Convertase-dependent regulation of membrane-tethered and secreted ligands tunes dendrite adhesion

Animal behavior relies on functional neuronal networks, which in turn rely on sensation of sensory stimuli and their coordinated processing. Defects in sensory processing of both auditory and mechanosensory modalities have been implicated with neuropsychiatric disease (Antoine et al., 2013; Orefice et al., 2019). The single building block of neural networks are neurons, which receive input through dendrites and pass on information through axonal projections. Crucial functions during neural development are played by cell adhesion molecules that provide adhesive forces to mediate controlled and directed growth of developing neurites. Much is known about the interactions between cell adhesion molecules of adjacent cells as well as their adhesion-independent functions (Moreland and Poulain, 2022), but how adhesive interactions are negatively regulated is a much less understood, yet equally important aspect to guarantee coordinated neuronal development.

Conserved mechanisms coordinate dendrite formation in neurons of both vertebrates and invertebrates (Dong et al., 2015; Lefebvre et al., 2015; Lefebvre, 2021). For example, the pair of multidendritic PVD somatosensory neurons in the nematode Caenorhabditis elegans are a well-established model for studying dendrite development (reviewed by Inberg et al., 2019; Sundararajan et al., 2019). These neurons each cover one side of the body surface of the nematode, with the exception of the head and neck region, which is covered by a structurally related somatosensory neuron called FLP. In addition to a single axon that both neurons send into the right ventral nerve cord, a stereotypic dendritic tree develops with primary, secondary, tertiary and quaternary dendrites branching at characteristic right angles (Fig. 1A) (Oren-Suissa et al., 2010; Smith et al., 2010; Albeg et al., 2011). After initial outgrowth of the primary dendrite along the lateral nerve tract, orthogonal secondary dendrites branch off in a ventral and dorsal direction. When the secondary branches reach the sublateral line where the lateral epidermis abuts the muscle quadrants, tertiary branches form orthogonally to the secondary branch and in parallel to the primary branch. Eventually, orthogonal quaternary branches emanate, sandwiched between the epidermis and the muscle cells towards the dorsal and ventral midlines without ever crossing them (Fig. 1A). Owing to their similarity to candelabras, the individual units of the dendritic trees have been termed menorahs (Oren-Suissa et al., 2010). The formation of these menorah-like dendritic trees requires the Menorin adhesion complex, which is formed by the skin-derived cell adhesion molecules SAX-7 (also known as L1CAM) and MNR-1 (or Menorin) (Dong et al., 2013; Salzberg et al., 2013). SAX-7 is produced by the epidermis and localized along the sublateral line as well as in a striped pattern in the epidermis along the sarcomers of the muscle, essentially prepatterning the path for growing tertiary and quaternary dendrites (Dong et al., 2013; Salzberg et al., 2013; Liang et al., 2015). Together with the muscle-derived, diffusible LECT-2 (also known as Chondromodulin II) chemokine (Diaz-Balzac et al., 2016; Zou et al., 2016), these three factors are believed to form an adhesion complex with the DMA-1 leucine-rich transmembrane receptor on PVD neurons to establish the elaborate dendritic trees (Fig. 1A) (reviewed by Inberg et al., 2019; Sundararajan et al., 2019).

Proprotein convertases are a class of proteases that cleave proteins at characteristic dibasic motifs (reviewed by Artenstein and Opal, 2011; Seidah, 2011; Seidah and Prat, 2012). They are synthesized as proproteins and undergo autocatalytic cleavage and activation in appropriate cellular compartments. For example, the furin proprotein convertase has been shown to process insulin-like peptides in mammals (Smeekens et al., 1992). The closest homolog to furin in C. elegans is the proprotein convertase KPC-1 (Thacker and Rose, 2000), which also has been shown to process insulin-like peptides in nematodes (Hung et al., 2014), suggesting a conservation of function. Furin has been implicated in dendrite development of mice, in which it is believed to facilitate the cell-autonomous maturation of pro-BDNF to BDNF (Zhu et al., 2017). In C. elegans, KPC-1 affects dendrite development of multidendritic neurons, including that of PVD (Schroeder et al., 2013), and functions genetically as a negative regulator of the Menorin pathway (Salzberg et al., 2014). Subsequent studies suggested that KPC-1 antagonizes the function of the Menorin complex by decreasing DMA-1 receptor levels on the PVD membrane of kpc-1 mutant animals (Dong et al., 2016). The observed overabundance of the DMA-1 receptor on the PVD membrane in kpc-1 mutants was suggested to be the result of defects in KPC-1-dependent endocytosis (Dong et al., 2016). The endocytic function of KPC-1 was shown to require self-activation through autocatalytic processing (Dong et al., 2016). However, it remained unresolved (1) whether proteolytic activity of KPC-1 is required beyond autocatalytic activation and (2) how KPC-1 regulates DMA-1 function or localization.

Using a combination of genetic and biochemical approaches, we discovered that the kpc-1-dependent recruitment/retention of DMA-1 on the PVD membrane is dependent on the putative cell adhesion molecule MNR-1. Intriguingly, KPC-1 from PVD dendrites negatively regulates MNR-1 function in the epidermis, which most likely dissociates the Menorin ligand-receptor complex. We propose that the resulting reduced adhesiveness of PVD dendrites and recycling of the DMA-1 receptor facilitates neurite outgrowth. Our findings provide the first instance of a proprotein convertase functioning across tissues to regulate cell-cell interactions during neural development.

