Dedicated mechanisms exist to maintain the architecture of an animal's nervous system after development is completed. To date, three immunoglobulin superfamily members have been implicated in this process in the nematode Caenorhabditis elegans: the secreted two-Ig domain protein ZIG-4, the FGF receptor EGL-15 and the L1-like SAX-7 protein. These proteins provide crucial information for neuronal structures, such as axons, that allows them to maintain the precise position they acquired during development. Yet, how widespread this mechanism is throughout the nervous system, and what other types of factors underlie such a maintenance mechanism, remains poorly understood. Here, we describe a new maintenance gene, dig-1, that encodes a predicted giant secreted protein containing a large number of protein interaction domains. With 13,100 amino acids, the DIG-1 protein is the largest secreted protein identifiable in any genome database. dig-1functions post-developmentally to maintain axons and cell bodies in place within axonal fascicles and ganglia. The failure to maintain axon and cell body position is accompanied by defects in basement membrane structure, as evidenced by electron microscopy analysis of dig-1 mutants. Expression pattern and mosaic analysis reveals that dig-1 is produced by muscles to maintain nervous system architecture, demonstrating that dig-1 functions non-autonomously to preserve the proper layout of neural structures. We propose that DIG-1 is a component of the basement membrane that mediates specific contacts between cellular surfaces and their environment through the interaction with a cell-type specific set of other maintenance factors.

The complex and interconnected structure of a nervous system is largely determined by patterning events during embryonic development. A question that has received little attention in the past is whether the architecture of the nervous system is maintained simply by those factors that have initially patterned it or by dedicated mechanisms that ensure the sustained integrity of the nervous system throughout life.

This question can be addressed in the nematode Caenorhabditis elegans. The architecture of the C. elegans nervous system is exceptionally well described at the single neuron level(White et al., 1986), is largely invariant from animal to animal, and can be readily visualized at different developmental and post-developmental time points with the help of fluorescent markers (Fig. 1). The nervous system develops mainly during embryogenesis, when most neurons are born, reach their final positions, organize into ganglia and axonal fascicles,and connect to their targets (Durbin,1987; Sulston et al.,1983). However, after hatching, the nervous system of larvae and adults faces a variety of challenges. During larval development, the size of the body of a worm increases considerably, as do neuronal structures such as axons. Moreover, tissues underlying some neuronal structures, such as the hypodermis, are remodeled (Podbilewicz and White, 1994). In addition, the locomotory movements of the entire body, the foraging movements of the head and the pumping movement of the pharynx, which neighbors all major head ganglia, conceivably exert significant pressure on neuronal structures. Conceptually, these sorts of challenges are faced by most nervous systems. Genetic analysis in C. elegans has revealed that dedicated maintenance mechanisms keep neuronal structures intact. Through the identification of mutants in which axon and cell body position fails to be maintained, the need for genetically-encoded maintenance factors was first recognized in the nerve ring(Zallen et al., 1999) and,subsequently in the ventral nerve cord (VNC)(Aurelio et al., 2002). The first two genes molecularly characterized as having a dedicated function in maintaining nervous system architecture, ZIG-4, a two-Ig domain containing protein and a specific isoform of the EGL-15/FGF receptor, maintain axon positioning within the fascicles of the VNC, but are not required for maintaining cell body position (Aurelio et al., 2002; Bülow et al.,2004). By contrast, SAX-7, a homologue of the cell adhesion molecule L1, is required for the maintenance of cell bodies in several neuronal structures (Sasakura et al.,2005; Wang et al.,2005). Thus, it appears that distinct mechanisms and genes underlie the maintenance of different neuronal structures (e.g. ganglia versus VNC). We wished to determine how widespread the phenomenon of maintenance is throughout the nervous system, and whether the mechanisms mediating it are shared by different neural structures, ranging from cell bodies within ganglia to axons within fascicles.

To address these questions, we analyzed the neuroanatomy of animals that carry mutations in sax-8, a previously uncloned gene, that was identified in a screen for mutants with sensory neuron defects(Zallen et al., 1999). We show that sax-8 mutants are defective in maintaining the position of cell bodies in several regions of the nervous system and of axons in the VNC. We find that sax-8 corresponds to dig-1, a previously uncloned gene originally implicated in gonad positioning in the early larva(Thomas et al., 1990), and that it codes for a large secreted protein. We show that dig-1 is required for proper basement membrane structure and that it functions post-developmentally and non-autonomously to maintain the integrity of ganglia in the head and axonal tracts in the VNC of C. elegans. We propose that DIG-1 is a component of the basement membrane that ensheathes ganglia and fascicles, mediating specific interactions between the neurons and their extracellular environment that are necessary for their structural maintenance.

### Strains and transgenes

Strains were maintained as described(Brenner, 1974). n1480,nu336, nu52, nu345, n2467 and nu319ts were generously provided by E. Ryder, J. Kaplan and R. H. Horvitz. All dig-1/sax-8 alleles were outcrossed at least three times, and then crossed into gfp-based reporter strains that label aspects of C. elegans anatomy. Rescuing array rhEx40, carrying cosmids K07E12 and R05H11, was generated by R. Proenca and E. Hedgecock and provided by E. Ryder.

