A collagen IV fluorophore knock-in toolkit reveals trimer diversity in C. elegans basement membranes

Basement membranes (BMs) are dense, sheetlike extracellular matrices that surround and support most tissues (Fidler et al., 2018; Jayadev and Sherwood, 2017a). Type IV collagen is a triple helical protomer and a core BM component that forms a self-associating network in BMs. The type IV collagen BM network confers resistance to mechanical loads placed on tissues (Fidler et al., 2018; Keeley et al., 2020; Morrissey and Sherwood, 2015), and type IV collagen genetic loss leads to embryonic lethality at the onset of muscle contraction in mice, Drosophila, and Caenorhabditis elegans (Borchiellini et al., 1996; Gupta et al., 1997; Pöschl et al., 2004). Precise deposition and remodeling of type IV collagen is also required to properly shape tissues, such as the Drosophila egg chamber, central nervous system, and wing disk, and the mouse salivary gland (Crest et al., 2017; Harunaga et al., 2014; Isabella and Horne-Badovinac, 2016; Morrissey and Sherwood, 2015; Pastor-Pareja and Xu, 2011). Additionally, type IV collagen networks serve as scaffolds that bind signaling ligands and other BM components that direct cell differentiation and guide cell migration (Brown et al., 2017; Morrissey and Sherwood, 2015; Parkin et al., 2011; Sherwood, 2021; Wang et al., 2008). Further emphasizing its importance to tissue function, pathogenic variants in type IV collagen are linked to at least 10 human genetic disorders that affect muscle, kidney, brain, and vasculature (Fidler et al., 2018). Additionally, increased type IV collagen BM accumulation is associated with the decline of vascular function in diabetes, in Alzheimer’s disease, and during aging (Baum and Bigler, 2016; Candiello et al., 2010; Fidler et al., 2018; Howe et al., 2020; Karttunen et al., 1986; Marshall, 2016; Thomsen et al., 2017; Uspenskaia et al., 2004).

The triple helical type IV collagen protomer is composed of three ⍺-chains, which have a cysteine-rich 7S domain at the N terminus, a central interrupted Gly-X-Y repeat region (X and Y can be any amino acid (aa), but is most often proline), and a noncollagenous NC1 domain at the C terminus (Hudson et al., 1993, 2003). Reconstitution studies have shown that the NC1 domain mediates α-chain selection for type IV collagen trimer assembly, which occurs in the endoplasmic reticulum (ER) (Boutaud et al., 2000; Dölz et al., 1988; Fidler et al., 2018). In vitro studies have also revealed that lysyl and prolyl hydroxylases, along with the cofactor vitamin C, are needed for proper folding and trimerization (Kivirikko and Pihlajaniemi, 1998; Lamandé and Bateman, 1999). Depleting prolyl 4-hydroxylase in Drosophila larvae results in the secretion of type IV collagen in its monomeric form and an absence of BM localization, suggesting BM incorporation requires trimerization (Pastor-Pareja and Xu, 2011). Upon secretion into the extracellular space, two type IV collagen protomers connect through the C-terminal NC1 domains and four protomers connect through the N-terminal 7S domains (Añazco et al., 2016; Cummings et al., 2016). The interfaces of the NC1 and 7S connections are reinforced with covalent crosslinks (Cummings et al., 2016; Vanacore et al., 2009). Together, the NC1 dimer and 7S tetramer connections establish a polygonal patterned network of type IV collagen in the BM. Electron microscopy studies have also revealed higher order lateral twisting interactions between neighboring protomers, which are thought to provide further stability to type IV collagen networks (Barnard et al., 1992; Ruben and Yurchenco, 1994; Yurchenco and Ruben, 1987).

Based on sequence similarity, type IV collagen α-chains have been divided into α1-like and α2-like groups. Most vertebrates have six α-chain–encoding genes (COL4A1–COL4A6 genes encoding the α1–α6 proteins). The α1-like chain group contains COL4A1, COL4A3, and COL4A5 genes, and the α2-like group includes COL4A2, COL4A4, and COL4A6 (Khoshnoodi et al., 2008). C. elegans and Drosophila melanogaster each possess two type IV collagen α-chains, encoded by emb-9 and let-2 in C. elegans and vkg and Cg25C in Drosophila (Guo and Kramer, 1989; Natzle et al., 1982; Sibley et al., 1993; Yasothornsrikul et al., 1997). Sequence comparisons with mammalian NC1 domains have suggested that emb-9 and vkg are more similar to ⍺1-like chains and let-2 and Cg25C are more similar to ⍺2-like chains (despite Cg25C officially being named Col4a1) (Guo and Kramer, 1989; Summers et al., 2023).

Chromatography-based protein isolation studies in vertebrates have revealed approximately twice as much α1 compared with the α2 chain in several tissues (Mayne and Zettergren, 1980; Siebold et al., 1988; Trüeb et al., 1982), which is consistent with the existence of the α1α1α2 heterotrimer. Characterization of type IV collagen hexamers (the NC1 domains of two cross-linked protomers) isolated from in vivo tissues has also supported the existence of α1α1α2, as well as α3α4α5 and α5α5α6 type IV collagen trimers (Borza et al., 2001; Boutaud et al., 2000; Khoshnoodi et al., 2008). These and other biochemical studies have led to the proposition that only three (out of a possible 56) specific combinations of vertebrate type IV collagen are assembled in vivo—α1α1α2, α3α4α5, and α5α5α6—with α1α1α2 being the most abundant and ubiquitous (Borza et al., 2001; Boutaud et al., 2000; Hudson et al., 1994, 2003; Khoshnoodi et al., 2008).

