Angiotensin-converting enzyme Ance is cooperatively regulated by Mad and Pannier in Drosophila imaginal discs

Angiotensin-converting enzyme (ACE) is an evolutionarily conserved peptidyl dipeptidase. Mammalian ACE converts angiotensin I to the active vasoconstrictor angiotensin II, thus playing a critical role for homeostasis of the renin-angiotensin system. In Drosophila, the ACE homolog Ance is expressed in specific regions of developing organs, but its regulatory mechanism has not been identified. Here we provide evidence that Ance expression is regulated by a combination of Mad and Pannier (Pnr) in imaginal discs. We demonstrate that Ance expression in eye and wing discs depends on Dpp signaling. The Mad binding site of Ance regulatory region is essential for Ance expression. Ance expression in imaginal discs is also regulated by the GATA family transcription factor Pnr. Pnr directly regulates Ance expression by binding to a GATA site of Ance enhancer. In addition, Pnr and Mad physically and genetically interact. Ance null mutants are morphologically normal but show genetic interaction with dpp mutants. Furthermore, we show that human SMAD2 and GATA4 physically interact and ACE expression in HEK293 cells is regulated by SMAD2 and GATA4. Taken together, this study reveals a cooperative mechanism of Ance regulation by Mad and Pnr. Our data also suggest a conserved transcriptional regulation of human ACE.

example, Dpp is critical for the progression of the morphogenetic furrow in the eye imaginal disc 21,22 and for wing disc growth signaling from the anterior-posterior boundary 15,23 . Dpp signaling is also required for wing venation during pupal development [24][25][26] . In eye discs, Pnr is known to be involved in establishment of the dorsoventral (DV) boundary by regulating the dorsal-specific gene expression 27 , although the relationship with Dpp signaling is unknown. As in embryogenesis, Dpp signaling affects pnr expression in wing discs. The dorsal region of the wing disc is a primordium for the notum, the dorsal part of the thorax. Dpp signaling in this region regulates Wg expression by inducing pnr 28,29 . Interestingly, previous studies have shown that Ance is specifically expressed in the dorsal region of the eye disc and the notum region of the wing disc 30 . We have noticed that the sites of Ance expression in these discs may correspond to the territories of pnr expression in the eye disc. These observations raise questions about whether the Ance expression in these discs is related to the Dpp signaling and the Pnr function.
Despite the region-specific patterns of Ance expression in imaginal discs, little is known about how Ance expression is regulated and what role Ance plays in development. It has been reported that male accessory glands secrete Ance proteins 31,32 and that males homozygous for Ance hypomorphic alleles are infertile due to defective spermiogenesis 33 . Moreover, some of the Ance mutant alleles are lethal 10 , indicating the requirement of Ance for development. Unexpectedly, new deletion Ance alleles generated in this work are fully fertile and viable, indicating that Ance is not critically required under normal condition. However, it is possible that Ance may become important when an interacting pathway is compromised. Thus, we have addressed two main issues in this study: the regulation of Ance expression and the possible genetic interactions between Ance and its regulators. We also asked if the mechanism for Ance gene regulation is shared in human cells.
Here we demonstrate that Ance expression in the eye and wing discs is regulated by Dpp signaling through Mad binding to the Ance regulatory region. Ance expression in these discs also depends on Pnr binding to one of the GATA sites in the Ance regulatory region. Furthermore, Mad and Pnr show physical interaction. Ance null mutants are fertile and morphologically normal. However, the Ance mutations show genetic interaction with dpp mutants. We also show that human ACE gene expression is regulated by SMAD2 and GATA4 transcription factors. Taken together, Ance is regulated by a combination of Dpp and Pnr and is involved in modulating some of Mad functions, providing insights into the regulation and function of mammalian ACE.

Results
Expression patterns of Ance-lacZ and Ance protein in eye and wing discs. The imaginal disc is a sac-like structure consisting of two epithelial layers, a disc proper (DP) and a peripodial epithelia (PE). While the disc proper is a columnar epithelium, peripodial cells form a thin squamous epithelium. Ance mRNA expression was shown to be enriched in the dorsal region of eye and wing discs 30 . We have found that this pattern of Ance mRNA is similar to that of pnr expression. Since pnr is selectively expressed in the peripodial cells of eye discs 34,35 , we investigated whether Ance is also preferentially expressed in the peripodial epithelium.
We utilized an Ance-lacZ reporter that reflects the dorsal expression of Ance mRNA in embryos 11 . First, we found that Ance-lacZ was dorsally expressed in eye and antenna discs (Fig. 1a,b), which resembled the pattern of Ance mRNA expression shown in a previous report 30 . Apical confocal sectioning of eye discs indicated that Ance-lacZ expression was preferentially localized to the peripodial epithelium ( Fig. 1a' ,b') like that of pnr 34,35 . Immunostaining with membrane marker anti-Discs-large (Dlg) antibody showed that Ance-lacZ expressing cells are large peripodial cells ( Fig. 1a"' , b"'). We also checked the pattern of Ance protein expression using an anti-Ance polyclonal antibody 36 . Ance protein was found in most peripodial cells of the eye disc in contrast to the dorsally localized Ance-lacZ reporter expression (Fig. 1a", b").