The following gfp transgenes were used: evIs111 Is[F25B3.3::gfp],jcIs1: Is[ajm-1::gfp], oyIs14: Is[sra-6::gfp], oxIs12 Is[unc-47::gfp], mgIs25 Is[unc-97::gfp], bwIs2 Is[flp-1::gfp], kyIs179 Is[unc-86::gfp], ccIs4251 Is[pmyo-3mitochondrial::gfp + pmyo-3nuclear::gfp](Fire et al., 1998) and stEx30 Ex[myo-3::gfp] (Campagnola et al., 2002). Reporter transgenes generated for this study were: otEx2576, 2577, 2578 (three lines of Ex[dig-1prom4.9kb::gfp; rol-6(d)]); otEx2293, 2294,2295 (three lines of Ex[dig-1prom3.3kb::gfp;rol-6(d)]); and otEx2289, 2290, 2291, 2292 (four lines of Ex[dig-1translational::gfp; rol-6(d)]).

### RNA interference

RNAi was performed in a genetically sensitized, rrf-3 mutant background (Simmer et al.,2002). rrf-3(pk1426);oyIs14 L4 hermaphrodites were placed on bacteria harboring a plasmid to express dsRNA corresponding to the dig-1 gene (J. Ahringer library, clone 3H03), or the empty vector(L4440), as described (Fraser et al.,2000). F1 animals were scored for neural defects. No phenotype was observed in P0 animals or in adult P0s placed as embryos on the RNAi plates. Similar experiments were performed in the ky188;oyIs14 and nu345;oyIs14 genetic backgrounds.

### Anatomical analysis

Unless otherwise indicated, animals were grown at 20°C and scored under a Zeiss Axioplan 2 microscope. To obtain freshly hatched L1 larvae, embryos were picked and allowed to hatch and develop no longer than 30 minutes post-hatching. Young adults have just molted from L4, have a slightly protruding vulva and no embryos in their uteri. Three-day-old adults were selected as L4 picked 3 days earlier.

All phenotypes were scored as percent animals defective and results are shown with error bars representing the standard error of proportion. Statistical significance was calculated using the z-test to compare the proportion of abnormal animals of two genotypes. When using the same control for multiple comparisons, the P value was multiplied by the total number of comparison (*P<0.05, **P<0.001). Only when a PVQ axon directly contacted the contralateral PVQ axon, as examined at 1000× magnification, was the animal scored as displaying an axon flip-over' phenotype, and thus the penetrance reported by this stringent criterion is probably an underestimate. When one or more chemosensory neuron cell bodies were located anterior to, or on top of the nerve ring, the animal was scored as mutant. Amphid chemosensory neurons ADL, ASH, ASI, ASJ, ASK and AWB were filled with DiI (10 μg/μl Molecular Probes, Eugene, OR) as previously described(Hedgecock et al., 1985). The position of the gonad relative to the P5/6 hypodermal cells was determined in freshly hatched L1 larvae using DIC microscopy as described(Thomas et al., 1990).

### Electron microscopy (EM)

Adult animals were fixed by immersion in cacodylate-buffered glutaraldehyde for 1 hour, washed and post-fixed in buffered osmium tetroxide, and stained in buffered uranyl acetate (Glauert and Lewis, 1998). All reactions were carried out at room temperature in the dark in presence of 1% CaCl2. After dehydration through alcohol and propylene oxide, samples were embedded in EMbed-812. Thin sections were examined on a Philips CM10 electron microscope after post-staining in UAc and PbCit. All other manipulations were as previously described(Hall, 1995). Several hundred sections were analyzed per genotype.

### Axon maintenance defects in sax-8 mutants

A previous screen for mutants that affect sensory axon pathfinding in the nerve ring identified an as yet uncloned mutant, sax-8, in which the relative position of a subset of nerve ring axons and cell bodies fails to be maintained (Zallen et al.,1999). To examine neuron anatomy more broadly in this mutant, we used a pan-neuronal gfp reporter (F25B3.3::gfp) and found that the overall morphology of other major fascicles appears largely normal. This was also the case in eight other sax-8 mutant alleles examined(data not shown). We examined the morphology of the VNC in more detail, using cell-specific gfp reporters to label the axons of the neurons PVQ,AVK, RMEV and HSNL. The axons of the PVQ interneurons develop normally in ky188 mutant animals but fail to maintain their position(Fig. 2A,B). Specifically,freshly hatched larvae in which the embryonic nervous system has long completed its development (∼300 minutes post-axon outgrowth) display normally positioned axons. By the adult stage, however, a significant proportion of the animals exhibit axon position defects, characterized by the inappropriate placement of axons across the ventral midline(Fig. 2A,B). We have previously termed this phenotype axon flip-over defect' or axon maintenance defect' to distinguish it from developmental defects during phases of axon outgrowth(axon cross-over defects'). Similar to other maintenance mutants(Aurelio et al., 2002; Bülow et al., 2004), these defects can occur along the entire anteroposterior axis of the VNC and usually once per animal, i.e. the axon flips from one fascicle into the other and back into the original fascicle. An axon flip-over can originate from either the right or the left fascicle of the VNC. We also examined other sax-8mutant alleles and found that PVQ axon maintenance is mildly affected in ky199, ky201 and nu345 animals, and unaffected in other sax-8 mutant alleles (Fig. 2B).

The PVQ axon maintenance defect of sax-8 mutants is locomotion dependent, as it can be suppressed by paralysis, induced pharmacologically(levamisol treatment) or genetically (in unc-36, unc-13, unc-54 and unc-97 mutant backgrounds) (Fig. 2C). This indicates that mechanical stress resulting from the locomotion of the worm largely contributes to the failure of maintaining axonal position in sax-8 mutants, a situation similarly observed in other axon maintenance mutants (Aurelio et al., 2002; Bülow et al.,2004).

