Drosophila Mmp2 Regulates the Matrix Molecule Faulty Attraction (Frac) to Promote Motor Axon Targeting in Drosophila

CG7526 interacts with Mmp2 and is related to vertebrate cbEGF-domain-containing proteins

We previously demonstrated that the catalytic activity of Mmp2 promotes motor axon fasciculation (Miller et al., 2008), implying that its substrate also functions in axon targeting. As a route to substrate identification, we performed a yeast interaction screen using a substrate-trapping E258A mutant of Mmp2 as bait (see Materials and Methods). The screen yielded six transmembrane or extracellular proteins representing potential substrates. We examined their RNA expression profiles via in situ hybridization to determine whether any were expressed in a manner consistent with regulation by Mmp2. Mmp2 is predicted to have a glycosylphosphatidylinositol (GPI) anchor (Llano et al., 2000, 2002; Page-McCaw et al., 2003) and is expressed by exit glia, which are tightly associated with motor axons (Sepp et al., 2001; Sepp and Auld, 2003; Freeman, 2006; Miller et al., 2008). Hence, we anticipated relevant substrates would be expressed in a pattern consistent with cleavage by exit glia-associated Mmp2, namely the glia themselves, neurons, or mesoderm. We identified one gene, CG7526, with salient mesodermal expression (see below and Fig. 2).

Sequence analysis indicates that CG7526 codes for a predicted extracellular protein, including three EGF and seven cbEGF repeats (Fig. 1A). CG7526 shares a conserved domain structure with extracellular proteins containing arrays of cbEGF repeats (Fig. 1A). Calcium binding is thought to impart structural rigidity to these proteins, and the cbEGF domains also participate in protein–protein interactions (Handford et al., 1995; Downing et al., 1996). Best characterized among related vertebrate proteins are Fibrillins, which serve both structural and signaling functions in the ECM. Fibrillins are modular glycoproteins that compose microfibrils and thus function as structural components of the ECM (Sakai et al., 1986; Zhang et al., 1995). They are characterized by arrays of cbEGF domains interspersed by TGFβ-binding (TB) domains. In addition to their structural role, Fibrillins regulate TGFβ superfamily bioavailability (Charbonneau et al., 2004; Ramirez and Rifkin, 2009). Underscoring the major functional relationship between Fibrillins and TGFβ signaling, human Marfan syndrome (MFS) results from haploinsufficiency of Fibrillin-1 and is strongly linked to elevated levels of active TGFβ (Ramirez and Dietz, 2007). Although CG7526 is structurally related to Fibrillins, the presence of CCP and hyalin domains, coupled with the absence of TB domains, argues that it is not orthologous (Hynes and Zhao, 2000). Based on the axon guidance phenotype described here, we have named CG7526 faulty attraction (frac).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

frac gene structure and allele generation. A, Domain structure of Drosophila Frac and Homo sapiens Fibrillin-1, Fibulin 1, and Fibulin 7. Two Fibulins were included to account for the variation between family members. Domains pictured include EGF-like repeats (green), cbEGF (orange), CCP/SUSHI (blue), hyalin (yellow pentagon), TB (gray hexagon), and fibulin (gray circles). B, Schematic of the frac gene region and span of frac LOF alleles. Exons are represented by rectangular black boxes and introns by thin lines. Position of Minos insertion is displayed.