While there is strong support for the preferential formation of α1α1α2, α3α4α5, and α5α5α6 type IV collagen trimers in mammalian tissues, there is also evidence that other combinations could exist. For example, isolation and characterization of type IV collagen hexamers have not ruled out the existence of α1 homotrimers or α1α2α2 heterotrimers (Boutaud et al., 2000; Johansson et al., 1992; Siebold et al., 1988). Indeed, reconstitution experiments using bovine and recombinant NC1 α1-α5 monomers have found that α1 NC1 domains can form a homohexamer and that the α2 NC1 can assemble in a promiscuous manner with other NC1 domains (Boutaud et al., 2000). Further, rat carcinoma cells generate and secrete α1 homotrimers (Haralson et al., 1985). No studies have yet addressed the overall α-chain levels and possible trimer combinations found in C. elegans. Mutations in the emb-9 or let-2, however, result in intracellular retention of the other (nonmutant) α-chain (Gupta et al., 1997), suggesting a preference for heterotrimer assembly prior to secretion. The recombinant expression of forced predetermined NC1 trimers in Drosophila larvae has shown that only the Cg25C-Cg25C-Vkg NC1 heterotrimer combination is recruited to the BM. Yet, NC1 domains expressed as individual monomers in mammalian cells assemble in almost all possible combinations. Additionally, analysis of endogenous hexamers did not rule out the formation of homotrimers (Summers et al., 2023). Together, these observations suggest that flexibility in the α-chain composition of type IV collagen trimers might exist in BMs in vivo.

Endogenous tagging of BM components with genetically encoded fluorophores allows quantitative comparisons of component abundance at the tissue, cell, and subcellular level (Jayadev et al., 2022; Keeley et al., 2020). A key challenge for tagging type IV collagen is identifying a fluorophore insertion site common to all α-chains that does not disrupt type IV collagen assembly and function. To date, endogenous tagging in mice, Drosophila, and C. elegans has predominantly used the Drosophila Vkg-GFP protein trap insertion site, which inserts GFP within the start of the 7S domain at the N terminus (Morin et al., 2001). While insertion of fluorophores at this site is viable for Drosophila vkg, C. elegans emb-9, and mouse COL4A2 (Keeley et al., 2020; Wuergezhen et al., 2025), insertion at this site is not viable in C. elegans let-2 or mouse COL4A1 (Jones et al., 2024; Keeley et al., 2020). A site within an interruption of the Gly-X-Y repeats in the collagenous domain of C. elegans emb-9 has also been reported to be viable (Matsuo et al., 2019), but the health of this strain has not been fully explored.

Here, we use genome editing, fitness assays, and live-cell imaging to compare sites for fluorophore insertion into the C. elegans emb-9 and let-2 type IV collagen α-chains. We show that fluorophore insertion at the Vkg-GFP site (renamed internal site 1, IS1) and an internal site within the collagenous domain of EMB-9 (internal site 2, IS2) lead to accumulation of the tagged α-chains in the ER and reduced organism fitness. In contrast, we use multiple fluorophores and show that insertion at the C terminus of EMB-9 and LET-2 proteins leads to dramatically less intracellular retention and phenotypically wild-type fitness. Using these new strains with mass spectrometry on whole animals, we reveal there is approximately twofold more LET-2 protein than EMB-9, suggesting predominantly a LET-2-LET-2-EMB-9 (LLE) heterotrimer. However, using quantitative live imaging of mNeonGreen (mNG)-tagged strains, we find that the LET-2/EMB-9 protein ratio present at various BMs is not precisely 2:1, but instead is ∼1.7 on most tissues, and at the spermathecal BM is over 2.2. These results suggest a preference for the LLE heterotrimer, but that there is also flexibility in α-chain trimer makeup. Using tissue-specific translational reporters, RNAi knockdown, and overexpression, we provide evidence suggesting that α-chain availability during assembly can drive trimer diversity. Finally, we show that the C-terminal site for type IV collagen allows endogenous tagging of type IV collagen variants found in the COL4A1/A2 syndrome (also known as the Gould syndrome), a genetically dominant heterogeneous multisystem disorder predominantly caused by substitutions of glycine aa in the Gly-X-Y motifs of the triple helical domain (Jeanne and Gould, 2017; Labelle-Dumais et al., 2024). Through localization and photoconversion analysis, we find that in addition to defects in type IV collagen secretion, these mutant forms affect extracellular accumulation and turnover. Together, our findings identify a permissive site for tagging type IV collagen α-chains in C. elegans, which will allow wide-ranging studies.

The inability to comprehensively label and visualize type IV collagen α-chains in animals has limited our understanding of type IV collagen regulation, function, and composition in the BM. Fluorophore insertion at the C terminus through genome editing has successfully labeled 37 of the 56 BM components tagged in C. elegans to date, but has not yet been attempted for the type IV collagen α-chain–encoding genes emb-9 and let-2 (Fan et al., 2020; Hastie et al., 2019; Jayadev et al., 2022; Keeley et al., 2020; Naegeli et al., 2017). Thus, we knocked in genetically encoded fluorescent proteins with an 18-aa linker within the endogenous emb-9 and let-2 genes at the C terminus (C site) of the protein-encoding regions using CRISPR/Cas9 gene editing (Fig. 1 and Fig. S1 A). The resulting gene-edited animals were homozygous-viable and fertile. To compare the health of these new strains to animals with fluorescent proteins inserted at IS1 (Drosophila Vkg-GFP site) and IS2 of type IV collagen (Fig. 1), we plated four L4 larvae from each strain onto a predetermined amount of food and recorded how long it took for the food to be depleted (time to starvation). This served as a collective estimate of growth, feeding, and fertility. Eight strains with six distinct knock-ins at the C site of emb-9 and let-2 were analyzed: emb-9 and let-2 each tagged with mNG or mRuby2, strains harboring both let-2 and emb-9 tagged with mNG and mRuby2 and vice versa, emb-9 tagged with photoconvertible mEos2, and let-2 tagged with photoconvertible mMaple (Heppert et al., 2016; McEvoy et al., 2012; Zhang et al., 2012) (Fig. 1; and Fig. S1, A and B). All strains were indistinguishable from wild-type animals, apart from let-2 tagged with mMaple, which showed a slight delay in the time to starvation (Table 1). In contrast, emb-9 strains tagged at IS1 (mRuby2, mScarlet-I, mEos2, mNG) and IS2 (mCherry) (Bindels et al., 2017; Heppert et al., 2016) all showed delays to starvation (Table 1; Fig. 1; and Fig. S1, A and B). It is notable that all fluorophore insertions at IS1 (Vkg site), which has been used most extensively to label type IV collagen (Keeley et al., 2020; Morin et al., 2001; Wuergezhen et al., 2025), caused delayed growth.