We also examined the anatomy of the axons of the AVK interneurons (using reporter flp-1::gfp) and of the RMEV motoneuron (using reporter unc-47::gfp) and found that they are abnormally placed in the opposite fascicle of sax-8 mutant animals(Fig. 2D,E). These defects can be suppressed by paralysis (Fig. 2E), indicating that these are also maintenance defects that result from mechanical stress exerted onto the VNC. Axons of the HSN motor neuron class are also affected in sax-8 mutants (see Fig. S1 in the supplementary material).

### Cell body maintenance defects in sax-8 mutants

Using a panel of sax-8 mutant alleles (described in more detail below), we recapitulated and expanded the previously reported maintenance defects of nerve ring anatomy (Zallen et al., 1999). The initial position of the amphid chemosensory neurons with respect to the nerve ring fascicle is entirely normal in L1 larvae (Fig. 3A,B), which indicates that they initially develop normally. Later, the chemosensory neuron cell bodies become progressively misplaced, displaying only partial mutant phenotypes by the fourth larval stage, and reaching full penetrance and expressivity by the third day of adulthood(Fig. 3A,B). This progressive loss of proper positioning is observed with similar severity in all sax-8 alleles examined. Frequent loss of fasciculation of the nerve ring was also observed in all sax-8 mutants (data not shown).

Examining the position of neuronal cell bodies located in other ganglia, we found that the cell bodies of the PVQL/R neurons also fail to maintain their normal position in sax-8 mutants. PVQ cell bodies are located in the lumbar ganglion in the wild type (Fig. 1), and are normally positioned in sax-8 mutants until the end of embryogenesis. However, by the first larval stage and adulthood,the cell bodies of one or both PVQ neurons became frequently misplaced along the anteroposterior axis in sax-8 mutants(Fig. 3C,D). The cell body displacement and axon flip-over of PVQ (described above) in sax-8mutants are independent of each other, as both defects are rarely found simultaneously in an animal (data not shown). The occurrence of cell body maintenance defects of the PVQ and the amphid chemosensory neurons is temporally distinct. Although both cell body types are initially placed normally, PVQ cell body maintenance defects are already evident at the first larval stage, while amphid cell body maintenance defects are manifested only at later larval stages.

The cell bodies of the AVK interneurons, which are located in the ventral ganglion (Fig. 1), are also misplaced in the majority of sax-8 mutants(Fig. 4A,B). In contrast to the AVK axon flip-over defects described above, the AVK cell body displacement defects are already strongly evident at late embryonic stages (data not shown). The AVK cell body displacements at the L1 stage may therefore be a reflection of either mispositioning during development or of a failure to maintain correct cell position at earlier embryonic stages, which we could not assess as the gfp marker for the AVK neuron is not activated early enough. As the AVK cell body defects increase in severity between the L1 and adult stage, it appears that, at least, there is a larval maintenance component to the defects.

### Function of sax-8 in other cell types

We examined the morphology of three major non-neuronal tissue types, the gonad, hypodermis and muscles. sax-8 mutants often display an anterior or mild posterior displacement of the gonad with respect to other unaffected landmark cells in the worm (the row of P and V cells, and the nuclei of body wall muscles) (Fig. 4C). However, none of the 10 sax-8 alleles displays any morphological defects in larvae or adults that could be indicative of abnormal hypodermis (knobs, lumps, deformities, interrupted and abnormal alae). We also examined in detail the hypodermal cell junctions in ky188 and nu345 mutants, using the adherens junction marker ajm-1::gfp, and found that the hypodermal tissue is normal in commastage embryos, freshly hatched larvae, and adults (data not shown). Other structures labeled by this reporter, such as the vulva, the excretory duct and pore, and the anus, appeared completely normal. Only the junction of the pharynx to the tip of the nose appeared slightly twisted in about a third of the animals in each of the two alleles. The overall shape of muscle cells along the body and of muscle arms (projections sent by the muscle toward neuronal partners), observed using myo-3prom::gfp, is also normal in larvae and adults (data not shown). Details of muscle subcellular structure, including the morphology of nuclei and mitochondria (visualized with transgene ccIs4251), dense bodies (mgIs25) and myofilaments (stEx30) are also normal (data not shown).

### Timing of sax-8 action

The post-developmental onset of most, if not all, sax-8 nervous system defects strongly hints at a late, post-developmental role of sax-8. To corroborate this notion, we carried out temperature shift experiments using a temperature-sensitive allele of sax-8, nu319ts. Adult animals of this mutant genotype display completely wild-type amphid chemosensory neuron position when grown at the permissive temperature of 15°C, but display a maintenance defect at the non-permissive temperature of 25°C (Fig. 5A).

Animals were grown at 15°C up to a given developmental stage, shifted to 25°C at defined developmental stages and scored as 3-day-old adults(Fig. 5B; red line). Any shift after the first larval stage, even as late as the fourth larval stage, leads to a mutant phenotype in adults, indicating that sax-8 is required after embryonic development is completed in order to maintain neuron position. In the reciprocal experiment, animals were grown at 25°C, shifted to 15°C and scored as 3-day-old adults(Fig. 5B; blue line). In order to see a significant defect, animals had to be grown at least to mid-larval stages at the non-permissive temperature. The most severe defects were observed only if animals were kept at the non-permissive temperature throughout all larval stages into adulthood. These experiments demonstrate that: (1) the phenotype of sax-8(nu319ts) animals is independent of normal sax-8 activity during embryogenesis and early L1; and (2) sax-8 activity is required post-developmentally, throughout the larval stages, for neurons to maintain their position.