Complete loss of KPC-1 function causes defects in dendrite branch extension, branch number and self-avoidance of tertiary dendrites in PVD neurons, likely owing to increased adhesion of the Menorin complex (Schroeder et al., 2013; Salzberg et al., 2014; Dong et al., 2016). To identify additional genes that function with kpc-1 during dendrite morphogenesis, we performed a genetic modifier screen of a kpc-1 partial-loss-of-function allele (gk333538, hereafter named R265Q). This allele changes a conserved arginine residue in close proximity to the catalytic histidine 262, and results in a significant increase in the number of secondary branches and defects in self-avoidance of tertiary branches (Fig. 1B, Fig. 2) (Salzberg et al., 2014). Of ten enhancer and suppressor mutants isolated in this screen, one allele suppressed, whereas nine alleles enhanced the kpc-1 loss-of-function phenotype (Table 1). We also identified 16 modifier mutations, which showed distinct phenotypes and have been (Tang et al., 2019; Tang et al. 2021) or will be described in the future. Of the enhancer mutations, eight alleles were either intragenic enhancers of kpc-1 or closely linked, and resulted in kpc-1-null-mutant-like phenotypes in PVD dendrites in combination with the kpc-1(R265Q) partial loss-of-function allele, suggesting that they likely affected the kpc-1 locus (Table 1). The remaining two alleles likely represented extragenic modifiers (one enhancer and one suppressor) of kpc-1(R265Q). Here, we focus on the dz213 allele, which completely suppresses all kpc-1(R265Q) mutant phenotypes.

Using whole-genome sequencing, mapping and transformation rescue, we identified the molecular lesion of the dz213 suppressor allele as a missense mutation in mnr-1 (Table 1, Fig. 2A; Fig. S1A,B, Fig. S2A,B). This missense mutation changes leucine 135 to a phenylalanine, a residue within the Menorin domain (previously DUF2181), which is conserved from choanoflagellates to humans (Fig. 2A). The L135F change in MNR-1 suppressed both the branching and self-avoidance defects of PVD in kpc-1(R265Q) mutant animals (Fig. 2B-D,F). Several observations indicate that mnr-1(dz213) [hereafter referred to as mnr-1(L135F)] is a partial loss of function allele. First, mnr-1(L135F) failed to complement the putative null allele mnr-1(dz175), both with regard to the number of branches and the dendrite self-avoidance defects (Fig. 2E-G). Second, two copies of the mutant allele (L135F) were required for suppression of the kpc-1(R265Q) PVD phenotype, showing that mnr-1(L135F) is recessive (Fig. 2D,E). Third, transgenic expression of a wild-type mnr-1 cDNA in the epidermis rescued the mnr-1(L135F) mutant, i.e. it reversed the suppression of the kpc-1(R265Q); mnr-1(L135F) double mutant, and overexpression of a mnr-1(L135F) mutant cDNA partially rescued mnr-1(dz175) null mutant animals (Fig. S2A,B). Some MNR-1 function is required for efficient suppression of partial loss of kpc-1 function because (1) the suppression is dosage sensitive (Fig. 2G); (2) the mnr-1(dz175) putative null allele could not suppress the kpc-1(R265Q) hypomorphic phenotype (Fig. S2C); and (3) overexpression of the mnr-1(L135F) mutant cDNA driven by an epidermal promoter completely rescued the mnr-1 null phenotype in kpc-1(R265Q); mnr-1(dz175) double-mutant animals (data not shown). Finally, the mnr-1(L135F) allele is temperature sensitive because mnr-1(L135F) mutant animals displayed PVD phenotypes at the non-permissive temperature of 25°C that were indistinguishable from the mnr-1 null phenotype, but no phenotype at all at the permissive temperature of 20°C (Fig. S2D). All suppression conferred by the mnr-1(L135F) allele in kpc-1(R265Q); mnr-1(L135F) double mutants was lost at the non-permissive temperature compared with at the permissive temperature (Fig. S2D), also supporting the notion that (1) mnr-1(L135F) is dosage sensitive and (2) that some MNR-1 function is required to suppress a partial loss of KPC-1 function. Thus, we conclude that mnr-1(L135F) is a recessive, partial-loss-of-function allele of mnr-1, which can suppress the defects of a partial-loss-of-function allele of kpc-1. Importantly, only reducing but not eliminating mnr-1 function can compensate for reduced kpc-1 function.

