approximated encodes a DHHC palmitoyltransferase that regulates Fat signaling and the subcellular localization and activity of Dachs

approximated is required for patterning and normal PCP

Crossvein spacing and tarsal leg segmentation are extremely sensitive to changes in Ft activity; they are disrupted in weak Ft pathway mutants that have no obvious growth or PCP defects. Similar disruption occurs in app¹ homozygotes: the distance between the anterior crossvein (ACV) and posterior crossvein (PCV) is reduced (Figure 1B) and one or more tarsal leg joints are lost or reduced (Figure S1B). app¹ hemizygote wings also have weak PCP defects (Figure 1H) [12]. We used EMS to generate additional mutations that failed to complement app¹. Of these, app^e6 was semi-lethal in homozygotes and hemizygotes, and escaper adults had more extensive wing PCP defects, both proximally and in a distal region between the third and fourth longitudinal veins (Figures 1C, 1D and 1J). They also had abdominal PCP defects (Figure 1L): in the anterior compartment most hairs point in the normal, posterior direction, but polarity was disturbed around anterior-posterior (A/P) boundary (a6 and p3 in the nomenclature of [13]) and extensively reversed in the posterior compartment (p3). app^e6 appears amorphic, as the defects were not noticeably stronger in hemizygotes.

The development of PCP in the pupal wing is accompanied by the polarized redistribution of the “core” planar polarity proteins to the proximal, distal, or proximal and distal faces of single wing cells [8]. PCP mutants can be separated depending on their effects on this polarization. Mutations in the core PCP proteins reduce the levels and block the polarization of the other core PCP proteins, while changes in ft, ds or fj, expression can reorient core PCP protein polarization along inappropriate axes [14-16]. We found that the levels of the core PCP protein Flamingo (Fmi, also known as Starry night) [17, 18] were not reduced in app^e6 mutant clones (over 30 examined), and that Fmi polarization was in some cases reoriented (Figure 1M). This further supports App’s involvement in Ft signaling.

Identifying app

We mapped app to a portion of 69A2-A4 containing seven known or predicted genes (Figure S2A). app¹, app^e1, app^e3 and app^e6 all contained mutations in the 5′ coding exons of one of these, the CG5620 Flybase gene prediction (Figure 2B). We constructed a UAS-driven RNAi transgene corresponding to the 5′ end of CG5620 and expressed it in developing dorsal wings using ap-gal4; this produced app-like wing and PCP defects on the dorsal surface (Figures 1E and 1I)

However, the 3′ end of the CG5620 coding prediction is in error (Figure S2B), as our 3′RACE products (and a recent Drosophila Genome Project EST [19]) use instead the 3′ exon of the adjacent CG17144 prediction. We call this transcript app-RA and show below that the corresponding App-PA protein is produced in flies. Another EST predicts a shorter transcript we call app-RB. We did not find transcripts covering the entire final coding exon of CG5620 in embryonic or larval cDNA libraries or by RACE. However, the full CG5620 prediction is conserved in Drosophila pseudoobscura suggesting that it might be utilized; we term this app-RC (predicted).

We rescued the wing and leg defects of app homozygotes by expressing UAS-app-RA, UAS-app-RB or UAS-app-RC with either act-gal4 or en-gal4 (act-gal4 UAS-app-RA in Figures 1F and S1F). Overexpression of higher levels of UAS-app with strong drivers such as ap-gal4 or tub-gal4 also disrupted PCP in the proximal wing and abdomen (see Figure 4K).

App is a DHHC protein

app encodes a member of the DHHC protein family, responsible for adding palmitates to cytoplasmic proteins [20, 21]. Eukaryotes contain multiple members of the DHHC family, with 8 predicted in yeast, 23 in mammals [22] and 20 in Drosophila. The region common to all predicted App isoforms contains four predicted transmembrane domains, with a DHHC cysteine-rich domain (DHHC-CRD) located between the second and third transmembrane domains (Figures 2A and 2B). It is likely that the DHHC-CRD is located on the cytoplasmic side of the membrane as in the yeast DHHC protein AKR1 [23].

