Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2012 Aug 16.
Published in final edited form as: Dev Cell. 2011 Aug 16;21(2):375–383. doi: 10.1016/j.devcel.2011.06.025

Shaping BMP morphogen gradients through enzyme-substrate interactions

Carolyn E Peluso 1, David Umulis 2, Young-Jun Kim 1, Michael B O’Connor 3, Mihaela Serpe 1,*
PMCID: PMC3175245  NIHMSID: NIHMS318103  PMID: 21839924

SUMMARY

Bone Morphogenetic Proteins (BMPs) regulate dorsal/ventral (D/V) patterning across the animal kingdom; however, the biochemical properties of certain pathway components can vary according to species-specific developmental requirements. For example, Tolloid (Tld)-like metalloproteases cleave vertebrate BMP-binding proteins called Chordins constitutively, while the Drosophila Chordin ortholog, Short gastrulation (Sog), is only cleaved efficiently when bound to BMPs. We identified Sog characteristics responsible for making its cleavage dependent on BMP binding. “Chordin-like” variants that are processed independently of BMPs, changed the steep BMP gradient found in Drosophila embryos to a shallower profile, analogous to that observed in some vertebrate embryos. This change ultimately affected cell fate allocation and tissue size and resulted in increased variability of patterning. Thus, the acquisition of BMP-dependent Sog processing during evolution appears to facilitate long-range ligand diffusion and formation of a robust morphogen gradient, enabling the bistable BMP signaling outputs required for early Drosophila patterning.

INTRODUCTION

Spatially non-uniform distributions of secreted morphogens guide tissue development in a highly reproducible and robust manner. In the early Drosophila embryo, Decapentaplegic (Dpp) a BMP-type ligand is key in assigning identity to all dorsal structures. Dpp is transcribed uniformly throughout the dorsal domain, yet it forms an activity gradient via a cascade of extracellular regulation that restricts Dpp availability laterally, while simultaneously amplifying Dpp activity near the dorsal midline. Dpp initiates signaling by binding to type I and II receptors and forming a higher-order complex that transduces the signal by phosphorylating the downstream effector Mad. Phospho-Mad (P-Mad), together with Co-Smad, accumulates in the nucleus where in conjunction with other transcription factors, regulate target gene expression. Prior to gastrulation Dpp signaling at the dorsal midline induces specification of amnioserosa cell fate; embryos with amnioserosa defects continue to develop, but fail to hatch.

Formation of the Dpp activity gradient requires several secreted modulators. In the early Drosophila embryo, laterally secreted Sog binds Dpp in a complex that inhibits Dpp signaling locally and facilitates long-range ligand diffusion, shuttling Dpp from the lateral domains towards the midline (Francois et al., 1994; Shimmi et al., 2005; Wang and Ferguson, 2005). Sog plays both positive and negative roles in regulating BMP activity, a phenomenon previously referred to as the “Sog paradox” (Biehs et al., 1996; Holley et al., 1995; Marques et al., 1997). The negative role comes from blocking access of ligands to receptors. The positive effect comes from its ability to facilitate Dpp diffusion. This facilitated diffusion also requires Twisted gastrulation (Tsg), a protein secreted in the dorsal domain of the early embryos. Tsg and Sog both diffuse from their initial regions of expression and bind to form Sog-Tsg, a complex that is more effective at binding Dpp than either Sog or Tsg alone (Chang et al., 2001; Ross et al., 2001; Scott et al., 2001; Shimmi et al., 2005). This shuttling process appears to favor BMP heterodimers for the long-range transport, in this case Dpp and Screw (Scw) (Shimmi et al., 2005). A key component that helps create flux is the processing of Sog by Tld, a metalloprotease of the BMP-1 family (Ashe and Levine, 1999; Marques et al., 1997; Piccolo et al., 1997; Shimell et al., 1991). The net movement of Dpp dorsally is generated by reiterated cycles of complex formation, diffusion and destruction by Tld.

The BMP shuttling mechanism is highly conserved throughout the animal kingdom and relies on the spatial distribution of a BMP ligand (Dpp/BMP2/4) and a BMP antagonist (Sog/Chordin). Although Chordin is thought to be the functional homolog of Sog, when introduced into Drosophila it only acts as an inhibitor, and cannot promote long-range Dpp signaling (Decotto and Ferguson, 2001). Here we use biochemical, genetic, quantitative image analysis, and computational modeling studies to explore the molecular etiology of these differences. One biochemical difference between these two molecules is that processing of Sog by Tld requires the BMP ligand as an obligatory co-substrate, while Chordin does not (Marques et al., 1997; Piccolo et al., 1997). In this study, we propose that Sog’s ability to function as a more efficient BMP transporter that results in long-range BMP signaling resides, in molecular terms, in the co-substrate requirements for Tld-mediated Sog degradation.