We next asked whether, conversely, suppression of the kpc-1 partial-loss-of-function phenotype in the mnr-1(L135F) allele is dependent on some kpc-1 function. Existing kpc-1 alleles form an allelic series as follows: gk8=gk779937>dz185>gk333538>dz254, where gk8 represents the putative null phenotype with the strongest phenotype and dz254 represents the weakest phenotype (Salzberg et al., 2014; Rahman et al., 2022) (Fig. S2C,E-G). The strong-loss-of-function-alleles display entrapment of the dendrites in the lateral epidermis, whereas hypomorphic alleles display primarily self-avoidance defects (Salzberg et al., 2014; Dong et al., 2016). We found that the mnr-1(L135F) allele completely suppressed the weaker kpc-1 alleles (dz254 and gk333538), but only partially suppressed the intermediate kpc-1(dz185) allele (Fig. S2C). The putative kpc-1(gk8) null allele (catalytic domain deletion) and the kpc-1(gk779937) (G263R) strong-loss-of-function allele were not suppressed by the mnr-1(L135F) allele (Fig. S2C). In conclusion, the suppression of kpc-1 by loss of MNR-1 function is (1) sensitive to both the dosage of MNR-1 and KPC-1, and (2) requires residual function of both genes.

How could loss of MNR-1 function alleviate defects due to reduced KPC-1 function? It has previously been suggested that KPC-1 can negatively regulate extracellular adhesion of dendritic growth cones to the growth substrate (the epidermis) by promoting endocytosis of the DMA-1 receptor (Dong et al., 2016). Our genetic results raised the possibility that increasing the expression of mnr-1 could have similar effects: adhesive forces exerted by overactivation of the Menorin complex from the epidermis could prevent dendrite extension. To test this hypothesis, we first expressed a fosmid containing mnr-1 under control of most, if not all, regulatory regions at low doses (1 ng/µl) in mnr-1(dz175) null mutants and found that it, as expected, completely rescued the PVD defects of mnr-1 mutant animals (Fig. 3A,D). Interestingly, the same mnr-1-expressing transgene in a wild-type background resulted in self-avoidance defects and a reduced number of fully extended 4° branches, reminiscent of kpc-1 partial-loss-of-function alleles (Fig. 3B,E). Moreover, transgenic overexpression of mnr-1 at high doses (20 ng/µl) resulted in a dramatic transformation of the PVD dendritic tree from a mnr-1-like phenotype to a kpc-1-like mutant phenotype (Fig. 3C). Specifically, we observed a strong increase in the number of short, immature secondary branches and self-avoidance defects that were indistinguishable from kpc-1 null mutants (Fig. 3C,D,F). Note that when kpc-1 function was completely lost [kpc-1(gk8)], dendrites became essentially trapped in the vicinity of the primary dendrite and self-avoidance defects could not be detected (Fig. 3C). Similar results were obtained when mnr-1 was expressed under control of an epidermis-specific promoter (Fig. 3C). Together with the loss-of-function studies, these results suggest a tightly regulated antagonistic balance between the functions of MNR-1 and KPC-1 (Fig. 3G).

A hallmark of the kpc-1 loss-of-function phenotype in PVD is the increased level of DMA-1 on the membrane. This has been suggested to result from diminished endocytosis of the DMA-1 receptor, based on colocalization in endocytic vesicles and co-immunoprecipitation experiments indicating that both KPC-1 and DMA-1 are part of a biochemical complex (Dong et al., 2016). DMA-1 accumulation on the neuronal surface is thought to result in increased adhesive forces, leading to dendrite trapping and reduced branch extension. An alternative scenario arose from our transgenic experiments with mnr-1: the similarity of kpc-1 loss-of-function and mnr-1 overexpression phenotypes raised the possibility that an excess of MNR-1 from the epidermis could retain DMA-1 on the plasma membrane of PVD dendrites. Interestingly, we previously found that the mnr-1; kpc-1 double-null mutant more closely resembled the mnr-1 null mutant than the kpc-1 mutant (Salzberg et al., 2014), suggesting that mnr-1 is epistatic or, in other words, that MNR-1 is required for the kpc-1 mutant phenotype. We therefore analyzed the localization of a DMA-1::GFP reporter in single and double mutants of mnr-1 and kpc-1. As previously reported (Dong et al., 2016), we detected a significant increase in diffuse DMA-1::GFP reporter levels in PVD in the absence of KPC-1 (Fig. 4A,B). Diffuse staining of the DMA-1::GFP reporter is considered to represent the cell membrane-bound fraction of the receptor (Taylor et al., 2015; Zou et al., 2015; Dong et al., 2016). Interestingly, the levels of the DMA-1::GFP reporter in kpc-1 mutants were reduced to wild-type levels in the absence of MNR-1 (Fig. 4A,B), demonstrating that MNR-1 is necessary for increased membrane localization of the DMA-1::GFP reporter in the absence of the KPC-1 proprotein convertase.

We next asked whether MNR-1 was sufficient to increase DMA-1 levels in the PVD membrane. We used mnr-1(dz175) null mutant animals carrying (1) a transgene that ectopically expresses MNR-1 in muscle cells, (2) a transgene that expresses a DMA-1::GFP fusion in PVD and (3) a transgene that expresses myristoylated mCherry in PVD. We calculated the GFP (DMA-1) to mCherry (membrane) fluorescence ratio in tertiary branches, normalized to that in secondary branches, and compared it with the ratio in a control strain expressing only the DMA-1::GFP fusion protein and myristoylated mCherry in PVD (Fig. 4C-F; see Materials and Methods for details on quantifications). Exclusive expression of MNR-1 in the muscle results in higher-order branches reminiscent of baobab trees, as opposed to the stereotyped candelabra-like structures (Salzberg et al., 2013). We found that the muscle-derived, overexpressed MNR-1 resulted in much higher accumulation of the DMA-1::GFP reporter in tertiary branches of baobab-like dendrites compared with DMA-1::GFP accumulation in tertiary branches of wild-type animals, suggesting that MNR-1 is sufficient to recruit/retain the DMA-1::GFP reporter on the PVD membrane surface in trans (Fig. 4C-F; Fig. S3C,D). Collectively, these findings show that MNR-1 is necessary and sufficient to non-cell-autonomously control DMA-1 surface levels on PVD somatosensory dendrites.