### sax-8, dig-1 and K07E12.1 are the same gene

As mentioned earlier, a previous screen for mutants that affect sensory axon morphology (SAX genes) identified two genes, sax-7 and sax-8, in which the position of the amphid chemosensory neurons relative to the nerve ring fails to be maintained(Zallen et al., 1999). sax-7 was previously found to encode an Ig- and FnIII-domain-containing protein homologous to the vertebrate L1 protein(Sasakura et al., 2005; Wang et al., 2005). The two other genes that have been implicated in VNC maintenance, zig-4 and egl-15, also encode Ig-domain containing proteins(Aurelio et al., 2002; Bülow et al., 2004). sax-8 mapped to a 2.18 cM interval on LGIII(Zallen et al., 1999). We further mapped sax-8(ky188) by three-point mapping with genetic markers to a 0.9 cM region (Fig. 6A). Among the genes in this region, a single predicted gene codes for an Ig- and FnIII-domain containing protein, K07E12.1(www.wormbase.org; Fig. 6B,C). Reduction of K07E12.1 activity by RNA interference (RNAi) phenocopied the sax-8mutation as it led to neuroanatomical defects of the PVQ axon and amphid chemosensory sensory neurons (Table 1). Genomic DNA that spans the K07E12.1 locus, and only one other gene (Fig. 6B), rescues defects of the PVQ axons and of the amphid neurons of sax-8 mutants(Table 1). Sequencing of K07E12.1 in sax-8 mutant animals revealed an 11 bp deletion in sax-8(ky199).

The predicted K07E12.1 gene also contains mutations in dig-1mutant animals (Burket et al.,2006), a previously uncloned gene initially identified because of its involvement in gonad placement (dig=displaced gonad)(Thomas et al., 1990). We found mutations ky188 and dig-1(n1321) to be allelic as they fail to complement (Table 1). In addition, ky188, ky199 and ky201 mutants display the displaced gonad phenotype reported for dig-1(n1321)(Thomas et al., 1990)(Fig. 4C). Reciprocally, dig-1 mutant alleles display similar neuronal defects to sax-8 mutants (Figs 2, 3, 4). Furthermore, the defects of chemosensory and PVQ neurons in dig-1(nu336) and dig-1(n1480) mutant animals can be rescued with wild-type copies of K07E12.1 (Table 1). We conclude that sax-8, dig-1 and K07E12.1 define the same gene. From this point onwards, we refer to sax-8 as dig-1.

### dig-1 codes for a giant novel secreted protein

The dig-1 gene spans 48 kb and encodes a predicted polypeptide that is 13,100 amino acids long. Consistent with the predicted gene structure,a single transcript of ∼39 kb is detectable by northern blot analysis(Burket et al., 2006), and mutant alleles lie within the 5′ or 3′ end of the gene (see below). The predicted DIG-1 protein can be roughly divided into three regions(Fig. 6C). An N-terminal region contains numerous conserved domains that are implicated in cell-cell interactions and adhesion, including immunoglobulin (Ig), fibronectin type III(FnIII), Sushi/Complement control protein (CCP), and epidermal growth factor(EGF) domains (Fig. 6C). A large central region is characterized by the presence of a large number of dig-1-specific repeats, which are rich in β-strands(http://cubic.bioc.columbia.edu/pp/predictprotein.html),therefore suggesting that these repeats may form individual, globular domains. A C-terminal region contains Ig, EGF and several von Willebrand factor A (vWA)domains, which are frequently part of proteins implicated in cell adhesion,and in components of the extracellular matrix. DIG-1 contains a signal peptide, but no hydrophobic sequences indicative of a potential transmembrane domain or GPI anchors, suggesting that DIG-1 is secreted. Database searches revealed that DIG-1 is the largest secreted protein predicted in the entire C. elegans proteome or any other known prokaryotic or eukaryotic proteome.

Orthologous proteins with similar domain composition and over 90% identical sequences can be found in two other nematode genomes (C. briggsae and C. remanei). No clear homologues of DIG-1 have been detected in other organisms, perhaps because gigantic genes are notoriously difficult to predict correctly in higher organisms with larger intron size. However, two structural similarities of DIG-1 with proteoglycans are notable (E. Ryder, personal communication). First, the central, repetitive region of DIG-1 contains a large number of Ser-Gly peptide motifs in an acidic environment, a feature shared by glycosaminoglycans attachment regions(Lindahl and Hook, 1978). Second, DIG-1 and the hyalectan class of proteoglycans show a similar domain architecture with N- and C-terminal Ig, EGF and Sushi/CCP domains, and a central glycosaminoglycan attachment region(Bandtlow and Zimmermann,2000). However, these hyalectan-type proteoglycans tend to be significantly smaller than DIG-1.

### Characterization of dig-1 mutant alleles

To identify the molecular lesions in dig-1 mutants, we sequenced large parts of the dig-1 gene in all 10 dig-1 mutants. As we were unable to reliably PCR amplify and sequence the repetitive central portion of the gene, we were limited to sequencing regions at the 5′(position 9892 to 19576 of cosmid K07E12), 3′ (54489 to 58025) and center (35374 to 35983) of the gene. We identified mutations in four alleles(Fig. 6C, Table 2). The molecular nature of mutations n1321, nu52 and ky199, all of which introduce premature stop codons, suggests that they reduce the activity of dig-1. This is corroborated by RNAi experiments, in which dig-1(RNAi) animals display phenotypes similar in nature and severity to the dig-1 mutants, including misplaced chemosensory neurons in adults but not in larvae (Table 1). dig-1(RNAi) also caused axonal flip-over defects like those observed in animals carrying the ky188 mutation, a molecularly uncharacterized allele (Table 1). This indicates that the ky188 phenotype is also due to reduction of dig-1 function.