We next wanted to compare protein trafficking and localization of the type IV collagen tagged in distinct sites. Type IV collagen is expressed in several tissues in C. elegans, but the predominant site of production is the C. elegans body wall muscle, where type IV collagen is made, secreted into the BM, and also trafficked from the body wall muscles to many other tissue BMs (Graham et al., 1997; Morrissey et al., 2016). Confocal imaging of L1-staged larvae (0–12 h after hatch) showed large and extensive body wall muscle intracellular aggregates of EMB-9 tagged at IS1 with mNG, mEos2, mRuby2, or mScarlet-I, and aggregates of EMB-9 tagged at IS2 with mCherry (Fig. 2 A and Table S1). In contrast, EMB-9 and LET-2 tagged with mNG or mRuby2 at the C site possessed few if any aggregates (Fig. 2 A and Table S1). Previous immunolocalization studies also found limited intracellular aggregates of EMB-9 and LET-2 in wild-type L1 animals (Graham et al., 1997; Kang and Kramer, 2000), suggesting tagging at the C site permits normal intracellular trafficking. Imaging of worms tagged at the C site from the L1 through L4 stage and into the gravid adult revealed the gradual appearance of small type IV collagen aggregates starting at the L3 larval stage that continued to the adult (Fig. S2 A). Given the wild-type health of these strains, this might represent natural intracellular accumulation.

Type IV collagen folding and trimerization occur in the ER, which has quality control mechanisms that retain misfolded proteins (Ellgaard and Helenius, 2003). We thus hypothesized that the type IV collagen aggregates found in strains tagged at IS1 and IS2 were likely localized to the ER. Colocalization of type IV collagen aggregates tagged at IS1 (EMB-9::mScarlet-I [IS1]) with an ER marker (ELO-1::mNG [Park et al., 2024]) revealed that the EMB-9::mScarlet-I (IS1) aggregates were always enclosed by the ER membrane (Fig. 2 B). Similarly, EMB-9::mCherry (IS2) aggregates were also contained within the ER (Fig. S2 B). This suggests that fluorophore insertion at IS1 or IS2 interferes with proper protein folding or trimerization, leading to ER accumulation and possibly reduced secretion. Consistent with this idea, measurements at the pharyngeal BM in L1 larvae showed that less EMB-9::mNG (IS1) localized to the BM compared with EMB-9::mNG (C) (Fig. 2 C). Decreased EMB-9::mNG (IS1) in the BM might also be caused by changes in the rates of addition to the BM, removal from the BM, or both (often referred to as turnover) (Stramer and Sherwood, 2024). However, fluorescence recovery in a photobleached region of EMB-9::mNG (IS1) compared with EMB-9::mNG (C) and LET-2::mNG (C) revealed similar recovery dynamics: LET-2::mNG (C) recovered ∼32%, and both EMB-9::mNG (C) and EMB-9::mNG (IS1) recovered ∼25% of bleached fluorescence over a 5-h period (Fig. S2 C). Taken together, the wild-type health and greatly reduced intracellular aggregation offer compelling evidence that the C terminus is a permissive site for fluorescently tagging type IV collagen α-chains in C. elegans.

We next wanted to determine the overall levels of EMB-9 and LET-2 proteins in the extracellular matrix of whole animals using the C site–tagged strains as material for liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). Comparisons of abundance of distinct proteins using LC-MS/MS are usually challenging because differences in the physical and chemical properties of distinct peptides can affect processing and detection (Bantscheff et al., 2007; Xie et al., 2011). However, the tagging of LET-2 and EMB-9 with the same fluorophore allowed us to use the fluorophore proteins as a proxy of LET-2 and EMB-9 abundance to determine their levels using label-free quantitative proteomics. We collected mixed populations of L1-L3 larval–staged worms expressing the knock-in proteins EMB-9::mRuby2 (C); LET-2::mNG (C) (Strain 1) and EMB-9-mNG (C); LET-2::mRuby2 (C) (Strain 2) and chemically fractionated them to enrich for extracellular matrix proteins as described previously (Morais et al., 2022, 2023). Measuring the protein abundance of mRuby2 in the extracellular matrix–enriched fraction of these strains using LC-MS/MS revealed an approximately twofold higher level of LET-2 compared with EMB-9 (Fig. 3 A). mNG was not detected in the matrix fraction, which may be due to differing trypsin susceptibilities between the two fluorophores since mNG is only 27% similar to mRuby2 (Fig. S1 B). Nonetheless, we found that in the cellular fraction, the levels of mNG indicated LET-2 was ∼1.7-fold higher than EMB-9::mNG (Fig. S3 A). Similar results were observed in the cellular fraction for LET-2 and EMB-9 tagged with mRuby2 (Fig. S3 B). Finally, comparison of EMB-9 with EMB-9 and LET-2 with LET-2 protein levels between Strain 1 (EMB-9::mRuby2 [C]; LET-2::mNG [C]) and Strain 2 (EMB-9-mNG [C]; LET-2::mRuby2 [C]) matrix fractions confirmed that the protein abundance between the two strains was comparable (Fig. S3, C and D). Taken together, these studies reveal that LET-2 is present at approximately twice the levels of EMB-9 in larval stage C. elegans.