Drosophila Frac is expressed in the embryonic mesoderm

We characterized the expression profile of frac to ask whether it is expressed in a pattern consistent with regulation by Mmp2. At embryonic stage 13, frac RNA is expressed strongly in ventral mesoderm adjacent to the CNS and at lower levels in more dorsal mesoderm (Fig. 2A). At stage 15, the frac RNA expression domain expands to include all embryonic musculature (Fig. 2B). To characterize Frac protein expression, we generated anti-Frac antibodies (Fig. 1A). Confirming antibody specificity, anti-Frac antibodies do not recognize protein in frac^Δ1 mutant embryos (see Fig. 4A,B). The dynamic pattern of frac RNA expression in the mesoderm is recapitulated by Frac protein. At stage 14, Frac expression is apparent in wedges of ventral mesoderm immediately adjacent to the nerve cord and then expands to encompass all embryonic muscle fibers at stage 15 (Fig. 2C,D). Frac continues to be expressed throughout the embryonic musculature until late stage 17, when its expression concentrates to slivers at hemisegment boundaries in a pattern characteristic of muscle attachment sites (Fig. 2E) (Becker et al., 1997). Confirming that Frac is mesodermal, it is coexpressed with myosin heavy chain, a well-characterized muscle marker (Fig. 1F) (Kiehart and Feghali, 1986). Furthermore, we verified that Frac expression in late stage 17 embryos is restricted to muscle attachment sites by double labeling with Shortstop/Kakapo (Fig. 2G) (Strumpf and Volk, 1998). Because the wedges of Frac-positive ventral mesoderm appear to be in close proximity to Mmp2-positive exit glia, Frac may represent an Mmp2 substrate. Hence, we analyzed the relative positions of Frac protein, Mmp2-positive glia, and motor axons in more detail. To compare the position of Frac protein to extending motor axons, we double-labeled wild-type embryos with anti-Frac antibodies and anti-FasII antibodies, which marks motor projections (Van Vactor et al., 1993). Motor axons extend through the periphery at stage 14, a time when high levels of Frac are expressed in the muscles through which these nerves traverse (Fig. 2H). Because Mmp2 is expressed by exit glia and is membrane bound (Llano et al., 2002; Miller et al., 2008), if mesodermal Frac is an in vivo substrate of Mmp2, it is predicted to be present immediately adjacent to these glia. In support of this hypothesis, Repo-positive glia align along the edge of the Frac-positive wedge of ventral mesoderm (arrows point to exit glia in Fig. 2I). This striking juxtaposition of Frac-positive mesoderm and Mmp2-positive glia argues that Mmp2 is poised to cleave Frac during embryogenesis. To examine the relative positions of Frac-positive mesoderm and glial processes in the periphery at later embryonic stages, we colabeled repo>actin–GFP embryos with anti-Frac and anti-GFP. Glial cell processes extend through Frac-expressing mesoderm (Fig. 2J), indicating that membrane-tethered Mmp2 is appropriately positioned to cleave Frac.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

frac RNA and Frac protein are expressed in the mesoderm and muscle attachment sites during embryogenesis. A, B, Stage 13 and stage 15 wild-type embryos hybridized with antisense frac RNA probe. C–E, Stage 14 (C), stage 15 (D), and stage 17 (E) wild-type embryos labeled with anti-Frac. F–I, Wild-type embryos colabeled with Frac in green and indicated antibodies in red. F, At stage 15, Frac colocalizes with the muscle marker MHC (red). G, During late stage 17, Frac protein colocalizes with Kakapo/Shortstop, marking muscle attachment sites (red). H, At stage 14, Frac protein is directly adjacent to the VNC from which axons, marked with 1D4 (red), are extending (dashed white lines mark the border of the VNC). I, Frac borders the Mmp2-expressing exit glia (arrows), a subset of glia marked by anti-Repo (red). J, repoGAL4>actinGFP embryos have glial processes marked by GFP (red) adjacent to Frac protein (green) expression in the periphery (arrows). A–E, Anterior is up; F–J, anterior is left, and midline is at the bottom. Scale bars, 100 μm.