Non-null alleles of genes are often enhanced when placed over a chromosomal deficiency. To test whether such a test would reveal that any of the dig-1 alleles are not null alleles, we used deficiency nDf16that removes dig-1, and examined the phenotype of dig-1/nDf16 animals that carry only one mutant copy of the gene. The PVQ axon flip-over phenotype of ky188/nDf16 and ky188homozygous animals is similar in penetrance and expressivity(Table 1), indicating that ky188 results in a very strong if not complete loss of dig-1function in the VNC, despite ky188 being less severe than other alleles for other phenotypes (i.e. maintenance of cell body position, Table 2). However, nu345/nDf16 animals displayed a strong PVQ axon flip-over phenotype(42%, Table 1), compared with the weak phenotype in nu345 homozygous mutants(Fig. 2B, Table 2). Thus, despite the overall severity of the nu345 mutation for other phenotypes, it only partially and selectively reduces the function of dig-1. Furthermore,RNAi in the ky188 or nu345 mutant backgrounds did not enhance the mutant phenotypes (Table 1). In addition, dig-1(RNAi) and all 10 mutant dig-1 alleles are completely viable and appear morphologically wild type. Taken together, none of the 10 alleles lead to a complete loss of all individual functions of dig-1, but in specific cellular contexts,individual alleles may exhibit a very strong, and possibly even complete,loss-of-function phenotype.

### dig-1 functions non-autonomously for neuronal maintenance

To determine potential sites of action of dig-1, we generated transcriptional gfp reporter fusions that include up to 4.9 kb of genomic sequence upstream of the first exon of the dig-1 locus, as well as a translational reporter fusion that includes the first three,relatively large introns of the gene (see Fig. S2 in the supplementary material). All reporters show qualitatively similar expression patterns. In transgenic embryos, gfp expression is detected in the developing gut by the comma stage and up to the twofold stage(Fig. 7A). By the late threefold stage of embryogenesis, dig-1 reporter constructs become strongly expressed in mesodermal cells(Fig. 7A). In larvae of all stages and adults, dig-1 reporter genes remain strongly expressed in mesodermal cells, including body wall, head, vulval, uterine, pharyngeal muscles, and in enteric and anal depressor muscles. Additional mesodermal cells expressing the dig-1 reporter construct include support cells in the head (GLR cells), the head mesodermal cell, coelomocytes and sex myoblasts. Expression could also be observed in the hypodermis. No expression could be observed in the nervous system.

The absence of dig-1 reporter gene expression in neuronal cells suggests that dig-1 may function non-autonomously to maintain of nervous system architecture. To provide firmer evidence for this, we carried out genetic mosaic analysis (Herman,1995), using a transgenic strain harboring a rescuing extrachromosomal array that carries wild-type copies of the gene dig-1, along with reporters that are expressed in muscles(myo-3::gfp) or pan-neuronally (F25B3.3::DsRed2)(Fig. 7B). We selected animals that had lost the rescuing extrachromosomal array at the division of the zygote into the two blastomeres, AB (from which almost all neurons are derived) and P1 (from which almost all muscles are derived). Animals that lost the extrachromosomal array in the AB blastomere, but retained it in the P1 blastomere and its muscle descendants, displayed a wild-type phenotype for the PVQ axon and for the amphid chemosensory neurons(Fig. 7C). Conversely, animals that lost the extrachromosomal array in the P1 blastomere, but retained it in the AB blastomere, displayed a mutant phenotype of the PVQ axons and of the amphid neurons. These results indicate that dig-1 is not required within the PVQ or amphid neurons themselves, or within the nervous tissue as a whole for its role in maintenance. Rather, dig-1 is required in the muscles, and perhaps also in the few hypodermal descendants of P1. Given the nature of the cell lineage of C. elegans, it is difficult to distinguish between the contribution by the muscles and part of the hypodermis, but clearly dig-1 functions non-autonomously to ensure maintenance of nervous system morphology.

In light of the muscle (and also hypodermal) site of expression of dig-1, we note that no defects in muscle or hypodermal development or fine structure were observed by the gfp reporter-based analysis(described above) and by EM analysis described below. This indicates that the neuronal defects of dig-1 mutants are not merely a consequence of gross developmental or morphological defects of neighboring tissue.

### dig-1 function is required for proper basement membrane structure

Basement membranes are extracellular structures composed of proteins and proteoglycans, including laminins, collagens type IV and XVIII, nidogen, and often unusually large molecules [e.g. UNC-52/Perlecan(Rogalski et al., 1993)]. We addressed whether dig-1 mutations affect basement membranes integrity using transmission electron microscopy. We observed frequent basement membrane abnormalities in dig-1 mutant animals. Specifically, in addition to the layer of basement membrane normally found to lie directly apposed to the plasma membrane of muscle, hypodermal and pharyngeal cells(Kramer, 2005)(Fig. 1), dig-1mutants exhibit multiple supplementary layers of basement membrane that form stacks and loose whorls (Fig. 8). All three alleles of dig-1 examined (ky188,nu345 and n1321) show similar phenotypes in the head region(Fig. 8). We could not observe any clear basement membrane defects in the VNC, but it is possible that fine ultrastructural or molecular abnormalities went undetected at this resolution of analysis.

To determine whether the basement membrane abnormalities were already present in two- and three-fold embryos, we sectioned and analyzed three ky188 embryos and found no obvious defects in basement membrane structure. In contrast to adults, tissues are much more closely packed in embryos, perhaps leaving little room for basement membrane delamination. This would be consistent with the idea that the function of dig-1primarily lies in allowing mature animals to deal with challenges relating to post-embryonic growth, tissue expansion and mechanical stress.