Alignments using the region common to the App isoforms indicate that App is in the same sub-family as the human ZDHHCs 9, 14 and 18 [22] and is the Drosophila protein most similar to yeast ERF2 [24] (Figures 2A, S2C and S3). The similarity between App and these proteins drops off at the C-terminal end of the common region (Figure S2C). The divergent C-termini of App-PA and App-PC (predicted) have no significant similarity to each other or to other proteins in the database outside the Drosophilids, except for a short region at the C-terminus of App-PA that is similar to predicted App homologs from the insects Tribolium and Apis. App-PB has a much shorter C-terminus. Since any of the putative app transcripts rescued the app leg and wing phenotypes (see above), the different C-termini are dispensable for these phenotypes.

app¹ contains a missense mutation N-terminal to the DHHC-CRD domain and an aberration that introduces a frameshift predicted to truncate the C-terminal end of App-PA (Figure 2B). app^e3 contains a missense mutation changing a conserved cysteine in the DHHC-CRD. Since the DHHC-CRD is required for palmitoyltransferase activity [25], this supports a role for palmitoylation in Ft signaling. app^e6 and app^e1 contain nonsense mutations predicted to truncate the protein prior to or towards the end of the DHHC-CRD, respectively, and therefore app^e6 is likely null for App function.

App is concentrated at the apical cell cortex but does not affect the levels or distribution of Ft or Ds

We generated two antisera, one specific for App-PA and one to the common region (Figure 2B). Both antisera uniformly stained embryos, imaginal discs and pupal wings, and staining was lost from mitotic homozygous app^e6 clones, confirming the expression of the App-PA isoform (anti-App-PA in Figures 2C-2F and S4; similar results with anti-App-common). There is no obvious asymmetric distribution of the App protein along the proximo-distal or anterior-posterior axes of imaginal discs or pupal wings. However, staining was especially strong in the apical cell cortex (Figures 2D and S4), and this concentration did not extend more basally to the adherens junction marker DE-cadherin (Figure 2E) or the septate junction marker Discs large 1 (data not shown). This is similar to the distribution of Ft and Ds [15] and there is overlap between the regions where App, Ds and Ft are concentrated (Figure 2F). This result is surprising as human ZDHHC9, 14 and 18 and yeast ERF2 are concentrated in the Golgi or ER; only a few, less similar, ZDHHCs have been detected at the plasma membrane [24, 26-28]. We did not observe any significant overlap between App and Golgi or ER markers in wing discs (data not shown). While App must traffic through the ER and Golgi, these results suggest that App is active in the plasma membrane, in or near the apical region where Ft and Ds are concentrated.

However, App does not detectably regulate Ft and Ds levels or their binding. The binding between Ft and Ds stabilizes both proteins at the cell surface in wing discs [14-16], but app^e6 clones did not affect Ft or Ds levels or distribution (Figures 2G-2H). Creating artificial boundaries of ft or ds expression also strongly polarizes wing PCP [14-16], as do boundaries of the Fj kinase that phosphorylates Ft and Ds and modulates their levels [14, 15, 29-31]. If App affected Ft-Ds levels or binding we would also expect App boundaries to affect PCP. However, small to moderately sized app^e6 clones did not affect PCP, while large clones only affected PCP in the regions of the wing where defects were observed in app^e6 homozygotes (Figure S5). There was no tendency to reorient hairs at clone boundaries (Figures 2I and S5), and only rarely did regions with altered PCP affected PCP in adjacent wild type cells; these may be due to altered cell interactions mediated by the core polarity proteins. Sharp boundaries of UAS-app-RA misexpression (e.g. driven with the posterior driver en-gal4) also had no effect on PCP (Figure 2J). Anti-App staining was not altered in wing discs by ft or ds clones (data not shown). Thus, despite their co-localization, there is no evidence that App physically interacts for Ft or Ds.

