Regulation of Dpp activity by tissue-specific cleavage of an upstream site within the prodomain

Abstract

BMP4 is synthesized as an inactive precursor that is cleaved at two sites during maturation: initially at a site (S1) adjacent to the ligand domain, and then at an upstream site (S2) within the prodomain. Cleavage at the second site regulates the stability of mature BMP4 and this in turn influences its signaling intensity and range of action. The Drosophila ortholog of BMP4, Dpp, functions as a long- or short-range signaling molecule in the wing disc or embryonic midgut, respectively but mechanisms that differentially regulate its bioactivity in these tissues have not been explored. In the current studies we demonstrate, by dpp mutant rescue, that cleavage at the S2 site of proDpp is required for development of the wing and leg imaginal discs, whereas cleavage at the S1 site is sufficient to rescue Dpp function in the midgut. Both the S1 and S2 site of proDpp are cleaved in the wing disc, and S2-cleavage is essential to generate sufficient ligand to exceed the threshold for pMAD activation at both short- and long-range in most cells. By contrast, proDpp is cleaved at the S1 site alone in the embryonic mesoderm and this generates sufficient ligand to activate physiological target genes in neighboring cells. These studies provide the first biochemical and genetic evidence that that selective cleavage of the S2 site of proDPP provides a tissue-specific mechanism for regulating Dpp activity, and that differential cleavage can contribute to, but is not an absolute determinant of signaling range.

Introduction

Bone morphogenetic protein4 (BMP4), and its Drosophila ortholog, decapentaplegic (Dpp), play diverse roles during development (Nakayama et al., 2000). Many of the developmental functions of BMP4 and Dpp, including establishment of the dorsal-ventral axis, patterning of appendages, and induction and/or patterning of the heart, gut and other organs, are evolutionarily conserved.

BMP4/Dpp functions as a morphogen in many tissues, meaning that the ligand is secreted from a restricted source, forms a concentration (or activity) gradient in the target tissue and activates target genes to specify distinct cell fates in a dose dependent fashion (Kicheva and Gonzalez-Gaitan, 2008). Numerous studies have shown that either too much, or too little BMP leads to defects in embryonic patterning, indicating that strict control of BMP dosage is essential for normal development.

The bioactivity of BMP4/Dpp is regulated at multiple levels, including post-translationally, at the level of proteolytic activation. BMP4/Dpp is synthesized as an inactive precursor that is cleaved by members of the proprotein convertase (PC) family (Kunnapuu et al., 2009; Nelsen and Christian, 2009) to yield the active, mature protein. In mammals, seven members of the PC family have been identified (Thomas, 2002). The best-characterized PC, furin, activates proproteins following the preferred consensus sequence -RXR/KR-, but can also cleave following the minimal sequence –RXXR-. In Drosophila, two furin genes, DFur1 and DFur2, have been identified but substrate preferences have not been defined (Roebroek et al., 1991; Roebroek et al., 1992).

Our previous work has shown that BMP4 is sequentially cleaved at two sites within its prodomain and that ordered proteolysis regulates the steady state level of mature ligand that is available for signaling in vivo (Cui et al., 2001). ProBMP4 is initially cleaved at an optimal furin motif adjacent to the mature ligand domain (the S1 site) and is subsequently cleaved at a minimal furin motif (the S2 site) within the prodomain. In Xenopus embryos, mature BMP4 cleaved from an exogenous precursor engineered to carry a non-cleavable S2 site accumulates at lower levels and thus signals over a shorter range than does the exact same mature ligand cleaved from native precursor (Cui et al., 2001). Primary and upstream furin cleavage motifs are conserved in all known vertebrate BMP2 and BMP4 precursor proteins, and in Drosophila Dpp. Thus, it is likely that sequential cleavage of proBmp4/Dpp is a conserved mechanism for regulating BMP levels.

Biochemical analysis has revealed that differential cleavage of proBMP directs intracellular trafficking of the ligand to either degradatory or secretory/recycling pathways, thereby providing insight into how proprotein maturation regulates the signaling range of mature BMP4 (illustrated in Fig. 1) (Degnin et al., 2004). Specifically, we have shown that cleavage of proBMP4 at the S1 site, which is presumed to occur in the trans-Golgi network (TGN), generates a non-covalently associated prodomain/ligand complex (Degnin et al., 2004). Subsequent cleavage at the S2 site, which is likely to occur in a post-TGN compartment, liberates mature BMP4 from the prodomain. This generates a relatively stable ligand that can signal at a distance (Fig. 1A). By contrast, if the S2 site is not cleaved, the prodomain/ligand complex remains intact. Although this complex can still be secreted, and can signal in Xenopus embryos, it is targeted for rapid lysosomal degradation, either within the biosynthetic pathway prior to secretion from signal sending cells (Fig. 1B, upper panel), or following receptor mediated endocytosis into neighboring cells (lower panel). The relative bioactivity and signaling range of a given ligand is determined by the balance between the rates of production, diffusion (or transport), and degradation of the protein (Kicheva and Gonzalez-Gaitan, 2008). Although there is no evidence that failure to cleave proBMP4 at the S2 site affects the intrinsic activity (Kunnapuu et al., 2009) or diffusion of the mature ligand, it strongly enhances its degradation due to prodomain mediated lysosomal targeting (Degnin et al., 2004). This lowers the steady state level of ligand, and thus its ability to signal at a distance.

Fig. 1.

Model for regulation of BMP4 signaling by sequential cleavage. (A) When proBMP4 is sequentially cleaved at the S1 and then the S2 site, the mature ligand is released from the cleaved prodomain and is relatively stable, enabling it to signal to distal cells. (B) When proBMP4 is cleaved at the S1 site alone, the mature ligand remains noncovalently associated with the prodomain and this complex is targeted for degradation in the lysosome (blue shape) either directly within the biosynthetic pathway of synthesizing cells (upper panel) or via endocytic targeting in signal receiving cells (lower panel). This lowers the rate of ligand production at the source and/or enhances the rate of ligand degradation in receiving cells. As a result, BMP4 generated by cleavage at the S1 site alone signals is present at lower steady state levels.