In more basal sections, Dlg was detected in an array of differentiating photoreceptor cells in the disc proper ( Fig. 1c"') where Ance-lacZ was weakly detected in the anterior dorsal region (Fig. 1c'). Tangential sectioning at the disc proper level also shows a cross-section view along the margin of the disc. In this marginal area, Ance was highly localized in the lumenal space between the peripodial epithelium and the disc proper (Fig. 1c, c", c"'). Although Ance protein was not clearly detected in the differentiating retina posterior to the morphogenetic furrow ( Fig. 1c"), closer examination revealed weak levels of Ance in photoreceptor cells (arrowheads in Fig. S1e). Since Ance-lacZ is preferentially expressed in the dorsal peripodial cells, it can be inferred that the Ance protein detected widely in other areas of the eye disc probably originates from the secretion from cells expressing Ance.
Next, we examined the Ance expression in the wing disc. Ance-lacZ expression was abundant in the peripodial cells of the notum area ( Fig. 1d-e"'), which is similar to the pnr expression pattern 27,34,35 . Weak expression of Ance-lacZ was also detected in the notum disc proper cells (Fig. 1f '). In contrast to the high levels of Ance-lacZ expression in the notum region, Ance-lacZ was weakly detected in the peripodial cells that cover the wing pouch region (Fig. 1g'). Ance protein in the wing disc was detected more broadly than the region of Ance-lacZ expression (Fig. 1d,e,g). Ance was abundant in the lumenal space ( Fig. 1f ", h"), thus in contact with most disc proper cells including the notum and the wing pouch (Fig. S2). Taken together, the expression patterns of Ance-lacZ and Ance-lacZ expression in dorsal PE cells. (e") Ance protein. (e"') Dlg staining of dorsal PE cells. (f-f "') Dorsal region of the wing disc sectioned at the DP level. (f ') Ance-lacZ expression is weakly detected in the dorsal region of DP. (f ") Ance protein in the lumen (arrow). (f "') Dlg staining shows DP cells and a cross-section view of the lumen between the PE and DP layers in the disc margin. (g-g"') High magnification views of the wing pouch region at the PE level shown in (d-d"'). (g') Ance-lacZ expression (arrow). (g") Ance protein. (g"') Dlg staining of dorsal PE cells. (h-h"') High magnification views of the wing pouch region at the DP level shown in (d-d"'). (h') No Ance-lacZ expression in the pouch. (h") Ance protein localized in the disc lumen (arrow). (h"') Dlg staining of DP cells. Scale bar, 50 μm (a-h"').

Figure 2.
Dpp signaling is required for Ance expression in eye and wing discs. (a,b) Wing disc containing Mad 1-2 mutant clones stained for GFP clone marker and Ance-lacZ. Large Mad 1-2 mutant clones (GFPnegative) are shown in the wing pouch region (the box area in Fig. 1d'). Ance-lacZ expression is nearly absent in Mad 1-2 clones (arrow). (a) Merge of (a'-a") at the PE level. Ance protein suggest that Ance mRNA is highly expressed in the dorsal peripodial region, while Ance protein is secreted to neighboring peripodial cells, lumenal space, and most disc proper cells.
Ance expression is regulated by Dpp signaling during larval development. Ance expression is known to be regulated by Dpp signaling during embryogenesis 10,11 . Because Dpp signaling can activate different transcriptional target genes in different tissues and stages 18,19,37 , it is important to check whether Dpp signaling is required for the specific pattern of Ance expression in larval imaginal discs. We therefore examined the effects of loss of Mad, the transcription factor that mediates Dpp signaling, on Ance-lacZ expression. Ance-lacZ reporter expression was nearly undetectable in Mad mutant clones generated in the peripodial cells of wing discs (Fig. 2a,b and S3a). In addition, loss of Tkv, the receptor serine/threonine kinase for the Dpp ligand, resulted in reduced Ance-lacZ reporter expression in peripodial cells of eye disc (Fig. S4a,b).
We also tested the effects of loss of Mad on Ance protein expression. When Mad 1-2 mutant clones were generated in the regions distant from the dorsal domain of Ance-lacZ expression in eye disc, they showed normal level of Ance protein (arrow in Fig. 2c). This result suggests that Ance protein in these clones is provided non-autonomously from the Mad + Ance-expressing cells. Mad 1-2 clones located within the Ance-expressing region showed Ance proteins within the clones but at a reduced level (Fig. 2c,d).
We also checked whether ectopic Dpp signaling can activate Ance expression using Gal4-UAS system 38 . When GFP was overexpressed along the anterior-posterior boundary of wing disc using ptc-Gal4 driver (labelled as ptc > GFP), there was no ectopic Ance expression in the ptc-expressing area (Figs 2e and S5a). However, overexpression of tkv QD , a constitutively active form of tkv, led to ectopic expression of Ance in the ptc region of wing disc (Figs 2f and S5b). Interestingly, the ectopic Ance expression was mainly localized to two proximal spots in the ptc region of the wing pouch ( Fig. S5b), suggesting that Dpp signaling can activate Ance expression in specific regions. Together, these results suggest that Dpp signaling is necessary and sufficient for Ance expression in specific territories of imaginal discs.
