The proprotein convertase BLI-4 promotes collagen secretion prior to assembly of the Caenorhabditis elegans cuticle

Some types of collagens, including transmembrane MACIT collagens and C. elegans cuticle collagens, are N-terminally cleaved at a dibasic site that resembles the consensus for furin or other proprotein convertases of the subtilisin/kexin (PCSK) family. Such cleavage may release transmembrane collagens from the plasma membrane and affect extracellular matrix assembly or structure. However, the functional consequences of such cleavage are unclear and evidence for the role of specific PCSKs is lacking. Here, we used endogenous collagen fusions to fluorescent proteins to visualize the secretion and assembly of the first collagen-based cuticle in C. elegans and then tested the role of the PCSK BLI-4 in these processes. Unexpectedly, we found that cuticle collagens SQT-3 and DPY-17 are secreted into the extraembryonic space several hours before cuticle matrix assembly. Furthermore, this early secretion depends on BLI-4/PCSK; in bli-4 and cleavage-site mutants, SQT-3 and DPY-17 are not efficiently secreted and instead form large intracellular puncta. Their later assembly into cuticle matrix is reduced but not entirely blocked. These data reveal a role for collagen N-terminal processing in intracellular trafficking and the control of matrix assembly in vivo. Our observations also prompt a revision of the classic model for C. elegans cuticle matrix assembly and the pre-cuticle-to-cuticle transition, suggesting that cuticle layer assembly proceeds via a series of regulated steps and not simply by sequential secretion and deposition.


Introduction
Proteolytic cleavage is a common regulatory step in the assembly of extracellular matrices: Components initially enter the secretory pathway as soluble proproteins and then are cleaved in later secretory compartments or extracellularly to allow their assembly into higher order structures [1][2][3].For example, mammalian fibrillar collagens undergo both N-terminal and Cterminal proteolysis in order to convert procollagen to the mature collagen that is found in extracellular fibrils [4][5][6][7].For Type I procollagen, these cleavages depend (at least in part) on members of the ADAMTS and BMP-1/astacin proteinase families, respectively [5,[8][9][10], and failure to appropriately cleave procollagen leads to human connective tissue disorders such as Ehlers-Danlos syndrome Type VII (ED-VII) and Osteogenesis imperfecta [4,[9][10][11][12].Some other types of collagens, including transmembrane MACIT collagens (membrane associated collagens with interrupted triple helixes) and C. elegans cuticle collagens, are instead N-terminally cleaved at a dibasic site that resembles the consensus for furin or other proprotein convertases of the subtilisin/kexin (PCSK) family [13][14][15][16][17].However, evidence for the importance of specific PCSKs in collagen cleavage is lacking and the functional consequences of such cleavage are not well defined.
Appropriate cleavage of procollagens may facilitate formation of fibrils or other higher order matrix structures at the right place and time, in the presence of appropriate partners.In the case of transmembrane collagens, N-terminal cleavage also would release the ectodomain from the plasma membrane.Unfortunately, potential redundancy and uncertainty about the specific proteinases involved, combined with difficulties in visualizing collagen matrix in vivo, have made it challenging to dissect these regulatory mechanisms in most biological systems.For example, while C-terminal cleavage strongly promotes Type I collagen fibril assembly in vitro [5,18], there is still uncertainty about the specific role N-terminal cleavage plays in matrix assembly [3,6,7,19].Unprocessed Type I Pro(N)-collagen can be found extracellularly in cell culture [4,20] and in morphologically abnormal fibrils in ED-VII patients [21,22], leading to the widespread view that N-terminal processing is not essential for collagen secretion or incorporation into fibrils, but rather affects specific aspects of fibril structure.It remains unclear if this is generally true for other types of collagens, including those cleaved by PCSKs.Finally, some invertebrate fibrillar collagens and many non-fibrillar collagens do not undergo N-or Cterminal cleavage and must rely on other mechanisms to control the time and place of collagen matrix assembly [23,24].
The nematode Caenorhabditis elegans has an external body cuticle that consists primarily of collagens, many of which have a predicted transmembrane domain and/or an N-terminal consensus furin cleavage site (CFCS) that could be cleaved by a PCSK [13,[25][26][27].A new collagenous cuticle matrix is synthesized in the embryo and during each molt between larval stages, and it is always preceded by a transient "pre-cuticle" matrix that contains zona pellucida (ZP) domain proteins and other non-collagen components [28][29][30][31][32][33][34].Therefore, these two matrix types must be assembled and then disassembled in the proper sequence during the molt cycle.These dynamic matrix changes are controlled in part by oscillatory gene expression programs, with pre-cuticle genes peaking relatively early in each molt cycle and different cuticle collagen genes peaking at early, intermediate, or late timepoints, consistent with an extended period of cuticle synthesis and assembly [35][36][37].The molt cycle also is controlled by various post-transcriptional mechanisms such as regulated trafficking and proteolysis [38][39][40][41][42][43].A BMP1-related astacin proteinase, DPY-31, has been implicated in C-terminal processing of SQT-3 cuticle collagen [44][45][46], while one or more furin/PCSKs are thought to be responsible for N-terminal processing of many cuticle collagens at the CFCS.The functional importance of N-terminal cleavage is supported by the fact that CFCS mutations in several collagens cause disruptions to cuticle structure [13,14,25].
Here we investigate the roles of C. elegans BLI-4, a member of the furin/PCSK family [47], in cuticle assembly.PCSK proteinases cleave secreted or transmembrane substrates immediately following dibasic sites of consensus sequence (R/K) Xn (R/K), where Xn can be 0, 2, 4, or 6 amino acids [48].Mammals have 9 PCSK family members that differ primarily in their Cterminal domains, which are thought to confer different subcellular localization patterns and/ or substrate preferences [48][49][50].C. elegans has 4 PCSK family members, among which BLI-4 and KPC-1 appear to be the major ones expressed in external epithelial cells, while EGL-3 and AEX-5 are expressed primarily in neurons, muscle, or internal epithelia [51][52][53].kpc-1, egl-3, and aex-5 mutants are viable and do not have any reported defects in the molt cycle or cuticle [54][55][56].In contrast, bli-4 is an essential gene, and isoform-specific bli-4(e937) mutants have blistered adult cuticles, making BLI-4 an excellent candidate for cleaving cuticle collagens [47].By imaging formation of the first (L1) cuticle in developing wild-type embryos and bli-4 or cleavage site mutants, we provide evidence that BLI-4-dependent N-terminal processing of specific cuticle collagens promotes their efficient secretion several hours prior to matrix assembly.