Because MNR-1 is required for the kpc-1-mediated increase of DMA-1::GFP on the cell surface, we investigated the localization of a KPC-1 reporter in PVD. We found that a functional KPC-1::sfGFP reporter specifically expressed in PVD localized to higher-order dendritic branches of PVD neurons as well as to the PVD axonal compartment (Fig. 5A, Fig. S4A). We further found that the putative transmembrane domain of KPC-1 (Fig. S5) was necessary to direct PVD dendritic patterning because a construct lacking the predicted C-terminal transmembrane domain (S675 to S688) failed to rescue the kpc-1 mutant phenotype, whereas a construct containing a heterologous transmembrane domain of PAT-3/β-integrin (PAT-3-TM) rescued the defects (Fig. 5B-D).

We next asked whether a kpc-1 cDNA expressed in PVD neurons could rescue FLP dendrite morphogenesis defects in kpc-1 mutant animals or, conversely, whether expression in FLP neurons rescued PVD dendrite morphogenesis defects. In both experiments, kpc-1 functioned strictly cell-autonomously, i.e. it rescued the kpc-1 mutant defects only in the neurons where it was expressed (Fig. S6). We conclude that, in vivo, KPC-1 is localized to a membrane compartment in higher-order dendrite branches and is therefore potentially in proximity to both MNR-1 expressed by epidermal cells and the DMA-1 receptor in PVD. We further suggest that owing to the strictly cell-autonomous function of kpc-1, a putative proteolytic target may not be diffusible.

To better understand the proteolytic function of KPC-1 in PVD dendrite patterning, we conducted transgenic rescue experiments. We found that, consistent with previous studies, KPC-1 functions cell-autonomously in PVD neurons (Salzberg et al., 2014) and requires catalytic activity to pattern somatosensory dendrites (Fig. 6A,B) (Dong et al., 2016). Previous studies had further shown that expression of a mature, processed form of KPC-1 lacking the prodomain could effectively rescue kpc-1 mutant phenotypes, suggesting that the catalytic activity of KPC-1 is required to autoactivate KPC-1 by cleavage of the prodomain (Dong et al., 2016). However, it remained unresolved whether the catalytic activity of KPC-1 is necessary beyond autocatalytic cleavage and activation. We confirmed that transgenic expression of the mature form of the KPC-1 (lacking the prodomain) in PVD rescued the kpc-1 mutant phenotype in PVD (Fig. 6C,D) (Dong et al., 2016). However, a mature, processed form of KPC-1 carrying mutations in either of two conserved residues essential for convertase function (H262A or N363A in KPC-1) (Creemers et al., 1993) failed to rescue the kpc-1 mutant phenotypes (Fig. 6C,D). This failure to rescue was not the result of obvious defects in protein stability or trafficking of mutant KPC-1, because a reporter fusion of mutant KPC-1 with a fluorescent protein [KPC-1(H262A)::sfGFP] expressed in PVD displayed protein expression levels and localization in PVD dendrites comparable with those of the analogous wild-type KPC-1::sfGFP reporter (Fig. S4C). Taken together, our experiments suggest that the convertase activity of KPC-1 is required beyond autocatalytic cleavage and imply the existence of additional substrate(s) for KPC-1 (other than itself) during PVD dendrite morphogenesis.

Finally, we asked whether KPC-1 is required to regulate protein levels of the Menorin adhesion complex, including the putative cell adhesion molecules SAX-7, DMA-1 and MNR-1, and the secreted chemokine LECT-2. To this end, we used animals carrying transgenes or encoding genome-engineered versions of these proteins with immunotags. We found no obvious change in the amounts or banding patterns on western blots for SAX-7 in kpc-1 null mutants compared with those for control animals (Fig. S7A), in spite of a conserved putative furin cleavage site in the fibronectin domain 3, which is important for SAX-7 function during PVD development (Dong et al., 2013; Salzberg et al., 2013). Similarly, we found no obvious changes in banding patterns for DMA-1, consistent with a prior report for DMA-1 (Dong et al., 2016) (Fig. S7B). In contrast, we found that the absolute amount of endogenously tagged LECT-2 was decreased, whereas the amount of MNR-1 was increased in kpc-1 mutants compared with those in control animals (Fig. 7A-C). Although we could not detect obvious changes in banding patterns of MNR-1, our resolution may be insufficient to detect a possible change of 5 kDa resulting from cleavage from a predicted membrane-proximal furin cleavage site (Fig. 7A-C). Therefore, kpc-1 does not obviously affect SAX-7 or the DMA-1 receptor but is involved in the regulation of the overall levels of MNR-1 and LECT-2 either directly or indirectly.

