The C. elegans F-spondin family protein SPON-1 maintains cell adhesion in neural and non-neural tissues

The morphogenesis of epithelial tissues and organs is profoundly dependent on the extracellular matrix (ECM), especially the specialized forms of ECM known as basement membranes (BMs) (Miner and Yurchenco, 2004). Active remodeling of BMs by tissues is essential for many developmental events, and aberrant cell-ECM interactions underlie tumorigenesis and metastasis of epithelial tissues(Larsen et al., 2006). Developing axons often grow over BM substrates, and BM components play central roles both in tissue morphogenesis and in axon guidance(Hinck, 2004). Yet, rather than a passive scaffold for tissue morphogenesis, the ECM can be regarded as an active participant in tissue morphogenesis and cell signaling(Nelson and Bissell,2006).

C. elegans embryonic epidermal morphogenesis is an example of an organogenesis process that involves multiple interactions between an epithelial sheet, underlying muscle, and an intervening BM(Chisholm and Hardin, 2005). In late embryogenesis, epidermal cells undergo coordinated shape changes that lead to embryo elongation, converting the ovoid embryo into a worm-shaped larva. Forces for elongation are generated within the epidermis by actomyosin-based contraction of circumferential actin bundles(Priess and Hirsh, 1986). Nevertheless, epidermal elongation is also critically dependent on the BM,indicating the importance of cell-cell interactions in coordinating the development of embryonic tissues.

Epidermal elongation beyond the 2-fold stage requires underlying muscle(Williams and Waterston,1994). Mutants defective in body muscle function are paralyzed and arrest at the 2-fold stage of elongation - the Pat phenotype. The need for body muscle function in epidermal elongation may reflect a role for muscle in organizing the epidermal cytoskeleton. Body muscles lie underneath dorsal and ventral epidermis, and adhere to adjacent epidermis via the BM, which allows force transmission from muscle to epidermis. The muscle-epidermal BM itself also promotes the organization of muscle and of the overlying epidermis. Importantly, different ECM components contribute to distinct aspects of epidermal morphogenesis. The earliest BM component to be deposited, laminin,is required for assembly of the myofilament lattice and the localization of dense bodies. Animals that lack laminin completely arrest in early embryonic elongation with defective muscle morphogenesis(Huang, C. C., et al., 2003; Kao et al., 2006). The BM proteoglycan Perlecan/UNC-52 is essential for assembly of the myofilament lattice (Rogalski et al.,1993), whereas type IV collagen has a later role in muscle-epidermal attachment (Guo et al.,1991).

Other BM components, such as nidogen (NID-1) or type XVIII collagen(CLE-1), are not required for embryonic morphogenesis, but play crucial roles in axon outgrowth, guidance and synaptogenesis(Ackley et al., 2001; Kang and Kramer, 2000). Similarly, the laminin receptor dystroglycan (DGN-1) is not required in embryonic morphogenesis, but functions in axon guidance(Johnson et al., 2006). Integrin signaling is required for multiple aspects of neuronal development,including cell migration, axon fasciculation(Baum and Garriga, 1997) and guidance (Poinat et al.,2002). These findings underscore the role of BM as a central scaffold for the developing nervous system.

Spondins are a conserved family of ECM proteins that were originally identified as axon guidance factors in the vertebrate spinal cord(Klar et al., 1992). In vertebrates, spondins have context-dependent effects on axon outgrowth and cell migration, and can promote neuronal differentiation(Schubert et al., 2006). Despite intensive analysis, the mechanisms by which spondins affect cell behavior and their in vivo roles remain poorly understood. Here, we report that SPON-1, the sole C. elegans member of the spondin family, is essential for embryonic morphogenesis. We show that SPON-1 promotes muscle-epidermal adhesion and is required for the completion of epidermal elongation. In the nervous system, SPON-1 promotes axon fasciculation and guidance, and continuously maintains neural architecture. These results are the first demonstration of an essential role for spondins in development.