### dig-1 encodes a novel nervous system maintenance factor

Our work establishes dig-1 as a novel factor that contributes to the maintenance of the architecture of the nervous system. In dig-1mutants, a number of neuroanatomical features develop normally, but become displaced post-developmentally. Displacement of cell bodies progressively worsens through larval and adult stages, suggesting that it is the continued wild-type activity of dig-1 that ensures their maintenance. The timing of dig-1 action, determined using a temperature-sensitive allele of dig-1, further underscores the post-developmental activity of dig-1 and indicates that the defects are not just a secondary consequence of initial developmental defects that worsen over time.

Given that the ventral midline flip-over of the PVQ axon is suppressed in paralyzed worms, it appears that mechanical stress inflicted upon the VNC by the locomotory movements of the worm constitute a main cause for maintenance failure in dig-1 mutants. It is conceivable that the progressive misplacement of chemosensory neurons results from a combination of the active foraging movements of the head and of internal forces generated by pharyngeal pumping. These types of movements are less readily inhibited by paralysis of worms, which prevented us from directly testing this hypothesis.

The expression of dig-1 reporter gene constructs in muscles and hypodermis is detected only by the threefold stage of embryogenesis, which is after neuronal cell migration, axogenesis and fascicle formation have taken place to lay down the embryonic nervous system of the worm(Durbin, 1987). This timing of expression is again consistent with a post-developmental role for dig-1. The onset of expression is also later than that of principal components of the basal lamina, the laminin genes themselves, which begins during much earlier stages in morphogenesis(Huang et al., 2003; Kao et al., 2005). However, it remains possible that the expression pattern of dig-1 is not fully revealed by our reporter fusions. Combined with the post-developmental occurrence of dig-1 defects and the temperature-sensitive period of dig-1, we can nevertheless firmly conclude that dig-1functions post-developmentally.

The present analysis of dig-1 mutants adds another temporal dimension to the maintenance of nervous system architecture. Previous analysis of this phenomenon, based on loss of function analysis of egl-15 and zig-4, as well as microsurgical approaches established that the first larval stage presents a crucial stage in which maintenance factors are required to maintain axon position (Aurelio et al., 2002; Bülow et al., 2004). Although head chemosensory neurons appear to require maintenance factors at a similar time point, as well as later in life, our analysis of the cell body positioning of the PVQ tail neurons reveals that these neurons are already starting to become displaced at late embryonic stages. PVQ neurons are born slightly earlier than other head chemosensory neurons affected by the loss of dig-1(Sulston et al., 1983), which may provide one explanation for the earlier occurrence of PVQ defects. Alternatively, it could be envisioned that nervous system structures that show later onsets of maintenance defects are better anchored in their surrounding environment, therefore taking more time to be displaced by mechanical stress in the absence of dig-1. In either case, our observation reveals that maintenance factors are required during multiple post-developmental stages.

Mutations in dig-1 affect basement membrane structure, suggesting that the function of DIG-1 is intimately linked to the basement membrane. Basement membranes are ∼20-100 nm thick macromolecular assemblies of proteins and proteoglycans that are secreted from various tissues, including muscles and the hypodermis (schematically shown in Fig. 1)(Kramer, 2005; White et al., 1986). DIG-1 could be a component of the basement membrane that interacts with several other basement membrane proteins and proteoglycans via its multiple protein-protein interaction domains. It is possible that DIG-1 could act as a scaffold for late assembly or integrity of the basement membrane. However, it is unlikely that DIG-1 is essential for the initial assembly of the basement membrane, as the complete disruption of central components of nascent basement membrane (e.g. laminin α or β subunit, and collagen IV mutants)causes early and severe developmental, phenotypes that are not observed in dig-1 mutants (Huang et al.,2003; Kao et al.,2005). The function of dig-1 also differs from that of known basement membrane proteins in that the defects of dig-1 mutant animals are due to a failure in maintenance of axon and cell body position,while those observed in other mutants such as nid-1 (nidogen) and cle-1 (type XVIII collagen), for example, are developmental defects(Kim and Wadsworth, 2000; Ackley et al., 2001).

It is conceivable that DIG-1 may mediate specific interactions between neurons and their extracellular environment. Given the thickness of the basement membrane, an intriguing property of the DIG-1 protein that is shared by other basement proteins such as collagen IV, is its predicted gigantic size. If one assumes that the individual domains of DIG-1 form globular units and if one takes the 2 nm diameter of the FnIII domains(Chothia and Jones, 1997) as a rough guideline for domain size, a linear array of DIG-1 domains can extend far over 100 nm in length. In such an extended conformation, DIG-1 is easily long enough to bridge with its central repeat region across basement membranes and interact with its N- and C-terminal protein interaction domains with surface proteins that are located on cells on either side of basement membranes. Indeed, using some fixation techniques, apparent connections between the basement membranes of separate adult tissues are seen in C. elegans (D. H. H. Axang and M. Pilon, unpublished). The absence of such a connection may not disrupt basement membranes per se, but may lead to exactly the type of detachment from surrounding tissue that we observe in DIG-1 mutants. In any case, the domain composition of DIG-1 strongly indicates that DIG-1 can interact with diverse proteins of the basement membrane and cell membrane. DIG-1 may be viewed as a scaffold that mediates interactions among proteins within fascicles and ganglia in different regions of the nervous system and/or with the basement membrane that neighbor these structures(Fig. 1).