To further test how MNR-1 might be regulated by KPC-1, we devised a different assay. Previous studies showed that an N-terminally tagged version of MNR-1 (mCherry::MNR-1) accumulates in coelomocytes, a set of scavenger cells that uptake diffusible proteins from the pseudocoelom, raising the possibility that MNR-1 is shed from cell membranes or secreted (Salzberg et al., 2013). We therefore established a functional MNR-1 fusion with N-terminally fused tagRFP and C-terminally fused mNeongreen (Fig. S7C) and quantified accumulation of N-terminally tagged MNR-1 in coelomocytes. We discovered that the endocytic compartment of coelomocytes showed consistently higher red fluorescence from tagRFP (i.e. more N-terminally tagged MNR-1) in wild-type animals compared with that in kpc-1 mutant animals (Fig. S7D,E). These observations were not the result of defective endocytosis in coelomocytes, because accumulation of mCherry in the coelomocytes of kpc-1 mutant animals of a control strain was indistinguishable from that of wild-type animals (Fig. S6C). In contrast, we found no significant differences in the levels of the C-terminal mNeongreen fragment in the epidermis between control and kpc-1 null mutant animals (Fig. S7D-G). We next measured the colocalization between the red and green fluorescence signals, i.e. between the N-terminal and the C-terminal MNR-1 fragments. To our surprise, the red (N-terminal) and green (C-terminal) fluorescent MNR-1 fragments in the epidermis almost perfectly colocalized in both wild-type and kpc-1 null mutants (Fig. S7G). Lastly, biochemical experiments with western blotting using antibodies against the N-terminal tagRFP or the C-terminal mNeongreen revealed no obvious evidence for cleavage of the dually tagged MNR-1 transgene (not shown). We do not know the reason for this apparent discrepancy, but one possible explanation is that the amount of cleaved MNR-1 is too small to be detectable in the background of the overexpressed tagRFP::MNR-1::mNeongreen transgene. Alternatively, MNR-1 may not be a direct target of KPC-1. Regardless, taken together, our results show that the KPC-1 proprotein convertase directly or indirectly regulates the amounts of MNR-1 and LECT-2. Of note, based on the strictly cell-autonomous function of KPC-1 in PVD patterning (Salzberg et al., 2014; Dong et al., 2016) (this study), these observations suggest that KPC-1 from PVD dendrites regulates the non-cell-autonomously acting factors MNR-1 and LECT-2 from the epidermis and muscle, respectively, in a negative feedback loop of the Menorin pathway. Additional experiments will be required to determine the precise mechanisms of this regulation across cells.

In this study, we establish a mutually antagonistic relationship between the KPC-1 proprotein convertase and the putative MNR-1 cell adhesion molecule. Genetically, phenotypes owing to partial- but not complete-loss-of-function mutations in kpc-1 could be suppressed by partial but not complete loss of mnr-1. We further show that the previously known regulation of DMA-1 receptor localization in PVD by KPC-1 is dependent on MNR-1 in the epidermis. Additionally, we establish that enzymatic activity of KPC-1 is required beyond self-activation, implying the existence of a substrate for KPC-1 other than itself. Interestingly, we found elevated levels of MNR-1 as well as changed localization in kpc-1 mutant animals, suggesting that the mutually antagonistic relationship between KPC-1 and MNR-1 is also true at the protein level. In addition, we found decreased levels of the muscle-derived factor LECT-2 in kpc-1 mutants. Taken together, our findings suggest that KPC-1 functions autonomously in PVD neurons to regulate MNR-1 in the epidermis, which in turn regulates surface localization of the DMA-1 receptor on PVD. We propose that KPC-1 serves in a negative feedback loop as a functional off-switch for signaling by the Menorin adhesion complex during development.

It had previously been suggested that KPC-1 regulates PVD dendrite development through direct regulation of DMA-1 receptor accumulation on the membrane (Dong et al., 2016). It was further argued that the proteolytic activity of KPC-1 was required for self-activation, an event thought to be crucial for its function in endocytosis of DMA-1 (Dong et al., 2016). The endocytosis hypothesis for KPC-1 appeared to be supported by precedent of the mammalian convertase PCSK9 (pro-protein convertase subtilisin kexin type 9) in the clearance of low-density lipoprotein receptors from the cell surface (Cameron et al., 2006; Lagace et al., 2006). However, several arguments suggest that KPC-1 is functionally more similar to catalytically active furin than to the largely catalytically inactive PCSK9 convertase. First, KPC-1 has been shown to process insulin-like peptides in C. elegans (Hung et al., 2014), a function that is also performed by furin in mammals (Smeekens et al., 1992). Second, several structural aspects are also distinct between PCSK9 and KPC-1: (1) in contrast to PCSK9, the mature KPC-1 convertase neither requires nor associates with its prodomain (Dong et al., 2016) (Fig. S6); (2) KPC-1 contains a P domain, which, based on the kpc-1(dz254) allele we isolated, is functionally important, and likely requires calcium for its catalytic activity (Thacker and Rose, 2000); and (3) we demonstrate that KPC-1 requires a transmembrane domain for function (Fig. 5), a finding reminiscent of membrane-bound furin (Henrich et al., 2003) but distinct from the secreted PCSK9 (Piper et al., 2007). Taken together with our findings that KPC-1 requires the presence of the putative cell adhesion molecule MNR-1 to regulate DMA-1 localization, these observations argue against a simple role in endocytosis alone for KPC-1.