app mutants partially rescue the viability, growth and PCP defects of ft mutants

PCP defects, reduced crossvein spacing and lost tarsal leg joints can be caused by either gains or losses in Ft signaling [15, 16, 30, 32], so we examined the phenotypes of app ft double mutants. ft^fd and ft^G-rv are likely null alleles predicted to truncate Ft N-terminal to its transmembrane region [33]. ft^fd homozygotes and ft^fd/ft^G-rv die during late pupal stages with strongly overgrown imaginal discs (compare Figures 3A and 3E with Figures 3B and 3F) and disc-derived tissues; late pupal abdomens are not overgrown, but have strong PCP defects (Figure 3L). These phenotypes were partially rescued in ft^fd ; app^e6 and ft^fd/ft^G-rv ; app^e6 flies: overgrowth and extra folding of imaginal discs was suppressed (Figures 3D and 3H) and lethality and abdominal PCP defects were reduced (Figure 3M). PCP was still defective in proximal portions of the wing but was normal in the distal wing (Figures 3I-3K), in contrast to the distal defects in viable ft¹⁸ wings [33]. Thus, App acts genetically downstream of and in opposition to Ft in both growth control and PCP.

App affects the levels and distribution of Dachs

The effects of app mutations on Ft signaling are strikingly similar to those caused by reducing the function of the atypical myosin Dachs [11]. Like app mutations, dachs (d) mutations partially suppress the overgrowth and PCP defects caused by loss of Ft [10, 11]. The adult phenotypes are also similar, although the defects in d null mutants are stronger than those of app^e6. Null d^GC13 hemizygotes and hypomorphic d¹ homozygotes reduce tarsal leg segmentation (Figures S1H and S1I) and the distance between the ACV and PCV and cause mild wing PCP defects quite similar to those observed in null app mutants (compare Figures 4A and 4B to Figures 1B-1D; details in Figures S6 and S7). Like app^e6 clones, d^GC13 clones had PCP defects when in regions of the wing where defects occur in d homozygotes, and Fmi polarization was reoriented in d¹ pupal wings (Figure S5E). d mutants also had abdominal PCP defects similar to those in app mutants: polarity was almost normal in the anterior compartment, but abnormal near A/P compartment boundary and reversed in the posterior compartment (Figure 4D).

We therefore examined the effects of App on the levels and distribution of a V5-tagged Dachs protein. Dachs::V5 normally accumulates at higher levels in the apical cell cortex in wing imaginal discs [11], overlapping the region of high anti-App staining (Figures S4A, S10A and S10B). Apical Dachs accumulation was greatly reduced, although not completely eliminated, in app^e6 clones (Figures 4F-4G”). We did not detect changes in the levels of basolateral or cytoplasmic Dachs::V5. While App affects Dachs accumulation at the apical cell cortex, anti-App staining was normal in d mutant clones (data not shown).

Co-overexpression of App and Dachs::V5 greatly increased the accumulation of Dachs at the cell cortex compared with the expression of Dachs::V5 alone (compare Figures 4H to 4I). Co-expression of App also increased the efficacy of Dachs in growth and PCP. Even though overexpression of App-RA did not obviously increase growth, co-expression of App-RA and Dachs caused greater overgrowth than did the expression of Dachs alone (Figure S8). Co-overexpression of Dachs and App caused more extreme PCP defects in the wing and abdomen than did the overexpression of App alone, while overexpression of Dachs::V5 alone did not affect PCP (compare Figures 4J-4L; more examples in Figure S9).

It is likely that much or all of the app mutant phenotype is mediated by the reduction of effective Dachs at apical cell cortex. The effects of app and d mutants are not additive: double mutants for null app and d alleles resembled the stronger d null phenotype, as expected if App works by controlling Dachs activity (Figures 4B-4E). Since App affects Dachs post-transcriptionally it is unlikely that overexpressed Dachs would fully rescue the app null. Nonetheless, overexpression of UAS-d with ap-gal4 or en-gal4 rescued the wing PCP defects normally found in the distal wing of app^e6 mutants (Figures 4M and 4M’), and partly rescued the crossvein spacing and leg joint defects of app^e6 (Figures S1G and S6). That Dachs retains some activity in the absence of App is consistent with the low but significant levels of Dachs that remain at the apical cell cortex in app^e6 clones (Figures 4F-4G). Different DHHC proteins can palmitoylate the same target (e.g. [34]), so other Drosophila DHHC proteins may be supplying residual activity in the absence of App.