Collagen secretion begins several hours before the pre-cuticle to cuticle transition
The collagenous cuticle matrix of C. elegans is preceded by a molecularly distinct pre-cuticle apical extracellular matrix (aECM) that it eventually replaces during each molt cycle [57].To determine the precise timing and sequence of these events for the first pre-cuticle and cuticle, we imaged staged live embryos expressing fluorescently-tagged matrix factors (Fig 1).Tags were located either internally (int), at the N-terminus immediately following the signal sequence (ss), or at the extreme C-terminus, as described in S1 and S2 Table and schematized in the figures below.All of these matrix fusions were generated by CRISPR/Cas9-dependent tagging of the endogenous loci and were functional based on phenotypic assays (Methods).
Images were collected at the 1.5-fold stage and then at one-hour intervals thereafter, as the embryos elongated to their ultimate worm body shape.
The pre-cuticle or "sheath" matrix can be detected by the 1.5-fold stage and is important for proper embryo elongation beyond the 2-fold stage [28,32,58].Consistent with that, the precuticle ZP protein NOAH-1 tightly marked the pre-cuticle from before 1.5-fold to post-elongation (Fig 1A), as previously reported [32].The secreted lipocalin LPR-3 initially accumulated between the embryo and the eggshell, and then marked pre-cuticle beginning about an hour after NOAH-1 (Fig 1B).Both proteins then were endocytosed and cleared.mCherry fusions but not Superfolder (Sf) GFP fusions accumulated in large lysosome-like structures (S1 Fig) , suggesting that endocytosed protein eventually moved into an acidic endolysosomal compartment where acid-tolerant mCherry fusions (or cleaved mCherry) can continue to fluoresce but SfGFP fusions are quenched [59].Together these data indicate that pre-cuticle matrix assembly occurs in a stepwise fashion.Furthermore, pre-cuticle matrix clearance involves considerable endocytosis.
Assembly of the first cuticle traditionally has been thought to begin late in embryogenesis, about 4 hours after 1.5-fold [58,60].However, we found that tagged cuticle collagens DPY-17 and SQT-3 were secreted before the 1.5-fold stage, contemporaneously with pre-cuticle factors (Fig 1C and 1D).These collagens initially accumulated between the embryo and the eggshell, began to incorporate detectably into matrix by the 2-hour timepoint (when embryos had elongated to the 3-fold stage), and appeared fully incorporated by the 3-to-4-hour timepoints.DPY-17 also consistently marked a single large extracellular aggregate that appeared in the extracellular space concomitant with matrix incorporation, consistent with some change in its molecular properties at this time (Fig 1C).In summary, the transition from pre-cuticle to cuticle begins during embryo elongation and components of both types of matrices transiently coexist, with the pre-cuticle disassembling once embryo elongation is complete.Furthermore, collagen secretion occurs earlier than previously thought, yet detectable collagen matrix incorporation occurs 2-3 hours afterwards, consistent with post-transcriptional (and possibly postsecretory) controls of cuticle matrix assembly.

BLI-4 is widely expressed in external epithelia and localizes to intracellular compartments and the extraembryonic space
The PCSK BLI-4 is a strong candidate for cleaving cuticle collagens to promote matrix assembly [25,47].Existing single cell RNA sequencing (scRNAseq) data from embryos [52] revealed bli-4 expression in external epithelial cells (e.g.those lined by pre-cuticle/cuticle), including the epidermis and various interfacial tubes and glia, as well as in the foregut (pharynx), intestine, and germline (S2 Fig) .Consistent with this, a bli-4 transcriptional reporter [61] also showed widespread epithelial expression, including in both the lateral (seam) and major (hyp7) epidermis and in the excretory duct and pore tubes (Fig 2A ).A functional BLI-4:: SfGFP(int) fusion protein, tagged at the protease domain and expressed from the endogenous locus, marked sub-apical intracellular compartments of external epithelia at the 1.5-fold stage and beyond (Fig 2B).BLI-4::SfGFP(int) also was faintly visible in the extraembryonic space (Fig 2B ) and it appeared transiently within the lumen of the foregut at the 1.5+4 hr timepoint (Fig 2C).In summary, BLI-4 is expressed in external epithelia before and during cuticle assembly, and it appears both intracellular and extracellular.

