Four specific immunoglobulin domains in UNC-52/Perlecan function with NID-1/Nidogen during dendrite morphogenesis in Caenorhabditis elegans

Neurons receive signals through often elaborately sculpted dendritic arbors, first documented by Santiago Ramón y Cajal in the late 19th and early 20th century (Ramón y Cajal, 1894). To understand dendrite arbor morphogenesis we study the stereotypically patterned somatosensory PVD neuron in Caenorhabditis elegans (Fig. 1A) (Smith et al., 2010; Albeg et al., 2011). Many mechanisms required for dendritic morphogenesis are conserved during PVD dendrite patterning, including the combinatorial use of different levels of transcription factors (Smith et al., 2013) and a role for molecular transport (Aguirre-Chen et al., 2011; Albeg et al., 2011; Taylor et al., 2015; Zou et al., 2015). Other factors or processes involved in dendritic morphogenesis include the proprotein convertase KPC-1/Furin (Schroeder et al., 2013; Salzberg et al., 2014; Dong et al., 2016), the EFF-1 fusogen (Oren-Suissa et al., 2010; Zhu et al., 2017), the claudin-like membrane protein HPO-30 (Smith et al., 2013) and the unfolded protein response (UPR) (Wei et al., 2015; Salzberg et al., 2017). The netrin axon guidance signaling cascade, which comprises the secreted UNC-6/Netrin cue and the UNC-40/DCC and UNC-5/hUNC5 receptors, plays an important role in self-avoidance of PVD dendrites (Smith et al., 2012).

In addition to the above, largely cell-autonomous cues, the epidermal MNR-1/Menorin and SAX-7/L1CAM cell adhesion molecules and the muscle-derived diffusible LECT-2/Chondromodulin II act in a complex through a leucine-rich repeat transmembrane receptor DMA-1/LRR-TM on PVD dendrites (Liu and Shen, 2011; Dong et al., 2013; Salzberg et al., 2013; Díaz-Balzac et al., 2016; Zou et al., 2016). More recently, it has been suggested that muscle sarcomeres pattern PVD quaternary branches by localizing the basement membrane protein UNC-52/Perlecan to these muscle structures, which in turn pattern the hemidesmosome-like fibrous organelles in the epidermis to localize the SAX-7 cell adhesion molecule (Liang et al., 2015). Whether UNC-52 serves functions in PVD patterning independently of shaping quaternary dendrites or of its structural functions to localize fibrous organelles remained to be determined.

Since PVD dendrites are shaped by extracellular factors originating from muscle and epidermis, we systematically tested genes encoding components of the fibrous organelles, which physically link muscles and epidermis (Hresko et al., 1994), for their involvement in PVD patterning (Fig. S1). We included genes that function in muscle [e.g. pat-2/α-integrin (Williams and Waterston, 1994)], the extracellular matrix [e.g. the heparan sulfate proteoglycan unc-52/perlecan (Rogalski et al., 1993, 1995)] as well as the epidermis [e.g. the intermediate filament protein mua-6/IFA-2 (Hapiak et al., 2003)] (Fig. S1). Compromising the function of genes encoding components of the fibrous organelles significantly reduced the number of secondary (2°), tertiary (3°) and quaternary (4°) dendritic branches (Fig. S1C-E). However, the proportions of the menorah-like PVD dendritic arbors remained unchanged as determined by the number of 4° branches per 2° branch (Fig. S1F). This conclusion is further supported by limited effects on the average length of branches but a reduction in the aggregate length of 2°, 3° and 4° dendritic branches under most conditions (Fig. S1G-L). Thus, components of fibrous organelles are important for controlling the number of menorah-like dendritic arbors per animal, without significantly disrupting their overall structure.

A mutation in unc-52, suggested to be important for patterning of 4° PVD dendrites (Liang et al., 2015), showed a previously unappreciated reduction in the number of 2° branches (Fig. 1A,B). UNC-52/Perlecan is a conserved multidomain protein of the extracellular matrix (Rogalski et al., 1993, 1995) (Fig. 1C). Domain I is Perlecan specific, whereas domains II, III, and V comprise LDL (low-density lipoprotein receptor) modules, laminin EGF (epidermal growth factor) repeats, and laminin G domains, respectively (Fig. 1C). Domain IV is conserved from cnidarians to humans (Warren et al., 2015) and defined by a string of immunoglobulin (Ig) domains. The unc-52 locus in C. elegans is alternatively spliced to produce short (domains I-III), medium (domains I-IV), and long (domains I-V) UNC-52 forms (Rogalski et al., 2001). The medium forms show additional alternative splicing of four Ig domains within domain IV (Fig. 1C), where the Ig domains are assembled in different combinations to produce distinct medium isoforms (Rogalski et al., 1995) (Fig.S2A).