Analysis of mice carrying a point mutation that allows for cleavage at the S1 site (so that mature ligand is still generated) but prevents processing at the upstream S2 site has shown that cleavage of the S2 site is essential for normal development and suggests that this site might be selectively cleaved in a tissue-specific fashion (Goldman et al., 2006). These mice exhibit severe loss of BMP4 activity in some tissues, such as testes and germ cells, whereas other tissues known to be sensitive to BMP4 dosage, including the limb, dorsal vertebrae and kidney, develop normally. Mice carrying a single cleavage mutant allele of Bmp4 in a null mutant background die during embryogenesis due to defects in multiple organ systems.

Our studies in mouse and Xenopus support the hypothesis that tissue-specific cleavage of the S2 site influences BMP4 bioactivity in vivo. These analyses are limited, however, by the presence of a redundant ligand, BMP2, in vertebrates. Furthermore, analysis of cleavage site usage in various tissues has not been feasible due to the lack of antibodies capable of detecting endogenous cleaved prodomain on Western blots. In the current studies, we take advantage of the strengths of Drosophila genetics to test whether differential cleavages within the prodomain of Dpp contribute to the previously documented ability of Dpp to function at long- or short-range in the wing disc and gut, respectively.

In the embryonic midgut, Dpp RNA is detectable only within the visceral mesoderm of parasegment 7, while the protein is found within these same cells and also outside of adjacent endodermal cells (Panganiban et al., 1990; St. Johnston and Gelbart, 1987). Genetic analysis confirms that Dpp signals to its immediate neighbors in the midgut, where it regulates transcription of homeotic genes, such as labial, that are essential for proper gut morphogenesis (Immergluck et al., 1990; Panganiban et al., 1990; Reuter et al., 1990). By contrast, in the wing imaginal disc, Dpp is expressed in a stripe of cells adjacent to the anterior-posterior compartment boundary of the wing disc but is required for normal growth and patterning of cells in both compartments (Basler and Struhl, 1994). A Dpp activity gradient can be detected by assaying for expression of immediate target genes, or by visualizing the phosphorylated form of Mothers against Dpp (pMAD), a cytoplasmic transducer of Dpp activity (Lecuit and Cohen, 1998; Nellen et al., 1996; Tanimoto et al., 2000). An overlapping, long range Dpp ligand gradient has also been visualized using GFP-Dpp fusion proteins (Belenkaya et al., 2004; Entchev et al., 2000; Teleman and Cohen, 2000). Mutational analysis and immunostaining for endogenous Dpp confirm that Dpp can signal multiple cells away from its source in the wing (Bangi and Wharton, 2006; Gibson et al., 2002).

In the current studies, we demonstrate that Dpp synthesized in the embryonic mesoderm is cleaved at the S1 site alone, and that this is sufficient for Dpp function in the midgut. By contrast, both the S1 and the S2 site of Dpp are cleaved in the wing imaginal disc, and this is essential to generate high steady state levels of mature Dpp that are sufficient to support normal growth and patterning of this tissue. Cleavage of Dpp at the upstream, S2 site is required for both short- and long-range signaling in the wing, but is not required for short-range signaling in the gut. Our results demonstrate that differential cleavage of the S2 site provides a tissue-specific mechanism for regulating Dpp activity that can contribute to, but is not an absolute determinant of signaling range.

Materials and Methods

cDNA constructs

The PCR-based splicing by overlap extension technique (Horton et al., 1990) was used to insert an HA-epitope tag in frame within the prodomain of Dpp following the codon for 237^th amino acid (-LFNMK[HA]RPPKI-), and to insert 6 tandem Myc-epitope tags in the mature domain, 12 amino acids downstream of the cleavage site (-GGKGG[6Xmyc]RNKRQ-). PCR based mutagenesis was used to introduce point mutations at the S1 and/or S2 cleavage site. All cDNAs were subcloned into vectors suitable for in vitro transcription of RNA and transient transfection in mammalian cells (pCS2+) or for expression in Drosophila cell lines and embryos under the control of the GAL4 responsive promoter, UAS (pUASg).

Transient transfection of S2 cells

cDNAs encoding Dpp precursor proteins were transfected into Drosophila S2 cells along with a ubiquitous driver (Actin-Gal4) using Lipofectamine 2000 (Invitrogen). 72 hours post-transfection, cells were lysed and supernatants collected and TCA precipitated for Western blot analysis as described (Goldman et al., 2006).

Heparin binding assay

HEK cells were transiently transfected with myc-tagged Xenopus BMP4, BMP4 RKK (Degnin et al., 2004), or DPP expression plasmids using Lipofectamine 2000 (Invitrogen) and cultured in 1 ml of OptiMEM-I (Invitrogen) for 48 hrs. Proteins were TCA precipitated from 300 l of conditioned medium. The remaining 700 μl of medium was dialyzed in urea buffer (20mM Tris, pH 8.0, 4 M urea) overnight and then incubated with heparin-sepharose (GE Healthcare) at 4°C for 3hr as described (Ohkawara et al, 2002). Beads were washed with urea buffer three times, and then boiled in sample loading buffer to release bound proteins. Western blots of total and heparin bound proteins were probed with anti-myc antibody, 9E10 (Sigma) as described (Degnin et al., 2004) and immunoreactive proteins detected using Enhanced Chemiluminescence reagent (GE Healthcare).

Expression and analysis of Dpp in Xenopus embryos

Xenopus embryos were obtained, microinjected with synthetic capped RNA, and cultured as described (Moon and Christian, 1989). Embryonic stages are according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1967). Blastocoel fluid was aspirated and pooled from 10 embryos in each experimental group at stage 9, and residual embryos were homogenized in lysis buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 2.5% NP-40 and protease inhibitors) as described (Birsoy et al., 2005). Western blot analysis using antibodies specific for HA (12CA5) or myc (9E10) epitope tags was performed as described (Degnin et al., 2004).

In vitro cleavage assays

Drosophila or human furin protein was produced by infecting BSC40 cells with recombinant vaccinia virus harboring Dfur2 (Roebroek et al., 1992), the Dfur1-CRR splice form of Dfur1 (De Bie et al., 1995) or FLAG-tagged human furin (Molloy et al., 1994). Secreted forms of each enzyme were concentrated from the conditioned medium of infected cells as described (Cui et al., 1998). [³⁵S]Met/Cys-labeled proDPP and cleavage variants were synthesized in rabbit reticulocyte lysates and immunoprecipitated using myc-specific antibodies. In vitro digestion was carried out using recombinant human furin and Drosophila furins at pH 6.5 as described (Degnin et al., 2004).