Frac is processed in an Mmp2-dependent manner

Frac was identified in a yeast two-hybrid screen as an Mmp2 interactor, and its embryonic expression domain is in line with regulation by Mmp2. Hence, we investigated whether Frac represents a proteolytic substrate of Mmp2. Frac has a predicted size of 172 kDa. We find that Frac runs as a doublet at 170 and 250 kDa on Western blot of wild-type embryo extracts (Fig. 3A). The larger molecular weight band likely reflects a posttranslationally modified form of Frac, because many matrix molecules are heavily glycosylated (Comer and Hart, 2000). These two bands are not present in extracts from frac deletion embryos, demonstrating specificity (Fig. 3A). To determine whether Mmp2 is responsible for processing Frac, we analyzed Frac expression in Mmp2 LOF and GOF embryo extracts (Fig. 3A). In Mmp2 misexpression embryos (tub>Mmp2), full-length Frac is nearly absent, arguing that Mmp2 can process the full-length form. Consistent with this model, three small molecular weight Frac fragments are observed in extracts from Mmp2 misexpression embryos (arrowheads). Conversely, in Mmp2^W307* embryo extracts, we do not observe the small molecular weight bands and we consistently detect a significant increase in the amount of full-length Frac (Fig. 3A,B), arguing that Frac is not appropriately processed in the absence of Mmp2. The alterations in Frac processing observed in Mmp2 mutant embryonic extracts provide evidence that Frac is sensitive to Mmp2 levels.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Frac protein processing is Mmp2 dependent. A, Western blot with Frac antibody. Bottom shows GAPDH loading control. Full-length Frac is detected in wild-type embryo extracts at 170 and 250 kDa. Nonspecific bands are indicated by brackets in the frac deficiency column. In Mmp2 GOF embryos, cleavage products are present at 17, 30, and 33 kDa (arrowheads). B, Quantification of the full-length Frac bands (see box in A) in relation to the density of the 22 kDa background band (significance calculated using ANOVA; n = 9 blots; *p < 0.05, **p < 0.01). C, D, Stage 14 wild-type (C) and tub>Mmp2 (D) embryos labeled with Frac antibody. Anterior is up. Scale bar, 50 μm. E, Quantification of Frac protein expression in indicated genotypes. Stage 14 embryos were dissected and scored blindly on a scale of 0–3 for strength of Frac protein expression. The number of embryos analyzed is indicated below the histogram. Each genotype was individually compared with wild type using the Student's t test (***p < 0.001).

The marked reduction in full-length Frac in Mmp2 GOF protein extracts indicated that we might detect a visible reduction in Frac expression in these embryos if Frac is processed by Mmp2 and the cleavage fragments are degraded. In stage 14 wild-type embryos, Frac is expressed strongly in wedges of ventral mesoderm. We compared Frac levels in Mmp2 GOF embryos and wild-type embryos at stage 14, with the observer blinded to genotype. We found a 2.5-fold reduction in fluorescence intensity in embryos with Mmp2 misexpression relative to wild type (p < 0.001) (Fig. 3C,D), demonstrating that Mmp2 overexpression is sufficient to block accumulation of Frac. As controls, we scored Frac intensity in embryos with overexpression of a dominant-negative form of Mmp2, Mmp1 overexpression, or in embryos mutant for sidestep, a guidance molecule expressed in muscle (Sink et al., 2001). None of these genotypes show significant modifications to Frac levels (Fig. 3E). MMPs often display overlapping substrate specificities; however, these findings indicate that Frac is unlikely to be cleaved by Mmp1. These expression data support our biochemical data and demonstrate that Frac is processed in an Mmp2-dependent manner. Taken with the interaction between Frac and Mmp2 in yeast, we conclude that Frac likely represents an Mmp2 substrate.

frac LOF mutants display ectopic motor axon projections and loosely bundled axons

