Barrier epithelia are shielded from the external environment by their apical extracellular matrices (aECMs). The molecular complexity of aECMs has challenged understanding of their organization in vivo. To define the molecular architecture of a model aECM we generated a toolkit of 105 fluorescently tagged aECM components using gene editing in C. elegans, focusing on proteins secreted by the epidermis to form the collagen-rich cuticle. We developed efficient pipelines for modular protein tagging and rapid fluorophore swapping. Most tagged collagens were functional and exhibited exquisitely specific patterning across stages, cell types, and matrix substructures. We define multiple reference markers for key substructures including the little-understood cortical layer, as well as the helical crossed fiber arrays that function as a hydrostatic skeleton to maintain organismal shape. We further tagged 30 members of key aECM protein classes including proteases, protease inhibitors, and lipid transporters. Our standardized markers will allow dissection of the mechanistic basis of aECM spatiotemporal patterning in vivo.

Fig 1. Collagen processing and cuticle structure.

(A) Standard images of cuticle aECM in this and subsequent figures show 75 x 75 µm panels centered on the lateral alae in the anterior or mid body. Insets show magnified views of the annular and furrow regions. Right, cuticle cross section, longitudinal aspect showing one annulus, two furrows, and three struts, indicating furrow and fibrous/basal collagen localization in adult cuticle. Our analyses suggest furrow collagens localize to the basal layer in adults [24]. (B) Generic C. elegans collagen structure, based on [30]. An N-terminal propeptide (containing a predicted signal peptide or transmembrane domain) is cleaved by the BLI-4 subtilisin-like protease. A central collagen domain contains Gly-X–Y repeats with a variable number of interruptions. Conserved cysteine-rich domains (Cys) flank the collagen domain and are found between one of the Gly-X-Y sections of the collagen domain. Prolines and lysines in the collagen domain are hydroxylated (OH). Some but not all C. elegans collagens are processed in the C-terminal noncollagenous domain by the DPY-31 protease.

https://doi.org/10.1371/journal.pbio.3004059.g001

A modular pipeline for protein-tagging knock-ins

We first optimized parameters for Cas9 ribonucleoprotein complex (RNP) and linear repair template-based genome editing using a GFP knock-in into the 5′ end of the glh-1 gene. We chose glh-1 as a test case due to its use in optimization of editing efficiency and ease of scoring [31,32]. We tested Cas9 source and amount, RNA:Cas9 ratio, repair template, and methods to enrich for edits. IDT Cas9 at 250 ng/µl was most efficient, similar to previous reports (S1A Fig) [31]. While RNA:Cas9 ratios from 1:1–10:1 have been used [33–35], we found that a 1:1 ratio was most efficient with our reagents (S1B Fig). We did not find significant improvements from 5′ modifications [36] (S1C Fig), however, ssDNA templates [37] slightly increased efficiency (S1D Fig). We next tested methods to enrich for edits. In pilot experiments generating collagen knock-ins we were unsuccessful with pRF4 (rol-6(su1006)) or dpy-10 co-conversion [31,38], possibly due to genetic interactions between dominant collagen mutations and the target genes.

We next tested whether an mlc-1p::NLS::tagRFP co-injection marker could enrich for successful edits. Among marker-positive animals, 26% contained a GFP::GLH-1 knock-in, compared with 2.4% among marker-negative animals (S1E Fig). Comparable editing efficiencies were obtained when screening F1 progeny laid either 1 or 2 days post-injection (S1F Fig), although more total knock-ins were recovered from day 1 progeny (S1G Fig). We also tested whether myo-2::T2A::GFP could be used as a co-CRISPR marker. However, significantly fewer knock-ins were recovered among myo-2::T2A::GFP-positive progeny (S1H Fig). Based on these results, we used the mlc-1p::NLS::tagRFP co-injection marker for subsequent experiments, together with unmodified dsDNA repair templates and our optimized Cas9 concentration and RNA ratio. We prioritized speed in our goal of generating 100+ aECM knock-in strains and did not extensively optimize failed edits. We detail these failed edits in S1 Table and plan to revisit them to improve our editing pipeline.

We developed a modular tagging strategy in which an initial insertion allows rapid validation of phenotype, expression, and localization. We designed an mNG::3xFLAG cassette flanked by flexible 13 amino acid linkers that contain four crRNA target sites oriented for optimal editing efficiency (crRNAs 1–4; Fig 2A) [39]. Our initial insertion includes an mNG::3xFLAG flanked by the flexible linker sequences. Re-editing of the knock-in using the internal sgRNA sites (crRNAs #2 + 3) allows for removal of the mNG::3xFLAG and subsequent insertion of a desired cassette using a single, common repair template and two crRNAs (Fig 2A). This approach ensures that our toolkit strains can be modified by any researcher in an efficient, cost-effective manner. Use of different internal and external crRNA pairs allows for removal or retention of the linker sequences during re-editing. We first tested this system by attempting to replace the mNG::3xFLAG tag in glh-1 with an mSc::2xOLLAS tag using unmodified or biotinylated dsDNA or ssDNA repair templates. While biotinylated dsDNA and ssDNA repair templates did not boost glh-1 knock-in efficiency (S1C and S1D Fig), they dramatically improved color swap efficiency (Fig 2B), suggesting variation between targets that warrants further exploration. We also validated that the other crRNA pairs (crRNAs 1 + 4 and 2 + 4) allowed efficient tag swapping, except for crRNAs #1 + 3 (Fig 2C). These data may suggest that crRNA 3 is a suboptimal guide and edits using it might benefit from biotinylated dsDNA or ssDNA repair templates. We proceeded to generate knock-ins in aECM genes using either mNG or modular mNG::3xFLAG knock-ins, as mNG has been widely used to tag C. elegans BM proteins [7], and is frequently brighter than GFP [40,41]. In most cases, we initially inserted a tag at the C-terminus unless there was evidence for C-terminal processing (e.g., SQT-3) [42]; as discussed further below, C-terminal tagging appeared to be broadly tolerated

We successfully swapped mNG::3xFLAG to mSc::2xOLLAS in 9 strains. In color swap experiments, we also frequently observed mNG-negative animals in the progeny of injected animals, suggesting that the cassette was lost or inactivated (Fig 2D). Further optimization such as inactivation of nonhomologous end-joining [43] may be desirable. We were also able to color-swap three of the nonmodular mNG alleles (col-63, dpy-10, and dpy-13).

Our first pass knock-in attempts typically used melted PCR products with 35 bp homology arms. For targets that were unsuccessful using this approach, we developed a complementary selection-based strategy. We embedded an unc-119(+) selection marker in a synthetic intron in our modular mNG::3xFLAG cassette (S2 and S3 Figs), similar to a previous study [44], generating three strains (S2B Fig). Several unc-119-based knock-ins also failed, which are detailed in S1 Table. In the case of spia-1, we employed a self-excising cassette [45] due to concerns over the failure of C. briggsae unc-119(+) to rescue spore adhesion phenotypes [46]. We also generated a bli-5::mNG::3xFLAG knock-in using the SEC method prior to our generation of the unc-119-based selection system. Overall, plasmid-based repair templates with long homology arms and a selectable marker provided an effective alternative when simpler repair templates were unsuccessful but required substantially more experimental work to construct.