Generation of bli-4 null and isoform-specific mutants
The bli-4 gene has many splice isoforms that differ in their 3' exons; the resulting proteins all share the N-terminal peptidase domain but differ in the presence or absence of other domains (Fig 3A and 3B) [47].These isoforms may have different substrate specificities.For example, bli-4(e937) (hereafter bli-4(ΔBLI)) is a deletion removing exons unique to isoforms a, e, g, and h as well as part of the 3'UTR of isoform f and intronic sequences of the other remaining isoforms; these mutants are viable but have a blistered (Bli) adult cuticle, suggesting failure to process key substrates unique to that stage [47].BLI-4 isoforms c and d contain a cysteine-rich domain (CRD) similar to that found in mammalian furin, PCSK5, and PCSK6 (Fig 3A).The PCSK5 CRD can confer cell surface anchoring via binding to heparan sulfate proteoglycans [49,50], and was therefore proposed to affect substrate specificity.Based on an analysis of 3' end reads from embryo scRNAseq data [52], the bli-4a, d and f/g isoforms are detectably expressed in the embryo.bli-4d is by far the most highly expressed isoform in the embryonic epidermis and is also detected at lower levels in other tissues such as glial and excretory cells, the pharynx and intestine, and the germline (S2 Fig) .Isoform bli-4a is expressed primarily in the germline and isoforms f/g are expressed most strongly in pharyngeal epithelial cells, though both are also detected in epidermis to a lesser degree (S2 Fig) .The predicted sizes of these BLI-4 isoforms are consistent with the most prominent bands observed on Western blots of lysates from BLI-4::SfGFP(int)-expressing embryos (Fig 3C and 3D).
To address the overall roles of BLI-4, and roles of the two CRD-containing isoforms specifically, we used CRISPR/Cas9 to generate new bli-4 null and isoform-specific alleles in an isogenic strain background (Fig 3B, S2 Table and Methods).Two null alleles (cs281 and cs283, hereafter bli-4(-)) were made by targeting exon 2, upstream of the peptidase domain; these cause frameshifts that should remove all splice isoforms.Two isoform c-& d-specific mutants (cs302 and cs308, hereafter bli-4(ΔCRD)) and two isoform d-specific mutants (cs293 and cs295) were made by targeting exons unique to those isoforms.All mutants were recessive lethal but could be rescued with a fosmid-based transgene containing the entire bli-4 genomic locus (Figs 3E, 4 and 5).bli-4(-) failed to complement the isoform-specific alleles, as expected, whereas bli-4(ΔCRD) and bli-4(ΔBLI), which remove distinct sets of isoforms, complemented each other (Figs 3E and 4B).

bli-4 null mutants arrest as retracted embryos following elongation
bli-4(-) mutants were embryonic lethal (Emb), as previously reported for other null alleles (Fig 4A and 4C) [47].Mutant embryos arrested as disorganized masses with occasional excretory tube dilations and evidence of debris between the embryo and the eggshell (Fig 4A).Although  [61] is broadly expressed in external epithelia, including in hyp7, seam cells, and in the excretory duct (d) and pore (p).B) An endogenous BLI-4::SfGFP(int) translational fusion (syb5321) marks intracellular puncta within epithelia and is faintly detectable within the extraembryonic space (EES).Single channel images are shown in inverted grayscale for clarity.Asterisk, fusion protein detected in the extraembryonic space.Arrow, intracellular puncta.CRISPR/Cas9 was used to insert SfGFP between the BLI-4 prodomain (Pro) and peptidase domain, as indicated (S2 Table ).The schematic shows the short isoform BLI-4f (Genbank NP_001360008.1),but all isoforms should be tagged.ss, signal sequence.P, P domain.C) BLI-4:: SfGFP(int) transiently accumulated in the foregut at the 1.5 + 4hr timepoint.https://doi.org/10.1371/journal.pgen.1010944.g002The % identity at the amino acid level is listed between relevant domains.Like other PCSK family members, both proteins have a signal peptide (SP) and prodomain (Pro) that are removed during trafficking, followed by the peptidase domain and an associated P domain thought to assist with its folding and stability [48].BLI-4d and PCSK6 also share a cysteine-rich domain (CRD).BLI-4d also has a transmembrane (TM) domain but it lacks the EGF-like (EGFL) domain found in PCSK6.B) bli-4 gene isoforms and mutant alleles.Colors indicate encoded protein domains, as in A. Isoforms are arranged by mutant groups.cs281 and cs283 are 1-2nt deletion/ frameshift mutations in exon 2, which is shared among all bli-4 isoforms.e937 is a 3,325 bp deletion that removes intronic sequences and exons associated with isoforms a, e, g, and h [47].the arrested embryos appeared unelongated, timelapse imaging revealed that mutant embryos did elongate initially, but then retracted and collapsed soon afterwards, about 4 hours after the 1.5-fold stage (Fig 4D and S1 and S2 Video).Notably, retraction occurred near the time when the cuticle replaces the pre-cuticle (Fig 1) and specifically resembled that previously reported for mutants lacking the essential cuticle collagen SQT-3 [58].

bli-4(ΔCRD) mutants arrest as Dumpy larvae with abnormal cuticles
The four bli-4(ΔCRD) or bli-4(d) mutants all appeared less severe than the null but similar to each other, with variable larval arrest (Lvl) at the L1, L2, or L3 stage (