RESULTS AND DISCUSSION

Molecular characterization of Sog processing sites

To identify and characterize the Tld processing sites in Sog, we purified and sequenced the Sog cleavage fragments using tagged proteins generated in S2 insect cell cultures. We captured the intermediate Sog cleavage fragments using sub-optimal amounts of enzyme and Dpp, the obligatory co-substrate (Figure 1A). The three major processing sites in Sog are in close proximity to the Cys-rich BMP binding modules (Figure 1B). The positions of processing sites 1 and 3 correspond to the two major processing sites in Chordin. Sequencing of the N-termini revealed a conserved Asp residue at position P1′ (positions of residues after the scissile bond are marked primed), a hallmark of the astacin family of proteases that includes Tld, a conserved aliphatic residue (V) in position P3, and no significant homology with other Tld/BMP-1 substrates (Hopkins et al., 2007) (Figure 1C).

Figure 1. Characterization of the three major processing sites in Sog.

Figure 1

Open in a new tab

(A) Sog is processed at three major sites. Cartoon and Western blot depicting the products of Sog cleavage. (B) Cartoon comparison of processing sites in Sog and Chordin. (C) Sequence alignment of Sog and Chordin processing sites (arrow indicates the site of cleavage). (D) Western blot analysis shows that mutations at site 1 or at sites 1,2 slow the destruction of full length Sog in vitro. (E) Sog-1,2u (Sog-u) cannot be cleaved by Tlr when over-expressed in the wing. “u”-indicates mutations that render a particular processing site uncleavable. The wing phenotypes are 100% penetrant.

Replacement of all four residues at processing site 1 (V183ALD) with Ala rendered Sog virtually uncleavable at this site in vitro (Figure 1D, Sog-1u; “u”-uncleavable). Full-length mutant Sog was still degraded over time, likely due to processing at the remaining unmodified sites, but the speed of its degradation was reduced. Additional Ala replacements at site 2 (V728PGD) further slowed down the Sog destruction (quantified in Figure 1D). When we replaced only the conserved D (position P1′) with E at processing site 1 we observed similar effects: undetectable cleavage at site 1 and a slower overall degradation of the mutant Sog (Figure S1). In our system we could not determine the precise cleavage kinetics at individual sites, but the overall Sog destruction in various uncleavable Sog mutants was clearly slowed down, likely because of blocked/reduced cleavage at the modified site(s). Also, in our hands any mutations at site 3 induced constitutive cleavage at this site, thus we kept site 3 intact for these studies.

To test if these mutations (Sog-u) could render Sog uncleavable in vivo, we examined Sog gain-of-function phenotypes. Overexpression of Sog in the wing imaginal disc produces very mild phenotypes of venation defects (Figure 1E and (Yu et al., 1996). Sog together with Tsg produces a more potent BMP inhibitor; their combined overexpression inhibits BMP signaling and results in smaller wings with altered patterns of venation (Figure 1E and (Ross et al., 2001). Co-overexpression of Sog and Tsg with Tolloid-related (Tlr) is able to reverse the small wing phenotype and restore normal patterning in the case of wild-type Sog (Figure 1E and (Serpe et al., 2005), but not in the case of Sog-u. In fact, overexpression of Sog-u by itself produced a significant loss of posterior crossvein, a structure that requires peak BMP signaling, suggesting that Sog-u is a better BMP inhibitor than the wild-type Sog (Figure 1E). Moreover, the loss of posterior crossvein tissue was exacerbated when Tlr was co-expressed with Sog-u, indicating further reduction in the BMP activity (Figure S1D). This is likely due to Tlr degrading endogenous Sog but not Sog-u. Thus, Sog-u appears resistant to cleavage and degradation in vivo and may act as a dominant negative by prolonged binding of the BMP ligands.

Several residues are responsible for making Sog destruction dependent on BMP binding

Unlike in vertebrates, Drosophila Tld and Tlr process Sog only when bound to a BMP-type ligand (Marques et al., 1997; Serpe et al., 2005). The binding of Sog to Tld requires several Tld protein interaction motifs besides the protease domain (Childs and O’Connor, 1994). Nevertheless, the requirement for the obligatory co-substrate for Sog processing is thought to indicate a BMP-induced conformational modification that allows the Sog-BMP complex, but not Sog alone to fit into the catalytic pocket of the enzyme. In contrast, Chordin, which exhibits BMP-independent processing, should bind and fit into Tld’s catalytic pocket without the need for a BMP-induced conformational change. Indeed, in spite of limited conservation between Sog and Chordin (40% similarity, 22% identity), we found that Drosophila Tld can cleave the vertebrate Chordin in a BMP-independent manner (not shown).

To focus on the enzyme-substrate interactions for Sog and Chordin we modeled the Tld catalytic domain using the crystal structures available for related enzymes, the crayfish Astacin - the founder member of this zinc metalloprotease family, and the human Tld catalytic domain (Grams et al., 1996; Mac Sweeney et al., 2008) (ClustalW alignment in Figure S2A). As previously described, the catalytic pocket of Tld enzymes appears very tight in the proximity of the catalytic Zn, where scissile bonds align, and has a relatively wide cavity that accommodates residues P3 and P2 in the substrate (Figure 2A) (Grams et al., 1996; Mac Sweeney et al., 2008). Bulky, hydrophobic residues in the substrate, such as in Chordin, might facilitate enzyme-substrate binding. Indeed, when we analyzed the effect of changing processing site 1 in Sog to either Chordin site 1 (Sog-1Ch1) or Chordin site 2 (Sog-1Ch2) we found indications of Tld cleavage at these sites in the absence of Dpp, although this cleavage was extremely weak (Figure S2B).