Using the total number of 2° branches as a proxy for the number of menorahs, we found a reduction of 2° branches in several additional unc-52 alleles, except the gk3 deletion allele that deletes domain V (Fig. 1C,D, Fig. S2A). There are more PVD 2° branches located on the right and dorsal sides of the animals than on the left and ventral sides, respectively (Smith et al., 2010). These asymmetries persist in unc-52 mutants (Fig. S2B,C), suggesting that unc-52 is equally required for dorsal/ventral and left/right pattering of PVD 2° branches. The changes in the number of 2° branches in unc-52 mutants are developmental rather than maintenance defects because, in contrast to wild-type animals, the number of 2° branches did not increase in unc-52 mutants during development from the L3 to the L4 larval stage (Fig. S2D). We conclude that (1) unc-52 functions to establish the appropriate number of 2° branches during development; (2) four Ig domains in domain IV are important for this process; and (3) domain V and hence the long forms of UNC-52 are dispensable for this function. The latter conclusion is reminiscent of findings reported for the correct formation of 4° branches in PVD (Liang et al., 2015).

Null alleles of unc-52 display an embryonic lethal phenotype due to the failure of muscles to attach to the body wall at the twofold embryonic stage (Rogalski et al., 1993). We found that the unc-52(e998) allele behaved as a genetic null for the 2° branching phenotype of PVD dendrites because animals in which this allele was placed over a deficiency of the unc-52 region [unc-52(e998)/jDf2] did not display a more severe phenotype than unc-52(e998) homozygous animals (Fig. 1E). Moreover, transheterozygous animals between unc-52(e998) and unc-52(e669) or unc-52(ra515) also failed to display enhancement (Fig. 1E), suggesting that they are likely to act as genetic null alleles for the 2° PVD branching defect as well. Finally, morphometric analyses of unc-52(e998) and unc-52(ra515) revealed no statistical differences (Fig. 1F-I). We conclude that the PVD patterning defect is the null phenotype due to loss of specific medium UNC-52 forms.

Most unc-52 alleles show disorganized body wall muscle structure, resulting in progressive paralysis as the animals enter adulthood owing to compromised muscle attachment (Rogalski et al., 1995). Thus, the PVD dendrite defects in unc-52 mutant animals could be a secondary consequence of structural body wall muscle defects, as has been suggested for the defects in 4° PVD dendrites (Liang et al., 2015). Alternatively, the PVD defects in unc-52 mutant animals could be due to loss of specific UNC-52 splice variants. Intriguingly, unc-52(ra515) animals do not display locomotion or morphogenetic defects at any stage (Mullen et al., 1999; Merz et al., 2003), but do show defects in PVD dendritic branch morphology similar to those seen in other unc-52 alleles (Fig. 1F-I).

To determine whether PVD patterning defects correlate with structural defects in body wall muscle, we analyzed body wall muscle morphology in different unc-52 alleles. Using polarizing light microscopy, which identifies the A and I bands in muscle (Waterston et al., 1980), and phalloidin staining, which labels the F-actin-containing thin filaments, we detected the well-documented defects in sarcomere organization in unc-52(e998) mutants. By contrast, we never observed similar defects in unc-52(ra515) animals or wild-type controls (Fig. 2A-C). We also investigated patterning of fibrous organelles with antibodies against PAT-3/β-integrin (a muscle component of fibrous organelles), UNC-52/Perlecan (an ECM component of fibrous organelles), myotactin (an epidermal component of fibrous organelles) and a MUA-6::RFP reporter that labels the intermediate filaments in the epidermis (Hapiak et al., 2003) (Fig. 2A). As previously reported (Mullen et al., 1999), we observed severe disorganization of the body wall muscle in the uncoordinated unc-52(e998) allele when we stained with antibodies against β-integrin or Perlecan (Fig. 2D). However, we did not observe structural irregularities on the epidermal side of the fibrous organelles in unc-52(e998) animals, as myotactin staining and the MUA-6::RFP reporter appeared indistinguishable from wild type (Fig. 2D). This suggests that the epidermal part of the fibrous organelles remains unperturbed in these unc-52 mutant alleles. No comparable patterning defects of any type were observed in unc-52(ra515) or wild-type animals (Fig. 2D). Therefore, the reduction in the number of 2° branches in PVD dendrites in unc-52 mutants is unlikely to be secondary to structural muscle defects but instead the result of specific molecular interactions that involve the four Ig domains (#7-10) of UNC-52.