All strains were generated from Bristol N2, and were grown under standard conditions (Brenner, 1974)unless stated. We used the mutations emb-9(g23ts cg46), pat-3(st564),ina-1(gm144), dig-1(ky188), unc-71(ju156), zig-4(gk34) and egl-15(n484). We used the GFP-expressing balancer mIn1 mIs14(Edgley and Riddle, 2001) to balance spon-1 lethal alleles.

We isolated the mutations ju348, ju402 and ju430ts in a screen for EMS-induced mutants displaying defective elongation and lumpy epidermal morphology (Mei Ding, W.-M.W. and A.D.C., unpublished). Two F1 progeny of mutagenized parents were picked and placed onto a plate, and their F2 broods scored for malformed L1 larvae. e2623 was isolated by Jonathan Hodgkin (personal communication). nc30 was isolated in screens for mispositioned ventral nerve cords(Shioi et al., 2001).

Genetic mapping placed spon-1 between dpy-10 and unc-4 on chromosome II. SNP mapping placed ju430 between pkP2148 and pkP2150, an interval of 420 kb. A 5.6 kb DNA fragment that contains ∼1.4 kb 5′ and ∼ 0.3 kb 3′ to the F10E7.4 coding sequence (pCZ695, Fig. 2A) rescued ju402 mutants to wild-type levels (data not shown). We sequenced a spon-1 full-length cDNA yk260h10 and confirmed the gene structure in Wormbase. The spon-1 coding sequence is spread over 12 exons and encodes a primary polypeptide of 819 amino acid residues. In strain constructions, spon-1 mutations were followed using allele-specific SNPs: e2623 creates a TaqI site and ju430 eliminates a BsrDI site.

Four-dimensional microscopy was performed as described(Chin-Sang et al., 1999). To record the embryogenesis of ju430 at 25°C, the objective was enclosed in a copper solenoid infused with temperature controlled water to keep embryos at the desired temperature. At least 10 embryos were recorded for each genotype. Statistical tests used GraphPad Prism (La Jolla, CA).

To determine the temperature-sensitive period of ju430, we cut open gravid adult hermaphrodites grown at 15°C and transferred two- and four-cell-stage embryos to plates pre-equilibrated to 15°C. We shifted plates to 25°C at appropriate times and scored lethality after 24 and 48 hours. 10-15 embryos were recorded for each time point.

We analyzed D type motoneurons using the Punc-25-GFP marker juIs76 (Huang et al.,2002) or the DD marker Pflp-13-GFP ynIs37(Kim and Li, 2004), PVQs using the Psra-6-GFP marker oyIs14(Sarafi-Reinach et al., 2001),and muscle using the marker trIs10(Dixon and Roy, 2005).

Electron microscopy was performed as described(Woo et al., 2004). For the analysis of muscle attachment, we analyzed four ju430 animals (raised at 22.5°C) and four e2623 animals; we cut serial sections from the posterior of the pharynx to the anterior end of the gonad. For analysis of PVQ crossing, we sectioned two ju430 oyIs14 animals.

We used duplex PCR to generate Pspon-1-GFP reporters, using 3.3 kb of 5′ DNA sequence (primer sequences available on request). This DNA was injected into wild-type animals at ∼20 ng/μl with pRF4, generating juEx592 and juEx593. To make the SPON-1::GFP C-terminal fusion pCZ697, a 5-kb BamHI-StuI fragment of pCZ695 was subcloned into pPD95.75 (Fire lab vector kit), such that GFP was fused in-frame at residue 756 of SPON-1, truncating the protein after TSR4. pCZ697 was injected into spon-1(ju402)/mIn1 mIs14 animals at 1 ng/μl,with pRF4 as a marker. All 17 transgenic lines failed to rescue ju402. An outcrossed version of transgenic line juEx734 was used in immunostaining experiments.

To tag SPON-1 at the N terminus with Venus YFP, we used a modular strategy(Hudson et al., 2006). We injected SPON-1::GFP at 10 ng/μl with Pttx-3-RFP (50 ng/μl)into spon-1(ju402)/mIn1mIs14 hermaphrodites. Rescued ju402homozygotes were selected based on the absence of mIs14, yielding transgenic lines juEx1111 and juEx1112. To express SPON-1 in pharyngeal muscles, we used the myo-2 promoter from pPD118.3. Pmyo-2-GFP::SPON-1 was injected into spon-1(ju402)/mIn1mIs14animals at 50 ng/μl with Pttx-3-RFP. Viable ju402homozygous transformants were selected, yielding lines juEx1302-1304.