Figure 2. Several residues are required for making Sog processing dependent on BMP binding.

Figure 2

Open in a new tab

(A) A surface diagram of hBMP-1 showing the catalytic zinc as a pink sphere and the cavity important for binding P3 and P2 residues in the substrate (arrow). Key aromatic residues are colored by elements (C- green, O- red, N- blue). (B) Western blot comparisons of Sog-wt and Sog-1i processing by Tld. (C) Western blot comparisons of Sog-wt and Sog-1i processing by Tld at low Dpp concentration. (D) Co-immunoprecipitation analysis to show that Sog-wt and Sog-i bind to Dpp similarly. (E) Cell-based assay to demonstrate that Sog-i is as efficient at inhibiting Dpp signaling as Sog-wt. “i”-indicates mutations that render a particular site independent of BMP binding for processing.

In addition, an aromatic residue in position P3 in the substrate could potentially stack against the aromatic ring a key position near the active site (see Figure S2A, sIV) to further lower the substrate-enzyme binding energy and facilitate substrate binding.

We tested these predictions and found that indeed changing several residues at the processing site could alter the co-substrate requirements. For example, Sog processing at mutated site 1 (V183ALDV to FYGDP) occurred independently of the co-substrate (compare lanes 9–12 and 1–4 in Figure 2B, Sog-1i). This processing was enhanced when Dpp was added to the reaction (compare lanes 13–16 and 9–12 in Figure 2B) partly because cleavage at unmodified sites 2 and 3 could not happen in the absence of the co-substrate. Addition of Tsg similarly enhanced the processing of wild type and mutant Sog (not shown and (Shimmi and O’Connor, 2003). Nonetheless, when Dpp was in limiting amounts, the mutant Sog was processed more efficiently than the wild-type protein (Figure 2C). We confirmed by Edman degradation that processing occurred at the expected G185-D covalent bond in the mutated site, and that mutagenesis did not create any promiscuous cleavage. Similar changes at processing site 2, separately or with site 1, further enhanced the speed of Sog degradation (Figures S2D and S2E). The strongest effect was seen for a Sog variant in which both sites 1 and 2 were rendered BMP-independent for processing, designated Sog-i for “independent of BMP for cleavage”. At the sequence level Sog-i is very different from Chordin, but it resembles Chordin in how it is processed by Tld: Sog-i exhibits significant BMP-independent processing by Tld, which is enhanced in the presence of BMP ligands. To emphasize these similarities we refer to Sog-i as “Chordin-like” Sog.

The BMP-dependence does not affect Sog-ligand interactions, but could impact the Dpp transport

We tested if these changes impact Sog’s ability to bind BMPs and/or inhibit their signaling. We found that purified Sog and Sog-i were indistinguishable in their binding to Dpp homodimers and Dpp/Scw heterodimers in co-ip experiments (Figure 2D and not shown). Also, in both cases, addition of Tsg equally increased Sog binding to the BMPs. The inhibitory activities of Sog and Sog-i on BMP signaling were compared in a cell-based assay; in the presence of Tsg, Sog inhibits Dpp-induced signaling in a concentration dependent manner (Figure 2E and (Shimmi and O’Connor, 2003). Equivalent amounts of Sog-i and Tsg produced a similar inhibitory response.

While the BMP binding properties of Sog-i appeared to be largely unaffected, we predicted that this “Chordin-like” Sog would resemble Chordin when introduced into fly embryos, and be less efficient in promoting long range BMP signaling. To model this process, we modified a previously published spatiotemporal patterning model by adding the BMP-independent processing of Sog (see methods and (Umulis et al., 2010) for details). Briefly, the rate for Tld-mediated processing of Sog increases when the rate of BMP-dependent or BMP-independent cleavage increases. An increase in the rate of Tld processing will modify the Sog protein levels and the shapes of the Sog and Sog/Tsg distributions in the model. This results in a simultaneous reduction in the inhibition of Dpp signaling laterally and a reduction in the Dpp accumulation near the dorsal midline. To quantify the effect of BMP-independent cleavage of Sog-i on the net transport of BMP molecules towards the dorsal midline, we calculated the net diffusive flux (see methods) of BMP ligand in the embryo by summing the contributions of free BMP and Sog-bound BMP. The flux provides the magnitude and direction of transport driven by the gradient of concentration (Figure 3A, red arrows). As shown in Figure 3A, the Sog-i simulation clearly indicates a lower net transport towards the midline than the simulation with Sog-wt. We also investigated in this model whether increased Sog-i expression could improve the transport towards the midline and whether the reduction in transport is solely the result of a reduction in Sog levels. Increasing the production of Sog-i increases the total amount of Sog-i in the system, however transport of Dpp/Scw towards the midline is still reduced even with significantly increased levels of Sog-i greater than in Sog-wt embryos with normal patterning.

Figure 3. The BMP morphogen gradient profile is shallower in embryos with “Chordin-like” Sog (Sog-i).