Complementary LOF and GOF analyses established that Mmp2 promotes motor axon fasciculation during outgrowth (Fig. 4J,N) (Miller et al., 2008). Identification of Frac as an Mmp2 substrate prompted us to investigate whether frac was involved in motor axon targeting. Hence, we generated LOF alleles via imprecise excision of a Minos transposon situated in the first intron (Metaxakis et al., 2005) (see Material and Methods). Two alleles were generated with significant deletions of the frac locus (Fig. 1B). frac^Δ1 is a 1.2 kb deletion removing part of the second exon. Anti-Frac antibodies fail to recognize Frac protein in homozygous mutant embryos, thus confirming it as a protein null (Fig. 4A,B). The second allele, frac^Δ2, is a 2.1 kb deletion removing the first two exons and half of the third exon. However, although the start is deleted, a truncated protein is translated, because frac^Δ2 is not a protein null (data not shown). Both frac alleles are homozygous viable and display no overt behavioral defects.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

A, B, frac LOF mutants display misprojections and defasciculation defects similar to Mmp2 mutants. Stage 14 wild-type (A) and frac^Δ1 (B) embryos labeled with Frac antibody. C, D, Lateral view of stage 16 embryos marked with anti-FasII, labeling all motor axon projections, and anti-GFP. In addition to the marked genotypes, embryos carry viking::GFP to visualize the cadherin-positive basement membrane. Wild-type (C) and frac^Δ1 (D) homozygous mutant embryos all have Collagen IV (green) expression at the basement membrane surrounding the ventral nerve cord (red) and midgut. E, F, MHC antibody was used to visualize the segmentally repeated pattern of muscles in each embryo. In wild-type (E) and frac^Δ1 (F) homozygous mutant embryos, muscles are properly formed and attached to the epidermis at muscle attachment sites (arrow). G–V, In each micrograph, two abdominal hemisegments of stage 17 dissected embryos stained with anti-FasII to label the ISNb (G–J) or SNa (O–R) motor projections are shown. Below each image are schematics of the branching patterns with the motors axons in brown and the muscles represented by gray boxes. G, K, In wild type, the ISNb defasciculates from the main branch of the ISN and then innervates four muscle clefts. H, I, L, M, frac^Δ1 and frac^Δ2 homozygous mutants display frayed ISNb axons and axons that separate at incorrect sites (arrows). I, M, In wild type, the SNa extends dorsally to lateral muscles and then bifurcates into posterior and dorsal branches. P, Q, T, U, In frac^Δ1 and frac^Δ2 homozygous mutants, axons make projections to aberrant muscle targets and are loosely bundled. J, N, R, V, Mmp2^W307* mutants display both defasciculation and targeting errors. A, B, Anterior is up; E–V, anterior is left and dorsal is up. Scale bars: A–D, 100 μm; E, F, 50 μm; G–V, 15 μm.

We first assayed embryonic development in frac LOF mutants. Because frac is a component of the mesodermal ECM, frac mutants might display gross defects that preclude an interpretation of any axon pathfinding errors. However, our phenotypic analyses argue that this is not the case. First, both frac alleles are homozygous viable, demonstrating that it is not an essential gene. Second, frac mutants have an intact basement membrane as visualized with viking::GFP (Morin et al., 2001; Urbano et al., 2009). Collagen IV, or viking, is expressed by macrophages of the basement membrane, which surround most tissues, including the ventral nerve cord (VNC), brain, and gut (Yasothornsrikul et al., 1997). Both frac alleles display normal localization of viking::GFP, indicating that the basement membrane is not perturbed (Fig. 4C,D and data not shown). Third, appreciable defects in muscle size, number, or epidermal attachment are not observed in frac mutants as visualized with myosin heavy chain antibody (Fig. 4E,F and data not shown) (Kiehart and Feghali, 1986). Thus, we conclude that morphological defects in frac LOF mutants do not interfere with an analysis of motor axon guidance.