We raised antibodies against peptides corresponding to residues 499-547 in TSR2 and purified the antisera according to standard procedures. The antisera recognize recombinant SPON-1 C-terminal proteins in bacterial lysates (not shown). Whole-mount immunofluorescence was as described(Finney and Ruvkun, 1990),except that fixation was extended to 3.5-4 hours, an optimization for ECM antigens (J. R. Crew and J. M. Kramer, personal communication). Monoclonal supernatants MH3 (anti-Perlecan), and MH25 (anti-αPAT-2 integrin;Developmental Studies Hybridoma Bank, University of Iowa) were used at a dilution of 1:200; DM5-6 (anti-MHC A)(Miller et al., 1983) was diluted 1:200. Phalloidin staining was as described(Ding et al., 2003). To detect GFP, we used monoclonal 3E6 (Invitrogen, Carlsbad, CA) or a rabbit polyclonal(A11122) at 1:500 to 1:1000. Raw or purified anti-SPON-1 antisera were diluted 1:100 for embryo staining and 1:800 for mixed stage staining, unless stated. For anti-SPON-1 immunostaining, we used pCZ695 arrays juEx696 and juEx698, both in ju402 background. We acquired images on Zeiss Axioplan 2 or LSM 510 confocal microscopes.

To identify genes required for muscle-epidermal interactions, we screened for mutations causing late arrest in elongation and muscle detachment. From a screen of approximately 8000 haploid genomes, we identified three mutations(ju348, ju402 and ju430ts) that cause highly penetrant defects in epidermal elongation and muscle attachment. We found that these mutations were allelic to vab-13(e2623), found in screens for morphogenetic mutants (J. A. Hodgkin, personal communication), and ven-3(nc30), isolated in a screen for mutations defective in fasciculation of the ventral nerve cord(Shioi et al., 2001). All five mutations affect the same gene, here renamed spon-1. e2623 mutants are slightly dumpy and are egg-laying defective (Egl); they display a variably abnormal (Vab) epidermal morphology that is more pronounced in early larvae(Fig. 1B,C). Other spon-1 alleles cause fully penetrant embryonic or early larval lethality with aberrant epidermal morphology and detached body muscles(Table 1, Fig. 1E,F). ju348 and ju402 are nonconditional lethals, and ju430 is a temperature-sensitive allele that resembles e2623 at 15°C and is inviable at 25°C. In penetrance, the five alleles rank from strongest to weakest: ju402>nc30>ju348>ju430ts>e2623. In addition to the epidermal elongation, muscle attachment and axon guidance defects described here, spon-1 mutations also disrupt development of the excretory canal and of epidermal seam cells(Shioi et al., 2001).

By mapping and transformation rescue, we identified the gene affected by these mutations as being F10E7.4 (Fig. 2A), which encodes a member of the spondin family of ECM proteins. Spondins are defined by the spondin domain, of unknown function, and are classified into two subfamilies, the F-spondins and the Mindins(Feinstein and Klar, 2004). F-spondins also contain an N-terminal reelin domain and up to six thrombospondin type I repeats (TSRs), whereas Mindins lack a reelin domain. C. elegans SPON-1 most resembles vertebrate F-spondins(Fig. 2B). Like some Drosophila and Anopheles F-spondins, SPON-1 contains a Kunitz serine protease inhibitor domain between TSRs 3 and 4. In domain organization, SPON-1 most resembles the Drosophila gene CG17739; however, the SPON-1 spondin domain is more similar to that of Drosophila fat-spondin (49%identity, Fig. 2C).

We identified the DNA lesions of spon-1 alleles. The ju402 mutation results in a premature stop codon in the spondin domain, and is a candidate null mutation. nc30 alters the last base of exon 3 (Fig. 2A); this does not affect the coding potential but might cause a strong loss of function by altering usage of the intron 3 splice donor. The ju348, e2623 and ju430 lesions cause missense alterations in TSRs 2, 3 and 4,respectively (Fig. 2D),suggesting that these TSRs are crucial for SPON-1 function.