UNC-52 is expressed in muscle (Rogalski et al., 1993) but also serves essential functions in the epidermis during early development (Spike et al., 2002). Using cell-specific rescue experiments, we found that expression of an unc-52 cDNA from muscle but not epidermis consistently rescued the unc-52 mutant phenotype (Fig. S2E-I). Thus, UNC-52 may function from muscle to pattern PVD dendrites or, alternatively, UNC-52 could undergo post-translational modifications specific to body wall muscle, such as the well-known modifications of UNC-52/Perlecan by heparan sulfates, a type of extracellular glycosaminoglycan. In support of this idea, we found that mutating a crucial enzyme in heparan sulfate biosynthesis, the N-deacetylase-N-sulfotransferase hst-1/NDST, resulted in a PVD mutant phenotype indistinguishable from that of unc-52(ra515) (Fig. S2J).

Nidogen is a conserved multidomain ECM protein comprising three globular domains separated by a rod-like domain (Fig. 3A). Mouse nidogen and perlecan display biochemical interactions (Hopf et al., 1999, 2001). Thus, we examined whether the C. elegans nidogen ortholog NID-1, which serves important roles in neuronal (Kim and Wadsworth, 2000) and synaptic patterning (Ackley et al., 2003) in C. elegans, is also important for patterning of PVD dendrites. We found that the PVD phenotype of a nid-1 null mutant was (1) not statistically different from the unc-52 mutant phenotype and (2) not enhanced in a double mutant (Fig. 3B-D), suggesting that both genes act in the same genetic pathway.

To determine whether UNC-52 is required for the localization of NID-1, we stained unc-52(ra515) mutant animals, which lack Ig domains #7-10, with antisera against NID-1. We found that the localization of NID-1 was severely affected and failed to consistently colocalize with the M lines and dense bodies of the body wall muscle as visualized with an antibody against β-integrin (Fig. 3E). Conversely, the localization of UNC-52 or PAT-3/β-integrin was unaffected in the nid-1 null mutant (Fig. S3A,B). NID-1 localization also appeared completely normal in unc-52(gk3) animals (which lack domain V) (Fig. S3C), consistent with the absence of defects in PVD patterning in this allele (Fig. 1D). These results suggest that Ig domains #7-10 in domain IV, but not domain V, of UNC-52 are required for NID-1 localization. Surprisingly, NID-1 localization appeared normal in the nid-1(cg118) allele, which is predicted to encode a NID-1 protein with an in-frame deletion of the G2 globular domain (Fig. S3C) (Kang and Kramer, 2000). Thus, the G2 domain, which can bind mouse perlecan in vitro (Hopf et al., 1999, 2001), is individually not required for NID-1 and UNC-52 to interact in vivo.

The secreted UNC-6/Netrin ligand functions through the two receptors UNC-40/DCC and UNC-5/hUNC5 alone or in combination to control axonal guidance in both invertebrates and vertebrates (Bradford et al., 2009). Mutations in unc-6, unc-40 and unc-5 result in fewer 2° branches in PVD dendrites (Smith et al., 2012). We confirmed these results (Fig. 4A) and also found that the PVD 2° branching defects could be rescued by transgenic UNC-40 expression in PVD but not in motor neurons, where unc-40 functions to pattern commissures (Chan et al., 1996) (Fig. 4B). Thus, UNC-40 is likely to function cell-autonomously to regulate the number of 2° PVD branches, and the observed defects are not an indirect consequence of motor neuron commissure defects of unc-40 mutants. The phenotypes of unc-52; unc-6 and unc-40; unc-52 double mutants were not more severe than the single mutants, suggesting that unc-52 acts genetically in the netrin pathway (Fig. 4A). Moreover, nid-1 behaved genetically like an unc-52 mutant because it also failed to enhance a mutation in unc-40 (Fig. 4A). We conclude that unc-52 and nid-1 act in a common pathway with netrin signaling. These findings are consistent with studies showing that the netrin receptor unc-40 functions together with nid-1 to guide axons (Kim and Wadsworth, 2000), and a report that unc-52 is required for the unc-6-dependent subcellular localization of an UNC-40::GFP reporter in HSN neurons (Tang and Wadsworth, 2014).