We next analyzed motor axon development in frac mutants. In wild type, the intersegmental nerve branch b (ISNb) extends from the CNS bundled with the main branch of the ISN. ISNb axons defasciculate, or separate from, the ISN before the ventrolateral muscle field (VLM). Subsets of ISNb axons innervate the clefts between VLM muscles 7, 6, 13, and 12 (Fig. 4G,K). Embryos mutant for either frac allele have prominent guidance errors. Sixty-four percent of ISNb axons in frac^Δ1 embryos do not stay tightly bundled and display targeting errors in which individual axons circle back or project to neighboring nerve bundles. These were further classified as defasciculation and/or misprojection errors. Fifty-one percent of ISNb nerves in frac^Δ1 mutant embryos display hypofasciculated or frayed axons, and 29% include axons that make connections with other nerve branches or other ISNb axons (Fig. 4H,L; Table 1). Similarly, in frac^Δ2 mutants, 57% of hemisegments have aberrant ISNb branches. The same phenotypic classes are present as in frac^Δ1, with 44% of ISNb nerves exhibiting ectopic splintering of axons, and 22% displaying erroneous projections (Fig. 4I,M; Table 1). Both frac alleles are homozygous viable and fertile, so we tested whether maternal contribution of Frac contributes to motor axon behavior. The penetrance of ISNb guidance defects does not increase in the absence of both maternal and zygotic Frac protein (p ≥ 0.3 for all) (Table 1), indicating that maternal Frac does not contribute to motor axon pathfinding.

Transposon excisions can result in second site mutations if the transposon excises and briefly integrates elsewhere. Hence, we analyzed frac^Δ1/frac^Δ2 embryos and found that the penetrance remains at ∼61% (Table 1). Similarly, we tested each allele over a frac deficiency. Neither frac^Δ1/Df nor frac^Δ2/Df were statistically different from the respective homozygous mutants (p ≥ 0.7) (Table 1). These data indicate that both alleles are genetic nulls, although frac^Δ2 is not a protein null. The ISNb phenotypes observed in frac LOF mutants strongly resemble those in Mmp2 LOF mutants and support the hypothesis that Mmp2-dependent regulation of Frac contributes to motor axon targeting.

Mmp2 mutants display defects in the segmental nerve a (SNa) projection in addition to ISNb. To determine whether frac is necessary for SNa projections, we scored SNa guidance in frac mutants. In wild type, the SNa extends dorsally from the CNS to innervate the lateral muscle field. SNa axons divide into dorsal and posterior branches. The dorsal branch extends between muscles 22 and 23 and divides a second time to innervate muscle 24. The posterior branch extends to innervate muscle 8 (Fig. 4O,S). In control embryos, 11% of hemisegments display hypofasciculation or misprojection defects in the SNa (Table 1). In contrast, 49% of hemisegments in frac^Δ1 embryos and 45% in frac^Δ2 embryos contain axons that separate inappropriately and project ectopically (p < 0.001 for both compared with wild type) (Fig. 4P,Q,T,U; Table 1). The majority of defects are hypofasciculation errors in which axons inappropriately branch away from the SNa, occurring in frac^Δ1 and 43% in frac^Δ2 (Table 1). Most SNa nerves fray near choice points, in which nerve bundles divide and axons follow different pathways to innervate their specific targets. The SNa hypofasciculation phenotypes observed in frac mutants resemble those in Mmp2 mutants and suggest that Mmp2 regulates Frac processing to modulate motor axon fasciculation in multiple motor tracks (Fig. 4R,V; Table 1). A priori, Mmp2-dependent cleavage could lead to either substrate activation or inactivation. Because Mmp2 and frac mutants both display loosely bundled axons with ectopic projections, we propose that Mmp2-dependent cleavage activates Frac to promote motor axon bundling.