The 2:1 whole-body protein ratio of LET-2/EMB-9 suggested an LLE heterotrimer combination is assembled in the ER and is present in BMs (Fig. 3, B and C). We thus next wanted to use our endogenously tagged EMB-9 and LET-2 strains to determine whether a 2:1 LET-2/EMB-9 ratio is indeed present in BMs. We measured the average fluorescence intensity of LET-2::mNG (C) and EMB-9::mNG (C) in a specific region of the BM in the pharynx (feeding organ), the body wall muscle, the distal tip cell (DTC, the cell that leads gonad migration), and the spermatheca (where oocyte fertilization and ovulation occur) (Fig. 4 A) (Gieseler et al., 2017; Keeley et al., 2020; Kelley and Cram, 2019; Sherwood and Plastino, 2018). The ratio of LET-2 to EMB-9 across the BMs of these tissues in L4.8–L4.9 worms (1-h window in larval development) (Mok et al., 2015) revealed an average ratio of ∼1.7 within the pharynx, muscle, and DTC, but a significantly greater ratio of ∼2.2 in the spermathecal BM (Fig. 4, A and B; and Fig. S4 A). While all tissue BMs had a LET-2/EMB-9 ratio close to 2:1, strongly suggesting the presence and perhaps a preference for the LLE trimer, ratios of LET-2 to EMB-9 lower than 2:1 indicated the existence of LEE or EEE or both trimers. Further, the ratio of >2:1 in the spermathecal BM implied the presence of LET-2 LLL homotrimers (Fig. 4 C).

To confirm the validity of these measurements, we assessed both the stability of fluorophore association with EMB-9 and LET-2 C termini and the possibility of fluorophore self-quenching, as type IV collagen networks associate end to end at the C terminus (NC1 domain) possibly placing multiple fluorophores in close proximity (Bae et al., 2021). Western blot analysis using mNG antibodies confirmed that EMB-9 and LET-2 C-terminal mNG fusion proteins were present almost exclusively as full-length proteins (Fig. S4 B). To test for quenching, we reasoned that reducing the number of fluorophores by half in heterozygous animals would reduce quenching (if present) and result in >50% fluorescence of the homozygous strain. Fluorescence intensity measurements of EMB-9::mNG (C) and LET-2::mNG (C) L4.8- to L4.9-staged heterozygous worms, however, were not >50% of the levels of homozygous animals (Fig. S4 C), suggesting that fluorescence quenching was not occurring in homozygous animals. We conclude that mNG is stably associated with C-terminally tagged EMB-9 and LET-2 and that C-terminal–inserted mNG does not show quenching in the BM. Further, the levels of LET-2/EMB-9 are close to 2:1 for all tissue BMs, consistent with a predominant LLE heterotrimer composition. However, the diversity in α-chain ratios also suggests some flexibility in the α-chain makeup of trimers.

We hypothesized that differences in levels of translation (either from differences in transcription or from post-translational regulation) could alter the availability of LET-2 and EMB-9 α-chains for trimerization in the ER and influence the tissue-specific differences in α-chain ratios at the BM. We thus used CRISPR/Cas9 to insert a P2A peptide sequence followed by PEST::mNG immediately before the emb-9 and let-2 terminal codons (emb-9::P2A::PEST::mNG and let-2::P2A::PEST::mNG). The P2A peptide causes the ribosome to fail in making a peptide bond between the glycine and proline near the C terminus of the 2A peptide during translation, resulting in the separate synthesis of EMB-9 or LET-2 and mNG::PEST from the same mRNA (Lartey et al., 2024; Sharma et al., 2012). The PEST-destabilized mNG that remains within the cell cytoplasm can be used as a sensitive reporter of active EMB-9 and LET-2 translation levels (Li et al., 1998; Stevenson et al., 2023). We examined the average fluorescence intensity of PEST::mNG in the cytosol of three tissues that express and secrete type IV collagen—the body wall muscle, spermatheca myoepithelium, and the DTC (Fig. 5 A) (Graham et al., 1997). The ratio of LET-2 to EMB-9 protein translation in the body wall muscle and DTC was ∼1.5 (Fig. 5 B and Fig. S4 D), similar to the ratio of the LET-2 and EMB-9 proteins in the body wall muscle and DTC BMs (Fig. 4 B and Fig. S4 A). In contrast, the ratio of LET-2 to EMB-9 protein translation in the spermatheca was ∼2.5 (Fig. 5 B and Fig. S4 D), which mirrored the higher LET-2/EMB-9 protein ratio specifically found in the spermathecal BM (Fig. 4 B and Fig. S4 A). Together, these findings support the idea that the levels of α-chains available for trimerization influence the α-chain composition of trimers assembled and then incorporated into BM.

Our results suggest that although the ratio of LET-2 to EMB-9 is close to 2:1 in tissue BMs, and thus likely predominantly contains an LLE trimer, there is also flexibility in the α-chain composition of trimers assembled and incorporated into the BM based on α-chain availability during trimer assembly. We next wanted to directly test both the preference for heterotrimer formation and whether changing α-chain levels is sufficient to alter the ratios of α-chains deposited in the BM. We first reduced emb-9 expression in EMB-9::mNG (C) and LET-2::mNG (C) knock-in worms by feeding RNAi targeting emb-9 beginning at the L1 larval stage (Jayadev et al., 2019). We then measured the mean fluorescence intensity of EMB-9::mNG and LET-2::mNG in the body wall muscle BM of day 1 adults (72 h post-RNAi initiation). RNAi targeting of emb-9 resulted in a 75% loss of the EMB-9 protein at the body wall muscle BM (Fig. 6, A and B; and Fig. S5 A). LET-2 levels in the BM were decreased by ∼50%, and there was increased internal aggregation of LET-2::mNG in body wall muscle (Fig. 6, A and B; Fig. S5, A and C; and Table S1). The reduction in LET-2 at the BM and increase in intracellular accumulation are consistent with a preference for heterotrimer assembly. Yet, a greater proportion of LET-2 than EMB-9 was incorporated into the BM—the LET-2/EMB-9 ratio went from ∼1.8 to ∼3.5 after emb-9 knockdown (Fig. 6 C and Fig. S5 B), also indicating flexibility in trimer α-chain composition based on expression levels of the α-chains.