Gene families of the aECM and selection for toolkit

We focused on secreted protein families likely to comprise stable components of the aECM [10,47] and which were robustly expressed in public RNAseq datasets. The body cuticle is composed of collagens and noncollagenous proteins. The C. elegans genome contains ~180 collagen-encoding genes (see Methods and S1 File). We first focused on those for which phenotypic information existed (‘genetically defined’ collagens; bold face genes in Table 1). We next prioritized collagens based on expression as described above. These included collagens expressed in multiple larval stages (‘oscillating’) as well as stage-specific collagens, especially those upregulated in L4 or adult. These classes overlapped as some collagen genes are both oscillating and display stage specificity. Single-cell transcriptomic datasets such as (www.cengen.org) [48] indicate cuticle collagens are primarily expressed in the epidermis (seam cells or epidermal syncytia) or in interfacial glial cells that generate the overlying aECM.

While our imaging largely focused on collagens, we also tagged selected members of major noncollagen secreted protein families likely to be components of aECMs. We tagged the ZP domain protein DPY-1, the cysteine cradle domain proteins DPY-6 [49] and SPIA-1 [50], and the novel secreted protein LON-8 implicated in body cuticle morphology [51]. We also tagged the ZP domain proteins NOAH-1 and NOAH-2, components of the pre-cuticle [52]. We tagged the non-ZP cuticlin CUT-2 as a representative of the BME-insoluble cuticle fraction [53]. As aECM remodeling is important in molting, we tagged the extracellular peroxidase SKPO-1, astacin metalloproteases (NAS-8, NAS-37) [54], cathepsins (CPZ-1, CPZ-2, and CTSA-1.1), and putative protease inhibitors (BLI-5, PIIT-1, F30H5.3 (KUND-6), K10D3.4 (KUND-7)). As C. elegans has a number of poorly characterized putative protease inhibitors with Kunitz domains, we annotated this group of genes, classifying them as KUNitz protease inhibitor Domain containing (kund) (S3 Table). We tagged CLEC-180 as a representative of C-type lectins [55] and CPG-7 as a representative of chondroitin proteoglycans [56]. We also tagged the mucin SRAP-1, and proteins implicated in molting (ACN-1, QUA-1, MLT-8, MLT-9, MLT-10, and WRT-2) [57]. We also tagged several genes that may regulate extracellular lipids, including secreted lipid binding proteins such GM2AP/GMAP-1 [21], the nematode polyprotein allergen-related NPA-1 [58], and the low-density lipoprotein receptor LRP-1 [59]. GMAP-1::mNG was expressed in epidermis and localized to the cuticle, consistent with prior transgenic studies [21]. Other noncollagen proteins localized to a variety of tissues or compartments (Table 1; see S4 Fig), broadly consistent with their transcriptomic data [48].

Localization of genetically defined cuticle components

To test whether aECM components tolerate mNG tagging we first examined a set of 27 cuticle genes initially defined by mutant phenotypes [10,51,60,61]. Knock-ins for some have been previously reported (e.g., bli-1, bli-2, bli-6, dpy-13, dpy-17, and sqt-3) [24,62]; we generated knock-ins for the remaining genes. Most knock-in strains for such genetically defined cuticle genes were morphologically normal and displayed bright fluorescence in the cuticle aECM. In our initial survey of localization patterns, we examined lateral body cuticle regions containing circumferential furrows and annuli as well as longitudinal alae (Fig 1A). Six ‘furrow DPY’ collagens localized close to morphologically defined circumferential furrows (Fig 3A) [24,26,63]. Within this set, individual patterns differed in detail, consistent with recent observations [64]. DPY-2::mNG showed the most consistent labeling, extending around the circumference to the lateral midline or lateral alae, without breaks (Fig 3A). DPY-3::mNG furrows often did not extend to the lateral midline whereas DPY-9::mNG extended to the midline but had occasional gaps (Fig 3A). DPY-10::mNG extended to the midline with few gaps but frequent bright dots at the ends (Fig 3A). Animals homozygous for DPY-2::mNG and DPY-9::mNG displayed brighter expression than single knock-in homozygotes, but with frequent gaps or dots (Fig 3B); neither mNG signal extended to the midline, although morphological furrows extended to the midline as visualized by DIC (Fig 3B, lower panel). In contrast, DPY-9::mNG DPY-10::mSc ROL-6::mNG triple knock-ins displayed furrow morphology similar to DPY-9::mNG single knock-ins (Fig 3C). These observations suggest certain knock-in alleles or combinations may mildly affect protein function or localization (see below). Although overlapping in the xy plane, DPY-10::mSc localized more cortically than DPY-9::mNG (Fig 3C, right panels), suggesting furrow collagens may define distinct subcompartments. Furrow collagen mutants display defective aECM permeability barrier function [65]. Interestingly, BIS-1/COL-121, identified on the basis of permeability barrier defects in loss-of-function mutants [61], also localized to embryonic and L1 stage furrows (Fig 4D). Overall, these results in conjunction with other observations [64] suggest ‘furrow’ collagens may define distinct subregions within furrows.

Several collagen knock-ins, including DPY-4::mNG, DPY-5::mNG and DPY-13::mNG, were enriched in annuli, forming crossed fibrous arrays corresponding to the two fibrous layers of the cuticle [66] (Fig 4A); we classified these and similar patterns as ‘annular fibrous’. DPY-4::mNG displayed a more braided fibrous pattern within annuli (Fig 4A). These three collagens were not expressed in embryos but were upregulated in late L1 and visible as fibrous patterns in L2 onwards (S5 Fig), consistent with the lack of fibrous layers in L1 cuticle [67]. Other ‘annular fibrous’ collagens such as ROL-1::mNG were detected from late L4 stage onwards (Table 1, Fig 4A). In contrast, SQT-2::mNG and SQT-3::mNG were not overtly fibrous, instead forming smooth or punctate structures in annuli (Fig 4B); we term such patterns as ‘annular non-fibrous’. DPY-14::mNG was expressed beginning in comma stage embryos and localized to annuli in late embryonic and L1 cuticle (Fig 4C), consistent with the early larval Dpy phenotypes of dpy-14 mutants and similar to the previously reported localization of DPY-17::mNG [62]. Three collagens originally defined by male ray abnormal morphology (Ram) phenotypes [60], as well as the novel secreted protein LON-8 [51], localized to various compartments in the male tail and other structures, discussed in more detail below (Fig 9). These observations suggest the Ram phenotype can result from aberrant function of multiple proteins in the male tail aECM. Many of the above knock-ins were readily detectable in shed cuticles, consistent with their accumulation in aECMs shed at molts. Overall, these findings suggest the genetically defined collagens are primarily localized to furrows or to the fibrous layers (Figs 1A, 3 and 4; Table 1).

In the course of this analysis, we found that sc22, annotated as a rol-1 allele [68], did not contain a lesion in rol-1 and was weakly linked to the ROL-1::mNG knock-in. Analysis of the sc22 strain sequence revealed two other SNPs on chromosome II that would result in collagen missense alterations, including one causing R106H (RTAR > RTAH) in the col-71 dibasic subtilisin cleavage site [69]. A col-71(R106H) mutation edited into a wild-type background also caused an adult Rol phenotype, suggesting this mutation is causative. The rol-1 reference allele e91, which produces recessive left-hand rollers, also alters the dibasic cleavage site (RTAR > RTAC) [70]. Like ROL-1::mNG, COL-71::mNG displayed an annular fibrous localization (Figs 4A and 6A), consistent with it playing a similar role to ROL-1 in the adult cuticle.