Figure 3

Open in a new tab

(A) Computational modeling of Sog- mediated transport of BMPs (red arrows represent net flux of Dpp). (B) Cartoon of the sog transgene used and the Sog-HA protein expression. The pPelican-based transgenes contain: gypsy insulators (orange), sog lateral stripe enhancer (red), a minimal sog promoter and 5′UTR, sog open reading frame (blue), 3xHA (green) (only in the tagged constructs), and sog 3′UTR. The levels and distribution of Sog-wt (left panels) and Sog-i (right panels) are compared using anti-Sog (upper panels), and anti-HA antibodies (lower panels). Images were collected using the same gain and intensity settings. (C) Direct immunofluorescent labeling of P-Mad in a sog−/− mutant embryo visualized from the dorsal surface (left panel) and from a lateral position (right panel). (D–G) Effect of Sog-i on the BMP gradient profile visualized as direct immunofluorescent labeling of P-Mad in wild-type (D) sog heterozygotes (E) sog−/−;2xsog-wt (F), and sog−/−;2xsog-i (G). (D′-G′) Mean intensity of the P-Mad stripe across the indicated populations. (D″-G″) Cross-sectional analysis of the P-Mad intensity at 1/3 of the embryo (dashed line in D′). Blue lines represent the mean intensity at each point along the axis (middle line), +/− the standard deviation (upper and lower blue lines, respectively). Analyses of embryo populations were performed for sog-wt (n= 11) (D″), sog heterozygote (n= 15) (E″), sog−/−;2xsog-wt (n= 8) (F″), and sog−/−;2xsog-i (n= 25) (G″). (H–I) The effect of sog-i on the slopes of the P-Mad profiles. (H) Distribution of P-Mad slopes as spatial derivative of P-Mad intensities in the cross-sections of indicated genotypes. Green points on the 2x sog-wt line and blue points on the 2x sog-i line indicate a measurable difference when compared to wt (p<0.03). (I) Population distribution of local P-Mad slopes in various fly lines. (J) The effect of lowering Sog protein levels on the slopes of the P-Mad profiles. Red points on the sog+/− line and black points on the 2x sog-i line indicate difference from wt (p<0.03). The blue points on the 2x sog-i line indicate difference between the 2x sog-i when compared to sog+/− (p<0.03).

Replacement of endogenous Sog with “Chordin-like” variants changes the BMP gradient profile in early Drosophila embryos

To test the biological effect of these Sog variants on the BMP morphogen gradient profile, we constructed transgenic fly lines that allowed for normal spatial and temporal expression of Sog proteins at endogenous levels (Figure 3B). The neural-ectoderm expression of tagged and non-tagged Sog proteins, Sog-wt, Sog-wt-HA and Sog-i-HA, in all of the transgenic lines obtained, overlapped the sog mRNA endogenous pattern. The relative Sog levels in these transgenic lines were quantified by immunofluorescence using anti-HA antibodies, anti-Sog antibodies, or both. We found that indeed these transgenic lines have similar levels of Sog protein (Figure 3B and not shown). In addition, all of the transgenic lines expressing Sog-wt (either HA-tagged or not-tagged) rescued the sogYL26 mutants and trans-heterozygous combinations (sog−/−) to viable, and fertile adults.

We examined the profile of the BMP morphogen gradient in stage 5 embryos by following the accumulation of activated/phosphorylated Mad (P-Mad), the effector of the BMP signaling pathway. In the absence of Sog, the facilitated diffusion of BMP ligands does not occur and Dpp remains uniformly distributed over the dorsal domain. No gradient of BMP activity is generated, thus the P-Mad levels are low and constant over the entire dorsal domain of sog−/− mutant embryos (Figure 3C and (Wang and Ferguson, 2005). In contrast, stage 5 wild-type embryos have a sharp, step-gradient of BMP signaling, in which P-Mad levels are high in the dorsal most cells and rapidly drop off to undetectable levels in more lateral regions. The P-Mad positive domain is wider and slightly reduced in intensity in heterozygous (sog+/−) embryos (Figures 3D and 3E). Among the sog alleles tested, the sogYL26/+ heterozygous embryos showed the widest P-Mad profile (Figure S3). The HA-tagged or untagged sog-wt transgenes were equally effective in restoring the sharp P-Mad profile in sog−/− embryos when in two copies, suggesting that the tag did not alter Sog activity (Figures 3F and S3). In contrast, addition of two sog-i copies to any sog−/− background produced a wide P-Mad positive domain with reduced signal intensities (sogYL26 allele shown in Figure 3G). In the latter embryos the boundaries of P-Mad positive domain were more diffuse, with reduced slopes evident in the cross-section profile. Analogous studies (not shown) of race expression, downstream of BMP signaling in the presumptive amnioserosa, indicated a similar effect. This suggests that Sog-i is indeed less efficient in supporting an adequate Dpp/Scw-Sog/Tsg flux towards the midline and consequently the formation of the steep BMP distribution profile.

To quantify the differences in BMP signaling profiles between wild-type, sog+/−, and sog−/− embryos with 2x sog-i or 2x sog-wt transgenes, we decomposed the P-Mad fluorescent staining of each embryo into the product of an amplitude multiplied by the P-Mad distribution “shape” (see Methods). In brief, for each embryo, a region of interest was selected that encompasses a 4 cell-wide band centered at 33% embryo length. Each embryo was then processed through a Savitzky-Golay filter that removed noise while preserving the shape of the distribution and allowed for reliable calculation of the slope of the P-Mad profile.