BLI-4 promotes secretion and cuticle incorporation of collagens SQT-3 and DPY-17
Next, we examined our matrix fusions in bli-4 mutant backgrounds.Although many ZP proteins are cleaved at a C-terminal CFCS site before matrix incorporation [2], we did not note any obvious change in the pre-cuticle appearance of a NOAH-1 fusion between wild type and Therefore, protein trafficking was not generally disrupted.However, we did find significant differences in the appearance of both cuticle collagens (Figs 6 and 7).
SQT-3 is a predicted Type II transmembrane collagen with a cytosolic N-terminus and external C-terminus, plus a CFCS whose cleavage could release that collagenous C-terminus into the environment (Fig 6A).DPY-17 is a secreted collagen with an N-terminal CFCS (Fig 7A ).Quite dramatically, in bli-4(-) mutants, both SQT-3::mNG(int) and DPY-17::mNG were poorly secreted and failed to accumulate robustly in the extra-embryonic space or to incorporate efficiently into the cuticle (Figs 6B, 6C, 6E and 7B, 7C, 7E).Instead, both collagens formed large puncta at or near the apical plasma membrane, with most puncta appearing at least partly intracellular when compared to the cuticle surface or an mCherry::PH membrane marker (Figs 6D-6H and 7D and 7E).Puncta were visible by the 1.5-fold stage (Figs 6B and 7B), several hours before normal matrix incorporation (Fig 1C and 1D).These data indicate that BLI-4 promotes initial secretion of the soluble forms of these two collagens.
bli-4(ΔCRD) mutants had less severe defects in collagen secretion; SQT-3 and DPY-17 fusions still formed some intracellular puncta, but they were at least partly secreted and eventually incorporated into the cuticle (Figs 6B-6E and 7B-7E).SQT-3 is the only known cuticle collagen required for C. elegans embryo viability, being critical to maintain embryo shape after elongation [27,58]; its partial secretion and incorporation in bli-4(ΔCRD) mutants but not bli-4(-) mutants likely contributes to the different arrest points of these mutants.These data are consistent with a role for multiple BLI-4 isoforms in cleavage of these collagens.

SQT-3 and DPY-17 are mutually dependent on each other for secretion
Although dpy-17 null mutants have a less severe Dpy phenotype than sqt-3 null mutants, prior data suggested that DPY-17 and SQT-3 function together and that DPY-17 is required for efficient SQT-3 secretion [45,64].We were able to confirm this result; in dpy-17(-) mutants, SQT-3 accumulated cytosolically and in a halo pattern surrounding epidermal nuclei (Fig 8A and  8C), suggesting retention in the endoplasmic reticulum.Conversely, we found that SQT-3 is required for both DPY-17 secretion and overall protein accumulation; DPY-17 was barely detectable in sqt-3(-) mutants (Fig 8B and 8C).While the underlying mechanism remains to be investigated, decreases in protein stability are often observed after removal of a key binding partner [65].Our data support the model that DPY-17 and SQT-3 travel together through the secretory pathway and suggest that bli-4 loss could affect DPY-17 and SQT-3 both directly and indirectly via effects on the other (Fig 8D ).

CFCS mutations in SQT-3 and DPY-17 mimic loss of bli-4
The dramatic aECM defects described above are consistent with roles for BLI-4 in the N-terminal processing of multiple cuticle collagens.Unfortunately, because bli-4 mutants are lethal, it is difficult to collect large numbers of mutant embryos and we have not found appropriate bli-4 knockdown and Western blot conditions to directly test if BLI-4 is required for collagen cleavage.Instead, we used CRISPR/Cas9 to mutate the predicted BLI-4-dependent cleavage sites from RxxR to AxxA within the endogenous SQT-3::mNG(int) and DPY-17::mNG fusions (see Figs 6 and 7).These CFCS mutations mimicked loss of bli-4, causing intracellular retention of the mutant proteins (Fig 9A -9D).Some mutant collagen eventually incorporated into the L1 cuticle, but matrix structures appeared abnormal and larvae exhibited a severe Dpy phenotype (Fig 9E and 9F).Because most sqt-3 and dpy-17 CFCS mutants survive past hatch, unlike bli-4(-) mutants, the embryonic arrest phenotype of bli-4(-) likely reflects cumulative effects on multiple substrates.Nevertheless, these data strongly support the model that BLI-4 directly cleaves both SQT-3 and DPY-17 collagens to promote their secretion.

Discussion
N-terminal processing is an important step in the maturation of mammalian fibrillar collagens and is generally thought to affect fibril structure.Here we provide evidence for an alternative role of N-terminal processing in C. elegans cuticle collagens: to allow secretion of soluble forms of collagen prior to matrix assembly (Fig 10E).Using new collagen fusion knock-ins that allowed us to visualize C. elegans cuticle assembly in live embryos, we showed that collagen secretion into the extraembryonic space precedes cuticle assembly by several hours.Loss of bli-4 PCSK prevents efficient secretion of two early cuticle collagens, SQT-3 and DPY-17, causing them to accumulate within apical compartments of epidermal cells and reducing (but not completely blocking) their later assembly into cuticle matrix.Mutation of the predicted BLI-4-dependent cleavage sites causes similar defects.These data demonstrate a role for collagen N-terminal processing in intracellular trafficking and control of matrix assembly in vivo and prompt a revision of the classic model for C. elegans cuticle matrix assembly and the precuticle-to-cuticle transition.