To understand how SPON-1 functions, we first assayed the activity of the spon-1 promoter. Pspon-1-GFP reporters were expressed exclusively in body wall muscles from early elongation onwards(Fig. 3A-C). We constructed translational GFP fusions in which SPON-1 was tagged with GFP either at the N terminus following the signal sequence, or after the fourth TSR (see Materials and methods). N-terminal SPON-1::GFP transgenes rescued all phenotypes of spon-1(ju402); C-terminal GFP fusions showed similar localization but did not rescue spon-1 mutants, suggesting that C-terminal TSRs are required for the morphogenetic function of SPON-1. Rescuing SPON-1::GFP was expressed by muscles from the comma stage onwards(Fig. 3D, J) and showed localization to embryonic BM (Fig. 3I). In larvae and adults, SPON-1::GFP localized to dense bodies and M lines on muscle surfaces (Fig. 3E,F), as determined by colocalization with the dense body component vinculin (Barstead and Waterston,1989). Dense bodies and M lines are sites of integrin-based adhesions (Francis and Waterston,1991), and are also enriched for the BM components UNC-52/Perlecan(Rogalski et al., 1993) and EPI-1/αB laminin (Huang, C. C. et al., 2003). We consistently detected SPON-1::GFP on BM surrounding the pharynx (not shown), and within coelomocytes(Fig. 3H). As coelomocytes endocytose extracellular proteins (Fares and Greenwald, 2001), we infer that SPON-1::GFP fusions are secreted. As the spon-1 promoter is not active in the pharynx, SPON-1 can move from sites of synthesis to sites of localization, like type IV collagen (Graham et al.,1997). Unlike type IV collagen, we did not detect SPON-1::GFP at intestinal or gonadal BMs. We also generated antibodies that detect SPON-1 in whole-mount immunostaining of animals overexpressing SPON-1, but not in wild-type animals. The pattern of anti-SPON-1 staining in overexpressing animals was similar to that of SPON-1::GFP, including expression in embryonic and larval muscles (Fig. 3J,K),the excretory canal (Fig. 3K,arrowhead), pharyngeal BM (Fig. 3G) and coelomocytes (not shown). We conclude that SPON-1 is secreted from muscle and incorporates into some, but not all, BMs, and that it is enriched at integrin-based adhesion sites.

The accumulation of SPON-1 on pharyngeal basement membrane suggested that SPON-1 is recruited to BM distant from the cells that synthesize it. To test whether SPON-1 can function distant from its sites of synthesis, we expressed SPON-1::GFP in pharyngeal muscles using the myo-2 promoter(Okkema et al., 1993). Such transgenes rescued the epidermal elongation and lethality of spon-1(ju402)-null mutants (Table 1). Pharyngeally expressed SPON-1::GFP specifically accumulated at muscle-epidermal BMs during late embryogenesis(Fig. 3I). These results show that SPON-1 can be produced by cells that do not normally express it, and that it translocates within the ECM to function in specific BMs.

To determine when spon-1 functions in epidermal morphogenesis, we performed timelapse microscopy on spon-1 embryos. Most (18/19) spon-1(ju402) mutants developed normally past the 2-fold stage of elongation, then either elongated to the 3-fold stage and retracted, or did not reach the 3-fold stage. spon-1 embryos displayed muscle movements before arrest (Fig. 4E-H; see Movie 1 in the supplementary material): they twitched at 1.75-fold (∼430 minutes) and kept twitching until 3-fold stage. However, only three out of 18 embryos displayed vigorous rolling movements similar to those of the wild type. Muscle movements slowed down between the 2- and 3-fold stages, and then stopped (paralysis). During this period of slower muscle twitching, the epidermis retracted and became uneven in shape(Fig. 4G,H). ju348 and ju430 mutations caused similar or slightly weaker elongation phenotypes than did ju402. Thus, although many spon-1mutants had a body length comparable to that of a 2- or 2.5-fold stage embryo,they usually elongated beyond this stage and then retracted. Compared with other extracellular matrix mutants, spon-1 most resembles mutants such as emb-9, which lack type IV collagen.