Shape was quantified by measuring the spatial-derivative of P-Mad in the cross-section. Starting on the left of a P-Mad cross-section plot, the derivative will be positive and change in magnitude at each position along the D/V axis directly proportional to the slope of P-Mad. As the slope decreases near the dorsal midline, the value of the derivative is approximately zero and then negative for the right side of the distribution where P-Mad is decreasing (Figure 3H). The local average P-Mad slope for the population of 2x sog-wt embryos at each spatial location was virtually indistinguishable from the wt P-Mad slope. In contrast, the population average P-Mad slope for 2x sog-i embryos was noticeably shallower than wt and 2x sog-wt embryos with a lower magnitude of the spatial derivative near the midline and a higher magnitude in the lateral dorsal ectoderm (p <0.03). Moreover, differences in the overall intensity of P-Mad staining and/or in embryo-to-embryo variability do not account for this observation; even when we chose the scaling for each population so the population means would have the same peak P-Mad levels (Figure S3G), or when we used the absolute value of the local P-Mad slope for each individual in each population to calculate the population distributions of slopes (Figure 3I), the differences remain clear: the replacement of sog-wt with sog-i leads to broader, shallower P-Mad profiles.

The difference between the P-Mad profiles in wt and 2x sog-i was not equivalent to a decrease in the total amount of Sog protein in the system. Distributions of P-Mad slopes in both sog+/− and 2x sog-i differed from wt embryos (Figure 3J, black dots: 2x sog-i; red dots: sog+/−), but the perturbations were not equivalent. For sog+/−, Sog protein levels were reduced approximately 50%, and the position of the peak slope shifted laterally away from the dorsal midline; however, the magnitude of the slope was still significantly greater than the magnitude of the slope for 2x sog-i embryos, though slightly less than wt. This means that the P-Mad profile in sog+/− is wider, but the steepness of the BMP activity gradient is similar to the steepness of the wt gradient. In contrast, the slope of the P-Mad profile in 2x sog-i embryos was significantly lower than wt embryos near where their peaks overlap and significantly greater than wt in the lateral dorsal ectoderm (Figures 3J and S3).

Cell fate allocation and tissue size are altered in embryos with “Chordin-like” Sog

Prior to gastrulation high levels of BMP signaling at the dorsal midline in early Drosophila embryos specify amnioserosa, an extra-embryonic tissue required for gastrulation. The sharp and narrow BMP signaling domain in wt embryos induced formation of an amnioserosa field of approximately 200 cells in stage 13 embryos (Figure 4A). The sog+/− heterozygous embryos have a wider BMP signaling domain that produced a larger amnioserosa field, about 50% bigger than that of the wt embryos (314 amnioserosa cells in sog+/− versus 197 in wt embryos, Figure 4B). The spatial extent of the BMP signaling field above a certain threshold but below wild-type peak values appears to determine how many amnioserosa cells will be specified and consequently the size of the ensuing tissue.

Figure 4. Patterning is affected in embryos with Sog-i.

Figure 4

Open in a new tab

(A–D) Amnioserosa fields in embryos of the indicated genotypes. (E) Mean number of cells (+/− s.e.m) in the amnioserosa fields. (ANOVA, F (4, 115) = 39.12, p < 0.0001, Tukey’s HSD, ***p< 0.01, **p< 0.05). The n of each group is indicated at the base of its respective bar. (F–G′) The effect of multiple copies of sog-wt or sog-i on the BMP gradient profile and cell fate determination. (F–G) Direct immunofluorescent labeling of P-Mad. (F′-G′) Mean intensity of the P-Mad stripe (F′-G′). (H–I) Amnioserosa fields in sog−/−;4xsog-wt (H), and sog−/−;4xsog-i (I). (J–L) The effect of multiple sog copies on the slopes of the P-Mad profiles. (J) Distribution of P-Mad slopes for intensity-scaled systems of indicated genotypes. Green points on 4x sog-wt and blue points on 4x sog-i indicate significant difference from wt (p<0.03). (K) Population distribution of local P-Mad slopes for the indicated genotypes. (L) Distribution of P-Mad slopes for intensity-normalized systems of indicated genotypes. Green points on 4x sog-wt and blue points on 4x sog-i indicate significant difference from wt (p<0.03). (M–N) Computational modeling of the predicted BMP gradient profiles in embryos with 4x sog-wt (M), or 4x sog-i (N). (L) Computationally predicted BMP gradient profile when the Sog-Dpp binding rate was changed. (P) Spatial distribution of reaction rates that regulate extracellular Dpp. The net rate corresponds to the sum of the Tld reactions, which release Dpp and the rebinding to the available Sog. (Q) Coefficient of variation (standard deviation/width) for various genotypes at different thresholds/percentage of population mean for wt. Amnioserosa cells were labeled with an anti-Hindsight antibody.

When sog-wt transgenes replaced the endogenous sog, the amnioserosa cell numbers were rescued to wild-type levels (Figure 4C). However, addition of two sog-i copies to any sog−/− background tested produced statistically significant increases in the amnioserosa fields (294 cells) (Figures 4D and 4E).