Ance is dispensable for development. Since Ance is expressed in specific domains of discs, we tested whether Ance plays a role in the development of these organs. To examine the function of Ance, we generated two Ance-deletion mutants, Ance d110 and Ance d128 , by imprecise excision of a P-element inserted in 5′ UTR of the Ance gene. In both mutants, the first exon and adjacent regions were deleted (1247 and 976 bp deletion in Ance d110 and Ance d128 , respectively), which was confirmed by genomic PCR and sequencing (Fig. 3a,b). Ance mRNA and proteins were not detected in these deletion mutants, which were checked by RT-PCR (Fig. 3c), western blot (Fig. 3d), and immunostaining of imaginal discs (Fig. 3e), indicating that these two mutations are null alleles. These Ance-deletion mutants were viable and fertile with no obvious morphological defects in adult eyes and wings (Fig. 3f,g). This was unexpected because it was previously reported that Ance mutants show lethality or infertility 10,33 . Interestingly, Ance null mutants generated in this work fully complemented the Ance mutants (Ance 34Eb-1 and Ance 34Eb-2 ) described in earlier reports (data not shown). In contrast, the lethality of Ance 34Eb-1 and Ance 34Eb-2 failed to complement a Smg5 mutation (Smg5 e04233 ), which is located approximately 1 kb upstream from the 5′ region of the Ance gene. These results indicate that the lethality of Ance 34Eb-1 and Ance 34Eb-2 is likely due to a mutation(s) in Smg5 rather than the Ance gene (data not shown). Taken together, these data suggest that Ance is dispensable for development.

Ance mutations show genetic interaction with Dpp signaling. Although Ance deletion mutants and
Ance RNAi flies show no discernible phenotype in the normal condition, it is possible that Ance mutations might show synthetic phenotypes with other mutations in the functionally related genes. Since Ance expression is regulated by Dpp signaling, we reasoned that Ance might show genetic interaction with Dpp signaling components. To detect genetic interaction, we examined the effects of altering the Ance level on the phenotypes of dpp overexpression or dpp mutations. First, we used flies that overexpress dpp in the eye. As shown in Fig. 4a, overexpression of dpp under GMR-Gal4 driver (GMR > dpp) resulted in roughening of the eye surface. Under this background, reducing the Ance gene dosage by one copy of Ance mutation partially suppressed the rough eye phenotype, showing improved arrays of ommatidia (Fig. 4c). In third instar eye disc, dpp overexpression by GMR-Gal4 resulted in irregular wrinkling of developing retina, as shown by immunostaining for the pan-neuronal marker Elav and the cell membrane marker Dlg ( Fig. 4b-b"). One copy of Ance mutation significantly suppressed the abnormalities caused by dpp overexpression in eye disc ( Fig. 4d-d").
We also checked whether Ance mutations genetically interact with dpp hypomorphic mutation. Ance d128 mutation does not affect any vein (Fig. 4e). dpp s1 dpp d-ho homozygotes show loss of distal part of 4 th longitudinal wing vein (L4) (Fig. 4f), besides the partial loss of the L2 vein at a low frequency (7.3%) as seen in Fig. 4g. Interestingly, Ance d128 mutation strongly enhanced the frequency of L2 vein loss phenotypes of dpp s1 dpp d-ho to 70.1% of flies examined (Fig. 4g,k). This result indicates that there is a synergistic interaction between Ance and dpp mutations. In addition, we tested whether mutations in Ance and tkv a12 can show genetic interaction in the wing. tkv a12 /+ heterozygote showed normal veins (Fig. 4h). However, tkv a12 /Ance d110 trans-heterozygote resulted in a weak thickvein phenotype at the distal tip of the L4 vein ( Fig. 4i, 54.7%). tkv a12 /Ance d128 trans-heterozygote showed nearly identical genetic interaction as tkv a12 /Ance d110 (Fig. 4j, 49.2%). In addition, tkv a12 heterozygotes under Ance d110 (or Ance d128 ) homozygote condition showed increased frequencies of the thickvein phenotype tkv QD overexpression by ptc-Gal4. (f) Merge of apical section images (f '-f "). (f ') GFP in the ptc-expressing region. (f ") Ectopic Ance expression in the proximal sites (arrows) in the ptc area of the wing pouch. Scale bars, 50μm (a-f "). The G3223 line has a P-element inserted in the 5′ UTR of Ance. Ance d110 and Ance d128 mutants were generated by imprecise excision of G3223. Two mutants lack the first exon of Ance gene. Red arrows and green box indicate genomic PCR primers used in (b) and RT-PCR in (c), respectively. (b) Confirmation of deletions in Ance mutants. Genomic PCR results of w 1118 , Ance d110 , and Ance d128 show smaller PCR products in two Ance mutants compared to w 1118 control. (c) RT-PCR results of w 1118 , Ance d110 , and Ance d128 . No PCR product is detected in two Ance mutants. (d) Ance protein is not detected in Ance d110 and Ance d128 mutants. (e) Absence of Ance staining in Ance mutants. Eye discs of w 1118 , Ance d110 , and Ance d128 stained for Ance and Dlg. Scale bar, 100 μm. (f) Adult eye morphology of w 1118 , Ance d110 , and Ance d128 . w 1118 control shows the normal eye. Ance d110 and Ance d128 show normal eye morphology. (g) Adult wing morphology of w 1118 , Ance d110 , and Ance d128 . w 1118 control shows the normal wing. Ance d110 and Ance d128 show normal wing morphology.