Temporal control of cuticle assembly and the pre-cuticle to cuticle transition
The C. elegans pre-cuticle and cuticle are molecularly distinct matrices that coat external epithelia at different stages of development.The pre-cuticle is present at earlier stages and is required for initial embryo elongation, while the cuticle eventually replaces it and is responsible for maintaining embryo shape [28,32,58].Consistent with those defined roles, we showed here that the transition to cuticle matrix happens during and shortly following embryo elongation.Surprisingly, prior to this transition, at least some cuticle collagens are present in the extra-embryonic space for several hours without appearing to substantively aggregate or incorporate into matrix.Eventually, cuticle collagens gradually incorporate and then transiently co-exist in the matrix with pre-cuticle proteins.After the collagens have incorporated, pre-cuticle proteins are removed by endocytosis.
This sequence of events suggests a revision of the classic model for C. elegans cuticle assembly, in which sequentially-deposited distinct layers are pushed progressively further away from the plasma membrane [66,67].The pre-cuticle does not become the outer layer of the mature cuticle.Furthermore, since cuticle collagen matrix assembly does not immediately follow secretion, it could potentially occur external to or within the initial pre-cuticle layer rather than more membrane-proximally.This means that the earliest expressed collagens need not necessarily join the matrix first nor ultimately define more external cuticle layers.For example, despite the early SQT-3 secretion shown here, sqt-3 mutants are defective in formation of the basal striated layer of the L1 cuticle, which forms only after embryo elongation [58].We propose that the layered organization of the final cuticle structure is determined not only by the initial timing of cuticle collagen expression and extracellular release, but also by processing events and protein-protein interactions that occur in the extracellular environment or along the still poorly understood routes that these collagens take through the secretory pathway.We do not currently know if the early pool of secreted collagen has a function or if the collagen that joins the cuticle comes from that extracellular pool (an "outside in" assembly direction) or from a later wave of secreted protein (an "inside out" assembly direction).A very interesting recent study suggested that mammalian fibrillar collagens are initially secreted in soluble form and then re-endocytosed and recycled through a distinct secretory pathway before being competent for fibril elongation [68].Our observations could be consistent with such a model and set the stage for more detailed future studies of collagen trafficking and matrix assembly in the C. elegans system.
Further studies also will be needed to determine if subsequent rounds of new cuticle assembly in C. elegans larvae follow the same early secretory schedule as in the embryo.The embryo is surrounded by a relatively impermeable eggshell that keeps secreted proteins concentrated near the plasma membrane and therefore potentially available for later assembly.The old cuticle could serve a similar role in larvae.However, it is also possible that transcriptional regulators of oscillatory gene expression take on a larger role in the post-embryonic stages.

Transmembrane and secreted collagens function together to form the cuticle
Sequence predictions suggest that only about a third (56/173) of C. elegans cuticle collagens are secreted using a conventional N-terminal signal sequence (as for DPY-17), while the remainder (like SQT-3) are predicted to be secreted in type II orientation with a cytosolic N-terminus followed by a transmembrane domain [27].Like mammalian MACITs and other collagenrelated transmembrane proteins [16,17], these transmembrane cuticle collagens could potentially remain associated with cell surfaces or could be released to the environment by proteolysis.Our observations indicate that a substantial portion of SQT-3::mNG(int) is released from the cell surface at early timepoints, but it is possible that a later switch to membrane retention is one mechanism that helps initiate cuticle assembly.
Our imaging of DPY-17 and SQT-3 fusions also confirm and extend prior evidence that these two very different collagens rely on each other for trafficking through the secretory pathway [45].In the absence of DPY-17, SQT-3::mNG(int) appears stuck in the ER, while in the absence of SQT-3, DPY-17::mNG is barely detectable.A similar reduction in levels of another cuticle collagen, DPY-7, were reported after removal of collagens that co-localize with it in cuticle furrows [69], suggesting that many functionally-related collagens may require each other for trafficking and/or stability.It will be interesting to investigate if such dependencies reflect the existence of heteromeric collagen triple helices or some other higher order matrix structures that begin to assemble intracellularly.