Our results raise the possibility that MNR-1 is shed from the epidermal membrane, whereas other factors of the MNR-1 pathway for PVD dendrite patterning, such as the epidermal cell adhesion molecule SAX-7 or the DMA-1 receptor itself, appear to be unaffected by loss of KPC-1 (Fig. S7). Unexpectedly, we also observed a decrease of LECT-2 in kpc-1 mutants (Fig. 7B). A possible explanation is that an excess of available MNR-1 obscures putative binding sites for LECT-2 in SAX-7, thereby resulting in less binding opportunity and consequently degradation of LECT-2. Additional biochemical and genetic experiments will be required to resolve this question. Regardless, the catalytic activity of KPC-1 is required to regulate directly or indirectly the localization and the amount of the MNR-1 ligand in trans. It is therefore tempting to speculate that retention of DMA-1 by epidermal MNR-1 prevents endocytosis of the DMA-1 receptor in PVD and that KPC-1 negatively modulates this MNR-1 function, e.g. by promoting disassembly of the Menorin complex and the associated interactions. This KPC-1 function may be facilitated by DMA-1-dependent recruitment of the KPC-1 proprotein convertase, as shown previously (Dong et al., 2016). Adhesive signaling complexes face the conceptual challenge that their adhesive properties must be tightly regulated to allow coordinated growth during neural development (reviewed by Moreland and Poulain, 2022). Our genetic experiments reveal a tight antagonistic relation between MNR-1 and KPC-1. The phenotypes observed represent extremes of a continuum (Fig. 3): on one end of the spectrum, the mnr-1 null mutant phenotype represents strongly reduced adhesiveness of PVD branches, whereas on the other end, the kpc-1 null mutant phenotype represents excessive MNR-1 complex interactions and adhesiveness. Consistent with this interpretation, we find that both reducing and increasing the function of the Menorin pathway leads to defects in dendrite patterning. Therefore, fine-tuned control of the Menorin pathway is necessary to allow both extension and consolidation of developing dendrites. We propose that KPC-1 is a switch that locally controls adhesiveness of the Menorin pathway cell autonomously in PVD dendrites, by regulating the availability and or levels of the MNR-1 ligand in the epidermis in trans. Further experiments will be required to determine the precise mechanism (or substrate) by which KPC-1 convertase activity regulates dendrite morphogenesis in PVD neurons.

Experiments in heterologous systems have shown that proprotein convertases can process extracellular proteins, e.g. the glycoprotein reelin. However, this processing occurs in cis, i.e. with both furin and reelin functioning in the same cell (Kohno et al., 2015). In contrast, our observations could be consistent with a scenario in which KPC-1 can process/regulate substrates also in trans, i.e. on adjacent cells. Whereas, to our knowledge, there is no evidence in the literature of a proprotein convertase cleaving a substrate in trans, a similar scenario has been documented for ADAM metalloproteinases (Janes et al., 2005). Specifically, it was shown that ADAM10 associates with the ephrin receptor on one cell to cleave the cognate ephrin ligand in trans on another cell upon engagement of the ephrin ligand and the ephrin receptor, thereby providing negative feedback (Janes et al., 2005). Therefore, both proprotein convertases and ADAM family proteases, i.e. different classes of extracellular transmembrane proteases, appear to be associated with specific cell-surface receptors on the same cell to regulate a given signaling pathway. This may be a more general regulatory mechanism for signaling pathways, which provides cells receiving a signal from adjacent cells a powerful means for a feedback loop to negatively regulate the received signal.

C. elegans animals were grown on OP50 Escherichia coli-seeded nematode growth medium plates, usually at 20°C unless otherwise specified. Strains used in this work can be found in Table S1.

A complete list of transgenes and plasmids created for these studies is provided in Tables S2 and S3, respectively.

A fosmid containing the mnr-1 locus including its regulatory regions (WRM618aD06) was injected at 1 or 20 ng/μl into mnr-1(dz175); dzIs53 [Is(F49H12.4::mCherry)] animals with myo-2p::mCherry as an injection marker at 25 ng/μl and pBlueScript (Stratagene/Sigma-Aldrich) at 50 ng/μl. The same injection mix with 1 ng/μl of WRM618aD06 was injected into dzIs53 control animals to assess putative mnr-1 overexpression effects on PVD dendritic arbor morphology.