Ance expression is directly controlled by pannier. As described earlier (Fig. 1), the Ance-lacZ reporter shows preferential dorsal expression similar to the pattern of pannier (pnr) expression 27,34,35 . To check whether pnr is involved in regulating Ance expression during larval development, we generated clones of pnr null mutant cells in eye imaginal discs. Because Ance-lacZ is expressed near the dorsal margin of the eye disc, pnr VX6 clones outside the Ance-lacZ expressing region did not show any lacZ expression. Conversely, pnr VX6 mutant clones within the Ance-expressing dorsal area of the eye disc showed little or no expression of Ance-lacZ (Figs 5a-a" and S6). In contrast, variable levels of Ance protein were detected in pnr VX6 mutant clones (Fig. S7), consistent with non-cell autonomous distribution of Ance protein. We also tested the effects of Pnr overexpression using various Gal4 drivers but could not check the effects due to early lethality. Nevertheless, our analysis of mutant clones clearly indicates that pnr is required for the dorsal-specific Ance expression.
As Ance-lacZ expression was lost in pnr VX6 mutant clones, we hypothesized that Pnr might be directly required for Ance expression. To check this possibility, we tested whether Pnr can bind to the Ance enhancer. The Ance-lacZ reporter contains a 533 bp Ance regulatory region which has three potential GATA binding sites in the 3′ half. First, we carried out electrophoretic mobility shift assay (EMSA) using purified GST-Pnr-DBD (DNA binding domain) and the 533 bp Ance regulatory region. EMSA using GST-Pnr-DBD and the 533 bp Ance enhancer showed an upward band shift when GST-Pnr was added, indicating that Pnr binds to the Ance region (Fig. 5c). The 3′ region of the Ance enhancer containing three GATA sequences showed similar band shifting, but the 5′ region did not, suggesting that one or more of the GATA sites might be involved in binding with GST-Pnr-DBD (Fig. 5c). Additional EMSA assays with three DNA fragments containing different GATA sites showed that the second GATA site was strongly bound to GST-Pnr-DBD (Fig. 5e). These data suggest that Pnr directly regulates Ance expression at transcriptional level by binding to the second GATA site of the Ance enhancer. To confirm that the second GATA site in Ance enhancer is important, we generated a mutant form of GATA2 fragment, which has AAAA instead of GATA in the second GATA site. In EMSA assays, Pnr binding to the GATA2 DNA was strongly reduced by the addition of unlabeled wild-type GATA2 DNA but not by the mutated GATA2 DNA (Fig. 5f). This further supports the specific binding of Pnr to the GATA2-containing DNA.
In addition, we carried out luciferase assay in S2 cells using the wild-type and the mutated Ance GATA enhancer. Single transfection of either Mad or pnr weakly increased luciferase activity by 8.8% and 20.6%, respectively. Double transfection of Mad and pnr significantly enhanced luciferase activity to 40.3%, which was higher than the combined effects of Mad and pnr (Fig. 6a). Next, we performed similar luciferase assays with Ance GATAmt , the Ance regulatory region with a mutation in the GATA2 site. In this assay, Mad alone could weakly activate luciferase activity (11.4%), but Pnr alone did not significantly enhance the luciferase activity due to the GATA2 mutation. Interestingly, when both Mad and pnr were overexpressed, the effect of Mad was consistently inhibited (3.8% lower than the control level) (Fig. 6b). This suggests that overexpressed Pnr may interact with Mad but fail to bind to the mutated Ance GATAmt regulatory region, perhaps dominantly inhibiting the Mad effects.
Our data from the luciferase assays in S2 cells showed that Mad and Pnr can bind to the Ance regulatory region to activate transcription. However, the level of activation was relatively weak, and it was difficult to determine whether the combined effects of Mad and Pnr were additive or synergistic. Thus, we tested the effects of Mad and Pnr in vivo using transgenic reporter lines. Firstly, to determine the effects of Dpp signaling on the Ance expression, we utilized an MDel-lacZ reporter, which is a modified version of Ance-lacZ mutated in the Mad-binding sites 19 . As shown in Fig. 6e and f, MDel-lacZ expression was significantly reduced in both eye and wing imaginal discs compared to Ance-lacZ (Fig. 6c,d). In addition, we checked the possibility that Ance regulated by Dpp signaling might have a positive or negative feedback during development. As indicated in Fig. S8, there was no difference in the level of phospho-Mad (p-Mad) between wild-type and Ance-deletion mutant, indicating that there is no obvious feedback regulation between Dpp signaling and Ance.