An important question at the outset was whether collagens readily tolerate tagging. Most C. elegans cuticle collagens undergo N-terminal cleavage by subtilisin proteases [71] and otherwise largely consist of Gly-X-Y repeats that are critical for function (Fig 1B). Thus, we tagged most collagens at their C-termini; in a few cases, we inserted a tag within the protein. For example, SQT-3 is cleaved at its C-terminus [72]; therefore, we inserted a tag internally [62]. Several other collagens we tagged contain similar C-terminal cleavage sites (e.g., DPY-4, DPY-13, ROL-1; see S1 File for sequences), however C-terminal knock-ins in most cases appeared to preserve function. Of the cuticle collagens defined by mutant phenotype, we detected overt phenotypes in four knock-ins: DPY-4::mNG and DPY-8::mNG were mildly Dpy, and COL-34/RAM-4::mNG and COL-82/RAM-2::mNG males displayed Ram phenotypes, suggesting these insertions affect gene function. The SQT-1::mNG allele suppressed the Bli phenotype of bli-2(0), consistent with it impairing sqt-1 function [24]. As mentioned above, animals homozygous for DPY-2::mNG and DPY-9::mNG knock-ins displayed fragmented furrow collagen localization (Fig 3B) but were not overtly Dpy. While loss of furrow collagens results in a compromised aECM permeability barrier [65], we found that animals homozygous for DPY-3::mNG or DPY-10::mNG exhibited no such defects. These results indicate that the endogenous C-terminal tags do not significantly interfere with the physiological roles of these proteins. In contrast, animals homozygous for DPY-5::mSc and DPY-13::mNG resembled single homozygotes in expression and localization, suggesting synthetic effects may be specific to certain knock-in combinations. Although this analysis was not exhaustive, it suggests most cuticle collagens tolerate C-terminal knock-ins; in certain cases tagging may result in cryptic loss- or reduction-of-function that can be detected in sensitized backgrounds.

Most of our analysis relied on homozygous knock-in strains. To assess how tagged collagens compete with untagged proteins in the heterozygous state, we examined a set of 7 collagen knock-ins expressed at varying levels. Interestingly, the fluorescence intensities of cuticle mNG of most knock-in heterozygotes (5/7), when scaled according to copy number, were significantly above those of homozygotes; 2/7 (col-81, sqt-2) were not significantly different (when scaled) from homozygote intensity (S6A Fig). With the caveat that we have not assessed protein levels, these comparisons suggest tagged collagens can be incorporated as efficiently into the ECM as untagged alleles. Most KIs exhibited minimal recovery in FRAP experiments except for LON-8::mNG in the L4 vulva (S6B Fig), consistent with tagged proteins being stably incorporated into the matrix.

Many knock-ins were visible within underlying epidermal cells, localized to puncta that likely correspond to secretory organelles [24] (‘punctate’ in Table 1). Some mNG-tagged collagens accumulated in puncta within the epidermis and showed little cuticle fluorescence; in these cases, tagging may have interfered with secretion and the mNG-tagged proteins may be aberrantly retained in secretory compartments. Some mSc-tagged knock-ins showed intense fluorescence in filamentous structures that resemble epidermal lysosomes [73]. It is unclear if these reflect normal endocytosis or aberrant trafficking of tagged proteins.

Other collagens and cortical aECM compartments

In addition to the above genetically defined collagens, we tagged an additional 35 cuticle collagens and a noncuticle collagen (col-135). As mutant phenotypes were not known for most of these, functionality of the tagged proteins could not be verified. All were detectable using dissecting or widefield microscopy. While some showed patterns reminiscent of genetically defined collagens, many showed distinctive localization especially to cortical layers. We summarize these patterns by compartment:

Molecular architecture of apical ECMs and regulation of aECM expression

Among the C. elegans aECMs, the cuticle alone likely contains hundreds of components. We consider whether, comparable to basement membranes, the cuticle aECM may be dissected into ‘core’ components expressed at all or most stages and cuticle types, to which are added stage and region-specific elements. Of the genetically defined cuticle components, some were expressed in all stages, and may form a ‘core’ cuticle aECM to which additional stage or structure-specific components are added. These include components of annuli such as SQT-3, and several furrow collagens, consistent with the presence of annuli and furrows in all body cuticle types. However, our gene set was slightly biased toward collagens expressed in a subset of larval stages; of ~20 collagens expressed at all stages by RNAseq we have only tagged SQT-3, COL-12, and RAM-4/COL-34, which is found in interfacial cuticles.

Clear examples of stage-specific components include the L1 collagens DPY-14 and DPY-17, the L2-4 stage collagen COL-41, and the numerous ‘adult specific’ collagens. It should be clarified that stage-specific collagens are first expressed in the stage preceding the stage in which they are incorporated into matrix: i.e., many ‘adult specific’ collagens are first transcribed in L4, secreted in late L4, and form stable structures in the adult cuticle [24]. Our knock-in analysis was consistent with the L4 upregulation of four other collagens known to be targets of the LIN-29 transcription factor [75]. Some aECM transcripts are first detected in L4 and are present at higher levels in adults, whereas others are first detected in adults. Interestingly, some genes such as col-12 show an additional transcript peak early in adulthood [79], and our protein analysis was consistent with this observation. The mechanistic basis of the early adult peak is not known but could be related to larval transcript oscillations. Several proteins such as EPIC-1 [14] display distinct localization in adult versus L4 cuticle, suggesting the adult cuticle undergoes maturation well beyond the L4/adult molt.

We found several collagens that localize exclusively to the cortical compartment of the cuticle. In contrast, most genetically defined collagens localize to more basal or fibrous layers. The cortical compartment is an exciting area for investigation as it may contribute to surface properties involved in host-pathogen interactions or mate recognition [18]. Equally, the architecture of the fibrous layers is little-understood. Crossed fibrous arrays are integral to the strength of the hydrostatic skeleton in many nematodes and other animals [87], however, little is known about their biogenesis. Our markers reveal that the crossed fibrous arrays are variably angled with respect to the body axis, reminiscent of variable-angle composites that maximize load-bearing capacity [88]. Moreover, we identified collagens unequally distributed between the two fibrous layers. Aberrant patterning of such fibrous layers has long been hypothesized to underlie helically twisted ‘Rol’ mutant phenotypes [68]. Our resource provides tools to address how fiber angle is regulated and how the two fibrous arrays are laid down with consistent chirality [77].

The dauer cuticle is highly specialized, with unique structures such as a basal nanoscale mesh that may contribute to stress resistance. Transcriptomic studies show ~30 cuticle collagens are transcribed predominantly in dauer or predauer stages [89]. We did not prioritize such dauer-specific collagens as their low expression in nondauer stages made them more challenging for fluorescence-based screening. Two dauer-specific collagens tested, COL-43 and COL-51, indeed were only visible in dauer and/or predauer cuticle. Additionally, several proteins expressed in other stages showed distinctive localization in dauer larvae, e.g., COL-48 (dauer deirids) or LON-8 (dauer alae)(S13 Fig). An important future direction would be to focus on the dauer-specific aECM and our unc-119(+) pipeline may be best suited for these genes.