To complement RNAi-mediated reduction of EMB-9, we also increased EMB-9 protein levels by overexpressing a tandem-tagged emb-9p::EMB-9::mNG::mRuby2 in EMB-9::mNG (C) knock-in animals and we overexpressed emb-9p::EMB-9::mRuby2 in LET-2::mNG (C) knock-in animals. We used the mRuby2 tag to ensure that the overexpressed EMB-9 (marked with mRuby2) was being incorporated into the BM of these strains. The overexpression of EMB-9 increased EMB-9 levels ∼36% in the BM of day 1 adults, and there was no significant change in LET-2 levels (Fig. 7, A and B; and Fig. S5 D). The LET-2/EMB-9 ratio decreased from 1.7 to 1.3 (Fig. 7 C and Fig. S5 E). There was also an increase in EMB-9::mNG aggregation in the body wall muscle but not LET-2::mNG (Fig. S5 F and Table S1), consistent with a preference for heterotrimer formation. Taken together, these experiments support the idea that there is a preference for LLE heterotrimer assembly, but also indicate that differences in α-chain expression can drive flexibility in trimer α-chain makeup at the BM.

Pathogenic variants in COL4A1 and COL4A2 lead to the multisystem disorder known as Gould syndrome (Labelle-Dumais et al., 2024). The most common variants found in Gould syndrome are substitutions of glycine aa in the Gly-X-Y motifs of the triple helical domain, which lead to ER retention, reduced secretion, and defective BMs in mice and humans (Gould et al., 2007; Jeanne and Gould, 2017; Kuo et al., 2014; Murray et al., 2014). Similar to humans and mice, immunofluorescence studies in C. elegans have revealed that emb-9 and let-2 glycine missense mutant alleles that mimic Gould syndrome result in intracellular accumulation and decreased BM levels (Gupta et al., 1997). Treatment of mouse and human fibroblast cells with the chemical chaperone 4-phenylbutyrate (4PBA) increases secretion of mutant type IV collagen and increases incorporation into mouse BMs, but does not alleviate disease severity for all tissues, and can even increase skeletal myopathy for some COL4A1 mutant alleles (Gould et al., 2007; Jones et al., 2019; Kuo et al., 2014; Labelle-Dumais et al., 2019). This suggests that type IV collagen harboring glycine substitutions might not function normally extracellularly. How these mutant forms may alter type IV collagen extracellular regulation and function, however, is unknown.

We next wanted to assess whether the ability to endogenously visualize fluorophore-tagged type IV collagen with quantitative live imaging could reveal new insights into how glycine substitutions affect type IV collagen extracellular regulation. Previous immunolocalization studies in C. elegans glycine substitution mutants in emb-9 and let-2 have reported apparent decreased type IV collagen α-chain BM levels, but these levels were not quantified (Gupta et al., 1997). We thus used genome editing to tag the temperature-sensitive glycine substitution mutants emb-9 (b117) (EMB-9 G1173D) and let-2 (b246) (LET-2 G1287D) with mNG at the C terminus (Fig. 8 A) (Gupta et al., 1997; Sibley et al., 1994). Animals harboring emb-9 (b117) or let-2 (b246) are homozygous-viable at the permissive temperature of 15°C but mothers shifted to 25°C have progeny that die as embryos or arrest as larvae (Gupta et al., 1997; Sibley et al., 1994).

We examined type IV collagen localization at the body wall muscle BM, where most type IV collagen is generated and secreted, and at the pharynx—a tissue that does not produce type IV collagen, but instead recruits it from extracellular fluid (Fig. 8, B and C) (Graham et al., 1997; Morrissey et al., 2016). We examined mNG (C)–tagged EMB-9 G1173D and LET-2 G1287D mutants arrested at the L1 larval stage and compared localization with mNG (C)–tagged EMB-9 and LET-2 wild-type L1 larvae. In line with previous studies, we saw extensive intracellular accumulation in mutant strains at nonpermissive temperatures (Fig. S5 G and Table S1). Interestingly, we found that the levels of mutant EMB-9 (G1173D) and LET-2 (G1287D) protein were higher in the body wall muscle BM compared with nonmutant control proteins, but the mutant protein levels were lower in the pharyngeal BM (Fig. 8, D and E). Consistent with a preference for heterotrimers, we found that when wild-type mNG-tagged EMB-9 and LET-2 α-chains were crossed into the respective untagged let-2 (b246) and emb-9 (b117) mutant backgrounds, we observed similar intracellular accumulation, but also increased body wall muscle BM levels and decreased pharyngeal BM levels (Fig. 8, D and E; Fig. S5 G; and Table S1). As type IV collagen is trafficked from the body wall muscles to other tissues such as the pharynx, these results suggest that mutant type IV collagen might not be properly removed from the body wall muscle BM to allow for its distribution to other tissues.

To examine removal dynamics of mutant type IV collagen trimers from the body wall muscle BM, we crossed animals with endogenously tagged photoconvertible EMB-9::mEos2 and LET-2::mMaple into untagged let-2(b246) and emb-9(b117) mutants, respectively. We examined the removal of type IV collagen from the body wall muscle BM in wild-type and mutant animals by photoconverting the body wall muscle BM and assessing reduction in the photoconverted signal after 5 h. Approximately 46% and 56% of the photoconverted signal were removed in wild-type EMB-9 and LET-2 controls, respectively, after 5 h, while 30% and 33% were removed in EMB-9 (G1173D) and LET-2 (G1287D) mutants, respectively (Fig. 9, A–C). Together, these results indicate that glycine missense mutant α-chains cause BM accumulation defects that might result from decreased removal from BMs surrounding tissues that produce type IV collagen.