Secondly, to test whether the second GATA site in the Ance enhancer is important in vivo, we generated an Ance GATAmt -lacZ line, which consists of AAAA instead of GATA in the second GATA site and checked its expression pattern in eye and wing discs. As indicated in Fig. 6g and h, lacZ expression was significantly reduced compared to Ance-lacZ expression (Fig. 6c,d), indicating that the second GATA site bound by Pnr is required for Ance expression. Since loss of either Mad or Pnr binding site severely reduces Ance expression in imaginal discs, these data suggest that Mad and Pnr act together in vivo rather than acting independently.

Physical and genetic interaction between Pnr and Dpp signaling. Our data described above
show that Mad and Pnr might function together for Ance expression in imaginal discs. While pnr expression is restricted to the dorsal domain 27,34,35 , dpp mRNA expression in third instar eye disc is restricted to the morphogenetic furrow and part of the dorsal/ventral margins 21,39 . However, Dpp is secreted to broader regions of the eye disc. Therefore, although mRNA expression patterns of pnr and dpp are not identical in the eye disc, cells that receive the Dpp signal might overlap in the dorsal margin region of pnr expression. Thus, we examined the pattern of p-Mad expression in third instar eye disc. p-Mad expression was enriched along the MF, but it was also elevated at the dorsal margin of eye disc where pnr is expressed (Fig. 7a-a", arrow). Therefore, Pnr and Mad may function together to regulate Ance expression. To address this possibility, we tested whether Pnr and Dpp signaling genetically interact (Fig. 7b). When tkv a12 and pnr VX6 flies were crossed, trans-heterozygote progeny (tkv/+; pnr/+) showed extra vein formation from the posterior crossvein (Fig. 7b,c). Although the extra vein phenotype was mild, it was consistently observed in 60.4% transheterozygotes, while such phenotype was rarely detected in tkv/+ or pnr/+ heterozygotes (7.9% and 0%, respectively) (Fig. 7d).
The overlapping localization of Pnr and Mad and their genetic interaction suggested that they might be associated together in a protein complex. To test this possibility, we carried out co-immunoprecipitation between Pnr and Mad using larval extracts. As shown in Fig. 7e, Pnr was co-immunoprecipitated by anti-Mad antibody, suggesting that Pnr and Mad physically interact. We also checked whether their interaction is direct. GST-pulldown assay using bacterially purified GST-Pnr and MBP-Mad showed that N-terminal part of Pnr binds to MBP-Mad ( Fig. 7f), indicating that Pnr and Mad directly interact. Taken together, these data support physical and functional interaction between Pnr and Mad, as outlined in the scheme shown in Fig. 7g (See Discussion).

SMAD2 and GATA4 are required for ACE regulation in human cells. Mammalian Angiotensin-
converting enzyme (ACE) plays key roles in regulating renin-angiotensin system. Since our data showed that Drosophila Mad and Pnr are required for Ance expression, we hypothesized that mammalian SMAD and GATA factors, homologs of Drosophila Mad and Pnr, respectively, might be involved in ACE expression. To test this possibility, we examined the effects of RNAi of SMAD1/2 and GATA2/4 in HEK293 cells (human embryonic kidney 293 cells). SMAD1 and GATA2 siRNAs did not affect ACE expression (Fig. S9a), but SMAD2 and GATA4 siRNAs were effective in reducing ACE expression (Fig. 8a,b). The levels of knockdown for each siRNA for SMAD1/2 and GATA2/4 were approximately 70-90% (Fig. S9b,c). Furthermore, double siRNA treatments for both SMAD2 and GATA4 were more effective in reducing ACE expression compared to single siRNA for SMAD2 or GATA4 (Fig. 8c). We also tested whether overexpressing SMAD2 and GATA4 is sufficient to induce the expression of ACE. Overexpression of either SMAD2 or GATA4 showed slight increase in the ACE mRNA level, whereas overexpression of both SMAD2 and GATA4 resulted in further increase (Fig. 8d), which was similar to the effects of overexpressing Mad and Pnr in Drosophila S2 cells (Fig. 6a).
To check whether SMAD2 and GATA4 are directly involved in regulation of ACE expression, we performed chromatin immunoprecipitation (ChIP) assays (Fig. 8e). Our data showed that GATA4 and SMAD2 bind to the   Transcription start site (TSS) and primer sets used for qPCR are indicated. Right: Relative enrichment of SMAD2 and GATA4 to IgG control. (f) Reciprocal co-immunoprecipitation between SMAD2 and GATA4 in HEK293 cells. Overexpressed GATA4 (GATA4-V5) is co-immunoprecipitated with endogenous SMAD2 and vice versa. GATA4 is detected by anti-V5 antibody. (g) ACE overexpression suppresses the increased level of phospho-ERK (p-ERK) in SMAD2/GATA4-knockdown HEK293 cells. siSMAD2/siGATA4 in HEK293 cells increases p-ERK compared to siSCR control. ACE overexpression decreases p-ERK in siSMAD2/siGATA4treated cells. (h) Angiotensin II (Ang II) treatment reduces the level of p-ERK in SMAD2/GATA4-knockdown HEK293 cells. siSMAD2/siGATA4 in HEK293 cells increases p-ERK compared to siSCR control. Ang II treatment decreases p-ERK in a dose-dependent manner. (i) Ang II rescues the reduced cell viability in SMAD2/GATA4-knockdown HEK293 cells in a dose-dependent manner. All data represent the mean ± s.e.m. (error bars) and P-values were calculated by using Student's t-test. n.s, not significant, *P < 0.05. **P < 0.01. ***P < 0.001. ****P < 0.0001. first exon of ACE gene, indicating their transcriptional roles in the regulation of ACE expression. Taken together, both SMAD2 and GATA4 are required for ACE expression in HEK293 cells as in Drosophila.
