57.5 to 29.87% (Figure 1I).

Next, we looked at discs after 7 hr of culturing, which, at 29°, is a time when most discs are initiating epithelial dissociation by dismantling their AJs and are breaking down their BM. Discs were fixed and immunostained for E-Cadherin, Rhodamine-Phalloidin, and anti-Laminin to label AJs, F-Actin and BMs, respectively (Figure 1J-M). The 7hr results were consistent with overnight eversion. In control discs only 9.1% (n = 66) of discs showed an intact AJs compared to 52% (n = 50) in Sce.IR discs (P = 0.0001) – the remaining discs showing either a loss of AJs or small to large perforations in the PE (Figure 1N). Similarly, the proportion of discs with an intact BM was doubled from 32.8% of control discs to 58% of Sce.IR discs (P = 0.0085) (Figure 1O). Thus, there was overall inhibition of these processes in Sce.IR discs but no other obvious qualitative differences were detected.

Next, we tested whether two other key events in wing eversion were affected by loss of Sce: activation of the JNK pathway (Martín-Blanco et al. 2000; Pastor-Pareja et al. 2004; Srivastava et al. 2007), and downregulation of the Netrin receptor Frazzled (Manhire-Heath et al. 2013). However, expression of the JNK reporter Tre-RFP and Frazzled appeared normal (Fig. S3) suggesting that whatever genes were being misregulated, they were not involved in these pathways.

Targeted DamID identifies de-repression of a small set of genes

To find which genes were affected we used Targeted DAMID (TaDa) with Dam-Pol II (Southall et al. 2013) to examine the change in transcriptional profile when Sce was knocked down. UAS-mCherry-Dam-Pol II and UAS-mCherry-Dam were expressed in control and Sce.IR discs and the ratios between Dam-Pol II and Dam profiles calculated (see Materials and Methods). Reproducibility between replicates was good with pair-wise Pearson correlation coefficients for GATC values over the genome between replicates ranging from 0.52-0.7 for control discs and 0.59-0.68 for Sce.IR discs.

We determined the list of genes that were significantly expressed in both genotypes (FDR < 0.01; see Materials and Methods) and those whose expression was significantly increased or decreased in Sce.IR discs compared to control discs. 17 genes were significantly increased in Sce.IR discs (hereafter “de-repressed”) (Table 2; Figure 2; Fig. S4; Supp. File 1). This list included the well-known PcG targets abd-A and Abd-B. 110 genes showed a significant reduction in expression, with a fold-change of >1.3 (Supp. File 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Sce.IR derepression loci are also Dam-Pc binding sites. (A-E) Genome browser views of major de-repression loci. Traces show fold-change of Dam-Pol II fusion over Dam alone, and Dam-Pc fusion over Dam alone for control discs and Ubx > Sce.IR discs. The bithorax region encompassing abd-A and Abd-B (A) is markedly derepressed in Sce.IR discs. These regions also show clear binding of Dam-Pc indicating that they are regions of PcG repression. Note that Ubx expression is unaffected by loss of Sce. Similarly, loci for caudal (B), nubbin (C), mirror (D) and ocelliless (E) show varying degrees of increased Dam-Pol II binding but are all clearly regions of Dam-Pc binding.

We performed Gene Ontology enrichment analysis on the lists of significantly changed genes, and on significantly expressed genes in the two genotypes (see Materials and Methods). For the 17 de-repressed genes, the most significant terms for biological function are “epithelium development” (10/17 genes; P = 0.001178; Holm-Bonferroni correction used for all enrichment analysis; Table 2) and “anatomical structure morphogenesis” (12/17 genes; P = 4.3e-4). There is also significant enrichment of genes with molecular function of transcription factors (8/17 genes; P = 4.37e-4), seven of which contain homeodomains. In contrast, for genes whose expression significantly decreased in Sce.IR discs there is no GO Term enrichment in any category.

Similar results were obtained when analysis was expanded to the entire set of significantly expressed genes in the two genotypes. The most strongly enriched biological function in Sce.IR discs is “epithelium development” (333/2045 genes; P = 1.84e-45), whereas for control discs it is “cellular-metabolic-process” (917/1898 genes; P = 2.8e-11). In Sce.IR discs there is also an enrichment of “cellular component” for cell junction proteins (62 genes; P = 2.0e-11) and of “molecular function” for actin binding (51 genes; P = 3.1e-7) consistent with cellular changes impacting upon pEMT processes.