The biological consequences of replacing Sog with Sog-i could not be explained simply by quantity differences between the sog-wt and sog-i transgenes. First, we observed shallow BMP gradient profiles, broader target gene expression domains, and increased cell allocation/amnioserosa fields using multiple independent sog-i transgenic lines with expression levels comparable with those of the sog-wt lines. Secondly, additional copies of sog-wt and sog-i transgenes did not significantly impact the BMP gradient profiles or amnioserosa fields (Figures 4F–I). The P-Mad positive domains were more intense in either 4x groups, but in the sog−/−; 4x sog-i embryos the signaling domain remained wide, the boundaries diffuse and the slopes of the cross-section profiles reduced (Figures 4J, 4K and 4L); also, the sog−/−; 4x sog-i embryos had significant embryo-to-embryo variability (see below).

To further search for alternative explanations for the Sog-i effects, we optimized Sog-wt and Sog-i versions of the 3D embryonic model against the 4x population data and asked whether our experimental observations could be captured by changes in Sog affinity to Dpp or changes in the processing rate of Sog by Tld independent of Dpp. We found that the model with an enhanced processing rate achieved a greater fit: a modest increase in the processing of Sog by Tld without Dpp (increase BMP independent processing from about 8% to 19% of Tld processing rate in presence of Dpp), resulted in signaling profiles that matched very well the experimental data (Figures 4M and 4N).

In the 2x sog-wt simulations, the net reaction rate was negative in the lateral portions of the dorsal region and reached maximum near the dorsal midline (Figure 4P). In the simulations for the 2x sog-i, the peak rate of Dpp release occurred laterally halfway between the neural-ectoderm and dorsal region. In contrast, no models obtained by decreasing the binding between Sog and Dpp (10x or more) could capture the experimentally observed loss of sharp boundaries (Figure 4O). Thus, the shift in the net rate of cleavage, in conjunction with less effective net flux (Figure 3A), produced less accumulation of Dpp near the dorsal midline in simulations of sog-i embryos.

Evolving a new developmental function by modifying a substrate-enzyme interaction

In conclusion, we found that several residues at the Tld processing site make Sog dependent on a (BMP) co-substrate for processing. Mutating these residues reduced the transport range of Sog-BMP complexes in vivo and altered the shape of the BMP signaling profiles and consequent cell fate allocation. Interestingly, BMP-dependent Chordin cleavage was also a requirement in mathematical modeling for scale-invariance of Xenopus embryos (Ben-Zvi et al., 2008). Here the co-substrate requirement ensured transport of both ADMP and the BMP ligands and the re-establishment of a well-proportioned DV axis. How might shuttling of ligands persist in the absence of BMP-dependent cleavage of Chordin? An intriguing possibility is that Sizzled-mediated repression of Xolloid spatially restricts Chordin processing providing a non-uniform Chordin sink (Lee et al., 2006). In mathematical models of embryo patterning, lowering the processing rate of Tld results in signaling distributions that are sharper and result in a greater net transport of BMP ligands away from the Sog/Chordin source.

An interesting and unexpected outcome of the comparison between sog-wt and sog-i embryos suggests that BMP-dependent Sog processing reduces embryo-to-embryo variability in P-Mad levels. Both sog-wt and sog-i embryos show sensitivity of P-Mad to gene dosage as shown in Figures 3, 4 and S4. However when we calculated the coefficient of variation (standard deviation/width) within the each genotype, we found that embryos with one or two sog-wt copies showed less variability in signaling width than their 2x sog-i counterparts (Figure 4Q, blue line). The variability is greater at nearly all threshold positions, and the variability within this population increased dramatically at higher threshold levels. This suggests that BMP-dependent Sog destruction may reduce embryo-to-embryo variability between individuals in a population of the same genotype to provide robust patterning of the dorsal structures.

Altogether, our results indicate that a “Chordin-like” Sog is less able to reliably support patterning of the early Drosophila embryo. By modifying the Sog-Tld substrate-enzyme interaction with just a few residue changes, it appears that a new developmental function for Sog evolved that ensured reliable shuttling of BMPs and robust patterning. Further refinement of this shuttling mechanism, such as its speed (Serpe et al., 2005) or its directionality (O. Shimmi, personal communication), expanded the repertoire of cell fate specification by BMP morphogen gradients and was likely exploited for diversified patterning during natural evolution.

EXPERIMENTAL PROCEDURES

Details of fly stocks, plasmid construction, protein production, purification and analyses are provided as Supplemental Experimental Procedures.

Immunohistochemistry

Embryos were collected and fixed using standard procedures. Rabbit anti-P-Mad antibody was a gift from Ed Leof and was used at 1:100, followed by anti-rabbit conjugated with Alexa-488 (Molecular Probes). The HA-tag was detected using anti-HA antibody 3F10 (Roche) at 1:100, followed by biotinylated secondary (Vector Laboratories) and streptavidin conjugated to Alexa-488 (Molecular Probes). The anti-Sog-8B (Srinivasan et al., 2002) antibody was used at 1:100, followed by an anti-rabbit secondary antibody conjugated to alexa-488 (Molecular Probes). The anti-hindsight antibody (1G9) (Developmental Studies Hybridoma Bank) was used at 1:100, followed by biotinylated secondary and tyramide signal amplification (Perkin Elmer).