The pathfinding phenotype in frac LOF mutants argues that frac promotes motor axon fasciculation. Appropriate levels of motor axon bundling require a balance between interaxonal adhesion and attraction between motor axons and their targets (Winberg et al., 1998; Yu et al., 1998, 2000). The Sema1a/PlexA repulsive signaling pathway plays a central role in this process. LOF mutants in the pathway display increased motor axon bundling and fail to properly innervate their targets (Yu et al., 1998), a phenotype roughly opposite to that displayed by frac LOF mutants. We asked whether frac displays genetic interactions with Sema1a to obtain additional evidence that frac enhances fasciculation. Sema1a dominantly suppresses the ISNb pathfinding phenotypes from 64% in frac homozygotes to 18% in frac mutants that are also heterozygous for Sema1a^P1 (p < 0.001) (Table 1). This interaction is not allele specific because the frac LOF phenotypes are suppressed by two independent Sema1a alleles (Table 1). To test whether this genetic interaction reflects a specific relationship between frac and Sema1a, we asked whether frac mutants also display interactions with beat1a, another secreted guidance cue. Motor axons in homozygous beat1a embryos display increased fasciculation (Fambrough and Goodman, 1996). We find that beat1a dominantly suppresses frac defasciculation defects (Table 1), indicating that frac activity regulates the degree of attraction between motor axons. The genetic interactions between frac, Sema1a, and beat1a further argue that the pathfinding defects observed in frac mutants reflect a specific role for frac in guidance and are not the secondary consequence of morphological defects.

Frac and Mmp2 interact genetically

These genetic analyses establish that frac is required for axon targeting, and the biochemical data argue that Frac is cleaved in an Mmp2-dependent manner. To test whether Mmp2 and Frac interact in vivo to control axon guidance, we undertook genetic interaction experiments. Because genes acting in common genetic pathways often exhibit dose-dependent genetic interactions, we asked whether Mmp2 and frac display dominant genetic interactions. Embryos doubly heterozygous for Mmp2 and frac exhibit elevated levels of ISNb and SNa guidance defects (Fig. 5B; Table 2). For example, the incidence of ISNb misprojection rises from 1 and 2% in frac and Mmp2 heterozygotes, respectively, to 18% in Mmp2; frac double heterozygotes (p < 0.001 for both). These interaction data support the hypothesis that Mmp2 and Frac cooperate to regulate motor axon targeting, but they do not address the question of their relative order of action. To ask whether Mmp2 and frac act in a linear genetic pathway, we asked whether the increased motor axon fasciculation in Mmp2 overexpression embryos depends on frac activity. Pan-glial Mmp2 overexpression strongly increases ISNb and SNa fasciculation (Fig. 5E,G) (Miller et al., 2008), a phenotype distinct from the decreased fasciculation in frac LOF embryos. If Frac is an Mmp2 substrate, the Mmp2 overexpression phenotype is predicted to require Frac, and the frac LOF phenotype would be epistatic to the Mmp2 overexpression phenotype. To test this model, we overexpressed Mmp2 in a frac homozygous mutant background and found that the excess fasciculation in Mmp2 overexpressing embryos is completely suppressed (63% ISNb hyperfasciculation in repo>Mmp2 to 3% ISNb hyperfasciculation in frac^Δ1, repo>Mmp2; p < 0.001) (Fig. 5F,H; Table 2). Furthermore, these embryos display diminished motor axon bundling similar to that observed in frac homozygotes (Figs. 4H,L, 5F,H; Table 2). Hence, frac is genetically downstream of Mmp2, consistent with our biochemical data identifying Frac as an Mmp2 substrate.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

frac is genetically downstream of Mmp2 in axon guidance. Two abdominal hemisegments of stage 17 embryos labeled with anti-FasII are shown with a schematic of the phenotype shown below. A, C, Wild-type embryos have tightly bundled axonal projections. B, D, In Mmp2^W307*^/+; frac^Δ1/+ embryos, the ISNb displays defasciculation defects. E, G, repo>Mmp2 embryos exhibit ISNb axons that are hyperfasciculated and fail to reach their targets. F, H, frac^Δ1, repo>Mmp2 embryos have axons that misproject and are defasciculated. I, J, In 24B>frac embryos, the ISNb displays extra projections and misrouted axons (arrows). Anterior is left and dorsal is up in all. Scale bar, 15 μm.