Since direct targets of the PcG complexes would be expected to have increased expression we focused our attention on the 17 de-repressed genes. Changes in expression levels for these genes, averaged across the whole gene locus were relatively modest, ranging from 0.61 to 0.047 log₂ (i.e., fold-change of 1.5 to 1.03) averaged over the gene locus.

To confirm that these genes corresponded to regions of PcG repression we again used TaDa to examine the binding profile of the PcG component, Polycomb using UAS-myr-GFP-Dam-Polycomb and a UAS-myr-GFP-Dam control (Materials and Methods). The Dam-Pc ratio profile exhibited the expected genomic patterns of Polycomb binding for known PcG target areas, such as the engrailed/invected and the bithorax regions (Tolhuis et al. 2006) indicating that the method had worked. For each of the 17 genes we then calculated the average level of Pc-binding in control discs (Fig. S4B). The genes with the most significant fold-change in Sce.IR discs vs. controls (FDR < 1e-4) (Figure 2) also tended to have higher levels of Pc-binding (Fig. S4). We also examined the Dam-Pc profile in Sce.IR discs but found the pattern of binding largely unchanged from control discs though the average ratio levels across the genome were reduced (Figure 2; Fig. S4; and data not shown).

Thus, the loss of Sce has resulted in increased expression of a small number of genes in PcG-repression regions, and this is accompanied by a genome-wide change in genes from those associated with cellular metabolism to those involved in epithelial development, consistent with an inhibition of the PE pEMT.

Abd-B is upregulated in the peripodial epithelium of Sce.IR discs and required for eversion failure

Based on the expression profiles of the de-repressed genes, we conducted further tests on four of the genes that had a distinct change in expression profile and higher levels of Pc-binding: abd-A, Abd-B, cad and nub.

We first used immunostaining to determine if any of the four genes showed significant upregulation in the PE of Sce.IR discs. Of the four genes, only Abd-B showed a clear change in expression in PE cells with nuclear staining apparent in the Sce.IR discs but not in control discs (Figure 3A-F). We further confirmed that loss of Sce was responsible for Abd-B upregulation by examining MARCM clones for the null allele Sce^KO. Clones in both the PE and DP showed clear upregulation of Abd-B (Figure 3G-I). In addition, there was a morphological change in both PE and DP clones in that they showed a “segregation-phenotype” whereby they became more rounded and developed furrowing/invagination at the borders with wild type cells as previously reported for several PcG genes (Beuchle et al. 2001; Fritsch et al. 2003; Gandille et al. 2010; Curt et al. 2013).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Sce represses Abd-B. (A-F) Third instar wing discs stained for Abd-B and E-Cadh. In control discs (A-C) no expression of Abd-B is detected. In Sce.IR discs, Abd-B is expressed in nuclei throughout the PE (E, arrowhead), and the tracheal branch on the side of the discs, opposite to the PE (F, arrowhead). (G-I) MARCM clones of Sce show clear upregulation of Abd-B in both the disc proper (H’’, arrowhead; I’, arrowhead) and PE (I’, arrowhead). Sce expression is clearly lost from MARCM clones (H’). MARCM clones show distinct morphological changes with invaginations at the boundary with surrounding heterozygous cells (H, arrows), and a smoother, more rounded profile in the PE (I, green).

Although no obvious change in abd-A, Nub or Cad expression/localization was seen in Ubx > Sce.IR PE cells, a subset of Sce^KO MARCM clones also showed clear upregulation of abd-A, though the levels were variable (Figure 4). We speculate that while Abd-B is directly controlled by PcG complexes, abd-A faces more complex regulation and may be being suppressed by Abd-B and/or the non-coding RNA mir-iab-8 which is also located in the de-repressed region between abd-A and Abd-B. In the case of Nub and Cad there was no nuclear expression though we cannot discount the possibility of a mild increase in cytoplasmic signal.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

abd-A is partially repressed by Sce. (A-F) Third instar wing discs stained for abd-B and E-Cadh. In control discs (A-B) there is no nuclear expression of abd-B though some cytoplasmic staining in PE cells was apparent. (C-D) Ubx > Sce.IR discs, appeared the same, though the cytoplasmic staining appeared somewhat stronger. (E-F) In Sce MARCM discs there was clearly some nuclear expression of abd-A in some clones (E, E’, F, F’, arrows) though this was of varying strength within a clone (F’, arrow), and some clones showed no expression (F’, arrowhead).