The effects of dig-1 are remarkably cell-type specific. Distinct alleles differentially affect the maintenance of individual neuronal structures, namely the cell body or the axon of different neuron classes located in various regions of the nervous system. Moreover, dig-1 is required in different parts of the nervous system, i.e. in several distinct fascicles and ganglia. The DIG-1 protein may interact with a specific set of different partner proteins in different cellular contexts. Consistent with this view is the fact that the 10 dig-1 alleles cannot be easily ordered into an allelic series. Rather, a given allele can simultaneously display the strongest defects for one aspect of the phenotype and have a very weak effect on other aspects. Candidates for cell-specific interaction partners of dig-1 include the previously identified axon maintenance factors encoded by the zig-4, egl-15(5A) and sax-7 genes(Aurelio et al., 2002; Bülow et al., 2004; Sasakura et al., 2005; Wang et al., 2005; Zallen et al., 1999). ZIG-4, a secreted protein, hypodermally expressed EGL-15(5A) and muscle-expressed DIG-1 function non-autonomously to maintain VNC axon position, while sax-7acts autonomously within neurons. Intriguingly, null mutations in sax-7,egl-15(5A) and zig-4 cause defined subsets of the dig-1mutant phenotype. A zig-4-null mutation specifically affects axon position of PVQ and AVK, but not HSN and RMEV(Aurelio et al., 2002), and a egl-15(5A) null mutation specifically affects axon position of PVQ and mildly HSN, but not RMEV and AVK(Bülow et al., 2004). Loss of dig-1 alone leads to axonal defects in PVQ, AVK, RMEV and HSN,revealing all the defects observed in zig-4 and egl-15single mutants. dig-1 functions to maintain not only axonal positioning in the VNC but also cell body positioning in ganglia of the head,a phenotype not observed in zig-4 or egl-15 mutants, but observed in sax-7 mutant (Zallen et al., 1999). DIG-1 may therefore interact with specific proteins in distinct cellular context to affect axon maintenance.

### Generality of the principle of maintaining tissue integrity

Nervous systems throughout the animal kingdom, including those of vertebrates, are under constant mechanical stress generated by the growth of the animal, the addition of new neurons and their axons throughout life, and displacement of nerves in moving structures of the body such as limbs, mouth and the optic nerve, which is constantly moving as the eye changes position to scan the environment. Conceivably, dedicated mechanisms also exist to maintain the architecture of these nervous systems. Myelination of axons is a common feature of vertebrate nervous systems and may provide the architectural stability that a nervous system requires, but not all axons are myelinated and unmyelinated axons may require dedicated maintenance mechanisms to ensure their structural integrity. In addition, maintenance of the integrity of cellular structures is not a phenomenon restricted to the nervous system. For example, maintenance of amphid organ integrity is actively controlled by the alr-1 transcription factor that acts cell autonomously in the amphid socket cell (Tucker et al.,2005), and attachment of the cuticle to the hypodermis is mediated post-developmentally by mua-3(Bercher et al., 2001). It will be interesting to determine the extent of evolutionary conservation of the molecular mechanisms of maintaining exquisite structural features in an organism.

We thank E. Ryder, J. Kaplan and R. H. Horvitz for generously providing mutant alleles; E. Ryder for generously sharing unpublished results, providing additional reagents, and discussions; the NIH-funded Caenorhabditis Genetics Center and members of the worm community for providing reporter gene constructs; H. Bigelow for protein database searches; Q. Chen for expert injection assistance; R. Proenca and E. Hedgecock for generating the rhEx40 array; and I. Greenwald, E. Ryder, W. Wadsworth and members of the Hobert laboratory for comments on the manuscript. This work was supported by a postdoctoral fellowship from the Natural Sciences and Engineering Research Council of Canada (C.Y.B.) and by a grant from the Muscle Dystrophy Association (O.H.). D.H.H. was supported by NIH RR 12596. O.H. is an Investigator of the HHMI.