Collagen localization determinants and timing

Our analysis indicates that collagen localization in the aECM shows a high degree of specificity, in that almost every collagen knock-in showed a slightly different pattern. Most C. elegans cuticle collagens are 300–400 aa; after cleavage, mature cuticle collagens mostly consist of Gly-X-Y domains and short N- and C-terminal regions. Cuticle collagens have been classified in several ways, such as the patterns of interruptions in the Gly-X-Y domain and the presence of a signal peptide or transmembrane domain [25]. It remains unclear which sequence features might direct localization. Interestingly, of four tagged collagens classified in cluster D [25], three were cortical (Fig 7E). Our toolkit provides a resource for the dissection of determinants of collagen localization.

Timing of aECM protein expression could provide additional information. Within each larval stage aECM transcription occurs in waves [79], resulting in oscillating transcript levels [28] and presumably phased secretion of aECM components. Such observations have led to models that the cuticle could be built in successive layers, with early-peaking proteins forming outer layers and later-peaking proteins forming inner layers. Such models are an oversimplification given the early expression of precuticle proteins [10] and the lag between collagen secretion and aECM assembly in the embryo [62]. Nevertheless, our findings confirm and extend previous evidence that transcription phases correlate with cuticle compartment, for example, precuticle proteins and furrow collagen transcripts peak prior to basally localized collagen [50]. Interestingly, transcripts for cortically localized collagens peak either very early (col-80, col-98, and col-149) or very late (col-12, col-91, and col-93) in the molting cycle. One possibility is that these are all deposited in the cuticle early in the molt cycle, consistent with the simple ‘outside-in’ model. Alternatively, the early-peaking cortical collagens may form an initial outer cortical layer, and the later-peaking cortical collagens may be somehow inserted after formation of basal layers. Our embryo imaging data (Fig 8) are consistent with this model as the cortical layer collagen, COL-125, is expressed late in embryogenesis, well after the basal layer collagens DPY-10 and SQT-3. COL-10 also follows this trend (Fig 8), though it is predominantly cortical with weak basal and medial expression (Fig 7), highlighting the complexity of collagen localization. EM images of the nascent adult cuticle in the late L4 stage suggest that a basal layer is present before the cortical layer reaches its full thickness [14]. Thus, cortical layers may be assembled in two or more phases. Electron microscopy studies suggest nascent furrow regions in the cortical and basal layers are compressed together, although they later become widely separated in the adult due to development of the intervening medial layer. An important future direction facilitated by this toolkit will be to more precisely define the order of assembly of aECM compartments and their relationship to the phased transcription of aECM genes.

For the current toolkit, we prioritized collagens, in view of the available genetic reagents and our finding that most cuticle collagens tolerate C-terminal tagging. Collagens make up about half the C. elegans cuticle by mass [66]. Less is known about noncollagenous proteins such as cuticlins or epicuticlins, components of the outer cortical layers [14,90]. Tagging of Hedgehog-related proteins has also revealed specific localization to substructures of the cuticle or precuticle [19,91,92]. Using collagen localization as a reference, tagging of additional noncollagen aECM components such as ZP domain cuticlins or lectins should be highly informative. Although C. elegans, like other nematodes [93] has an especially large set of collagen genes, mammals also have numerous collagens (~28 types, ~44 genes) [94] whose localization mostly remains little-explored.

Optimization of CRISPR for large-scale knock-in projects

Our aECM toolkit establishes numerous lessons to establish large-scale tagging projects. Our fastest screening approach involved RNPs, PCR-derived linear repair templates, and screening under a fluorescent dissecting scope. We were most successful with genes expressed at high levels (>1,000 transcripts per million in a single-cell RNA-seq data set [95]) and in large tissues such as the epidermis. We detail failed edits in S1 Table. While we prioritized speed and hitting our target number of strains, revisiting these failed edits will be valuable for further refining our editing pipeline. Interestingly, neither dpy-10 co-conversion nor a dominant rol-6(su1006) co-injection marker helped enrich for aECM knock-ins, perhaps because such cuticle mutants display genetic interactions with aECM target genes. KIs with spatially or temporally restricted expression were much easier to identify with our selection-based approach.

Approximately 10% of the genetically defined cuticle collagen strains displayed overt or cryptic phenotypes, suggesting C-terminal tagging of most collagens is mostly tolerated. Some collagens, such as SQT-3 undergo C-terminal processing by the protease DPY-31 [72]. Although we avoided tagging SQT-3 at the C-terminus, other collagens with predicted C-terminal cleavage sites (see Materials and methods) appeared to tolerate tagging. Some collagens proved challenging to tag due to lack of unique crRNA sites (see Materials and methods). Our C-terminal QUA-1 knock-in was likely processed and an internal tagging attempt broke QUA-1. While our RNP and unc-119(+)-based selection approaches are high-throughput, SEC-based selection [96] with longer homology arms could be a useful alternate approach to tagging proteins that have highly similar homologs, which poses a challenge when using 35–65 bp homology arms.

Several groups have recently reported parallel efforts to optimize C. elegans protein-tagging knock-ins, either using CRISPR/Cas9 or recombinases such as FLP or phiC31. CRISPR/Cas9 has been used to generate KIs in 30 distinct genes with 3 different FP tags inserted per injection [97]. This approach has the potential to accelerate the workflow with a tradeoff of needing to use green, red, and blue FP KIs in each injection. Recent recombinase-based systems create an initial landing pad that can be efficiently re-edited [98,99]. While similar to our CRISPR-based modular knock-in approach, recombinases are not adversely affected by the size of the insertion, so can insert larger cargo, and additionally the recombinase methods allow tag swapping by genetic crossing. Ideally, a combination of an initial fluorescent knock-in carrying recombinase sites could allow functional initial strains combined with efficient modular re-editing.

Uses of the resource and prospects for future tagging resources

Our resource provides the first large-scale overview of aECM protein localization. Our resource can address key questions in aECM biology such as how the aECM is re-established in wound healing, how it is maintained during normal adult aging, how proteases and protease inhibitors coordinate its remodeling, and how specific aECMs are assembled in a cell type-specific manner [82]. Importantly, pathogenic variants [100] can now be modeled using proteins expressed at physiological levels. The multiscale patterning of the ECM remains a major challenge in mammalian biology. Protein-tagging knock-ins have been reported for a small number of mammalian collagens, primarily those of the basement membrane [101]. Our studies suggest tagging of other collagens and matrix components could reveal unanticipated aspects of ECM architecture and function. The human genome contains ~45 collagen-encoding genes in ~28 families; pathogenic variants have been found in at least 17 collagen families [102]. Imaging collagen patterning in normal and disease states may inform understanding of such collagen disorders.

The C. elegans matrisome includes ~700 members of conserved matrix families [25], including receptors. Other potential matrix components are not highly conserved in primary sequence but are conserved in terms of intrinsic disorder propensity or low sequence complexity. For example, the C. elegans pharyngeal cuticle contains numerous intrinsically disordered proteins [103], and epicuticlins consist solely of low complexity repeats [14,104]. Taking 700 as a lower bound of protein species in the C. elegans ECM, at most 25% have been tagged in this work and other projects [7]. The matrisome itself is a small fraction of the ‘secretome’ defined as all proteins normally secreted into extracellular compartments, ~3500 proteins in humans [105]. The distinction between matrisome and secretome proteins is not rigid, and proteins may form a continuum of stability or diffusibility within extracellular space. Localization of most secretome proteins has not been investigated using protein tagging, but has been proposed as a priority for genome-scale tagging [106]. Such approaches could be highly informative for understanding the architecture of the extracellular environment in development, adult homeostasis, wound healing, and aging.