Drosophila stocks

dpp^blk1-Gal4, MS1096-Gal4, and 24B-Gal4 stocks are described in Flybase. Ubx-Gal4 flies (Pallavi and Shashidhara, 2003) were obtained from Dr. Sashidhara, dpp^d8 and dpp^d10 strains (Zecca et al., 1996) were a gift from Konrad Basler, dpp^S4 strains were obtained from Mariann Bienz and CyO Dfd-YFP(w⁺) and TM6 Sb Tb Hu Dfd-YFP(w⁺) balancers were from the Bloomington stock center (Greg Beitel, unpublished).

Wing rescue

yw,dpp^d10/CyO;dpp^(blk1)-Gal4/MKRS were crossed to yw;dpp^d8/CyO;UAS>proHA-Dpp-Myc^WT/ MKRS or yw;dpp^d8/CyO;UAS>proHA-Dpp-Myc^S2KK/MRKS. The dpp^d8/dpp^d10;dpp^blk1-Gal4/UAS>proHA-Dpp-Myc adults were identified by the absence of CyO and MKRS. Larvae were reared at 30°C, for maximal Gal4 mediated protein expression. Photographs were collected using a Leica MZFL-III stereomicroscope and photographed with an Optronics Magna Fire CCD Camera.

Generation of marked clones

The MARCM (Mosaic analysis with a repressible cell marker) technique (Lee and Luo, 2001) was used to mark by GFP-expression clones of cells expressing proHA-Dpp-Myc^WT or proHA-Dpp-Myc^S2KK in the genotype MS1096-Gal4; y⁺ck FRT40A / tub>GAL80 FRT40A; UAS>GFP/UAS>Dpp-Myc^{WT or S2KK}. Clones were induced by 1h of heat shock at 38 C during first and second instar and wing imaginal discs were dissected from wandering 3^rd instar larvae for immunostaining.

Immunostaining and imaging of embryos

Flies of the genotype dpp^s4/CyO Dfd-YFP; UAS>proHA-Dpp-Myc^{WT or S2KK}/TM6 Sb Hu Dfd-YFP and dpp^s4/CyO Dfd-YFP; 24B-Gal4/TM6 Sb Hu Dfd-YFP were crossed and 12–18 hour embryos collected and stained with rabbit anti-Labial antibody (1:100; a kind gift from Thomas C. Kaufmann) and chicken anti-GFP (1:5000, Aves Laboratories), followed by Alexa488 and Alexa546 secondary antibodies, respectively. Embryos were processed and stained as described in Mahaffey and Kaufman (1987) with modifications as described in Diederich et al (1989). Embryos of the genotype dpp^s4/dpp^s4; UAS>proHA-Dpp-Myc^{WT or S2KK}/24B-Gal4 were identified by the absence of YFP expression in the head region (Greg Beitel, unpublished). Embryos were examined using an Olympus Fluoview FV1000 Confocal Laser Scanning Microscope.

Immunostaining and imaging of wing discs

Immunostaining of wing discs was performed as previously described (Wehrli et al., 2000). Primary antibodies were used at the following dilutions: mouse anti-myc, 9E10 1:500 (Sigma), rabbit anti-phosphoSMAD, 1:2500 (gift of Ed Laufer). Secondary antibodies (Alexa488, Alexa546 and Alexa633; Molecular Probes, Eugene, OR) were used at 1:5000. Images were collected using a Zeiss Axiovert LSM5 Pascal laser-scanning microscope.

Western blot analysis of wing discs and embryos

Wild type and mutant Dpp precursor proteins were expressed in the wing pouch under the control of MS1096-Gal4. Wing discs were dissected from 400 third instar larvae of transgenic or nontransgenic flies and lysed in RIPA buffer (50 mM Tris-Cl, pH 8, 150 mM NaCl, 1% NP-40, 0.5% deoxycholate, 0.1% SDS, with protease inhibitors). Proteins were deglycosylated using PNGaseF (New England Biolabs), with one duplicate sample from Dpp transgenic wing discs left untreated to verify the identity of bands corresponding to each cleavage product. Dpp precursor proteins were expressed in the mesoderm under the control of the 24B-Gal4 driver. Embryos (~100 μl) were collected 14–18 hours after egg laying, dechorionated, and lysed in RIPA buffer. Precursor proteins and cleaved mature domains were immunoprecipitated using anti-myc (9E10) antibody conjugated to protein A-agarose beads. Prodomain cleavage products were then immunoprecipitated from the supernatant using mouse monoclonal anti-HA antibody, 12Ca5 (Roche). Immunoprecipitated proteins were deglycosylated with PNGaseF. Western blots of wing disc extracts and immunoprecipitates from embryos were probed with rat monoclonal anti-HA antibody, 3F10 (Roche) and with anti-myc antibody, 9E10 (Sigma) as described (Degnin et al., 2004). Bands on Western blots were scanned and semi-quantitative analysis of relative expression levels of precursor proteins and cleavage products in different fly lines was performed using the Macintosh IP Lab Gel program.

Results

Dpp is cleaved at the S1 and S2 sites in vivo and in vitro

Sequence analysis of regions flanking the cleavage sites in BMP2, BMP4 and Dpp reveal a high degree of conservation across species (Fig. 2A). To determine whether both sites of proDpp can be cleaved by Drosophila PCs, we expressed cDNAs encoding epitope-tagged wild type Dpp (Dpp^WT) or a variant in which the furin consensus motif at the S2 site had been disabled (Dpp^S2KK, illustrated in Fig. 2B) in Drosophila S2 cells and analyzed cleavage products on Western blots. ProDpp^WT was efficiently cleaved at both sites whereas the Dpp^S2KK precursor was only cleaved at the S1 site to generate the same mature ligand but a more slowly migrating prodomain fragment (Fig. 2C). Recent studies have identified a third site within the mature domain of Dpp that is cleaved, but is not essential for function in vivo (Kunnapuu et al., 2009). This site is located immediately downstream of the myc-tag in our constructs, and thus we do not detect the cleavage fragment generated by use of this site.

Fig. 2.

Dpp is cleaved at two sites within the prodomain. (A) Alignment of sequences flanking the cleavage site(s) of BMP4 and BMP2 from human, chick and Xenopus (Xen), and DPP from Drosophila (fly). (B) Schematic illustration of cleavage sites in wild type and cleavage variant forms of proDpp. Shaded bar represents the prodomain, white bar the region between the two cleavage sites and black bar the mature ligand domain. (C) Dpp precursors were expressed in S2 cells and Western blots of cell media were probed with antibodies specific for the HA-tag in the prodomain or the myc-tag in the mature-domain. Bands corresponding to uncleaved precursor, prodomain following cleavage at the S1 site, prodomain following cleavage at the S2 site, and mature ligand are indicated schematically to the right of the gel. The same bands were observed in Western blots of cell lysates. (D) Radiolabeled native or cleavage mutant forms of proDpp were incubated with recombinant Dfurin or Hfurin for the times indicated. All results were reproduced in at least three independent experiments.