These studies argue that Mmp2 cleaves Frac to promote motor axon fasciculation and raise the question of the identity of the Mmp2-generated pro-fasciculation signal. A matrix molecule such as Frac may influence signaling by a variety of means (Hynes, 2009). We can envision three likely relationships between Frac and the guidance signal. First, it is possible that full-length Frac sequesters a soluble guidance cue. In this case, Mmp2 cleavage liberates the signaling molecule from full-length Frac. frac LOF would then be characterized by elevated guidance factor signaling (no Frac to sequester cues), and Mmp2 LOF would be characterized by decreased guidance factor signaling (increased Frac to sequester cues). The finding that the guidance phenotypes displayed by Mmp2 LOF and frac LOF mutants strongly resemble each other argues against the model that Frac is simply a sink for an active guidance factor. Alternatively, the Frac fragment may have intrinsic signaling activity and act as a repulsive mesoderm-based cue to keep motor axons tightly bundled. As a third possibility, the Frac cleavage fragment may be a cofactor in a signaling complex with a guidance cue.

To distinguish between the second and third signaling mechanisms, we overexpressed Frac throughout the mesoderm. If the Frac cleavage fragment has intrinsic signaling activity, then we expect frac overexpression to lead to elevated levels of axon bundling because the axons are exposed to more pro-fasciculation signal. However, if the Frac cleavage fragment represents only part of a signaling complex, then frac overexpression would not lead to hyperfasciculation because the Frac fragment is insufficient to signal alone. We find that frac overexpression does not promote hyperfasciculation; in fact, the motor axons in 24B>frac embryos are poorly fasciculated and project ectopically (Fig. 5I,J; Table 2). This phenotype is observed with independent UAS–frac lines (Table 2), indicating that it is not a secondary effect of insertion site. The hypofasciculation observed in these embryos suggests that frac overexpression overwhelms normal regulation by Mmp2 and attenuates an active Frac fragment/signaling cue complex. This phenotype demonstrates that Frac is not sufficient to promote fasciculation and hints at the existence of a signaling cue associated with the Frac cleavage fragment.

Frac regulates a LIMK-dependent BMP signaling pathway

The motor axon phenotypes associated with frac overexpression indicate that frac is not sufficient to promote fasciculation and suggest the presence of a frac-regulated pathfinding cue. Studies in vertebrate systems provide ample evidence that the related Fibrillin family controls the bioavailability of growth factors, most notably TGFβ/BMP family members (Neptune et al., 2003; Charbonneau et al., 2004; Chaudhry et al., 2007; Sengle et al., 2008). To test whether frac modulates BMP signaling, we investigated genetic interactions between frac and the BMP pathway. With the exception of the activin-like protein Dawdle (Parker et al., 2006; Serpe and O'Connor, 2006), there is a lack of published evidence for BMP involvement in motor axon guidance. However, many of the BMP pathway mutants have pleiotropic embryonic defects, complicating LOF analysis (Khalsa et al., 1998; Wharton et al., 1999).