Limitations of the study

Our toolkit is based on tagging endogenous loci with fluorescent protein tags, and thus conclusions are based on the localization of the tag itself rather than the endogenous protein. We detected an effect of tagging on gene function in a small minority of cases; however, it is possible that many more tags have subtle effects on function not detectable in the single mutants. Fluorescent tags could alter stability or localization of the tagged protein. In most cases, we used a single tag site and have not validated multiple sites. We have validated numerous knock-ins at the site of insertions but have not searched exhaustively for aberrant editing. Off-target effects of CRISPR/Cas9 editing in C. elegans are very rare, but complex alterations at the editing site are sometimes seen [107]. In any event, conclusions based on a single knock-in strain should be treated with caution.

C. elegans genetics

C. elegans were maintained at 20 °C on NGM plates seeded with E. coli OP50, using standard procedures unless stated. Mutations were confirmed by PCR or sequencing. Strains used in this study are listed in S4 Table.

Molecular biology

We designed flexible 39 bp 5′ and 3′ homology arms encoding N-SSSSSSGGSGGSA-C and N-AASSGSSSGGSGG-C linkers, respectively. These arms were added to the mNeonGreen::3xFLAG from pJW2172 [108] by PCR and the products were Gibson-cloned into pDONR221 to generate pJW2332. We also designed an mNeonGreen::AID*::3xFLAG equivalent for tagging proteins with nuclear or cytoplasmic domains using a pJW2171 template to generate pJW2333. These cassettes include 4 unique crRNA sites to allow re-editing. pJW2332 was linearized by PCR, introducing silent mutations in the PAM to disable the internal crRNA sites. We then cloned in SNAPTag and HaloTag cassettes to make pJW2335 and pJW2336, respectively. We also designed a set of mScarlet cassettes to replace the modular mNeonGreen knock-ins. First, we replaced the 3xMyc cassette from pJW2072 (30 amino acid linker::mScarlet(GLO)::3xMyc::10 amino acid linker) [108] with 3xHA and 3xOLLAS tags through Gibson cloning to make pJW2469 and pJW2470, respectively. We then linearized pJW2335 and cloned in the mScarlet(GLO)::3xHA to make pJW2471. We attempted the same with an mScarlet(GLO)::3xOLLAS cassette but were only able to obtain a plasmid with 2xOLLAS (pJW2558). To make a SapTrap compatible modular mNG cassette, we first Gibson-cloned the modular mNG::3xFLAG cassette with the 39 bp homology arms into pDONR221 with TCC and GGT connectors for SapTrap cloning to make pJW2474. We then linearized this plasmid by PCR and used Gibson cloning to insert a syntron+ C. briggsae unc-119(+) cassette from pMLS271 [44] such that the cassette is in an antisense orientation to the mNG cassette, yielding pJW2476.

dsDNA repair templates with 35 bp homology arms to the genome were generated for RNP-based genome editing by PCR amplification using DreamTaq (ThermoFisher). 8–12 50 µl PCRs were pooled, then purified using Ampure XP beads (Beckman Coulter) as described [32]. ssDNA repair templates with 35 nt homology arms were generated by PCR amplification similar to the dsDNA repair templates except that one primer was phosphorylated on the 5′ end. This phosphorylation was either performed during primer synthesis (Invitrogen) or in-house using T4 polynucleotide kinase (NEB). The phosphorylated strand was degraded using lambda exonuclease (NEB) and purified as previously described [37]. For some edits, ssDNA or 5′ biotinylated dsDNA repair templates drastically improved editing efficiency. Our cassette swaps particularly benefitted from using ssDNA or biotinylated DNA templates (Fig 2). Notably, we observed efficient editing with 35 nt homology arms with our ssDNA templates, whereas a previous report used 59–134 nt homology arms [37]. The shorter homology arms should substantially cut down reagent costs.

unc-119-selection based repair templates were generated by SapTrap cloning (S2 and S3 Figs) [44]. Oligos to generate a 20 bp sgRNA targeting sequence and 60 bp 5′ and 3′ homology arms were designed by SapTrap Builder [109]. We set the program to produce a C-terminal fusion and removed the “5′-CGC-3′” connector at the 5′ end of the hUS-B oligo with “5′-GGA-3′” to make this homology arm compatible with our modular toolkit. We performed similar SapTrap reactions as described [44,110] with minor modifications. To anneal the oligonucleotides, we added 1 μl of each 100 µM oligo stock (100 pmol/μl) to 8 μl of water and pipetted up and down to mix. We then added 0.5 μl of that dilution to 5 μl 10x annealing buffer (500 mM KOAc, 50 mM Tris pH 7.4) + 44.5 µl dH₂O. We then ran an annealing program on a thermocycler consisting of 95 °C for 5 min followed by dropping 1 ºC per minute down to 4 °C. Annealed oligos were stored at 4ºC before use. We made 10x SapTrap buffer by adding 2 µl of 2.5 M potassium acetate to 48 µl of 10x T4 ligase buffer. We thawed a fresh tube of 10x T4 ligase buffer and made ~50 µl aliquots. For both the 10x SapTrap buffer and 10x T4 ligase buffer we tally each freeze-thaw on the tube as part of our quality control pipeline. We made a 10x SapTrap enzyme mix by mixing 10 µl of SapI (NEB) (pipetting up and down repeatedly before transferring), 5 µl of T4 PNK (NEB) and T4 DNA ligase (NEB). If multiple stocks were present in the lab, they were marked and used oldest to newest. We then set up SapTrap reactions with ~20 fmol of pDD379 [111] (backbone with F + E sgRNA; ~50 ng), ~20 fmol of pJW2476 (modular mNG::3xFLAG+unc-119(+); ~75 ng) and 200 fmol of each annealed oligo pair (5′ homology arm, 3′ homology arm, sgRNA). While establishing an efficient editing pipeline we set up a control omitting pJW2476 and the annealed oligos. We ran the reactions on a thermocycler as follows: 37 °C for 10 min, 16 °C for 5 min followed by 30 cycles of 37 °C for 5 min, 16 °C for 5 min. There was a final 37 °C 10 min digestion followed by 65 °C for 20 min. To increase cloning efficiency, we added 1 µl of SapI (pipetting up and down before transferring) and incubated at 37 ºC for 2 hours followed by a 65 °C 20 min heat kill step. The entire reaction was transformed into DH10beta competent cells and plated on LB + Amp. We frequently screened reactions with colony PCR using M13-Rev (5′-CAGGAAACAGCTATGAC-3′) and 5′-AGCTCCTCGTATCCGTCGTTTG-3′. Colonies were picked into 50 µl of sterile water in a 12-well PCR strip or 96-well plate. 1 µl of this mixture was used as a template in a 10 µl PCR using homemade Phusion PCR and 5x Phusion HiFi buffer with green loading dye. We typically use a pJW2586 col-75 repair template as a positive control in colony PCR reactions.