Ordered cleavage of BMP4 is driven, in part, by the presence of optimal (-RXK/RR-) and minimal (-RXXR-) furin consensus motifs at the S1 and the S2 site, respectively (Degnin et al., 2004). The order of these motifs is reversed in Dpp (Fig. 2A), raising the possibility that the preferred motif for cleavage by Drosophila furin (Dfurin) is –RXXR-, or that Dpp cleaves the S2 site first, and then the S1 site, as recently suggested (Kunnapuu et al., 2009). To test these possibilities, we compared the relative rate of cleavage of each site on Dpp precursor proteins that contain native or altered cleavage motifs, as illustrated in Fig. 1B. We used in vitro assays to do so, since cleavage at the two sites is rapid and tightly coupled such that only the final, fully cleaved products are detected when assayed in vivo (Fig. 2C). Radiolabeled precursors were incubated with recombinant Dfurin1 or human furin (Hfurin), and cleavage products were analyzed at increasing time intervals. Products corresponding to cleavage at the S1 and S2 sites of proDpp^WT appeared simultaneously at ~3–10 minutes of incubation with either PC, and the same was true for Dpp variants that contained either two optimal (Dpp^S1R) or two minimal (Dpp^S2I) furin motifs (Fig. 2D). By contrast, in the case of ProDpp^S1R/S2I, in which the cleavage sites are reversed so as to resemble those in BMP4, a band corresponding to cleavage at the optimal motif at the S1 site was initially observed, followed by the appearance of an S2-cleaved product. The same cleavage patterns were observed when Dpp was incubated with recombinant Dfurin2 (data not shown). These results suggest that Dfurin1 and Dfurin2 have the same substrate selectivity as Hfurin, and prefer to cleave following an -RXK/RR- motif, at least in vitro. It is possible that the S1 and S2 sites of proDpp are cleaved simultaneously in vivo, as they are in vitro, or that the conformation of the protein and/or sequences adjacent to the cleavage site drive sequential cleavage of Dpp in vivo.

ProDpp containing a minimal or no cleavage motif at the S2 site generates less mature ligand

To begin to examine the roles of optimal and minimal cleavage motifs in generating biologically active mature Dpp in vivo, we compared the activity of ligands cleaved from wild type and variant precursors using a Xenopus embryo ventralization assay. Previous studies have shown that Dpp has the same activity as its vertebrate ortholog in mammalian bone forming assays, and the above data demonstrate that vertebrate and fly furin have the same substrate selectivity, thus validating the use of a simple vertebrate model system to assay bioactivity (Sampath et al., 1993). Equivalent amounts of RNA encoding native or cleavage variant forms of Dpp were injected into four-cell Xenopus embryos near the dorsal midline. At the tailbud stage, embryos were scored for Dpp-mediated loss of dorsal structures using the dorsoanterior index (DAI) scale (Kao and Elinson, 1988) in which five signifies normal patterning (no ectopic Dpp activity), while zero signifies loss of dorsal and anterior structures (maximal ectopic Dpp activity; illustrated above Table 1). The results of one representative experiment, in which individual embryos were scored and an average DAI value calculated, are shown in Table 1, and all results were reproduced in at least three separate experiments. First, we demonstrated that native Dpp can ventralize Xenopus embryos and that the presence of HA- and myc-epitope tags does not significantly interfere with in vivo bioactivity (Table 1). We also showed, using an in vitro heparin binding assay, that the myc tags present in the mature domain of Dpp do not interfere with the ability of mature Dpp to bind heparan sulfate proteoglycans (HSPGs) (Supplemental Figure S1), since this binding is important for ligand stability, transport and signaling activity (Belenkaya et al., 2004). Next we compared the in vivo activity of Dpp-myc generated from native and cleavage variant precursors. Dpp-myc cleaved from the native precursor, or from a precursor with two optimal furin motifs (Dpp^S1R) generated the highest ventralizing activity. Dpp-myc generated from a precursor with an optimal and then minimal furin motif (analogous to vertebrate BMP4; Dpp^S1R/S2I) showed slightly less activity, while that cleaved from Dpp^S2KK, which can only be cleaved at the S1 site, or proDpp^S2I, which contains minimal furin motifs at both sites, generated very little or no activity.

Table 1.

In vivo activity of DPP generated from wild type and cleavage variant precursors in Xenopus embryos.

RNA	Dose	DAI	n
DPPWT	250ng	2.2	41
DPPHA-MycWT	250ng	2.9	45
DPP HA-MycWT	500ng	2.4	71
DPP HA-MycS2KK	500ng	3.9	51
DPP HA-MycS2I	500ng	4.9	60
DPP HA-MycS2I/S1R	500ng	3.0	54
DPP HA-MycS1R	500ng	2.6	47

We next asked whether the differences in bioactivity generated by native and variant Dpp precursors reflect differences in the amount of ligand that is produced. Wild type or variant proDpp was expressed in Xenopus embryos and levels of precursor protein and cleavage products were analyzed by Western blot. All precursor proteins were present at similar levels but levels of mature Dpp varied in a fashion that correlated well with the relative in vivo bioactivity generated by each precursor (Fig. 3). Approximately equivalent levels of prodomain and mature ligand fragments were detected in embryos injected with proDpp^WT, proDpp^S1R or proDpp^S1R/S2I. By contrast, levels of S1 only-cleaved prodomain and mature ligand were significantly lower in embryos injected with proDpp^S2KK, suggesting that failure to cleave the S2 site of DPP targets the ligand for lysosomal degradation, similar to what is observed for BMP4. Steady state levels of prodomain and mature ligand cleaved from proDpp^S2I were also significantly reduced. The observation that ligand levels and vivo activity generated from proDpp^S2I, in which the S2 site is converted from an optimal to a minimal motif, are equally or even more severely reduced than those generated from proDpp^S2KK was unexpected. These findings demonstrate that not only the presence or absence, but also the identity of the cleavage motif at the S2 site can influence in vivo activity. It is possible that the presence of two minimal cleavage motifs in proDppS2I leads to simultaneous, rather than sequential processing and that this, in turn disrupts protein folding, trafficking and/or stability. Alternatively, the lack of an optimal furin motif at either the S1 or the S2 site may slow the kinetics of cleavage to the point that the precursor is recognized as aberrant and is degraded. Consistent with the latter possibility, the presence of a minimal rather than an optimal furin cleavage motif in two otherwise identical chimeric BMP7/4 precursors significantly reduces the amount of cleaved ligand that is generated (Sopory et al., 2006). Further analysis will be required to test these possibilities.