Type IV collagen is built as a helical trimer, composed of three interweaved α-chains. Comparative studies in animals suggest that type IV collagen is a founding and core BM component with crucial structural and signaling roles throughout animal life (Fidler et al., 2018; Pöschl et al., 2004). Most vertebrates encode six α-chains, while C. elegans and Drosophila each has two (Fidler et al., 2017). Genetic and regulatory defects in type IV collagen are implicated in numerous human diseases and aging (Gatseva et al., 2019; Mao et al., 2015; Uspenskaia et al., 2004; Zhang et al., 2022). Thus, there is broad interest in understanding type IV collagen regulation and function. Endogenous fluorophore tagging has powerfully advanced the understanding of BM protein regulation and function in Drosophila and C. elegans, and more recently in zebrafish and mice (Stramer and Sherwood, 2024). Endogenous tagging of type IV collagen α-chains in Drosophila, C. elegans, and mice has almost exclusively used the Drosophila Vkg-GFP protein trap insertion site, where GFP is inserted at the start of the 7S domain near the N terminus of the protein (Jones et al., 2024; Keeley et al., 2020; Morin et al., 2001; Stramer and Sherwood, 2024). Insertion of fluorophores at this site is viable for Drosophila vkg, C. elegans emb-9, and mouse COL4A2 α-chains (Keeley et al., 2020; Morin et al., 2001; Wuergezhen et al., 2025), but has not yet been attempted for Drosophila Cg25C, and is not viable for C. elegans let-2 or mouse COL4A1 α-chains (Jones et al., 2024; Keeley et al., 2020). As a result, there has been an inability to comprehensively examine type IV collagen regulation and composition using live imaging in any animal to date.

Using a health assay and localization analysis, we found that C. elegans emb-9 tagged at the Vkg-GFP site, which we have named IS1, causes reduced health and extensive aggregation of the fluorophore-tagged protein in the ER of body wall muscle cells, where type IV collagen is predominantly expressed and secreted (Graham et al., 1997). Tagging emb-9 at IS1 also reduced the amount of the EMB-9 protein localized to BM. We also found that emb-9 tagged at an internal site within an interruption of the Gly-X-Y repeats (IS2) (Matsuo et al., 2019) had reduced health and had ER-localized aggregation of the tagged EMB-9 protein. In contrast, we discovered that tagging the C. elegans type IV collagen α-chains EMB-9 and LET-2 at the C terminus with a linker and many different fluorophores resulted in animals that were indistinguishable from wild-type animals in our health assay. Even animals that were homozygous for both labeled α-chains (e.g., EMB-9::mRuby2 [C]; LET-2::mNG [C]) were similar to wild-type animals. Further, these type IV collagen strains showed little to no internal protein aggregation in early larval–staged worms. Western blot analysis also confirmed that the fluorophore mNG was stably associated with both EMB-9 and LET-2. One reason that the C terminus of type IV collagen α-chains might not have been selected for endogenous tagging previously is that the NC1 domain, which both selects for heterotrimer assembly and mediates dimerization of type IV collagen molecules extracellularly through sulfilimine bonds, is located at the C terminus (Fidler et al., 2018; Khoshnoodi et al., 2008). C-terminally tagged type IV collagen trimers could also cluster multiple fluorophores—up to six when both α-chains are tagged. Yet, we did not detect fluorophore quenching (Bae et al., 2021), suggesting that C-terminal site tagging, along with the 18-aa linker we used, does not cause quenching in native C. elegans BM.

One unexpected finding from our studies was the dramatic influence of different fluorophores inserted at IS1 (Vkg) on strain health. Insertion of the red fluorophores mRuby2 or mScarlet-I resulted in only mild delays in the time to starvation health assay for worm populations (1-day delay). In contrast, the insertion of mEos2 or mNG resulted in dramatically slower growth and long delays to starvation (5–8 days). All these genetically encoded fluorescent proteins are similar in size, share the characteristic β-barrel structure of the Aequorea GFP, and have similar mechanisms of fluorophore formation (Rodriguez et al., 2017). However, red fluorescent proteins and photoconvertible proteins have been constructed through extensive engineering of genes from a variety of anthozoan species and are thus encoded by significantly different aa (Bindels et al., 2017; Kredel et al., 2009; Thorn, 2017; Zhang et al., 2012). As a result, the folding rates and fluorophore formation rates vary significantly between fluorescent proteins (Balleza et al., 2018; Thorn, 2017). Our results suggest that these differences might impact the ability of type IV collagen α-chains to fold and trimerize in the ER, particularly when tagged at IS1 (Vkg), and negatively affect animal health.

Using our endogenous C-terminal fluorophore-tagged LET-2 and EMB-9 α-chains, our proteomics findings indicated that LET-2 is approximately twofold more abundant than EMB-9 in whole animals. This observation suggests that the C. elegans type IV collagen trimer is composed predominantly of two LET-2 (α2-like chain) and one EMB-9 (α1-like chain)—an LLE heterotrimer composition (Guo and Kramer, 1989). Notably, this composition does not follow the proposed mammalian rule of combining two odd-numbered (α1-like) and one even-numbered (α2-like) α-chain (Khoshnoodi et al., 2008). Drosophila also appears to not follow the mammalian rule. Hexamer isolation, NC1 recombination experiments, and transgenic NC1 domain expression suggest that the Drosophila type IV collagen trimer is predominantly Cg25C-Cg25C-Vkg (i.e., two α2 [Cg25C] and one α1 [Vkg] chain). This discrepancy, however, might simply be due to the difficulty of assigning clear homology of C. elegans and Drosophila NC1 sequences into vertebrate α1- and α2-like groups (Guo and Kramer, 1989; Summers et al., 2023).