Plasmids and oligonucleotides used in this study are detailed in S5 and S6 Tables, respectively.

CRISPR knock-ins

JDW strains were generated by the Ward lab using the following methods. crRNA, trcRNA (both from IDT), and Alt-R S.p. Cas9 Nuclease V3 (IDT 1081058) or Cas9-NLS purified protein (S. pyogenes WT and R691A, AsCas12a Ultra; UC Berkeley QB3 MacroLab) were mixed and incubated at 37 ºC for 15 min. Note that a range of protein and RNA concentrations were used in optimization (S1 Fig). Pre-optimization reactions used 700 ng of IDT Cas9 and a 2:5:2.5:1 ratio of crRNA:trcRNA:Cas9. Post-optimization reactions used 250 ng of IDT Cas9 and a 1:1:1 ratio of crRNA:trcRNA:Cas9. Repair template PCRs were melted as described [32] and used at 50 ng/µl. ssDNA repair templates were used at 110−250 ng/µl. We tested dpy-10 co-conversion and pRF4 co-injection markers using previously described approaches [31,38]. We also developed an alternate co-conversion approach using insertion of a T2A::GFP cassette at the 3′ end of myo-2 to create an endogenous promoter reporter [112]. Optimized editing used 10 ng/µl of a pSEM232 (mlc-1p::2xNLS::tagRFP::tbb-2 3’UTR) co-injection marker [113]. Animals were screened for mNG expression on a Leica M165 FC GFP/RFP stereomicroscope with 16.5:1 zoom ratio, 7.3-120x magnification range and GFP or GFP long pass filters, and an Excelitas X-Cite FIRE LED light source. For mSc color swaps, we similarly screened by RFP fluorescence using an mCherry filter. We generated 9 color-swapped strains as a proof of principle using a biotinylated linear dsDNA repair template and the internal crRNA sites. We chose this repair template as it boosted efficiency and we tended to get more injections out of a single prep compared to ssDNA. PHX strains were generated by SunyBiotech. We also performed 3 color swaps of PHX mNG knock-in strains.

unc-119(+)-based selection was performed by injecting repair template plasmids containing sgRNA (F + E) cassettes [111] at 65 ng/µl, 25 ng/µl pCFJ2474 [114] (smu-1p::Cas9(PATCs)::gpd-2 tagRFP-T(myr, PATCs)::smu-1 3’UTR), 10 ng/µl pSEM238 [115] (HisCl counter-selection marker), 25 ng/µl pSEM232 (mlc-1p::2xNLS::tagRFP::tbb-2 3’UTR) [113] into CFJ302, a temperature-sensitive unc-119(kst33) strain. We made use of a recently generated temperature-sensitive unc-119 allele which further facilitates adoption of this pipeline as labs can inject into animals with a WT phenotype. unc-119(kst33) mutants were maintained at the permissive temperature of 15 °C, at which they are phenotypically wild-type, prior to injection. After injection, animals were shifted to the nonpermissive temperature of 25 °C at which a fully penetrant Unc phenotype manifests, unless animals are rescued with unc-119(+). Plates were flooded with histamine when they starved and wild-type moving animals lacking the mlc-1p::tagRFP marker were picked as candidate knock-ins. The selective marker was excised through injection of 10 ng/µl pMDJ20 (smu-1p::Cre Recombinase (PATCs)::smu-1 3′ UTR]) [114], 10 ng/µl pCFJ782 [116] (HygroR marker), 25 ng/µl pSEM232. Unc animals were picked, outcrossed to remove unc-119(ts), and the knock-in recovered by either fluorescence or PCR. The spia-1:: mNG::3xFLAG and bli-5::mNG::3xFLAG knock-ins were generated using an SEC-based approach [50]. Modular mNG::3xFLAG knock-ins for bli-1 and noah-1 recapitulated independently generated knock-ins and have been described [117]. Reagents used for genome editing are detailed in S7 Table.

We observed an interesting variation in repair template efficiency that warrants further investigation. With our glh-1 GFP knock-ins, we did not observe a difference in efficiency with unmodified dsDNA, 5′ modified dsDNA, or ssDNA repair templates (S1C and S1D Fig). In contrast, for glh-1 mNG::3xFLAG to mSc::2xOLLAS color swaps the biotinylated dsDNA repair template was 10x more efficient and the ssDNA repair template was 34x more efficient than an unmodified dsDNA template, respectively (Fig 2B). As glh-1 is an efficient edit, modification might not further improve efficiency. Notably, unmodified dsDNA repair templates allowed efficient tag swapping with other combinations of crRNAs (Fig 2C). For large-scale tagging, we used unmodified dsDNA repair templates as they were cheaper than biotinylated repair templates and faster to produce than ssDNA repair templates. However, for smaller-scale editing or edits that prove challenging such alternative templates may be a better approach.

Sequence validation

All knock-ins were validated by PCR using primers that bind outside of the homology arms followed by Sanger sequencing. For knock-ins with unc-119(+) selection, we initially performed PCRs to amplify the 5′ and 3′ knock-in junctions and Sanger sequenced the PCR products. Following selectable marker excision, we PCR amplified the insertion and performed Sanger sequencing.

Editing efficiency and troubleshooting

Our approach was to avoid lengthy troubleshooting of genes that were not tagged in an initial attempt (‘failed’) and therefore estimates of efficiency across the gene set are to be taken with caution. In total, we had 23 failed RNP-based edits and 8 failed unc-119-based edits, with one gene in common between the two approaches (lrp-2; S1 Table). Our initial unc-119-based knock-ins (col-75, wrt-2) were successful, so future efforts will focus on optimizing plasmid amount, purity, and homology arm length. Among RNP-based edits, three putative knock-ins could not be confirmed by genotyping (col-90, col-155, cpz-2) and will require additional analysis. For certain collagens, the presence of near-identical paralogs likely precludes efficient editing (e.g., col-154 and col-155, or col-10, and col-144).

Collagen domain organization, sequence classification, and cleavage sites

Cuticle collagen genes are defined as encoding proteins with a central block of 37−43 Gly-X-Y triplet repeats (with one or more short interruptions) and additional motifs such as a dibasic subtilisin cleavage site (Homology box A), three Cys-rich clusters, and a signal peptide or transmembrane domain (see S1 File for annotated sequences of C. elegans collagens). Thirty-nine collagens have potential C-terminal cleavage sites resembling that of SQT-3 (DGG motif) including BLI-6, DPY-4, DPY-13, ROL-1, and LON-3. ROL-8 has been classified as non-cuticular [25] however, it has several hallmarks of cuticle collagens, including an N-terminal dibasic cleavage site. COL-135 has an atypically large region of ~200 triplets, most of which are Gly-X-Lys [25]. COL-55 has a short Gly-X-Y region and is not represented in RNAseq data; it contains a dibasic cleavage site and a Cys-rich motif. COL-70 lacks Gly-X-Y repeats but instead contains a repeated motif GGPG(N/T)P. We excluded col-182 as it is a pseudogene in the domesticated N2 strain [118]. We excluded known noncuticular collagens such as LET-2, EMB-9, CLE-1, MEC-5, and transmembrane proteins such as COL-99, UNC-122, and COF-2.