Fig. 3.

Effect of optimal and minimal cleavage motifs on production of mature Dpp in Xenopus embryos. RNA (2ng) encoding Dpp precursor proteins was injected into two-cell Xenopus embryos. Levels of precursor and cleavage products were examined by Western blot of embryonic extracts (lysate) or blastocoele fluid (BF) collected at the blastula stage from uninjected (UI) embryos or Dpp-expressing embryos. Bands corresponding to uncleaved precursor, prodomain fragments and mature ligand are indicated to the right of the gel. Results were reproduced in three independent experiments.

Cleavage of the S2 site is required for DPP function in wing or leg development

To test whether cleavage of the S2 site of Dpp is required in the context of its normal function in wing development, we compared the ability of mature ligand cleaved from Dpp^WT or Dpp^S2KK precursors to replace Dpp in this process. Dpp precursors were expressed in their endogenous domain under the control of a disc-specific Dpp-Gal4 driver in dpp^d8/dpp^d10 mutant flies. Adults of this genotype have truncated legs, tiny winglets and small eyes (Fig. 4D–F) relative to wild type flies (Fig. 4A–C). Dpp-Myc cleaved from wild type precursor completely rescued eye and leg development (Fig. 4G, I), and partially restored growth and patterning of the wing (Fig. 4H) in all animals examined, whereas that cleaved from proDpp^S2KK never rescued leg or wing development to any extent (Fig 4J, K). By contrast, Dpp-Myc cleaved from proDpp^S2KK partially, but reproducibly rescued the small eye phenotype observed in the dpp^d8/dpp^d10 animals (compare Fig. 4F and L). This was an unexpected observation and raises the possibility that cleavage at the S1 site alone is sufficient to support development of the eye disc. It is equally plausible, however, that eye development is merely less sensitive to Dpp levels than is wing development. Because the range of Dpp signaling in the eye has not been as well characterized as in the wing disc, we have not pursued further analysis in the eye disc in the current studies, but these are interesting questions for future studies. Collectively, these results show that cleavage of proDpp at the S2 site is required for normal growth and patterning of the wing and leg disc.

Fig. 4.

Cleavage of the S2-site is required for normal wing development. Normal leg (A), wing (B) and eye (C) formation in wild type (wt) flies is contrasted with that in dpp^d8/dpp^d10 mutant flies that have truncated legs (D) tiny winglets (E) and ventrally receding eyes (F). Expression of UAS-Dpp^WT-myc under the control of dpp-Gal4 fully rescued leg (G) and eye (I), and partially rescued wing development (H) in all dpp^d8/dpp^d10 mutants. UAS-Dpp-myc^S2KK did not rescue the truncated legs (J) or missing wings (K) in any flies but partially rescued ventral eye development in almost all flies examined (L). Double arrows (C, F, I and L) indicate the distance from the ventral eye margin to the bristles in wild type, mutant and rescued animals. Arrows (A, G) indicate the claws whereas arrowheads (D, J) highlight distal truncation with loss of tarsal segments. Co, cox; fe, femur; ti, tibia; ta, tarsus with segments I–V.

To further test whether cleavage at the S2 site is required for Dpp to signal in the wing we expressed Dpp^WT or Dpp^S2KK precursors under the control of Ubx-Gal4, which drives expression in all cells of the wing disc proper in first instar larvae, but becomes restricted to cells in the overlying layer of squamous epithelium (the peripodial epithelium) by the second instar (Pallavi and Shashidhara, 2003). Expression of wild type Dpp led to robust activation of pMAD in all cells of the disc proper (Supplementary Fig. S2A) and the peripodial epithelium (A’), and caused a massive overgrowth of both cell layers. By contrast, ligand generated from Dpp^S2KK induced only a minimal increase in pMAD in the disc proper (Fig. S2B) and no change in the growth properties of either cell layer.

Cleavage of Dpp at the S2 site is required for activation of pMAD in most cells of the wing disc

Dpp normally signals across multiple cells in the wing disc. Our observation that proDpp^S2KK cannot rescue wing development in mutants is consistent with the possibility that S2 cleavage is required for non-cell autonomous signaling, and that this is essential for normal growth and patterning of the wing. By contrast, S1 cleavage may be sufficient to enable Dpp to signal cell autonomously, but this may be inadequate for wing development. To examine these possibilities, we used the MARCM system to generate GFP-marked clones of cells ectopically expressing proDpp^WT or proDpp^S2KK in the wing disc. Wing discs were immunostained for pMad as a sensitive indicator of Dpp activity. Dpp^WT induced a robust increase in pMAD both within, and also multiple cells outside of the borders of precursor-expressing clones (Fig. 5A’, arrows). By contrast, ectopic expression of proDpp^S2KK led to minimal or no induction of pMAD either within or outside of clones located in central portions of the wing disc (Fig. 5B’). The absence of pMAD even in clones that ectopically express proDpp^S2KK at high levels demonstrates that ligand activity is very tightly regulated and that signaling at both short- and long-range is entirely dependent on S2 cleavage in the wing. Notably, when large clones of cells expressing proDpp^S2KK were generated using the flip-out technique, localized overgrowth of the wing disc was observed, but only when clones abutted the extreme periphery of the disc (Supplementary Fig. S3) where expression of the Dpp receptor thick veins is highest, and cells are sensitized to low doses of Dpp (Lecuit et al., 1998). In analogous experiments, flip out clones expressing proDpp^WT induced massive overgrowth of the entire wing blade, regardless of where the clones were located (data not shown). Thus, ligand synthesized from proDpp^S2KK is able to signal in a highly sensitized environment, but always at a lower level than that cleaved from proDpp^WT. Discs were also immunostained using an intracellular staining protocol that detects primarily the myc-tagged precursor protein (Supplementary Fig. S4) or using an extracellular staining protocol to detect secreted myc-tagged mature Dpp (data not shown). No differences were detected in signal intensity within cells made to express wild type or S2-mutant precursor suggesting that both precursors were expressed at a similar level (also see the more quantitative analysis in Fig. 7) but we were unable to detect extracellular protein expressed from either precursor. These results suggest that cleavage at the S2 site is essential to generate physiological levels of ligand that, at steady state, reach the threshold required to activate pMAD at both short- and long-range in cells that normally respond to endogenous Dpp. Notably, ligand synthesized from overexpressed proDpp^S2KK is capable of activating signaling, but only outside of the endogenous Dpp activity domain, in cells that are highly sensitized to pathway activation.