Image Analysis

Each fluorescent intensity image was rotated, aligned and scaled to correct for variability in mounting and population variability in embryo size. Each embryo intensity distribution was normalized by minimizing the difference between an individual and the population mean concentration as described previously (Gregor et al., 2007; Umulis et al., 2010). To compare between genotypes, each imaging run included a wild-type control sample collected, fixed and imaged with the same conditions, on the same day, and exposed to the same concentration of antibodies as the test cases. The population mean amplitudes were determined for each embryo by comparing the maximum 10% intensity pixels for a test case against the maximum 10% intensity pixels for the wt case.

Calculation of profile slopes and distributions

Following minimization of squared error between individuals in a population additional image processing steps are required to estimate the spatial derivative of each image P-Mad intensity profiles. Noise in the original images precludes accurate estimation of the slope (first spatial derivative) of the P-Mad profiles and classical smoothing filters such as Gaussian filters, and moving average filters flatten out the profiles and systematically alter the original image data. The ROI P-Mad intensity of each image was processed through a 3rd-order Savitzky-Golay smoothing filter with a window length of 51 pixels that preserved the shapes of the P-Mad distributions and eliminated high-frequency noise from the raw image. The derivative of P-Mad for each ROI was then numerically calculated using a five-point stencil approximation (a high order method with greater accuracy than 2-point methods).

Multiple independent experiments can be combined into large population ensembles by scaling wt P-Mad profiles to each other to systematically control for the differences between experimental runs.

Supplementary Material

01
02
Download video file (40.3MB, mov)

Highlights.

  • Tld cleavage of the BMP-binding protein Sog is constrained by substrate sequences

  • Long-range BMP diffusion requires BMP-dependence of Sog destruction by Tolloid

  • BMP gradients have shallower profiles in flies with BMP-independent Sog processing

  • BMP-dependent Sog degradation allows robust, rapid developmental patterning by BMPs