To address when spon-1 acts in epidermal elongation, we determined the temperature-sensitive period for spon-1(ju430)(Fig. 4R). When spon-1(ju430) embryos grown at the permissive temperature (15°C)were shifted to the restrictive temperature (25°C) at the 1.5-fold stage,92-100% arrested as embryos or early larvae. When these embryos were shifted up after the 2-fold stage, lethality progressively dropped to levels observed in animals raised at the permissive temperature. We infer that SPON-1 acts directly in elongation after the 2-fold stage.

Although SPON-1 does not appear to be essential for the early integrin-dependent assembly of the myofilament lattice, the localization of SPON-1 to integrin attachment sites suggested that SPON-1 might have a subtle or redundant role in integrin-mediated adhesion. We therefore used double-mutant analysis to test whether spon-1 interacts with integrin signaling. As animals lacking PAT-2 or PAT-3 arrest at the 2-fold stage, we focused on the α-integrin INA-1, which is not essential for embryonic elongation, but which colocalizes with PAT-3 in embryonic BMs(Baum and Garriga, 1997). Animals carrying the hypomorphic mutation ina-1(gm144) display normal epidermal elongation (not shown). spon-1(ju430) ina-1(gm144) double mutants were completely inviable (in contrast to either single mutant at 20°C) and arrested in early elongation(Fig. 4M-P), suggesting that SPON-1 and INA-1 may act redundantly in early elongation. Strikingly, spon-1/+ ina-1 animals resembled spon-1 homozygotes in that they showed a late elongation arrest (Fig. 4I-L). These results suggest that when integrin function is reduced, SPON-1 becomes dose dependent for embryonic elongation.

The phenotype of spon-1 mutants suggests that, unlike integrins or Perlecan/UNC-52, SPON-1 was not essential for initial assembly of the myofilament lattice. Similarly, the late onset of epidermal elongation defects in spon-1 implies that it is not essential for the assembly of trans-epidermal attachments, as these are required for early elongation(Bosher et al., 2003; Woo et al., 2004). Instead,the late block in elongation appears to be a result of muscle detachment,which in turn leads to failure to maintain epidermal cell shape. We examined epidermal and muscle morphology by using the adherens junction protein AJM-1 to mark epidermal cell boundaries and body muscle myosin heavy chain (MHC) to mark muscle cells (Miller et al.,1983). MHC staining was normal in early elongation(Fig. 5A,B), but, as elongation progressed, we observed frequent gaps along muscle quadrants(Fig. 5C,D) due to muscle detachment. MHC organization in the remaining muscle was essentially normal. Muscle-associated PAT-3/β-integrin and UNC-52/Perlecan likewise became fragmented after the 2-fold stage because of muscle detachment(Fig. 5E-H). Epidermal actin bundles became misoriented following elongation arrest in regions where muscles had detached (Fig. 5I,J). Epidermal attachment structure components, including intermediate filaments and Myotactin, appeared normal before elongation arrest, but became discontinuous in arrested embryos(Fig. 5K,L). These results confirm that SPON-1 is not essential for early muscle or epidermal organization, but becomes essential in later muscle-BM-epidermal adhesion.

To test whether SPON-1 was required in later muscle adhesion, we analyzed post-embryonic muscle anatomy using GFP markers and electron microscopy. In addition to defective muscle-epidermal attachment, spon-1 mutants also displayed defects in muscle-muscle adhesion within a muscle quadrant(Fig. 5M,N). Muscle-muscle adhesion became progressively more defective during larval development and was suppressed by levamisole paralysis, indicating that spon-1 mutant muscles are pulled apart by the force of muscle contraction(Fig. 5O). Muscle ultrastructure in spon-1 mutant larvae was disorganized; in particular, the BM between muscle and epidermis was thicker and invaginated into the muscle (Fig. 5P-R). We conclude that SPON-1 acts continuously throughout development to promote strong adhesion both at muscle-muscle and at muscle-BM interfaces.