We previously generated the dzIs49[Is(myo-3p::mnr-1)] integrated strain that expressed MNR-1 in the body wall muscle of a mnr-1 null mutant strain (Salzberg et al., 2013). The injection marker in this strain was rol-6(su1006). We used this strain to quantify DMA-1::GFP expression levels when MNR-1 was expressed in the muscle. To control for the presence of the rol-6(su1006) injection marker in dzIs49, we injected otherwise isogenic animals without the integrated transgene mis-expressing MNR-1 in muscle with 40 ng/μl of rol-6(su1006) and 60 ng/μl of pBlueScript. Transgenic animals and non-transgenic siblings were used as controls for fluorescence intensity quantification and the acquisition of comparative images in Fig. 4 and Fig. S3.

kpc-1 cDNA tagged at the C-terminus with sfGFP was cloned under control of a short version of the ser-2prom3 promoter (which we term ser-2p3s) or the rab-3p promoter, and injected at 5 ng/μl into dzIs53 [Is(F49F12.4p::mCherry)] animals either with myo-2p::mCherry or rol-6(su1006) as injection markers at 25 ng/µl together with pBlueScript at 50 ng/μl. A mutagenized version of the ser-2prom-3s::KPC-1::sfGFP plasmid carrying the H262A substitution in KPC-1 was used to assess rescue and protein expression of this catalytically dead form of KPC-1.

cDNA of kpc-1 containing the signal sequence of KPC-1 but lacking its prodomain (H34 to R146) was cloned downstream of the ser-2p3s promoter. Two catalytically dead versions of this preprocessed, prodomain-less construct were generated by site-directed mutagenesis, including H262A (which is part of the catalytic triad) and N363A (which constitutes the oxyanion hole residue that is essential for catalytic activity in other proprotein convertases) (Creemers et al., 1993).

mnr-1 plasmids doubly tagged with either N-terminal mCherry or tagRFP and C-terminal mNeonGreen were generated by digestion and ligation of individual N-terminal- and C-terminal-tagged mnr-1 (cDNA) constructs (Salzberg et al., 2013). The final fusions mCherry::mnr-1::mNeonGreen or tagRFP::mnr-1::mNeonGreen were expressed under control of a dpy-7p epidermal promoter, and injected into mnr-1(dz175) animals at 1 ng/μl together with the unc-122p::tagBFP marker at 25 ng/μl, the F49H12.4::tagBFP marker at 25 ng/μl and pBlueScript at 50 ng/μl.

A dpy-7p::mnr-1 rescuing construct was injected directly into kpc-1(gk333538); mnr-1(dz213) double-mutant animals or used for site-directed mutagenesis to introduce the L135F substitution. The L135F mutant construct was injected into kpc-1(gk333538); mnr-1(dz175) double-mutant or kpc-1(gk333538) or mnr-1(dz175) single-mutant animals. All constructs were injected at a concentration of 5 ng/μl together with 25 ng/μl of myo-3p::GFP or myo-3p::tagRFP injection markers and 70 ng/μl of pBlueScript.

Plasmids carrying the kpc-1 cDNA sequence under control of the ser2prom3s or the mec-7p promoters (Salzberg et al., 2014) were individually injected into kpc-1(gk8); dzIs53; hmIs4 animals in which the defective dendritic arbors of PVD and FLP were visible. Constructs were injected at a concentration of 5 ng/μl either with 25 ng/μl of myo-3p::tagRFP or ttx-3p::mCherry injection markers, and 70 ng/μl of pBlueScript DNA.

kpc-1(gk333538); dzIs53 animals were mutagenized with ethyl methane sulfonate and the progeny of cloned F1 animals was scored for suppression or enhancement of the kpc-1 hypomorphic dendrite branch defects (Fig. S1). In a separate screen for modifiers of a lect-2 hypomorphic phenotype (Rahman et al., 2022), an additional allele for kpc-1 was identified. This new allele, dz254, results in a L535P missense mutation in the P domain of the proprotein convertase (Fig. 1; Fig. S5). kpc-1(dz254) mutant animals displayed shorter 1° branches and self-avoidance defects but not an increased number of 2° branches, compared with those of wild-type control animals (Fig. S2F,H). dz254 is a hypomorphic allele of kpc-1 because its phenotype was less severe compared with the phenotype of the presumptive gk8 null allele and more severe when placed in trans with the kpc-1(gk8) deletion (Fig. S2F).

In all experiments, fluorescence images were captured in live C. elegans at the L4 larval stage using a Plan-Apochromat 40×/1.4 or 63×/1.4 objective on a Zeiss AxioImager Z1 Apotome. Worms were immobilized using 1 mM levamisole for 30 min and mounted on 4% agarose pads. At least 25 adult animals were scored per genotype. Optical sections were collected and, depending on the experiment, individual sections or maximum-intensity projections were used for further analysis.

For tracing, the region of interest (ROI) comprised 100 μm of the primary branch anterior to the PVD cell body. Morphometric analyses were conducted using the NeuronJ plugin of the FIJI software and branches were defined as follows: 2° dendritic branches as any neurite branching out of the primary dendrite in the ROI; 3° branches as any neurite coming from the tip of a 2° branch and located in close proximity to the sub-lateral nerve cords; and 4° dendritic branches as those originating from 3° dendritic branches and extending to the dorsal or ventral nerve tracts. The self-avoidance index was defined as the ratio of the number of gaps between adjacent 3° branches divided by the number of candelabras within the ROI.