A critical step in type IV collagen trimer maturation is catalyzed by prolyl 4-hydroxylase. In the absence of prolyl 4-hydroxylase or its cofactor ascorbate/vitamin C, type IV collagen α-chains cannot form trimers (Lamandé and Bateman, 1999). Work in Drosophila using the GFP-tagged Vkg protein showed that in prolyl 4-hydroxylase–deficient larvae, type IV collagen is not incorporated into BMs (Pastor-Pareja and Xu, 2011). These results strongly suggest that all type IV collagens within BMs are trimers and that fluorescence intensity of individual α-chains tagged with the same fluorophore at the same site can be used to assess the composite picture of BM trimers within the BM. Importantly, our quantitative fluorescence analysis of tissue-specific BMs revealed BMs did not have a strict LET-2/EMB-9 ratio of 2.0 (i.e., are not stringently LLE); instead, most were ∼1.7, whereas the spermathecal BM had a ratio of ∼2.2. These observations suggest a predominance of the LLE trimer, likely because LLE heterotrimer assembly is more efficient (i.e., higher molecular affinity), but also some flexibility in trimer composition. LET-2/EMB-9 ratios >2 suggest the existence of LLL homotrimers, while ratios <2 indicate LEE and possibly EEE trimers. Analysis of translational reporters within specific tissues and experimentally increasing and decreasing α-chain expression showed that trimer compositional variability within BMs can be driven by differences in α-chain expression levels. This implies that the relative availability of α-chains for trimerization in the ER influences trimer makeup. Notably, our studies also revealed dependence between α-chains for secretion—decreasing the levels of EMB-9 led to the retention of LET-2 in cells, while increasing EMB-9 expression led to EMB-9 intracellular accumulation. This observation is consistent with studies showing that mutations in emb-9 and let-2 result in intracellular retention of the other (nonmutant) α-chain (Gupta et al., 1997). Together, these results suggest a preference for the in vivo assembly of the LLE heterotrimer, but also some flexibility in building other trimers in C. elegans.

Based on α-chain abundance and hexamer associations of the six α-chains in vertebrates, it has been proposed that only three (out of a possible 56) type IV collagen trimers are built—α1α1α2, α3α4α5, and α5α5α6 (Khoshnoodi et al., 2008). Yet, studies on the α1α1α2 trimer, the most prevalent and most investigated type IV collagen in vertebrates, suggest that other trimer combinations could exist—carcinoma cells secrete α1 homotrimers (Haralson et al., 1985), reconstitution of NC1 domains has revealed many possible trimer combinations (Boutaud et al., 2000), and in vivo hexamer analysis has not excluded the existence of α1 homotrimers or α1α2α2 heterotrimers (Boutaud et al., 2000; Johansson et al., 1992; Siebold et al., 1988). Studies in Drosophila have also not ruled out variable trimer formation (Summers et al., 2023). Thus, flexible type IV collagen trimer combinations might occur in other organisms.

When assessing if trimer flexibility exists in other species, an important aspect to consider is the varied gene arrangement in C. elegans wherein the two type IV collagen α-chain–encoding genes are located on different chromosomes (Guo and Kramer, 1989). In other animals, including Drosophila and mammals, type IV collagen genes are arranged on chromosomes in a head-to-head orientation and share a common promoter (Khoshnoodi et al., 2008; Yasothornsrikul et al., 1997). The alternative arrangement in C. elegans may allow for unique regulatory mechanisms that permit variable transcriptional expression levels for each α-chain and therefore trimer flexibility. However, despite having a shared promoter, COL4A5 and COL4A6 show variable expression between tissues in humans. One possible mechanism driving this difference is that COL4A6 may have an alternative promoter as evidenced by the identification of two separate COL4A6 transcripts that vary in their 5′ regions (Khoshnoodi et al., 2008; Sugimoto et al., 1994). Additionally, different microRNAs can regulate the expression of distinct type IV collagen α-chains. For example, in vivo and in vitro RNA expression profiles indicate that microRNA 29a downregulates the expression of five type IV collagen α-chains (COL4A1-COL4A5), while microRNA 29b specifically downregulates COL4A2 expression (Ma et al., 2023; Wang et al., 2012). Therefore, type IV collagen genes that share a promoter may still be differentially expressed, potentially allowing for trimer flexibility in other species.

Trimer compositional variability likely expands the possible functions of type IV collagen and could allow for tissue-specific tailoring of type IV collagen trimer construction. For example, it has been hypothesized that the more restricted and tissue-specific vertebrate α3α4α5 heterotrimer might have increased mechanical strength relative to that of the α1α1α2 heterotrimer because of additional internal cross-linking (Gunwar et al., 1998; Naylor et al., 2021). Interestingly, we found that the spermathecal BM had a LET-2/EMB-9 ratio of ∼2.2, whereas other tissue BMs had a significantly lower ratio of ∼1.7. The spermatheca tissue is unique in that it undergoes repeated stretching every ∼20 min during peak egg laying in C. elegans (Kelley and Cram, 2019). The distinct type IV collagen composition found in the spermathecal BM might mechanically support the ability of the spermatheca to repetitively stretch. Trimer α-chain compositional diversity could also increase the variety and specificity of interactions with other BM components, signaling molecules, and cell surface receptors (Khoshnoodi et al., 2008; Wang et al., 2008) to support many tissue-specific functions.