Acknowledgments

We thank Osamu Shimmi for communicating unpublished information and to Stuart Newfeld for comments on the manuscript. We thank Chi-Hon Lee and Alan Hinnebusch for helpful discussions and suggestions. We are grateful to MaryJane Shimell for help in constructing the sog-HA transgene and to Seth Blair for Sog antibodies. We thank Peter Nguyen, Gabriel Peal, Jeremy Swan, and David Zhitomirsky for technical assistance. This work was supported in part by the Intramural Research Program at NIH. D.U. is supported by a grant from the Showalter Trust. M.B.O. is an Investigator with the Howard Hughes Medical Institute.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Ashe HL, Levine M. Local inhibition and long-range enhancement of Dpp signal transduction by Sog. Nature. 1999;398:427–431. doi: 10.1038/18892. [DOI] [PubMed] [Google Scholar]
  2. Ben-Zvi D, Shilo BZ, Fainsod A, Barkai N. Scaling of the BMP activation gradient in Xenopus embryos. Nature. 2008;453:1205–1211. doi: 10.1038/nature07059. [DOI] [PubMed] [Google Scholar]
  3. Biehs B, Francois V, Bier E. The Drosophila short gastrulation gene prevents Dpp from autoactivating and suppressing neurogenesis in the neuroectoderm. Genes Dev. 1996;10:2922–2934. doi: 10.1101/gad.10.22.2922. [DOI] [PubMed] [Google Scholar]
  4. Chang C, Holtzman DA, Chau S, Chickering T, Woolf EA, Holmgren LM, Bodorova J, Gearing DP, Holmes WE, Brivanlou AH. Twisted gastrulation can function as a BMP antagonist. Nature. 2001;410:483–487. doi: 10.1038/35068583. [DOI] [PubMed] [Google Scholar]
  5. Childs SR, O’Connor MB. Two domains of the tolloid protein contribute to its unusual genetic interaction with decapentaplegic. Dev Biol. 1994;162:209–220. doi: 10.1006/dbio.1994.1079. [DOI] [PubMed] [Google Scholar]
  6. Decotto E, Ferguson EL. A positive role for Short gastrulation in modulating BMP signaling during dorsoventral patterning in the Drosophila embryo. Development. 2001;128:3831–3841. doi: 10.1242/dev.128.19.3831. [DOI] [PubMed] [Google Scholar]
  7. Francois V, Solloway M, O’Neill JW, Emery J, Bier E. Dorsal-ventral patterning of the Drosophila embryo depends on a putative negative growth factor encoded by the short gastrulation gene. Genes Dev. 1994;8:2602–2616. doi: 10.1101/gad.8.21.2602. [DOI] [PubMed] [Google Scholar]
  8. Grams F, Dive V, Yiotakis A, Yiallouros I, Vassiliou S, Zwilling R, Bode W, Stocker W. Structure of astacin with a transition-state analogue inhibitor. Nat Struct Biol. 1996;3:671–675. doi: 10.1038/nsb0896-671. [DOI] [PubMed] [Google Scholar]
  9. Gregor T, Tank DW, Wieschaus EF, Bialek W. Probing the limits to positional information. Cell. 2007;130:153–164. doi: 10.1016/j.cell.2007.05.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Holley SA, Jackson PD, Sasai Y, Lu B, De Robertis EM, Hoffmann FM, Ferguson EL. A conserved system for dorsal-ventral patterning in insects and vertebrates involving sog and chordin. Nature. 1995;376:249–253. doi: 10.1038/376249a0. [DOI] [PubMed] [Google Scholar]
  11. Hopkins DR, Keles S, Greenspan DS. The bone morphogenetic protein 1/Tolloid-like metalloproteinases. Matrix Biol. 2007;26:508–523. doi: 10.1016/j.matbio.2007.05.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Lee HX, Ambrosio AL, Reversade B, De Robertis EM. Embryonic dorsal-ventral signaling: secreted frizzled-related proteins as inhibitors of tolloid proteinases. Cell. 2006;124:147–159. doi: 10.1016/j.cell.2005.12.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Mac Sweeney A, Gil-Parrado S, Vinzenz D, Bernardi A, Hein A, Bodendorf U, Erbel P, Logel C, Gerhartz B. Structural basis for the substrate specificity of bone morphogenetic protein 1/tolloid-like metalloproteases. J Mol Biol. 2008;384:228–239. doi: 10.1016/j.jmb.2008.09.029. [DOI] [PubMed] [Google Scholar]
  14. Marques G, Musacchio M, Shimell MJ, Wunnenberg-Stapleton K, Cho KW, O’Connor MB. Production of a DPP activity gradient in the early Drosophila embryo through the opposing actions of the SOG and TLD proteins. Cell. 1997;91:417–426. doi: 10.1016/s0092-8674(00)80425-0. [DOI] [PubMed] [Google Scholar]
  15. Piccolo S, Agius E, Lu B, Goodman S, Dale L, De Robertis EM. Cleavage of Chordin by Xolloid metalloprotease suggests a role for proteolytic processing in the regulation of Spemann organizer activity. Cell. 1997;91:407–416. doi: 10.1016/s0092-8674(00)80424-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Ross JJ, Shimmi O, Vilmos P, Petryk A, Kim H, Gaudenz K, Hermanson S, Ekker SC, O’Connor MB, Marsh JL. Twisted gastrulation is a conserved extracellular BMP antagonist. Nature. 2001;410:479–483. doi: 10.1038/35068578. [DOI] [PubMed] [Google Scholar]
  17. Scott IC, Blitz IL, Pappano WN, Maas SA, Cho KW, Greenspan DS. Homologues of Twisted gastrulation are extracellular cofactors in antagonism of BMP signalling. Nature. 2001;410:475–478. doi: 10.1038/35068572. [DOI] [PubMed] [Google Scholar]
  18. Serpe M, Ralston A, Blair SS, O’Connor MB. Matching catalytic activity to developmental function: tolloid-related processes Sog in order to help specify the posterior crossvein in the Drosophila wing. Development. 2005;132:2645–2656. doi: 10.1242/dev.01838. [DOI] [PubMed] [Google Scholar]
  19. Shimell MJ, Ferguson EL, Childs SR, O’Connor MB. The Drosophila dorsal-ventral patterning gene tolloid is related to human bone morphogenetic protein 1. Cell. 1991;67:469–481. doi: 10.1016/0092-8674(91)90522-z. [DOI] [PubMed] [Google Scholar]
  20. Shimmi O, O’Connor MB. Physical properties of Tld, Sog, Tsg and Dpp protein interactions are predicted to help create a sharp boundary in Bmp signals during dorsoventral patterning of the Drosophila embryo. Development. 2003;130:4673–4682. doi: 10.1242/dev.00684. [DOI] [PubMed] [Google Scholar]
  21. Shimmi O, Umulis D, Othmer H, O’Connor MB. Facilitated transport of a Dpp/Scw heterodimer by Sog/Tsg leads to robust patterning of the Drosophila blastoderm embryo. Cell. 2005;120:873–886. doi: 10.1016/j.cell.2005.02.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Srinivasan S, Rashka KE, Bier E. Creation of a Sog morphogen gradient in the Drosophila embryo. Dev Cell. 2002;2:91–101. doi: 10.1016/s1534-5807(01)00097-1. [DOI] [PubMed] [Google Scholar]
  23. Umulis DM, Shimmi O, O’Connor MB, Othmer HG. Organism-scale modeling of early Drosophila patterning via bone morphogenetic proteins. Dev Cell. 2010;18:260–274. doi: 10.1016/j.devcel.2010.01.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Wang YC, Ferguson EL. Spatial bistability of Dpp-receptor interactions during Drosophila dorsal-ventral patterning. Nature. 2005;434:229–234. doi: 10.1038/nature03318. [DOI] [PubMed] [Google Scholar]
  25. Yu K, Sturtevant MA, Biehs B, Francois V, Padgett RW, Blackman RK, Bier E. The Drosophila decapentaplegic and short gastrulation genes function antagonistically during adult wing vein development. Development. 1996;122:4033–4044. doi: 10.1242/dev.122.12.4033. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

01
NIHMS318103-supplement-01.doc (6.1MB, doc)
02
Download video file (40.3MB, mov)

RESOURCES