F-spondin has context-dependent roles in axonal outgrowth and fasciculation in vertebrates (Burstyn-Cohen et al.,1998; Tzarfati-Majar et al.,2001a). We therefore examined specific axon morphologies to determine the roles of SPON-1 in C. elegans nervous system architecture. As spon-1-null mutants arrest in embryogenesis, we examined animals mutant for the weak allele e2623 or the conditional allele ju430. spon-1 mutants displayed widespread defects in axon guidance and fasciculation of motoneurons and PVQ interneurons, described below, as well as defects in migration, fasciculation and guidance of mechanosensory neurons (not shown).

spon-1 mutants displayed extensive defasciculation of motoneuron processes within the ventral cord (Fig. 6A,B, Table 2). Motoneuron commissures also displayed defects in left-right choice of outgrowth (Table 2) and often deflected laterally upon reaching the subdorsal muscle quadrant, before eventually reaching the dorsal cord (Table 2A, Fig. 6C-E). To determine whether such guidance defects might be a secondary effect of muscle detachment, we examined motor commissure and muscle morphology simultaneously using the trIs10 marker. Of 20 commissures with dorsoventral guidance defects, only five were in regions of muscle detachment; conversely, in regions of muscle detachment, we frequently saw normal commissural guidance(Fig. 6F). We conclude that dorsoventral guidance defects appear to arise independently of muscle detachment.

spon-1 mutants raised at semi-permissive temperatures had a normal number of D neuron commissures, suggesting SPON-1 is not essential for the outgrowth of commissures (Table 2). To test whether a further reduction in SPON-1 function could affect outgrowth, we examined the progeny of ju430ts animals shifted to the restrictive temperature as L4s, and found that 100% (60/60) of DD commissures reached the dorsal midline, as in the wild type(Fig. 6I-J). spon-1(ju402) embryos also displayed a normal number of DD commissures (Fig. 6H). Although motoneuron outgrowth appeared normal in spon-1 single mutants, spon-1 mutations significantly enhanced the outgrowth defects of unc-71 ADAM mutants (Huang, X. et al., 2003) (Table 2B). We infer that SPON-1 has a minor role in outgrowth that is masked by redundancy with other pathways, such as UNC-71.

F-spondin regulates the crossing of midline axons at the floor plate of the vertebrate spinal cord (Burstyn-Cohen et al., 1999). To learn whether SPON-1 is involved in axon behavior at the C. elegans ventral midline, we examined the PVQ neurons, which undergo regulated midline crossing in the wild type. PVQ neurons in spon-1 mutants frequently displayed inappropriate crossing of the ventral midline (Fig. 7A,B). These defects became more severe during larval development, suggesting that SPON-1 actively maintains process positions. PVQ midline crossing errors were increased in ju430ts animals shifted from 15°C to 25°C in the L1 stage compared with unshifted controls(Fig. 7C), indicating that SPON-1 functions in larval development to prevent midline crossing. This post-embryonic increase in midline crossing errors was also suppressed by levamisole, indicating that crossing errors arise as a result of movement of the animal. Such post-embryonic `flip-overs' have been distinguished from crossing over during initial axon outgrowth, and imply a defect in the maintenance of process positions (Hobert and Bulow, 2003). For PVQ axons to `flip-over' they must traverse the ventral hypodermal ridge. To determine whether PVQ flip-overs might be secondary to an epidermal defect, we performed correlative light and electron microscopy of PVQ flip-over regions, and found that the overall structure of the ventral nerve cords and hypodermal ridge was normal(Fig. 7F-J). We infer that PVQ flip-overs are a direct result of reduced SPON-1 function in the microenvironment of the ventral cord.

Several other proteins act to maintain PVQ axon positions, including the secreted Ig domain protein ZIG-4 (Aurelio et al., 2002), the FGF receptor EGL-15(Bulow et al., 2004) and the giant ECM protein DIG-1 (Benard et al.,2006). To investigate how SPON-1 might interact with these midline maintenance factors, we constructed double mutants. Strikingly, zig-4and egl-15A significantly suppressed the midline crossing defects of spon-1 mutants, whereas spon-1 and dig-1 displayed additive effects on crossing (Fig. 7D). These results suggest that the SPON-1 pathway is distinct from the DIG-1 and ZIG-4/EGL-15 pathways, and moreover suggests that, although SPON-1 and ZIG-4/EGL-15 maintain PVQ axon positions, they do so using opposing mechanisms.