Roles for N-terminal and C-terminal processing in C. elegans cuticle collagen secretion and matrix assembly
C. elegans cuticle collagens resemble mammalian transmembrane collagens in having predicted N-terminal cleavage sites that match the consensus for furin/PCSKs rather than ADAMTS proteinases [13,14].A recent bioinformatic analysis found that 109 of the 173 predicted cuticle collagens contain an N-terminal CFCS that specifically matches the sequence RxxR [27].An RxxR sequence was found in most cuticle collagens with predicted secretion signal sequences (45/56) and in more than half of the predicted transmembrane cuticle collagens (64/117).N-terminal cleavage at CFCS sites is important for function of multiple collagens, since mutations in these sites cause cuticle abnormalities [13,14,25] (this work).BLI-4 previously was speculated to be the PCSK that performs these N-terminal cleavages [25,47] and our data strongly support that model for the two collagens tested here, SQT-3 and DPY-17.
Surprisingly, loss of BLI-4 or mutation of its predicted collagen target sites caused a substantial portion of SQT-3 and DPY-17 to accumulate within an apical compartment rather than being released in soluble form to the external environment.This defect cannot be attributed simply to failure to release a transmembrane form of SQT-3 collagen, since a CFCS mutation in DPY-17 (a secreted collagen) also caused retention within intracellular puncta.Our interpretation of these apical puncta is that they correspond to a secretory compartment from which collagens can't be released without N-terminal cleavage.Another possible (not mutually exclusive) interpretation is that they correspond to insoluble aggregates that form as the unprocessed procollagens move through the secretory pathway and encounter other partners or environments that allow them to initiate matrix assembly prematurely (Fig 10E).This latter model implies that N-terminal processing of SQT-3 and DPY-17 inhibits a key step of matrix assembly during trafficking, in contrast to the more typical scenario where proprotein cleavage facilitates matrix assembly.
In humans, failure to remove the Type I procollagen N-terminus leads to Ehlers-Danos syndrome type VII, a matrix disorder characterized by frequent joint dislocations and tissue fragility [4,9,11].In that case, the collagen molecules that retain their N-termini incorporate into abnormal fibrils, but to our knowledge no defects in secretion or in the timing of fibril formation have been reported [21,22].Part of this difference could be technical based on the in vivo imaging vs. in vitro biochemical approaches used, but it is also clear that there are many biological differences in collagen regulation between different families of collagens.Indeed, in mammalian fibrillar collagens, triple helix formation initiates at the C-terminus and then proceeds towards the N-terminus, whereas in MACIT collagens the opposite is true [15,70].The direction of triple helix assembly is not known for C. elegans cuticle collagens, but it has been noted that some have N-terminal coiled-coil regions that could serve as oligomerization domains [71]. C. elegans and mammalian collagens also differ in other processing events, for example C. elegans cuticle collagens display tyrosine-based crosslinking rather than hydroxylysine based crosslinking [72,73].It is not unreasonable to hypothesize that N-terminal cleavage may serve different roles in such different contexts.
Our data also suggest that the role of C-terminal cleavage could vary among cuticle collagen subtypes.Although SQT-3 contains a predicted BMP1/astacin cleavage site and requires DPY-31 for timely matrix assembly [44,45] (this work), we note that many other cuticle collagens, including DPY-17, do not have recognizable sites for C-terminal processing.Furthermore, Cterminal tags are retained on DPY-17 and multiple other cuticle collagens in matrix structures [63,74], suggesting that those collagens may not undergo such processing.Instead, there may be other unknown modifications, partners, and/or environmental conditions that must be present in order for those collagens to initiate matrix assembly.

Intracellular vs. extracellular processing of collagens
Despite the clinical importance of Type I procollagen processing, there is still debate in the literature about when and where this happens.C-terminal procollagen cleavage is thought to occur in a late secretory compartment or at the plasma membrane, while N-terminal cleavage may occur either before or after that [3,7].Growing evidence suggests N-terminal processing occurs at least in part within an ER or Golgi compartment, since the cleaved collagen and isolated N-terminal region can be detected in those locations and since the golgin Giantin is important for collagen processing [19,[75][76][77].However, uncleaved procollagen and N-proteinase activity also can be readily detected extracellularly in cell culture [4,20] and the best understood N-terminal proteinases, of the ADAMTS family, are secreted proteins also found extracellularly [78,79].These latter observations led to the traditional view that N-terminal processing occurs outside the cell.It is possible that processing normally occurs in both locations [3] and/or that partially processed and secreted collagens traffic through endocytic recycling compartments before final processing and matrix assembly [68].
Our data are more consistent with intracellular N-terminal cleavage of C. elegans cuticle collagens, since BLI-4 is mainly detected intracellularly and its loss leads to intracellular retention of SQT-3 and DPY-17.In most cases, the intracellular puncta appear to extend near to and potentially across the apical plasma membrane, suggesting that they could be present in tubulovesicular compartments that have access to the outside environment.Future identification of the BLI-4-and collagen-containing cellular compartments, combined with direct assays for cleavage and other modifications, should more precisely define the site and order of cuticle collagen processing.Ultimately, these studies should reveal how processing events help control the time and place of collagen assembly to construct the various cuticle structures and layers observed in vivo.

Strains and animal husbandry
See S1 Table for a list of all strains used in this work.C. elegans N2 was used as the wild-type strain.Unless otherwise indicated, strains were grown at 20˚C under standard conditions [80].See S2 Table for specific mutant lesions and a list of all sgRNAs and primers used for genome editing or transgenics.Tagged collagen and BLI-4::SfGFP strains and CFCS mutants of sqt-3 and dpy-17 (RxxR to AxxA) were made by Suny Biotech (Fuzhou, China).The endogenous fusions are functional based on viability and normal body morphology of the homozygotes.Rescue transgene csEx919 (bli-4+) was generated by microinjection of fosmid WRM069bE05 (20 ng/ul) with sur-5::GFP (30 ng/ul) and bluescript SK+ (50 ng/ul).

Generation of bli-4 alleles
Although bli-4 null alleles had already been described [47,81], they were generated on chromosomes carrying other markers that could complicate analysis; therefore, we opted to generate new alleles in an N2 background.Mutant alleles were generated by CRISPR-Cas9 genome editing using methods described in [82] and the sgRNAs listed in S2 Table .N2 hermaphrodites were injected with sgRNAs (IDT), Cas9 (University of California Berkeley), and the marker pRF4, and F2 progeny were screened for expected embryonic or larval lethal phenotypes.Mutant alleles were recovered over a genetic balancer (either hT2 (I;III) or szT1 (I;X)) and then rescued with bli-4+ transgene csEx919.Mutant lesions were identified by PCR amplification and Sanger sequencing.We originally identified 4 putative bli-4 null alleles, 8 putative bli-4(c/ d) alleles, and 3 putative bli-4(d) alleles; however, only the alleles described here had small (<50 bp) deletions that permitted PCR-amplification with our methods, while the remaining alleles appeared to have larger deletions or rearrangements and were not further characterized.