The regulation of ACE expression by SMAD2 and GATA4 suggests that SMAD2 and GATA4 might be associated in a protein complex to regulate their target gene expression as in Drosophila. To test this possibility, we carried out co-immunoprecipitation experiments between SMAD2 and GATA4 using GATA4-overexpressing HEK293 cell lysates. Overexpressed GATA4 was co-immunoprecipitated by endogenous SMAD2 immunoprecipitation, and endogenous SMAD2 was also co-immunoprecipitated by overexpressed GATA4 (Fig. 8f). These results suggest that SMAD2 and GATA4 form a protein complex in human as seen in the Mad-Pnr co-immunoprecipitation in fly.
To look for clues to the function of SMAD2 and GATA4, we examined their effects on the extracellular signal-regulated kinase (ERK) signaling that regulates a variety of cell physiology. First, we checked the level of phosphorylation of ERK (p-ERK) and found that the treatment of siSMAD2/siGATA4 increased p-ERK compared to the siScramble (siSCR) control (Fig. 8g). Interestingly, the increased p-ERK level in SMAD2/GATA4-knockdown HEK293 cells was suppressed when their target gene ACE was overexpressed. We also tested whether an addition of Angiotensin II (Ang II) can have similar effects. Indeed, the increased p-ERK level in SMAD2/GATA4-knockdown cells was antagonized by Ang II in a dose-dependent manner, indicating that ACE overexpression and Ang II treatment could rescue the effects of silencing SMAD2 and GATA4. Next, we checked viability in HEK293 cells. As shown in Fig. 8i, SMAD2 and GATA4 siRNAs decreased cell viability compared to control siRNA. An addition of Ang II did not affect the viability in siSCR-treated HEK293 cells, but it partially rescued reduced cell viability of SMAD2/GATA4-knockdown cells. Taken together, we conclude that SMAD2 and GATA4 directly regulates ACE expression as in Drosophila and loss-of-function phonotypes of SMAD2 and GATA4 can be restored in part by ACE overexpression or Ang II treatment.

Discussion
Expression of the Drosophila ACE homolog Ance is developmentally regulated in larval imaginal discs. Here we have provided evidence that Mad and Pnr transcription factors play key roles in Ance regulation.
Previously, it has been shown that the Ance expression during embryogenesis is activated by Dpp signaling from the ectodermal dorsal midline 10,11 . Our data show that Dpp signaling is essential for Ance expression in imaginal discs, consistent with the role of Dpp for Ance expression in embryogenesis. Transcriptional activation in eukaryotes is often regulated by synergistic effects of other enhancer binding proteins 40 . It has been shown that Mad is a partner of the Fos transcription factor for a context-dependent transcriptional activation in Drosophila 41 . Mammalian Smad proteins can also promote different TGFβ target gene expression in cooperation with cell type-specific transcription factors 42,43 . In this regard, it is worth noting that both Mad and the GATA factor Pnr are required for Ance expression during larval development. In embryogenesis, Ance and pnr are independent targets of Dpp signaling with different thresholds for Dpp activity. In contrast, Ance expression in imaginal discs is regulated by a combination of Dpp and Pnr.
According to the clonal analysis, loss of either Mad or Pnr almost completely abolishes the Ance-lacZ expression in the eye and wing discs. This indicates that Mad and Pnr regulate Ance expression in a cooperative manner, consistent with their physical interaction. We also found a weak but consistent genetic interaction in wing venation between tkv and pnr mutations (Fig. 7b-d). Based on these results, we propose a model in which Mad activated by Dpp signaling can interact with Pnr for tissue-specific Ance expression (Fig. 7g). Secreted Dpp activates Mad in broad regions of imaginal discs. In contrast, Ance expression is enriched in the dorsal area of the eye and wing discs, indicating that additional dorsal factors are required for the restricted Ance expression. In our model, one of such dorsal factors might be Pnr, thus allowing context-dependent expression of the Ance gene in the dorsal-specific pattern.