The founding member of the spondin family, F-spondin, was identified by its expression in the floor plate of the rat embryonic spinal cord(Klar et al., 1992). Since the discovery of F-spondin, a family of spondin-related genes has been defined,containing two subfamilies, F-spondins and the Mindins(Higashijima et al., 1997). Most animals encode at least one member of each subfamily. Genetic analysis of the spondins in other animals has been hampered by putative redundancy between family members, which are often expressed in overlapping patterns(Higashijima et al., 1997). For example, the loss of M-spondin function in Drosophila does not cause detectable phenotypes (Umemiya et al., 1997), and mindin mutant mice have a normal nervous system(He et al., 2004). By contrast, we find that the sole C. elegans spondin, SPON-1, is essential for embryonic development. The essential nature of SPON-1 function suggests that it may combine the functions performed separately by F-spondins and mindins in other animals.

F-spondins contain a reelin domain, a spondin domain and multiple TSRs. Our genetic and transgenic analysis suggests the TSRs of SPON-1 are crucial for its roles in morphogenesis and axon guidance. Three spon-1 alleles cause missense alterations in TSRs, including the strong allele ju348and the weak allele e2623. These two mutations affect equivalent residues in the WXXW motifs of TSRs 2 and 3; as these residues are not highly conserved, TSR2 may be more critical for the embryonic morphogenesis function of SPON-1 than TSR3. The temperature-sensitive allele ju430 affects a highly conserved cysteine in TSR4, and could render SPON-1 thermolabile as a result of disruption of one of the disulfide bridges important in TSR tertiary structure (Tan et al., 2002). Although all of these mutations could cause a global disruption of SPON-1 folding, they suggest that the TSRs are required for SPON-1 function in vivo. TSRs 5 and 6 of vertebrate F-spondin mediate interactions with ECM proteoglycans; plasmin-mediated cleavage within TSRs 5 and 6 can release the rest of the protein from the ECM(Tzarfaty-Majar et al.,2001b). As the motifs in TSRs 5 and 6 that mediate proteoglycan interactions are not present in invertebrate spondins, SPON-1 may be less stably attached to the ECM than are vertebrate spondins, consistent with our findings that SPON-1 can accumulate on BMs far from its site of synthesis.

Several lines of evidence suggest that SPON-1 acts either in integrin-mediated adhesion or in a closely related parallel process. SPON-1 is localized to integrin-containing sites and SPON-1 function becomes dosage sensitive when α-integrin INA-1 function is reduced, indicating that SPON-1 and INA-1 have closely related functions in embryonic development. C. elegans expresses two integrin heterodimers,αPAT-2/βPAT-3 and αINA-1/βPAT-3(Brown, 2000; Gettner et al., 1995). PAT-2 is a member of the RGD-binding subfamily of α-integrins and is likely to bind Perlecan. INA-1 is a member of the laminin-binding subfamily; it is expressed in embryonic BM but is not required for embryonic morphogenesis(Baum and Garriga, 1997; Poinat et al., 2002). As spon-1 ina-1 double homozygotes arrest earlier in elongation than does either single mutant, SPON-1, at least in part, acts in parallel to INA-1. We speculate that SPON-1 might function in a subset of both INA-1- and PAT-2-dependent adhesion processes. Interestingly, F-spondin has also been linked to integrin signaling (Terai et al., 2001), as has Mindin (Jia et al., 2005; Li et al.,2006). An important goal for the future is to determine whether spondins interact directly with integrins in vivo, or whether they indirectly influence integrin-mediated adhesion via candidate receptors, such as APP or ApoER2 (Ho and Sudhof, 2004; Hoe et al., 2005).