Microscopy and image processing
For timelapse imaging, ~24 cell stage embryos were mounted in egg buffer/methyl cellulose with 20μM beads as spacers [83], incubated at 20˚C for 4 hours, and then imaged using a stage temperature controller at 12˚C and a Leica TCS SP5 confocal (20 z-planes at 0.5 μM spacing and 15 minute time spacing, total 16 hours).To immobilize animals for still imaging, embryos or larvae were suspended in M9 buffer with 10mM levamisole and mounted on 5% agarose pads supplemented with 20mM sodium azide.DIC and epifluorescence images were obtained with a Zeiss Axioskop (Carl Zeiss Microscopy) fitted with a Leica DFC360 FX camera with Qcapture (Qimaging) software.Confocal images were captured with a Leica TCS SP8 confocal microscope, except for images in Figs 9C and 9D, and 10C and 10D, which were captured with a Zeiss LSM780 confocal microscope.Images were analyzed and processed in FIJI [84].To quantify fusion protein accumulation in the extraembryonic space, fluorescence intensity was measured in a 3x3 μm region within a single medial confocal slice of each specimen.To quantify puncta number and size, maximum intensity projections were first thresholded with uniform settings and then the "analyze particles" function was used to identify particles between 0.05 and 3 microns 2 in area.For permeability assays, L1 larvae were incubated in 2 μg/ml Hoescht dye 33258 (Sigma) in M9 buffer for 15 minutes at room temperature, then washed twice with M9 before imaging.

Statistical analyses
Statistical analyses were performed using GraphPad Prism.In all dot-plots, lines and error bars indicate the median and range, respectively, and each dot represents a measurement from a single animal.To perform statistical analyses on quantitative measurement data, genotypes were compared using a non-parametric Mann-Whitney test.To perform statistical analyses on categorical data, phenotypes were classified as either normal or abnormal and then proportions compared using a two-tailed Fisher's Exact test.Raw data for all graphs can be found in Supporting data files 1-8.

bli-4 isoform analysis from single-cell RNA-seq data
scRNA-seq reads from [52] were remapped to each exon of the protein coding genes in Wormbase genome build WS277 using cellranger software (10x genomics).Only reads mapping to isoform-specific exons or 3' UTRs were considered in this analysis.The resulting UMI counts per exon for each tissue were combined in VisCello software with the existing C. elegans embryo scRNA-seq atlas [52].Pseudobulk expression levels of each bli-4 exon in reads per million were calculated in cells annotated as each cell type, and in major tissue classes as annotated in [52].Code used for these analyses are available at Github: https://github.com/jisaacmurray/bli4_paper.

Fig 2 .
Fig 2. BLI-4 localizes to intracellular and extracellular compartments.A,B,C) Transcriptional and translational reporters reveal bli-4 expression throughout pre-cuticle and cuticle assembly.Animals also express the epithelial junction marker DLG-1::RFP (mcIs46, magenta) to aid in cell identification and Z-depth assessment.Images are maximum intensity projections from confocal Z-stacks and representative of at least n = 8 animals examined per genotype per stage.A) A bli-4pro::GFP transcriptional reporter (sEx11763, green)[61] is broadly expressed in external epithelia, including in hyp7, seam cells, and in the excretory duct (d) and pore (p).B) An endogenous BLI-4::SfGFP(int) translational fusion (syb5321) marks intracellular puncta within epithelia and is faintly detectable within the extraembryonic space (EES).Single channel images are shown in inverted grayscale for clarity.Asterisk, fusion protein detected in the extraembryonic space.Arrow, intracellular puncta.CRISPR/Cas9 was used to insert SfGFP between the BLI-4 prodomain (Pro) and peptidase domain, as indicated (S2 Table).The schematic shows the short isoform BLI-4f (Genbank NP_001360008.1),but all isoforms should be tagged.ss, signal sequence.P, P domain.C) BLI-4:: SfGFP(int) transiently accumulated in the foregut at the 1.5 + 4hr timepoint.