Fig. 5.

Cleavage of Dpp at the S2 site is required for activation of pMAD in the wing disc. Wing discs with clones of cells (marked by GFP) expressing proDpp^WT (A–A’’) or proDpp^S2KK (B–B’’) were immunostained to detect pMAD. Endogenous pMAD staining in the absence of ectopic Dpp is also shown (C’, C’’). Representative clones outside of the endogenous dpp-expression domain are outlined. Arrows denote induction of pMAD outside of clones.

Fig. 7.

The S2 site of proDpp is selectively cleaved in the wing disc, but not in embryonic mesoderm. Western blots of proteins from pooled wing disc extracts or from embryo immunoprecipitates derived from non-transgenic flies (none) or from transgenic flies expressing proDPP^WT or proDPP^S2KK were probed with antibodies specific for the HA-tag in the prodomain or the myc-tag in the mature domain of each precursor protein. Bands corresponding to uncleaved precursor, prodomain fragments and mature ligand are indicated to the right of the gel. The top and middle panels on the left are short and long exposures, respectively, of the same Western blot. The upper mature ligand bands detected with the anti-myc antibody are most likely due to incomplete deglycosylation (based on predicted molecular weight of S1 cleaved mature Dpp and comparison with duplicate protein extracts that had not been deglycosylated with PNGaseF, as indicated at the top of the gel). Bands indicated by an asterisk are most likely precursor degradation products. The arrowhead indicates the faint band corresponding to S1-only cleaved prodomain generated from DPP^WT in the wing disc.

Cleavage of the S2 site is not essential for Dpp to signal from the embryonic mesoderm to the gut epithelium

We next asked whether cleavage at the S2 site is also required in the embryonic midgut, where Dpp expressed in the visceral mesoderm (VM) travels to the adjacent cell layer to induce expression of labial in immediately adjacent endodermal cells (Panganiban et al., 1990; Reuter et al., 1990). We ectopically expressed Dpp^WT or Dpp^S2KK precursors throughout the mesoderm of wild type flies under the control of the 24B-Gal4 driver, since we were unable to generate appropriate driver lines to express dpp within the endogenous expression domain. Embryos were then immunostained for Labial at stage 13–14. Consistent with previous studies (Staehling-Hampton and Hoffmann, 1994), endogenous Labial was expressed in a stripe encompassing 2.99 (±0.64) cells within the endoderm, and this domain was expanded to encompass 4.55 (±0.66) cells in embryos made to express proDpp^WT (Fig. 6A–B, D). Expression of proDpp^S2KK had an identical effect, expanding the stripe of labial expression to encompass 4.68 (±0.78) cells (Fig. 6C–D).

Fig. 6.

Cleavage at the S2 site is not required for ectopic expression of labial in the midgut primordium. (A–C) Stage 14–15 nontransgenic embryos (WT), or those expressing Dpp^WT or Dpp^S2KK precursors using the mesodermal driver 24B-Gal4 were immunostained to detect Labial expression in the midgut endoderm. (D) Quantification of labial stripe width. Values represent the mean number (+/− s.d.) of labial expressing cells in a stripe. The width of the labial stripe in WT embryos was significantly different than that in embryos expressing Dpp^WT or Dpp^S2KK precursors (p<0.05) but there was no significant difference in the number of labial cells between embryos expressing Dpp^WT or Dpp^S2KK precursors. (E,F) dpp^S4 homozygous embryos expressing either proDpp^WT or proDpp^S2KK immunostained to detect Labial expression in the endoderm.

Dpp is known to positively regulate its own expression, raising the possibility that ligand cleaved from overexpressed precursor proteins induced ectopic Labial indirectly, via induction of endogenous dpp expression within the VM, as opposed to directly, by traveling from the VM to adjacent endodermal cells. To distinguish between these possibilities, we repeated the above experiment in dpp^S4 mutant flies that lack expression of Dpp in the VM and consequently do not express labial in the midgut epithelium (Immergluck et al., 1990). Labial expression was rescued in transgenic mutants expressing either proDpp^WT or proDpp^S2KK in the mesoderm (Fig. 6E, F) indicating that both constructs are functional in the gut. Collectively, these data show that cleavage of Dpp at the S1 site alone generates sufficient ligand to activate physiological target genes in neighboring cells in the gut.

ProDpp is cleaved at both sites in the wing disc but only at the S1 site in the gut

The above results are consistent with the hypothesis that the S2 site of proDpp is cleaved tissue-specifically, in the wing disc but not in the embryonic mesoderm. To test this possibility, we compared cleavage of Dpp expressed in the wing disc or in the embryonic mesoderm by Western blot. Dpp^WT and Dpp^S2KK precursor proteins were ubiquitously expressed in the wing pouch under the control of the MS1096-Gal4 driver, or in the embryonic mesoderm, under the control of 24B-Gal4. Western blots of cell lysates from dissected wing discs or embryos were probed with antibodies specific for the HA-tag present in the prodomain, and for the Myc-tag present in the mature domain of Dpp. In both the wing disc and in the embryo, proDpp^S2KK was cleaved to generate a single band that serves as a size control for S1-only cleaved prodomain, along with the mature ligand (Fig. 7). In the wing disc, proDpp was cleaved at both the S1 and the S2 sites to generate a predominant, more rapidly migrating prodomain band, although a trace amount of S1-only cleaved prodomain was also detected (Fig. 7, arrowhead). By contrast, in the embryonic mesoderm, native Dpp precursor protein was cleaved primarily at the S1 site alone to generate a prodomain fragment that co-migrated with that cleaved from proDpp^S2KK. Only a minor fraction of wild type Dpp was cleaved at both the S1 and the S2 sites in embryonic mesodermal cells. These results provide the first biochemical evidence that cleavage of the S2 site of Dpp/BMP4 is regulated in a tissue-specific fashion in vivo. Strikingly, although equivalent levels of wild type and S2-mutant precursor proteins were detected in the wing disc, steady state levels of cleaved prodomain and mature ligand generated from proDpp^S2KK were approximately four-fold lower than those generated from the native precursor, as determined by densitometric analysis of band intensity in three separate experiments. This result is consistent with the possibility that failure to cleave the S2 site of proDpp in the wing disc leads to targeted degradation of a prodomain/mature ligand complex, as has been shown for vertebrate BMP4 (Degnin et al., 2004). By contrast, in the embryonic mesoderm, proDpp^S2KK and proDpp^WT generate relatively equivalent steady state levels of mature ligand, when normalized to the level of precursor in each pool. This is as predicted since wild type Dpp is cleaved predominantly at the S1 site alone in these cells, as is proDpp^S2KK and thus cleavage products generated from each precursor are expected to turn over at the same rate. Notably, while this analysis provides a comparison of relative levels of cleavage products generated by each precursor in a given tissue, it is not valid to compare steady state levels of cleavage products observed in the wing versus those in the gut.