SPON-1 has distinct roles in axon development depending on the neuron type. Such context dependence is reminiscent of the known roles for F-spondin in the vertebrate nervous system. F-spondin promotes the fasciculation of commissural axons after they cross the floor plate(Burstyn-Cohen et al., 1999),and repels the growth cones of motoneurons that do not cross the midline(Tzarfati-Majar et al.,2001a). F-spondin also promotes the outgrowth of sensory and hippocampal neurons in vitro(Burstyn-Cohen et al., 1998; Burstyn-Cohen et al., 1999; Feinstein et al., 1999). In C. elegans, SPON-1 promotes the fasciculation of motoneuron axons in the ventral nerve cord, and acts partly redundantly to promote motoneuron outgrowth. By contrast, SPON-1 appears to prevent midline crossing(fasciculation) of PVQ interneurons.

The deflections of motor commissures at the muscle boundary in spon-1 mutants suggest that SPON-1 promotes the entry of growth cones into the muscle-epidermal BM. Defects in entry of the commissural growth cone at the dorsal muscle quadrant are also found in integrin mutants(Poinat et al., 2002),indicating that integrins act at this choice point. The similarity between spon-1 and integrin defects in motor axon development is consistent with SPON-1 promoting integrin-mediated adhesion or a closely related process in axon guidance.

Midline crossing defects could result from lack of repulsion from a midline repellent, or from lack of axonal adhesion to the normal environment(Hobert and Bulow, 2003). Because we have not seen differential localization of SPON-1 at the ventral midline, we favor the interpretation that the inappropriate crossing of PVQs in spon-1 mutants reflects a lack of adhesion. As dig-1 and spon-1 show additive effects on crossing over, these two ECM molecules may act in distinct adhesive pathways. The ZIG-4 and FGFR(EGL-15A)pathways also maintain PVQ positions, but unexpectedly zig-4 and egl-15A mutations suppress spon-1 midline crossing defects,suggesting that spon-1 and zig-4/egl-15 maintain PVQ positions by opposing mechanisms. We speculate that PVQ maintenance involves a balance of adhesive and repulsive mechanisms. SPON-1 could promote the normal adhesion of PVQs in their respective fascicles, so that a lack of SPON-1 leads to defasciculation and flip-overs. Several models have been proposed for how ZIG-4 might prevent midline crossing(Hobert and Bulow, 2003). One possibility is that ZIG-4 interferes with axon-BM adhesion at the midline, so that axons are unable to cross the midline region. As depicted in Fig. 7E, SPON-1 in the adjacent BM (green) may promote adhesion of PVQ axons (blue) to their appropriate locations, whereas extracellular ZIG-4/EGL-15 (red) might act locally to inhibit axon-BM adhesion at the midline. Thus, impaired axon-BM adhesion in spon-1 mutants might be compensated for by the loss of the ZIG/FGFR anti-adhesive pathway. ZIG/FGFR signaling might directly inhibit SPON-1 based adhesion, or could inhibit a parallel adhesion pathway. Our results underscore the view that the maintenance of process position in the nervous system involves a balance of multiple interacting forces.

In conclusion, our results demonstrate that an F-spondin-related protein promotes tissue adhesion in multiple contexts. SPON-1 functions in parallel to integrin-mediated adhesion in embryogenesis, and is antagonized by other extracellular axon maintenance factors in post-embryonic growth. These results should be useful in elucidating the in vivo cellular receptors for spondins.

We thank Mei Ding for isolating spon-1 mutations, Christie Emigh and Jian Cao for mapping, and Chris Suh and members of the Chisholm and Jin laboratories for help and support. Some reagents used in this work were obtained from the Caenorhabditis Genetic Center, which is supported by the NIH NCRR. We thank Yuji Kohara for cDNA clones, Andy Fire for vectors,and Jim Kramer for advice on immunostaining. We thank Lindsay Hinck for discussions. W.-M.W. was supported by NIH training grant T32 GM08646 and by a University of California President's Dissertation-Year Fellowship. E.C.B. was supported in part by Diversity Awards from the UC Santa Cruz Center for Biomolecular Sciences. A.G. is supported by HHMI funding to Yishi Jin. This work was supported by the Whitehall Foundation and the US Public Health Service (NIH R01 GM54657).