Next, we tested whether RNAi knockdown of any of the four genes could suppress the Sce.IR phenotypes. We utilized the Sce.IR ^V106328 RNAi line but used a temperature shift regime to restrict the period of GAL4 expression to third instar stages, thereby avoiding the excessive early pupal lethality. Two independent RNAi lines were used for each gene (Figure 5). Knockdown of any of the four genes was able to partly rescue the defects while co-expression of an arbitrary UAS construct, UAS-GFP, had no effect (normal progeny = 4.8%, n = 165, P = 0.65). Of the four genes loss of Abd-B had the strongest effect increasing the proportion of normal eversion from 4% in Sce.IR discs (n = 379) to 47.15% in Sce.IR;Abd-B.IR discs (n = 397, Figure 5; Table 3; P < 0.0001). The results suggest that the inhibition of eversion may not be due to any one of these genes, but rather to a genome-wide change in transcriptional profile toward an epithelial state. The other implication is that the maintenance of epithelial/BM integrity in Sce.IR discs is relatively unstable, since knockdown of any of the four PcG targets was enough to substantially restore successful eversion.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Knockdown of de-repressed loci substantially represses Sce.IR eversion phenotypes. Effects on adult eversion failure when Ubx-GAL4 knockdown of Sce is accompanied by expression of the indicated UAS RNAi lines, or UAS-GFP control. Co-expression of GFP does not significantly decrease the rates of eversion failure in Ubx > Sce.IR discs, but co-expression of UAS RNAi lines for Abd-B, abd-A, cad, and nub all repress eversion failure. Ubx-GAL4 expression of Abd-Bm produces a high proportion of weak phenotypes in which the thorax is normal, but wings are deformed or mispositioned (56%, n = 222). Expression of abd-A has no effect (n = 84). Error bars = 95% confidence interval (Wilson score method).

Finally, we tested whether over-expression of either of the two genes with strongest rescue, Abd-B and abd-A, could phenocopy loss of Sce. Ubx-GAL4-driven expression of Abd-B in the PE did not block eversion, though a high proportion of adults had reduced/misplaced wings (Figures 5, 6A-C). Clonal expression of Abd-B did, however, recapitulate the epithelial invagination/segregation phenotype, as has previously been described (Gandille et al. 2010) (Figure 6F-I). Ubx-GAL4-driver expression of abd-A had no effect, however clonal expression of abd-A also created invaginations suggesting that this phenotype is a conserved ability of Hox genes to regulate epithelial morphology (data not shown). Since sole expression of Abd-B was not able to recapitulate the Sce.IR phenotypes we conclude that while the epithelial morphology changes induced by Abd-B, and to a lesser extent, abd-A, may contribute to eversion failure, they are not sufficient.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Abd-B affects epithelial cells and wing morphology. (A-C) Pharate adults showing eversion phenotypes. The primary phenotype was for one or more malformed and mispositioned wings (A’, A’’). (B) An adult with both wings affected, and legs malformed. (C) Adult with a more severe phenotype in which the thorax is disrupted and only one wing has everted, but is malformed. (D) Expression of Abd-Bm with the Ubx-GAL4 driver induces strong expression of Abd-B throughout the PE (D’) but some regions of expression in the DP cells in the wing blade also robustly express Abd-B. Wing discs show a range of morphological disruption ranging from mild depressions (D, arrow) to more substantial folding (E, arrows). (F-I) Clonal expression of Abd-Bm creates regions of epithelial invagination. (F) A small clone of Abd-Bm expressing cells has folded inwards to produce a depression. (G) Cross-section of dotted line in F’. (H) A clone of Abd-Bm expressing cells in the PE creates a depression of the underlying DP epithelium. (I) Cross-section of dotted line in H. Error bars = 95% confidence interval (Wilson score method).

Overall, our results imply that the eversion failure of Sce.IR discs is due to a genome-wide change in gene expression toward an epithelial state, and that Abd-B likely plays the major role in this change.