A structural model of kpc-1 was created using the I-TASSER server (Zhang, 2008; Yang et al., 2015), based on the crystal structure of human furin (PDB: 5JXG). The chosen model had a confidence score of −0.48 and an estimated TM-score (structural similarity) of 0.65. The pro-domain was removed from the sequence prior to simulation. The editing of the structure was performed using Chimera (Pettersen et al., 2004). The catalytic domain extends from N22 to K364 (shown in cyan, Fig. S5A), the P domain from H370 to V498 (shown in yellow), and the transmembrane domain from S529 to S542 (shown in purple). Wild-type residues corresponding to amino-acid substitutions in the mutated alleles used in this study are shown in red (R265 and L535). In the catalytic domain, the catalytic triad formed by the amino acids D221, H262 and S436 are shown in dark blue.

The structure of PCSK9 was downloaded from PDB (2PMW) and edited using Chimera. The pro-domain extends from T61 to Q152 (shown in orange; Fig. S5B), the catalytic domain from S153 to S447 (shown in cyan), and the V domain from G452 to H683 (shown in green). Note that contrary to other proprotein convertases, the prodomain of PCSK9 is cleaved only once and remains associated with the protease after cleavage.

For immunoprecipitation, five full plates of DMA-1::GFP transgenic worms were washed in RIPA buffer pH 7.0 (10 mM Tris-HCl, pH 7.0, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS ,140 mM NaCl) and lysed for 15 min in a Biorupter water bath, as previously described for whole-worm protein extraction (Li and Zinovyeva, 2020). Post lysis, 20 µl of Protein A/G Plus Agarose beads (Santa Cruz Biotechnology, sc-2003) and 1 µl of undiluted anti-GFP (Roche, 11814460001) or anti-HA (Roche, 11867423001) antibodies were used to pull down DMA-1::GFP overnight at 4°C. For SAX-7::GFP::FLAG and LECT-2::mNeonGreen::FLAG western blots, ten and twenty gravid adult animals, respectively, were boiled in loading buffer and loaded directly into the gels. Gradient gels (4-12%, GenScript) were used for all experiments. To detect endogenous MNR-1::HA, 10 full plates of C. elegans were washed with HBS (25 mM HEPES, pH 7.5, 150 mM NaCl) and collected in a 15 ml Falcon tube. Following centrifugation at 1000 g for 5 min, the worm pellet was resuspended in HBS at 4× the volume of the pellet and 10% SDS at 1× the volume of the pellet. The suspension was sonicated with a probe sonicator at 40% power for 1 min with a 5 s on/5 s off cycle and centrifuged at 20,000 g for 10 min. The supernatant was combined with 4× SDS loading buffer and incubated at 60°C for 15 min, before loading onto gels for SDS-PAGE (10-20 µl for MNR-1::HA detection, 1-5 µl for α-tubulin detection). Standard SDS-PAGE and wet-transfer western blotting were followed with 5% milk used for blocking. Antibodies for western blots were used at the following concentrations: for anti-FLAG blots, 1:800 anti-Flag (Sigma-Aldrich, F1804) and 1:5000 HRP-conjugated anti-mouse IgG (Millipore, AP308P); for anti-GFP blots: 1:500 anti-GFP (Roche, 11814460001) and 1:5000 HRP-conjugated anti-mouse IgG (Millipore, AP308P); for anti-HA blots, 1:500 anti-HA (Roche, 11867423001) and 1:10000 HRP-conjugated anti-rat IgG (Invitrogen, 69-9520); and 1:5000 anti-α tubulin (12G10 monoclonal antibody from the Developmental Studies Hybridoma Bank created by the National Institute of Child Health and Human Development).

Statistical tests were applied as described in each legend. Statistical significance of the number of 2° branches, 4° branches or self-avoidance index were calculated with one-way ANOVA using Sidak's correction for multiple comparisons. Statistical differences in self-avoidance and 4° branch number in mnr-1 fosmid overexpression experiments were calculated by comparing transgenic and non-transgenic siblings using an unpaired, two-tailed t-test. DMA-1::GFP fluorescence ratio differences for epistasis analysis and for mis-expression of mnr-1 experiments were calculated using a one-way ANOVA with Sidak's correction for multiple comparisons. For proportions, statistical significance was calculated using the z-test with Bonferroni correction for multiple comparisons where applicable. For comparisons of coelomocyte fluorescence intensity ratios, a non-parametric Mann–Whitney test was performed. All tests were performed using the Prism 7 Statistical Software suite from (GraphPad). Statistical significance is indicated as ns, not significant; *P≤0.05; **P≤0.01; ***P≤0.001; and ****P≤0.0001.

We thank members of the Bülow laboratory for comments on the manuscript and discussions throughout the course of this work; and Ryan Peer and William Corman for their initial help with the modifier genetic screen. We acknowledge the Genomics Core facility and the Advanced Imaging Facility at Albert Einstein College of Medicine for help during these studies. We are grateful to Kang Shen, David Miller and the Caenorhabditis Genetics Center (which is funded by National Institutes of Health Office of Research Infrastructure Programs P40OD0104400) for some of the strains used in this study, and Lhisia Chen for the anti-SAX-7 antibody.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201208.reviewer-comments.pdf.