Fig 3 .
Fig 3. Generation of bli-4 null and isoform-specific alleles.A) C. elegans BLI-4d (Genbank NP_001021543.1)protein schematic and comparison to human PCSK6 (Genbank BAA21625.1).The % identity at the amino acid level is listed between relevant domains.Like other PCSK family members, both proteins have a signal peptide (SP) and prodomain (Pro) that are removed during trafficking, followed by the peptidase domain and an associated P domain thought to assist with its folding and stability[48].BLI-4d and PCSK6 also share a cysteine-rich domain (CRD).BLI-4d also has a transmembrane (TM) domain but it lacks the EGF-like (EGFL) domain found in PCSK6.B) bli-4 gene isoforms and mutant alleles.Colors indicate encoded protein domains, as in A. Isoforms are arranged by mutant groups.cs281 and cs283 are 1-2nt deletion/ frameshift mutations in exon 2, which is shared among all bli-4 isoforms.e937 is a 3,325 bp deletion that removes intronic sequences and exons associated with isoforms a, e, g, and h[47].cs302 and cs308 are 4-19 nt indel/frameshift mutations in the first exon unique to isoforms c and d. cs293 and cs295 are identical 5nt deletion/frameshift mutations in the first exon unique to isoform d.See S2 Table for specific allele sequences.C) Predicted sizes of major embryonically-expressed BLI-4:: SfGFP(int) fusion proteins before and after removal of the Pro domain.Sizes were estimated based on the isoform sequence using https://www.bioinformatics.org/sms/prot_mw.html.See also S2 Fig for data regarding isoform expression in the embryo.D) Western blot of lysates from BLI-4::SfGFP(int) expressing embryos.Arrowheads indicate four major bands between 80 and 130 kD.Blot is representative of 3 replicates.E) Summary of complementation test results.bli-4(cs281) failed to complement both bli-4(cs302) (n = 59) and bli-4(e937) (n = 120) for the larval lethal (Lvl) and adult Blister (Bli) phenotypes, respectively, while bli-4(cs302) complemented bli-4(e937) (n = 186).Balancer hT2 [bli-4(e937) let-?(q782) qIs48] (I;III) was used as the bli-4(e937)-containing chromosome.
Fig 3. Generation of bli-4 null and isoform-specific alleles.A) C. elegans BLI-4d (Genbank NP_001021543.1)protein schematic and comparison to human PCSK6 (Genbank BAA21625.1).The % identity at the amino acid level is listed between relevant domains.Like other PCSK family members, both proteins have a signal peptide (SP) and prodomain (Pro) that are removed during trafficking, followed by the peptidase domain and an associated P domain thought to assist with its folding and stability[48].BLI-4d and PCSK6 also share a cysteine-rich domain (CRD).BLI-4d also has a transmembrane (TM) domain but it lacks the EGF-like (EGFL) domain found in PCSK6.B) bli-4 gene isoforms and mutant alleles.Colors indicate encoded protein domains, as in A. Isoforms are arranged by mutant groups.cs281 and cs283 are 1-2nt deletion/ frameshift mutations in exon 2, which is shared among all bli-4 isoforms.e937 is a 3,325 bp deletion that removes intronic sequences and exons associated with isoforms a, e, g, and h[47].cs302 and cs308 are 4-19 nt indel/frameshift mutations in the first exon unique to isoforms c and d. cs293 and cs295 are identical 5nt deletion/frameshift mutations in the first exon unique to isoform d.See S2 Table for specific allele sequences.C) Predicted sizes of major embryonically-expressed BLI-4:: SfGFP(int) fusion proteins before and after removal of the Pro domain.Sizes were estimated based on the isoform sequence using https://www.bioinformatics.org/sms/prot_mw.html.See also S2 Fig for data regarding isoform expression in the embryo.D) Western blot of lysates from BLI-4::SfGFP(int) expressing embryos.Arrowheads indicate four major bands between 80 and 130 kD.Blot is representative of 3 replicates.E) Summary of complementation test results.bli-4(cs281) failed to complement both bli-4(cs302) (n = 59) and bli-4(e937) (n = 120) for the larval lethal (Lvl) and adult Blister (Bli) phenotypes, respectively, while bli-4(cs302) complemented bli-4(e937) (n = 186).Balancer hT2 [bli-4(e937) let-?(q782) qIs48] (I;III) was used as the bli-4(e937)-containing chromosome.https://doi.org/10.1371/journal.pgen.1010944.g003 Fig 5A).Despite the accumulation of debris during embryogenesis (Fig 5B), the majority of these mutants elongated and hatched (Fig 5A).Mutants appeared only mildly Dumpy (Dpy) as L1 larvae, but they completely lacked the alae ridges typical of the L1 cuticle (Fig 5C).The L1 cuticle retained its barrier function to exclude Hoechst dye, but the gut permeability barrier appeared defective (Fig 5D), suggesting that one or more components of that barrier could be substrates for gutexpressed BLI-4d.bli-4(ΔCRD) mutants became more severely Dpy by the time of L2 or L3 arrest (Fig 5A), suggesting that the CRD isoforms are particularly important for processing substrates during the early larval stages.This strong Dpy phenotype resembles that of many cuticle collagen mutants[26,62,63].

Fig 5 .
Fig 5. bli-4 CRD isoform mutants arrest as Dumpy larvae with cuticle defects.A-D) bli-4(ΔCRD) isoform mutant phenotypes.Only images of bli-4(cs302) and rescued siblings are shown.Phenotype quantitation and rescue data for all alleles are shown below.The bli-4+ rescue transgene is csEx919.*P<0.01,**P<0.001,***P<0.0001,Fisher's Exact test.A) Most mutants arrest as Dumpy L2 or L3 larvae.Larvae in image are 48hr after egg lay (AEL).B) Embryos accumulate extracellular debris (small arrows) between the embryo and the eggshell.Embryos are 1.5 fold + 3-4 hours old.C) L1 larvae are slightly Dumpy and entirely lack cuticle alae ridges.Bracket indicates position of alae in the rescued sibling.D) Whereas wild-type L1 larvae have a permeability barrier that excludes Hoechst dye (left), many bli-4(ΔCRD) mutants had strong staining in the gut epithelium.Staining was not apparent in the epidermis or pharynx, suggesting a gut-specific barrier defect.Note that bli-4d is expressed in each tissue (S2 Fig) but likely has different substrates in each location since these tissues vary widely in their aECM composition.https://doi.org/10.1371/journal.pgen.1010944.g005