Whole–Genome sequencing, sanger sequencing, and editing of col-71

We performed whole–genome sequencing (WGS) of the BE22 strain, containing the roller allele sc22. Genomic DNA was prepared as described [119] and resequencing performed at BGI Americas (San Jose, CA). We inspected BE22 sequence-specific variants for genic changes on chromosome II relative to N2 wild-type. Sanger sequencing of alleles of genetically defined collagens used standard procedures. To edit the sc22 missense change in col-71 we used the melting method [32]. We injected N2 animals with 2 µM col-71 crRNA and 4 µM of repair template along with 50 ng/µl Pinx-6-RFP co-injection marker and Cas9 (UC Berkeley MacroLab). F₁ progeny were picked based on the co-injection marker and the ju2038 SNP mutation verified by sequencing. Positive edits were retrieved as heterozygotes; we picked Rol progeny and verified the mutation again by sequencing.

Imaging

We assessed overall strain brightness and localization using Leica MZ FLIII or M165 fluorescence dissection microscopes. For widefield and confocal imaging, we immobilized animals using 10 mM levamisole (Sigma) and imaged them on a Zeiss Axio Imager M2, a Zeiss LSM800 confocal, or a Zeiss LSM900 confocal with Airyscan capability, as described [14]. We varied laser power depending on the brightness of the specific knock-in strain. For the brightest strains, we reduced intensity by cutting the emission window. Confocal z stacks were generally made from 5–7 planes 0.3 µm apart to image the entire adult cuticle thickness. For imaging of the fibrous layers, we used z sections 0.17 µm apart. Airyscan speeds varied from 7 (fast) to 5 (slow); samples were first imaged using the fast setting, then more slowly for higher resolution as needed. Unless stated, animals were staged in mid L4 based on vulval morphology, then imaged 24 h later; images in Figures are representative of at least 10 animals imaged per strain. FRAP experiments were performed as described [14].

To measure the angles of fibers in the fibrous layer arrays with respect to the body axis, we examined maximum intensity projections of the midbody in day 1 adults. Angles were measured manually using line scans in Fiji at multiple points spaced 10 µm apart along the alae ridges (laterodorsal (LD) and lateroventral (LV)), and at multiple points 10 and 20 µm further dorsal or ventral (dorsolateral (DL) and dorsal (D); ventrolateral (VL) and ventral(V)).

Localization in L4 or adult males was generally analyzed in mNG/+ heterozygous cross progeny from crosses of N2 males with mNG/mNG hermaphrodites. For quantitation of knock-in expression intensity in heterozygous animals, we first crossed N2 wild-type males with mNG/mNG knock-in homozygotes, then crossed mNG/+ heterozygous F₁ males with N2 hermaphrodites. Fluorescence intensity was measured in at least 3 ROIs in the midbody lateral cuticle of F₂ animals 24 h post mid L4 stage. Background was subtracted and data normalized such that the homozygous strain mean = 1.

To image embryonic expression and localization, 100–150 embryos were picked from plates and mounted together on a 2% agarose pad in a droplet of 5–10 µl of M9 + 0.05% gelatin. Embryos were imaged on a Zeiss Axio Imager M2. We varied light exposure time depending on the brightness of the specific knock-in strain. A minimum of 20 embryos were observed for each developmental stage and knock-in strain.

Quantitation and statistical analysis

Statistical analyses were performed using GraphPad Prism 10. Data were tested for normality, and parametric or nonparametric tests were used as appropriate. Statistical significance was determined using Fisher’s exact test, unpaired t test or Mann–Whitney test. For multiple comparisons, we used one-way ANOVA or Kruskal-Wallis tests followed by a post test. Data are represented as mean ± SEM unless noted. Sample sizes are shown inside or above columns in Figures or legends.

S1 Fig. Optimization of knock-in editing parameters.

(A–D) Editing optimization using GFP::GLH-1 knock-ins as a test case. (A) Cas9 from the indicated sources at the indicated amounts with a 2.5:1 RNA:protein ratio. IDT Cas9 at 250 ng/µl was most efficient, similar to previous reports [31]. (B) The indicated RNA:protein molar ratios (crRNA and trcRNA at equal concentration, RNA amount varied to 250 ng/µl Cas9). (C and D) The indicated repair templates with 250 ng/µl Cas9, 1:1 RNA:protein ratio). (E) Number of GFP::GLH-1 knock-ins found in animals lacking (-) or containing (+) a mlc-1p::NLS::tagRFP co-injection marker. (F) Percent of mlc-1p::NLS::tagRFP co-injection marker positive animals laid on day and day 2 post-injection with myo-2::T2A::GFP knock-ins. (G) Number of mlc-1p::NLS::tagRFP co-injection marker positive animals laid on day and day 2 post-injection with myo-2::T2A::GFP knock-ins. (H) Number of GFP::GLH-1 knock-ins found in animals lacking (−) or containing (+) a myo-2::T2A::GFP co-CRISPR knock-in. Statistical significance was determined using a two-tailed unpaired Student t test. P < 0.05 was considered statistically significant. * Indicates P < 0.05, ** indicates P < 0.005. Two independent experiments were performed for all optimization tests. Each dot in the graphs A–D and F–H represents the parameter scored on the Y-axis in the F1 of a singled, injected P0 animal and the line indicates the mean. See S2 Data for the data underlying the graphs.

https://doi.org/10.1371/journal.pbio.3004059.s009

(PDF)

S2 Fig. An unc-119(+)-selection based modular mNeonGreen::3xFLAG knock-in pipeline.

(A) A schematic of the knock-in cassette with the C. briggsae unc-119(+) rescue cassette being transcribed from the opposite strand as the mNeonGreen::3xFLAG cassette. The selection cassette is flanked with loxP sites and is within an intron. Knock-ins rescue the uncoordinated phenotype of unc-119 mutants and the cassette can be excised through Cre expression. (B) Strains generated with the unc-119(+) knock-in pipeline. Our RNP-based pipeline was most efficient with genes expressed in larger tissues (hyp7 syncytium, seam cells, pharyngeal cells) and at high levels (>1,000 transcripts per million in a single cell L2 stage RNA-seq dataset [95] as these criteria allowed facile screening under a fluorescence dissecting microscope. unc-119(+) selection is more suitable for high-throughput pipelines than alternatives such as self-excising cassettes, as the 60 bp homology arms can be made using oligo annealing, bypassing need for PCR. We developed a robust assembly pipeline and successfully tagged two collagens (col-75 and col-97) and one warthog gene (wrt-2) as proof-of-principle. We chose these test cases as: i) col-75 is predicted to be restricted to socket cells and excretory duct or pore cells at moderate levels; ii) an RNP-based attempt at col-97 tagging had failed; and iii) wrt-2 was expressed at moderate levels.

https://doi.org/10.1371/journal.pbio.3004059.s010

(PDF)

S3 Fig. A SapTrap assembly pipeline for unc-119(+)-selection based modular mNeonGreen::3xFLAG repair templates.