The Ance deletion mutants generated in this study develop normally to fertile adults. This result was somewhat unexpected because Ance mutants were reported to be lethal or sterile 10,31,33 . Our molecular and genetic analysis of Ance null mutants shows that Ance is dispensable for development and fertility. It is possible that other Ance homologs might be able to compensate the loss of Ance. Ance and its homolog, Angiotensin-converting enzyme-related (Acer), can catalyze the same peptide, although they have different affinities for the substrates 44 . Therefore, there is a possibility that the Acer function could compensate for loss of Ance during development. However, double mutants for Ance and Acer were also found to be developmentally normal and fertile (data not shown). Analysis of the Ance function is more complicated due to the presence of four other Ance-related genes in the fly genome; Ance-2, Ance-3, Ance-4, and Ance-5. Due to their potentially redundant functions, loss of one or more Ance homologs may not affect animal development, unless the functions of all the Ance homologs are disrupted. Alternatively, other peptidases may also be used in the development of imaginal discs to circumvent the absence of Ance peptidase activity. Imaginal discs contain a variety of peptidases that hydrolyze a wide range of peptides and proteins 45 . The hydrolysis of peptides and proteins by some of these peptidases might be important for the production of dipeptides and free amino acids for protein synthesis required for morphogenesis. Active forms of signaling peptides might also be generated by peptidases for development of imaginal discs 44,45 . Thus, the absence of Ance can be compensated by other peptidases that can hydrolyze unidentified substrates of Ance to produce necessary peptides for development.
Although Ance null mutants do not exhibit obvious developmental defects, it is important to note that loss of Ance can modulate the phenotypes of impaired Dpp signaling. It has been shown that overexpression of dpp by GMR-Gal4 causes severe roughness and bulging of the adult eye 46 . We showed that reduced Ance partially suppressed the effects of Dpp overexpression (Fig. 4a-d). These results suggest that the effects of dpp gain-of-function in the eye are mediated at least in part by Ance, consistent with the induction of Ance gene expression by Dpp signaling. Because Ance overexpression alone is insufficient to affect the eye, Ance seems to modulate the function of an unknown target of Dpp signaling to modify the Dpp-induced eye phenotype (Fig. 7g). Because Ance protein is detected in the eye disc lumen as well as disc proper cells, it may produce peptide factors that modify the effects of Dpp shown in Fig. 4c,d. Furthermore, our data show genetic interaction between hypomorphic dpp (or tkv) mutations and Ance mutation in wing veins. Ance-lacZ is enriched in the dorsal part of wing disc, but high levels of Ance protein are broadly detected in the lumen of the wing pouch area (Fig. 1d' ,d", Fig. S2k,l). Ance protein abundant in the wing pouch, either in the lumen or disc proper cells by uptake, may produce active peptides to affect vein differentiation, thus providing a possible explanation for the genetic interactions in vein phenotypes. These genetic interactions in dpp Ance double mutants (dpp Ance) are synergistic, since dpp or Ance single mutants show little deficiency in the second vein ( Fig. 4e-g,k). Under dpp (or tkv) hypomorphic conditions, the Ance expression level is expected to be reduced partially. The introduction of an Ance null allele under this condition further reduces the Ance level. However, because Ance null mutants develop normally, the vein phenotype in dpp Ance double mutants cannot simply be explained by a reduction of the Ance level. Instead, Ance may modulate a Mad function in vein patterning. When Dpp signaling is fully functional, the Mad function in vein development may not be affected by loss of Ance alone. In tkv heterozygotes or dpp hypomorphic conditions, however, the reduction of Ance may critically impair such pathways as seen in the double mutant conditions. Thus, despite the lack of visible phenotypes in Ance null mutants, Ance seems to play modulatory roles when Dpp signaling is partially compromised. These genetic results raise a possibility that Ance expression is not only regulated by Mad but also interacts with specific Mad signaling outputs that affect diverse tissues, as proposed in Fig. 7g. The identification of related substrates for Ance dipeptidase activity will help to understand the mechanism underlying these genetic interactions.
The human somatic ACE, a zinc-dependent dipeptidyl carboxypeptidase, plays an important role in the regulation of the renin-angiotensin system. ACE has long been targeted for effective treatment of hypertension and related cardiovascular diseases through ACE inhibitors [47][48][49] . TGFβ is also associated with activity of ACE in renal fibrosis and disease progression. Thus, anti-TGFβ antibody can improve renal fibrosis in mouse model 50 . Kidney diseases can be treated more effectively when both ACE and TGFβ are blocked by inhibitor and antibody, respectively 51 , suggesting an association between ACE and TGFβ signaling. In addition, GATA4 activates atrioventricular canal-specific enhancer that has synergistic effects with Smad during heart development 52 , although the mechanism underlying the synergy is unknown. Our data show that human SMAD2 and GATA4 are not only required for ACE expression but can also induce it in HEK293 cells. Our ChIP data and co-immunoprecipitation data suggest that SMAD2 and GATA4 physically interact each other and bind to the first exon of ACE gene to play a transcriptional role in ACE expression.
The Ras/Raf/ERK pathway plays oncogenic roles in many tumors 53,54 . However, ERK activation can also lead to anti-proliferative events depending on the conditions of signal activity 55,56 . According to our data, downregulation of SMAD2 and GATA4 in HEK293 cells reduces cell viability, possibly due to increased cell death. Increased p-ERK levels in SMAD2/GATA4-knockdown cells suggest that cell viability may be reduced by ERK-dependent apoptosis. These data are consistent with recent studies in which ERK signaling induces apoptosis in HEK 293 cells 57 . Interestingly, Angiotensin II treatment on SMAD2/GATA4-knockdown cells can rescue reduced cell viability and restore the phosphorylation level of ERK. In addition, ACE overexpression decreased p-ERK in SMAD2/GATA4-knockdown cells. Together, our work in Drosophila and human cells provides insight into the possible regulation of ACE expression by these two transcription factors in mammals.