Ackley, B. D., Crew, J. R., Elamaa, H., Pihlajaniemi, T., Kuo,C. J. and Kramer, J. M. (
2001
). The NC1/Endostatin domain of Caenorhabiditis elegans type XVIII collagen affects cell migration and axon guidance.
J. Cell Biol.
153
,
1219
-1232.
Aurelio, O., Hall, D. H. and Hobert, O. (
2002
). Immunoglobulin-domain proteins required for maintenance of ventral nerve cord organization.
Science
295
,
686
-690.
Bandtlow, C. E. and Zimmermann, D. R. (
2000
). Proteoglycans in the developing brain: new conceptual insights for old proteins.
Physiol. Rev.
80
,
1267
-1290.
Bercher, M., Wahl, J., Vogel, B. E., Lu, C., Hedgecock, E. M.,Hall, D. H. and Plenefisch, J. D. (
2001
). mua-3, a gene required for mechanical tissue integrity in Caenorhabditis elegans, encodes a novel transmembrane protein of epithelial attachment complexes.
J. Cell Biol.
154
,
415
-426.
Brenner, S. (
1974
). The genetics of Caenorhabditis elegans.
Genetics
77
,
71
-94.
Bülow, H. E., Boulin, T. and Hobert, O.(
2004
). Differential functions of the C. elegans FGF receptor in axon outgrowth and maintenance of axon position.
Neuron
42
,
367
-374.
Burket, C. T., Higgins, C. E., Hull, L. C., Berninstone, P. M. and Ryder, E. F. (
2006
). The C. elegans gene dig-1 encodes a giant member of the immunoglobulin superfamily that promotes fasciculation of neuronal processes.
Dev. Biol.
(in press).
Campagnola, P. J., Millard, A. C., Terasaki, M., Hoppe, P. E.,Malone, C. J. and Mohler, W. A. (
2002
). Three-dimensional high-resolution second-harmonic generation imaging of endogenous structural proteins in biological tissues.
Biophys. J.
82
,
493
-508.
Chothia, C. and Jones, E. Y. (
1997
). The molecular structure of cell adhesion molecules.
Annu. Rev. Biochem.
66
,
823
-862.
Durbin, R. M. (
1987
).
Studies on the development and organisation of the nervous system of Caenorhabditis elegans. PhD thesis
, University of Cambridge.
Fire, A., Xu, S., Montgomery, M. K., Kostas, S. A., Driver, S. E. and Mello, C. C. (
1998
). Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans [see comments].
Nature
391
,
806
-811.
Fraser, A. G., Kamath, R. S., Zipperlen, P., Martinez-Campos,M., Sohrmann, M. and Ahringer, J. (
2000
). Functional genomic analysis of C. elegans chromosome I by systematic RNA interference.
Nature
408
,
325
-330.
Glauert, A. and Lewis, P. (
1998
).
Practical Methods in Electron Microscopy.
London:Portland Press.
Hall, D. H. (
1995
). Electron microscopy and three-dimensional image reconstruction.
Methods Cell Biol
48
,
395
-436.
Hedgecock, E. M., Culotti, J. G., Thomson, J. N. and Perkins, L. A. (
1985
). Axonal guidance mutants of Caenorhabditis elegans identified by filling sensory neurons with fluorescein dyes.
Dev. Biol.
111
,
158
-170.
Herman, R. K. (
1995
). Mosaic Analysis. In
Caenorhabditis elegans-Modern Biological Analysis of an Organism
. Vol.
48
(ed. H. F. Epstein and D. Shakes), pp.
123
Hobert, O. (
2002
). PCR fusion-based approach to create reporter gene constructs for expression analysis in transgenic C. elegans.
Biotechniques
32
,
728
-730.
Huang, C. C., Hall, D. H., Hedgecock, E. M., Kao, G., Karantza,V., Vogel, B. E., Hutter, H., Chisholm, A. D., Yurchenco, P. D. and Wadsworth,W. G. (
2003
). Laminin alpha subunits and their role in C. elegans development.
Development
130
,
3343
-3358.
Kao, G., Huang, C. C., Hedgecock, E. M., Hall, D. H. and Wadsworth, W. G. (
2005
). The role of the laminin beta subunit in laminin heterotrimer assembly and basement membrane function and development in C. elegans.
Dev. Biol
.
290
,
211
-219.
Kim, S. and Wadsworth, W. G. (
2000
). Positioning of longitudinal nerves in C. elegans by nidogen.
Science
288
,
150
-154.
Kramer, J. M. (2005). Basement membranes. In WormBook (ed. The C. elegans Research Community). doi/10.1895/wormbook.1.16.1, http://www.wormbook.org.
Lindahl, U. and Hook, M. (
1978
). Glycosaminoglycans and their binding to biological macromolecules.
Annu. Rev. Biochem.
47
,
385
-417.
Podbilewicz, B. and White, J. G. (
1994
). Cell fusions in the developing epithelial of C. elegans.
Dev. Biol.
161
,
408
-424.
Rogalski, T. M., Williams, B. D., Mullen, G. P. and Moerman, D. G. (
1993
). Products of the unc-52 gene in Caenorhabditis elegans are homologous to the core protein of the mammalian basement membrane heparan sulfate proteoglycan.
Genes Dev.
7
,
1471
-1484.
Sasakura, H., Inada, H., Kuhara, A., Fusaoka, E., Takemoto, D.,Takeuchi, K. and Mori, I. (
2005
). Maintenance of neuronal positions in organized ganglia by SAX-7, a Caenorhabditis elegans homologue of L1.
EMBO J.
24
,
1477
-1488.
Simmer, F., Tijsterman, M., Parrish, S., Koushika, S., Nonet,M., Fire, A., Ahringer, J. and Plasterk, R. (
2002
). Loss of the putative RNA-directed RNA polymerase RRF-3 makes C. elegans hypersensitive to RNAi.
Curr. Biol.
12
,
1317
.
Sulston, J. E., Schierenberg, E., White, J. G. and Thomson, J. N. (
1983
). The embryonic cell lineage of the nematode Caenorhabditis elegans.
Dev. Biol.
100
,
64
-119.
Thomas, J. H., Stern, M. J. and Horvitz, H. R.(
1990
). Cell interactions coordinate the development of the C. elegans egg-laying system.
Cell
62
,
1041
-1052.
Tucker, M., Sieber, M., Morphew, M. and Han, M.(
2005
). The Caenorhabditis elegans aristaless orthologue, alr-1,is required for maintaining the functional and structural integrity of the amphid sensory organs.
Mol. Biol. Cell
16
,
4695
-4704.
Wang, X., Kweon, J., Larson, S. and Chen, L.(
2005
). A role for the C. elegans L1CAM homologue lad-1/sax-7 in maintaining tissue attachment.
Dev. Biol.
284
,
273
-291.
White, J. G., Southgate, E., Thomson, J. N. and Brenner, S.(
1986
). The structure of the nervous system of the nematode Caenorhabditis elegans.
Philos. Trans. R. Soc. Lond. B Biol. Sci.
314
,
1
-340.
Zallen, J. A., Kirch, S. A. and Bargmann, C. I.(
1999
). Genes required for axon pathfinding and extension in the C. elegans nerve ring.
Development
126
,
3679
-3692.