We first asked whether an activated form of the type I receptor saxophone (sax) modulates the frac LOF phenotype (Haerry et al., 1998). Pan-neuronal overexpression of sax^act does not yield appreciable motor axon phenotypes (Fig. 6B,E; Table 3), yet sax^act suppresses the frac LOF phenotype, reducing the incidence of defasciculation from 64 to 30% (p < 0.001) (Fig. 6A,C,D,F; Table 3), suggesting that BMP signaling is compromised in a frac LOF mutant background. Similar to frac LOF, frac overexpression embryos are characterized by reduced levels of ISNb fasciculation. In this case, we argue that Frac overexpression may overcome Mmp2-dependent regulation and effectively sequester a guidance cue. If this cue is a BMP, then expression of sax^act in a frac overexpression background is predicted to suppress ISNb defasciculation. Indeed, motor axon misprojections and hypofasciculation in frac overexpression is partially suppressed by co-overexpression of sax^act (55 to 27%; p < 0.001) (Table 3). Suppression of the frac pathfinding phenotypes via neuronal activation of the BMP pathway argues that frac activity normally promotes BMP pathway activation in motoneurons. In line with this view, mesodermal overexpression of the BMP ligand gbb suppresses the phenotype observed with frac overexpression (48 to 28%; p < 0.001) (Table 3) (Khalsa et al., 1998), providing additional genetic evidence that frac overexpression attenuates signaling via a Frac/BMP complex. However, the incomplete suppression in these assays suggests that frac regulates additional guidance cues.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Frac signals downstream through an LIMK-dependent BMP signaling mechanism. A–C, J–L, Two abdominal hemisegments of stage 17 embryos stained with anti-FasII to mark motor axon projections. A, D, frac^Δ1 mutant embryos display hypofasciculation and misprojection errors. B, E, In embryos expressing constitutively active Sax, the ISNb innervates its proper targets. C, F, Homozygous frac^Δ1 mutant embryos expressing elav>sax^act reach their targets with some defasciculation errors. Stage 16 wild-type (G), frac^Δ1 homozygous mutant (H), and 24B>frac (I) embryonic nerve cords labeled with pMad antibody. J, M, LIMK1^EY08757 homozygous mutants exhibit defasciculation errors. K, N, In embryos with neuronal RNAi knockdown of LIMK1 with concurrent neuronal misexpression of UAS–Dicer2, the ISNb displays targeting errors. L, O, The ISNb of LIMK1^EY08757; frac^Δ1 double mutant embryos is defasciculated and has ectopic projections. A–F, J–O, Anterior is left and dorsal is up. Scale bar: 15 μm. G–I, Anterior is up. Scale bars, 20 μm.

Classical BMP signaling results in phosphorylation of the transcription factor Mad, which translocates to the nucleus and activates gene transcription (Marqués et al., 2002). BMP signaling is required for neuromuscular junction growth in Drosophila, and pMad has been shown to accumulate in motoneurons. Motoneuronal pMad is apparent during embryogenesis, raising the possibility that this pathway plays a role in targeting. To test whether Mad-mediated BMP signaling is regulated by frac, we examined pMad expression in frac LOF and GOF embryos. We do not detect an obvious difference in the number of pMad-positive neurons or in pMad intensity in either LOF or GOF frac mutants (Fig. 6G–I). These results suggest that frac regulates a noncanonical BMP signaling pathway. LIMK1 is downstream of the BMP type 2 receptor Wishful thinking at the neuromuscular junction, in which it is essential for synaptic stability (Eaton and Davis, 2005). Thus, LIMK1 represented an attractive candidate to be involved in a Frac/BMP-dependent, pMad-independent pathway in motor axons. Indeed, LIMK1 homozygous mutants display defasciculation defects qualitatively similar to that of frac LOF mutants (Fig. 6J,M; Table 3). In support of a neuronal function for LIMK1 in guidance, neuronal-specific expression of LIMK1 RNAi constructs also drives ISNb defasciculation (Fig. 6K,N; Table 3). To test whether frac and LIMK1 are likely to act in a common genetic pathway, we quantified defasciculation in LIMK1; frac double homozygotes. Consistent with this hypothesis that LIMK1 is downstream of a Frac/BMP signaling complex, the frequency of defasciculation in LIMK1; frac double mutants is not significantly elevated over that observed in frac homozygotes (64 to 58%; p > 0.3) (Fig. 6L,O; Table 3). Moreover, embryos doubly heterozygous for LIMK1 and frac display an increase in misprojections relative to either heterozygote alone (9 to 1%; p < 0.05) (Table 3). Thus, we conclude that Frac acts via an LIMK1-dependent noncanonical BMP pathway to promote proper motor axon pathfinding.