Electrophoretic Mobility Shift Assays (EMSA). EMSA was carried out with GST-Pnr-DBD, PCR products of Ance 533 bp enhancer, 5′ Ance enhancer, 3′ Ance enhancer, and biotin-labeled oligonucleotides with potential GATA binding sites. DNA binding domain of pnr cDNA was PCR-amplified and cloned into pGEX-4T1 vector using one-step SLIC method 58 . After transformation of pGEX-4T1-Pnr-DBD into BL21 competent cells, GST-Pnr-DBD protein was purified by using GST bead. 533 bp full-length enhancer, 5′ enhancer, and 3′ enhancer of Ance-lacZ enhancer were amplified by PCR reaction. 1 ng of Ance enhancer PCR products was incubated with 0.5 or 1 μg of GST-Pnr-DBD proteins in binding buffer (24 mM HEPES pH 7.9, 8 mM Tris pH 8, 2 mM EDTA, 12% glycerol, and 50 ng/μl Salmon sperm DNA) for 30 min at room temperature. After incubation, DNA-protein complexes were resolved on a 6% native PAGE gel for 1 h at 70 V in 0.5X TBE. DNA-protein complexes on native gel were visualized by RedSafe.
For EMSA with biotin-labeled oligonucleotides and GST-Pnr-DBD, biotin-labeled probes with potential GATA sites were synthesized from Bioneer (Korea). EMSA was carried out according to the method described above. After electrophoresis, DNA-protein complexes were transferred to nylon membrane. Further steps were proceeded using LightShift Chemiluminescent EMSA kit (Thermo Scientific) according to the manufacturer's instruction.
Immunoprecipitation. Drosophila larval lysates were prepared in CHAPS buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% CHAPS, and protease inhibitor cocktail), and the lysates were cleared by centrifugation. The cleared lysate was incubated with 1 μg of goat anti-Mad antibody (Santa Cruz) for 1 hour at 4 °C. The mixture of lysate and antibody was incubated with 30 μl of Protein G Dynabead (Thermo Scientific) for overnight at 4 °C. After several rounds of washing in CHAPS buffer, proteins were eluted in SDS sample buffer by boiling at 94 °C for 5 min. Mouse anti-Pnr antibody and secondary anti-mouse antibody conjugated with HRP (Jackson ImmunoResearch) were used for western blotting.
Cell culture and transfection. Drosophila S2 cells were cultured in M3 media (Sigma) with 10% insect medium supplement (Sigma). Transfection was carried out with Effectene reagent (Qiagen, Hilden, Germany) according to manufacturer's instructions.
HEK293 cells were cultured in Dulbecco's modified Eagles medium (DMEM) with 10% fetal bovine serum (FBS). For siRNA-mediated gene silencing, siRNA was transfected into HEK293 cells by reverse transfection using Lipofectamin 2000 to a final concentration of 10 μM. siRNAs were generated from Bioneer (Korea). siRNA sequences are shown in Table S1. Total RNA of siRNA-transfected cells was extracted 48 hours after transfection by Qiazol (Qiagen, UAS) according to the manufacturer's instruction. For plasmid transfection, pCEP4 SMAD2 (Addgene, #16484) and tetO-GATA4 (Addgene, #46030) were used. cDNA was synthesized using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Realtime PCR Master Mix (Toyobo, Japan) was used to perform qRT-PCR assays. qRT-PCR primer sequences are shown in Table S2.
ACE vector (Sino Biological, HG11598-UT) was used for overexpression experiment and western blot analysis, and Angiotensin II (Sigma-Aldrich, A9525) was treated for western blot analysis and cell viability assay. For western blot analysis, rabbit anti-ERK (Cell Signaling, 4595 T) and rabbit anti-phospho ERK (Cell Signaling, 3101 S) were used.
ChIP assays. ChIP assays were performed by using a standard protocol (Millipore) with mouse anti-GATA4 (Santa Cruz, sc-25310 X), control mouse IgG, rabbit anti-SMAD2 (Cell Signaling, 5339 S), or control rabbit IgG. Eluted DNAs were analyzed by qPCR. The primer sequences are provided in Supplementary Table S3.
Cell viability assay. Approximately 1000 cells were plated onto 96 well plates in 10% FBS media. The culture media were exchanged using 0.5% FBS media, and Ang II was treated at indicated concentration 24 hours after plating. Relative cell viability was measured 24 hours after Ang II treatment using CellTiter-Glo Luminescent Cell Viability Assay (Promega) according to the manufacturer's instructions. Statistical analysis. Statistical significance was determined by unpaired one-tailed Student's t-test. P-values of <0.05 were considered as statistically significant. All data represent the mean ± s.e.m. (error bars).