Discussion

Endogenous Dpp functions as a morphogen, to instruct cells of their fate in a concentration-dependent fashion. Many studies have described the critical role that extracellular factors play in shaping the Dpp activity gradient by facilitating or inhibiting Dpp movement and/or by modulating the ability of signal receiving cells to perceive a given concentration of ligand (O'Connor et al., 2006). The Dpp receptor Thick veins, for example, sensitizes cells to low levels of Dpp but also limits Dpp movement when present at high levels. Similarly, the secreted Dpp binding proteins Short gastrulation and Twisted gastrulation inhibit Dpp from binding to Thick veins, thereby blocking local signaling but facilitating long-range diffusion. The current studies demonstrate that proteolytic activation of the precursor in signal sending cells also contributes to differences in Dpp bioactivity in various tissues. Specifically, in the wing disc, Dpp is cleaved at both the S1 and the S2 site and the latter cleavage is essential to generate steady state levels of ligand that surpass the threshold for induction of pMAD at both short- and long-range, over much of the disc. By contrast, in the embryonic gut Dpp is cleaved only at the S1 site, and this generates sufficient ligand to activate physiological target genes in neighboring cells. Future studies will be required to determine whether cleavage at the S1 site alone is necessary to prevent aberrant spread of the ligand beyond immediately adjacent cells in the gut.

It is possible that the ability of proDpp^S2KK to rescue labial expression in the gut, and its inability to rescue growth and patterning in the wing reflect differences in the level of expression of precursor proteins due to the use of different drivers in the two tissues, rather than differences in use of the S2 site. This is unlikely, however, since precursor protein is always present in excess of cleavage products, while BMP4/Dpp signaling activity is determined by the rate of precursor folding, dimerization and cleavage, and by the rate of ligand turnover (Degin et al., 2004; Goldman et al., 2006; Sopory et al., 2006). In the current studies, when proDpp^S2KK was expressed in the wing using the MS1096-Gal4 driver, precursor protein was present at high levels and yet very low steady state levels of mature Dpp were detected. Furthermore, induction of pMAD was not observed in these wing discs, even when incubated at higher temperature to maximize GAL4 activity (data not shown). Thus, in the wing disc but not in the gut Dpp activity is tightly regulated by cleavage of the S2 site.

Several models have been proposed for how Dpp moves across cells in the wing, all of which invoke endocytic trafficking to generate or shape the gradient (Affolter and Basler, 2007). One model suggests that Dpp is actively transported by endocytosis, intracellular transport and exocytosis (Entchev et al., 2000) while other models postulate that Dpp moves by extracellular diffusion (Teleman and Cohen, 2000) and/or by extracellular transport facilitated by binding proteins such as Short gastrulation and Crossveinless (Shimmi et al., 2005; Vilmos et al., 2005). More recent studies argue that Dpp is passed from cell to cell after binding to the HSPGs, Dally and Dally-like (Dly) (Belenkaya et al., 2004). These studies reveal that endocytosis is required for signal transduction and for lysosomal degradation of Dpp, which shapes the activity gradient (Eldar et al., 2003; Teleman and Cohen, 2000). This latter finding provides a feasible explanation for why ligand generated by cleavage at the S1 site alone is unable to signal even at short range over most of the wing disc. Specifically, we propose that the ligand is preferentially targeted for degradation, and thus does not accumulate to high enough steady state levels to surpass those needed for signal activation.

Our previous studies have shown that the prodomain/ligand complex generated by cleavage of BMP4 at the S1 site alone is targeted to the lysosome, but it is not known whether this occurs directly from the biosynthetic pathway in signal sending cells, or following secretion and receptor-mediated endocytosis in signal receiving cells. Deletion of the heparin binding motifs on BMP4 partially stabilizes the ligand/prodomain complex generated by cleavage at the S1 site alone (Degnin et al., 2004), suggesting that degradation may occur following endocytosis in a process that is facilitated by binding to cell surface HSPGs, such as Dally. If Dpp is regulated in the same way, this might contribute to the ability of Dpp^S2KK to active pMAD when expressed in cells on the periphery of the wing disc since levels of Dally are at their lowest, and levels of Thick veins at their highest in these cells, thereby minimizing lysosomal targeting while maximizing signal activation. Although recent studies suggest that Dally disrupts, rather than enhances internalization of native Dpp in the wing (Akiyama et al., 2008), it may function differently in the context of a ligand/prodomain complex generated by cleavage at the S1 site alone. This complex might also be endocytosed after binding to crossveinless2/BMPER, which has been shown to direct receptor-mediated internalization and lysosomal targeting of BMP4 (Kelley et al., 2009). Alternatively, Dpp generated by cleavage of the precursor at the S1 site alone may be targeted directly for degradation within the biosynthetic pathway, prior to secretion, consistent with its inability to activate pMAD cell-autonomously in cells located in central regions of the disc. Analysis of Dpp maturation within or adjacent to clones of cells that are defective for components of the lysosomal trafficking machinery, or for HSPGs may shed light on this question and on the mechanisms by which the prodomain directs intracellular trafficking.