(A) Oligos encoding the target sgRNA, the 5′ and 3′ homology arms (HA) are annealed to produce the indicated 3 bp overhangs and combined with the pJW2476 vector containing the modular mNeonGreen::3xFLAG unc-119(+) cassette and an appropriate backbone (such as pDD379) in a SapTrap cloning reaction. (B) lists the settings used in SapTrap Builder. We set cloning for SapTrap, output as oligos + annotated, and 60 nt oligos for the upstream (hUS) and downstream (hDS) homology arms. (C) is a representative output for tagging col-75. The red box indicates the change to the output needed for a custom 3 nt overhang needed to use the pJW2476 cassette. For the hUS-B oligo the 5′-CGC-3′ at the 5′ end nees to be replaced with “5′-GGA-3′

https://doi.org/10.1371/journal.pbio.3004059.s011

(PDF)

S4 Fig. aECM markers in non-epidermal tissues.

Representative examples of mNG knock-in localization in nonepidermal tissues. (A) Pharynx (KUND-7 (K10D3.4)::mNG). Widefield imaging, embryo to L4 stages. Nontransgenic control N2 is shown above (B) Gut (CTSA-1.1::mNG). Widefield and Airyscan confocal images. Expression in the gut/ intestine is marked with yellow dashed lines; expression in the coelomocytes is marked with red arrows; expression in the muscles is marked by blue arrows. (C) Coelomocytes (CTSA-1.1::mNG). Confocal images of anterior, midbody and posterior coelomocytes. (D) Somatic gonad (SKPO-1::mNG). Confocal images of anterior and posterior gonad arms. Scales, 10 µm.

https://doi.org/10.1371/journal.pbio.3004059.s012

(PDF)

S5 Fig. Stage specificity of collagen knock-in expression.

(A) DPY-4::mNG, (B) DPY-5::mNG, and (C) DPY-13::mNG are not detectable in late embryos; fluorescence was first detectable in late L1 and remains visible through L4 stage. Widefield DIC and fluorescence. Autofluorescent gut granules are visible in the DPY-5::mNG L1 (yellow arrows). Scales, 10 µm.

https://doi.org/10.1371/journal.pbio.3004059.s013

(PDF)

S6 Fig. Knock-in dosage and mobility of fusion proteins.

(A) mNG fluorescence intensities in the midbody cuticle from homozygous and heterozygous strains, dot plots of intensities normalized to controls (see Materials and methods). Intensities of heterozygous strains were doubled to correct for copy number and compared to the homozygous strain intensity by Mann–Whitney test; *** P < 0.001. See S3 Data for the data underlying the graphs. (B) Fluorescence Recovery After Photobleaching (FRAP) of selected KIs in the body and vulval cuticles. Most mNG fusion proteins did not show significant recovery, apart from LON-8::mNG in the vulval cuticle. See S4 Data for the data underlying the graphs.

https://doi.org/10.1371/journal.pbio.3004059.s014

(PDF)

S7 Fig. Furrow and strut localization.

Single Airyscan focal planes showing dual localization to furrow and struts. COL-81::mNG localized to the midline of each furrow in cortical planes (S6) and to strut puncta in basal planes (S9). COL-103::mNG localized to furrow-flanking stripes in cortical planes (S6-8) and to strut-like puncta in basal planes (S9). Scales, 5 µm.

https://doi.org/10.1371/journal.pbio.3004059.s015

(PDF)

S8 Fig. Strut localization.

Single Airyscan focal planes of strut-related localizations. COL-98::mNG showed diffuse localization in more cortical planes (S9-10) and punctate strut localization in medial planes (S11). COL-93::mNG showed ring-like localization in multiple focal planes. Scales, 5 µm.

https://doi.org/10.1371/journal.pbio.3004059.s016

(PDF)

S9 Fig. COL-80::mNG localization.

(A) COL-80::mNG localization in adults (L4 + 24 h). COL-80::mNG localized to furrows and annuli in the L4 stage; additional strut-like puncta developed in adults. (B) In bli-1 mutants COL-80::mNG failed to localize to struts and remained in furrows and annular ridges on the cortical surface of blisters. Scales, 10 µm.

https://doi.org/10.1371/journal.pbio.3004059.s017

(PDF)

S10 Fig. Crossed fibrous array collagens and localization within the fibrous layer.

(A) Airyscan single focal plane images of COL-38::mNG and COL-138::mNG showing crossed fiber arrays in young adults. Scale, 10 µm. (B) Airyscan z-stacks (11 x 0.17 µm) and single focal planes of COL-20::mNG showing localization to outer (right-handed) oriented fibrous layer and basal fibrillar layer. Stacks from anterior body, midbody, and posterior body. Scales, 5 µm. Cartoons of COL-20::mNG localization in different focal planes.

https://doi.org/10.1371/journal.pbio.3004059.s018

(PDF)

S11 Fig. Collagen localization in blistered mutant adults.

Confocal images of collagen KIs in Bli mutant adults. Lateral views of head or midbody cuticle; anterior to the left and dorsal up. Blister cortical surfaces outlined with pink dashed lines, basal surfaces outlined by red dashed lines. (A) LON-3::mNG, DPY-13::mNG and ROL-8::mNG were basal to blisters. (B) COL-63::mNG, COL-125::mNG, and COL-130::mNG were cortical to blisters. (C) COL-20::mNG was primarily basal with faint localization to cortical surface. Patterning of the markers was otherwise unaffected in the blistered background. Scales, 10 µm.

https://doi.org/10.1371/journal.pbio.3004059.s019

(PDF)

S12 Fig. Double label imaging of collagens.

(A) DPY-5::mSc DPY-9::mNG double label knock-in strain. (B) DPY-5::mSc DPY-10::mNG double label. (C) DPY-10::mSc ROL-6::mNG double label. (D) ROL-6::mNG; DPY-5::mSc double label. Images are from L4 or day 1 adult as indicated. Left, single focal planes (0.2 µm slices) from cortical to basal. Scales 10 µm. Right, Magnified xz slices generated using Zeiss Zen 3D projection tool; z axis is exaggerated to illustrate differences in z plane. Scale in x axis, 2 µm.

https://doi.org/10.1371/journal.pbio.3004059.s020

(PDF)

S13 Fig. Localization patterns in adult lateral alae and in dauer stage larvae.

(A) Selected examples of adult alae localization patterns, showing localization to alae ridges (yellow arrows), ridge flanks (red lines), and valleys. Confocal Airyscan projections; scale, 5 µm. (B) L1 alae localization of DPY-14::mNG (yellow arrow) compared to exclusion from alae of DPY-9::mNG. (C) Dauer alae localization of COL-43::mNG, COL-103::mNG, COL-48::mNG, and LON-8::mNG. Scales, 10 µm.

https://doi.org/10.1371/journal.pbio.3004059.s021

(PDF)

S14 Fig. Localization of the atypical collagen COL-135:mNG.

Images of anterior spermatheca (left), uterus/eggs (middle), and posterior spermatheca (left) in L4 and young adult stages. (A) COL-135::mNG was expressed in spermathecal cells beginning in late L4 stage (L4.7 shown). (B) In young adults COL-135::mNG localized to puncta in spermathecal cells and/or the spermathecal lumen, and to uterine matrix surrounding newly fertilized eggs. COL-135::mNG was often deposited on eggs in extended folds or furrows on the egg surface (yellow arrows) and appeared to be retained on the eggshell after egg laying. Images are maximum intensity projections of z stacks of 23–49 x 0.5 µm slices. (C) Fluorescence and brightfield images of spermathecal and uterine expression in young adults. Scales, 10 µm.

https://doi.org/10.1371/journal.pbio.3004059.s022

(PDF)