We show that four specific Ig domains of the conserved basement membrane protein UNC-52/Perlecan function to localize the basement membrane protein NID-1/Nidogen and correctly pattern PVD 2° dendrites. The four Ig domains are part of domain IV, a domain common to all known perlecan proteins, which is composed of a string of Ig domains that are alternatively spliced in C. elegans and mice (Iozzo et al., 1994). Both unc-52 and nid-1 act genetically in the unc-6 pathway. Surprisingly, UNC-52 acts from muscle to pattern PVD dendrites. NID-1 has been shown to be concentrated at the edges of the muscle quadrants (Kang and Kramer, 2000), and we observe enrichment for some fibrous organelle components (e.g. MUA-6::RFP) and, possibly, UNC-52 along the edges of the muscle (Fig. 2D). Thus, molecular interactions between specific UNC-52 splice variants and NID-1 could shape the basement membrane lattice as part of specialized fibrous organelles along the edges of the muscle quadrants. Alternatively, muscle-derived UNC-52 could be incorporated into the epidermal basement membrane at a distance from its place of biosynthesis and modification in the muscle, similar to what has been shown for laminin α subunits (Huang et al., 2003).

How could two basement membrane proteins in conjunction with netrin signaling control the number of 2° branches? The Perlecan-Nidogen complex and netrin signaling normally promote the formation of 2° branches. Nidogen links collagen and laminin complexes (Mak and Mei, 2017) and thus may affect the distribution, availability or presentation of the secreted laminin-related UNC-6/Netrin morphogen. For example, UNC-6 could normally be incorporated in the basement membrane lattice in a manner that inhibits/shields its C-terminal domain, which is known to inhibit branching, likely in cis (Lim and Wadsworth, 2002; Wang and Wadsworth, 2002) (Fig. 4C). Compromising the specific UNC-52–NID-1 complex could release or change the presentation of UNC-6 and allow the C-terminal domain to exert its inhibitory function, resulting in fewer branches (Fig. 4C). Further work will be required to resolve this question. Altogether, our work provides insight into the function of the alternatively spliced Ig domains of UNC-52/Perlecan in localizing NID-1/Nidogen and, possibly, specific basement membrane signaling complexes during dendrite patterning.

Worms were cultured as previously described (Brenner, 1974). N2 (Bristol) was used as the control strain. Alleles used include e998, e1421, e699, ra515 and gk3 for unc-52, nid-1(cg119), nid-1(cg118), unc-6(ev400) and unc-40(e271). RNAi-mediated gene knockdown experiments were performed as described (Timmons and Fire, 1998). Transgenic animals were created using standard protocols (Mello et al., 1991). For additional information, including the primers used to construct the full-length unc-52A cDNA, see the supplementary Materials and Methods and Table S1.

Whole-mount freeze-fracture antibody staining was performed using the protocol described by Duerr (2006). Briefly, animals were fixed using methanol and acetone and immunodetection was accomplished using primary antibodies MH2, MH3 (both UNC-52/Perlecan), MH25 (PAT-3/β-integrin) and MH46 (myotactin) (Francis and Waterston, 1991) at 1:100 dilution. Secondary antibody Alexa Fluor 555 donkey anti-mouse (Thermo Fisher Scientific, A-31570) was used at 1:500.

L4/young adult animals were placed in 1 mM Levamisole and mounted on a 5% agarose pad. Animals were imaged using a Zeiss Axio imager Z.1 compound microscope under control of Zeiss Axiovision software. Maximum intensity projections were generated of each animal at 40× magnification with an exposure time of 120 ms and a slice separation of 0.6 µm. Images were exported in a 16-bit tiff format, converted into 8-bit images in ImageJ (NIH), and subjected to morphometric analysis using the NeuronJ plugin (ImageJ) as described (Salzberg et al., 2013). Statistical analysis was performed using one-way ANOVA with Tukey correction for multiple comparisons (GraphPad, Prism v7).

We thank J. Plenefisch and members of the H.E.B. lab for comments on the manuscript and helpful discussions; the Caenorhabditis Genetics Center (which is funded by the NIH, P40 OD010440), J. Plenefisch for the MUA-6::RFP reporter, and D. Miller III for the unc-40 rescuing strains.