A recent independent analysis of Dpp maturation agrees with our finding that cleavage at the S2 site is essential to generate sufficient mature Dpp to support wing development (Kunnapuu et al., 2009). In this study, the distribution of mature Dpp was imaged directly, by extracellular staining of wing discs, and ligand cleaved from an S2-mutant precursor was not detected outside of cells. This is consistent with biochemical analysis showing that steady state levels of ligand generated from this precursor are very low. These authors concluded that cleavage of the S2 site occurs first, and is an obligate requirement for cleavage of downstream sites such that mature ligand cannot be generated from proDpp^S2KK. This remains a feasible explanation, since it is not possible to definitively identify the order of cleavage of Dpp in vivo, as discussed earlier. However, our data showing that the S1 site can be cleaved independently of the S2 site in Drosophila cell lines and embryos, and that cleavage of the S2 site is dispensable for Dpp function in the embryonic midgut support an alternate interpretation. Specifically, we propose that failure to cleave the S2 site in the wing disc leads to rapid degradation of mature Dpp generated by cleavage at the S1 site, consistent with our biochemical analysis of BMP4 processing and degradation in vertebrate embryos. Since we are unable to visualize the small ligand fragment noted in the studies of Kunnapuu et al. (2009), which is generated by cleavage N-terminal to the myc epitope tag in our construct, we cannot rule out the possibility that mutation of the S2 site affects processing at this third site, leading to differences in signaling efficacy. This seems unlikely, however, since cleavage of the third site was shown to be dispensable for Dpp function, at least in the wing (Kunnapuu et al., 2009).

Although we currently favor a model in which cleavage at the S2 site is regulated in a tissue-specific manner and this in turn determines the rate of degradation of the mature ligand in the wing versus the gut, it is possible that both sites are cleaved in the wing and the gut, and that these tissues instead differ in their ability to traffic or degrade S1-cleaved DPP. For example, in the wing cleavage products generated by S1-cleavage may be recognized by a trafficking receptor that targets them to the lysosome, whereas in the gut, this receptor may be absent. This model offers an alternate explanation for the observation that steady state levels of S1-only cleaved prodomain generated from either the wild type or S2KK-mutant precursor protein are very low in the wing, but relatively high in the gut. Specifically, this may reflect differential stability of the S1-cleavage products in the wing versus the gut, rather than differential use of the S2 cleavage site in the wild type precursor. However, our data showing that steady state levels of S2-cleaved prodomain generated from the wild type precursor are barely detectable in the mesoderm, but abundant in the wing cannot be accounted for by the differential stability model, and are more consistent with selective, tissue-specific cleavage of the S2 site. Conversely, our finding that equivalent levels of cleavage products are generated and secreted from cultured Drosophila S2 cells regardless of whether the S2 site is cleaved (Fig. 2C) support the differential stability model, and suggest that S2 cells are unable to traffic or degrade the S1-cleaved prodomain/ligand complex. An analogous study in the same S2 cell line, however, reported the opposite result: that significantly lower steady state levels of mature Dpp are detected in the media of cells expressing proDpp^S2KK relative to those expressing Dpp^WT (Kunnapuu et al., 2009). A likely explanation for these contradictory results is our use of the actin-GAL4 driver to express proDpp in S2 cells, leading to supraphysiological levels of cleavage products that saturate the lysosomal trafficking machinery. In support of this proposal, steady state levels of cleavage products generated by wild type proBMP4 are significantly higher than those synthesized from S2-cleavage mutant variants in Xenopus embryos, oocytes, or mammalian cultured cells expressing low amounts of each precursor protein, but these differences disappear as the dosage is increased above a certain threshold (Degnin et al., 2004, Sopory et al., 2006, and data not shown).

Although our results show that cleavage of the S2 site of proDpp correlates with long-range signaling in vivo, S2 cleavage is not an absolute determinant of signaling range, but is instead one of many factors that influence the distance over which Dpp signals. For example, spatial and temporal expression of dpp is tightly controlled at the transcriptional level, and cells differ in their ability to respond to Dpp based on the level of receptors and other cofactors present at the cell surface (Affolter and Basler, 2007). Varying any one of these parameters perturbs the Dpp activity gradient. Importantly, our studies show that cleavage of the S2 site is required for long- and short-range signaling in cells that are normally responsive to this ligand in the wing, but not for short range signaling to cells that normally respond to endogenous Dpp in the gut. Thus, differential use of the S2 site can influence signaling range when assayed in the context of cells that normally respond to endogenous Dpp. As with all regulatory mechanisms, however, it is possible to bypass this control if the precursor is ectopically expressed at high levels outside of the endogenous Dpp signaling domain.

The current studies provide the first biochemical evidence that the S2 site of Dpp/BMP4 is cleaved in a tissue-specific fashion, and future studies will be required to determine how this is regulated. We have recently shown that two members of the PC family, furin and PC6 function redundantly to cleave the S1 and the S2 sites of proBMP4, whereas a third PC, possibly PC7, functions to selectively cleave the S1 site, possibly in a developmentally regulated fashion in vertebrate embryos (Nelsen and Christian, 2009). These studies raise the possibility that tissue-specific cleavage of proBMP4 is regulated by differential expression of a site-specific protease. The results of RNAi knockdown studies in S2 cells show that DFur1 and DFur2 function redundantly to cleave both the S1 and the S2 site of Dpp, although DFur1 may preferentially cleave the S2 site (Kunnapuu et al., 2009). The latter result raises the possibility that the S2 site is cleaved in all tissues that express DFur1, and not in tissues that express only DFur2. This simple hypothesis is not supported by the observations that DFur2 also contributes to S2 site cleavage (Kunnapuu et al., 2009) and that DFur1 is expressed in the embryonic mesoderm (Creemers et al., 1993), where the S2 site is not cleaved. Because PC activity is tightly regulated post-translationally, RNA expression does not necessarily indicate that functional protein is present and thus further in vivo analysis will be required to stringently test this hypothesis.

BMP4 and Dpp play highly conserved roles in patterning the limbs, and during specification and differentiation of the endoderm layer. Our studies suggest that Dpp/BMP4 maturation and activity may be regulated differently in these tissues in vertebrates relative to invertebrates. Specifically, the current studies show that S2 cleavage is essential for Dpp function in the wing disc, but not in the gut whereas analysis of mice carrying a point mutation that prevents cleavage of the S2 site of BMP4 demonstrates an opposite requirement for S2 cleavage in analogous tissues in mammals (Goldman et al., 2006). At present, nothing is known about which sites are cleaved in proBmp4 in the limb versus the gut in mammals. Further studies will be required to determine whether and how tissue-specific use of the S2 site differs between vertebrates and flies.