The most common human pathogenetic variants in the COL4A1 and COL4A2 genes are those that result in substitutions of glycine residues with other aa within the triple helical domain (Jeanne and Gould, 2017). These variants often impair the formation and secretion of COL4A1 and COL4A2 heterotrimers (α1α1α2) and lead to Gould syndrome, which is a multisystem disorder involving variable defects in many tissues, including cerebrovascular, eye, kidney, brain, skeletal, and neuromuscular tissues (Labelle-Dumais et al., 2024; Mao et al., 2015). Notably, not all glycine mutations affect type IV collagen secretion (Kuo et al., 2014). Further, pharmacological treatment of mice harboring COL4A1 mutations with the chemical 4PBA, which enhances secretion and BM incorporation of mutant forms of type IV collagen, can decrease, increase, or have no effect on tissue pathology depending upon the mutation and tissue (Hayashi et al., 2018; Jones et al., 2019; Labelle-Dumais et al., 2019). These observations highlight the complex roles of type IV collagen in tissue function and suggest that mutations in type IV collagen can affect its extracellular functions. Through C-terminal tagging with mNG, we show that temperature-sensitive mutants harboring glycine substitutions in emb-9 and let-2 can be successfully labeled with fluorophores. We found that mutant forms of both α-chains led to type IV collagen retention in cells at the nonpermissive temperature, as previously reported using immunofluorescence (Gupta et al., 1997). However, we also discovered increased accumulation of mutant type IV collagen in the BM on the body wall muscle, which is a site of type IV collagen production and secretion, and decreased accumulation of type IV collagen on the pharynx—a tissue that does not produce type IV collagen (Graham et al., 1997). Through optical highlighting of photoconvertible mEos2– and mMaple-tagged type IV collagen trimers, we found that type IV collagens harboring glycine substitution mutations were more stably associated with the body wall muscle BM. This decrease in the type IV collagen BM on-off rate might not only increase accumulation in BMs surrounding cells and tissues that secrete type IV collagen, but also decrease levels in BMs surrounding tissues such as the pharynx that depend upon type IV collagen trafficking away from secreting tissues (Duncan et al., 2020; Pastor-Pareja and Xu, 2011; Stramer and Sherwood, 2024). Consistent with this notion, a similar increase of type IV collagen in the body wall muscle BM and a decrease in the pharyngeal BM occur after loss of the extracellular collagen chaperone SPARC, which promotes type IV collagen trafficking from sites of synthesis (Morrissey et al., 2016; Pastor-Pareja and Xu, 2011). As type IV collagen levels can affect BM biomechanical and signaling properties (Crest et al., 2017; Jayadev and Sherwood, 2017b; Tian and Jiang, 2014), these alterations in type IV collagen BM accumulation could in part explain extracellular dysfunction of COL4A1 and COL4A2 variants.

Collagens are a major ECM superfamily composed of 46 α-chains that form 28 different collagens in mammals (I–XXVIII). All collagens share a triple helical domain, but they take on many forms, ranging from fibrils, to networks, and even membrane-spanning proteins, which together play essential roles in cell adhesion, mechanical support, and signaling (Ricard-Blum, 2011; Wakabayashi, 2021). Variants in collagens are implicated in nearly 40 human genetic diseases, most of which are glycine substitutions in the triple helical domain (Fidler et al., 2018). Although generally viewed as stable, static scaffoldings, work in Drosophila and C. elegans has revealed that type IV collagen has dynamic properties, including rapid polymerization, remodeling, and association and disassociation from the BM (Serna-Morales et al., 2023; Stramer and Sherwood, 2024). We expect that the tagging at the C terminus of mature α-chains (either extreme C terminus or prior to cleaved pro-peptide) will be broadly applicable to many collagens. Supporting this notion, the type XVIII collagen, the membrane-spanning COL-99, and the BLI-2 and BLI-6 cuticular collagens in C. elegans have been endogenously tagged with mNG at the C terminus and are homozygous-viable (Adams et al., 2023; Jayadev et al., 2022; Keeley et al., 2020). Endogenously tagging diverse collagen molecules with genetically encoded fluorophores offers to provide important insight not only into human disease, but also into the fascinating functions and likely underappreciated dynamics of this crucial ECM superfamily.

C. elegans were maintained according to standard procedures on nematode growth medium (NGM) agar plates seeded with OP50 E. coli at 16°C or 20°C (Stiernagle, 2006). Strains used in this study are listed in Table S2.

All genome-edited strains were generated using CRISPR/Cas9-mediated homologous recombination with a self-excising cassette (SEC) for drug selection (hygromycin) as previously described (Dickinson and Goldstein, 2016; Keeley et al., 2020). For all C-terminally tagged strains, we placed an 18-aa flexible linker in frame and directly upstream of the fluorophore. The flexible linker and fluorophore were inserted at the endogenous locus, just upstream of the stop codon. For emb-9::mScarlet-I, flexible linkers were placed on both sides of the fluorophore and the construct was inserted in frame at the same location as previously published IS1 constructs (Jayadev et al., 2022; Keeley et al., 2020). Primers used in CRISPR repair plasmid construction are listed in Table S3. We generated single guide RNA (sgRNA) plasmids by inserting the sequences into the pDD122 plasmid (Table S4) (Dickinson et al., 2013). Nucleic acid sequences of sgRNAs and aa sequences of flexible linkers are listed in Fig. S1.

A C. elegans codon-optimized P2A sequence was built using primers listed in Table S3. The PEST sequence was amplified from the plasmid pCFJ150 (Table S4) (Kaymak et al., 2016) using primers listed in Table S3. let-2::mNG (C) and emb-9::mNG (C) CRISPR repair plasmids were linearized by ClaI digestion. The P2A and PEST sequences were assembled with both linearized CRISPR plasmids by the Gibson assembly. C-site sgRNAs of emb-9 or let-2 were used to direct Cas9 cleavage for homologous insertion.

The ∼2-kb let-2 upstream regulatory sequence was amplified from wild-type genomic DNA using the primers listed in Table S3. The mNG sequence was amplified from the emb-9::mNG (C) CRISPR repair plasmid. Both fragments were ligated into the pAP088 starter repair plasmid (Pani and Goldstein, 2018) containing homology arms for the ttTi4348 site using the Gibson assembly. This single-copy transgene was inserted into the ttTi4348 transposon insertion site on chromosome I (Frøkjær-Jensen et al., 2012) using the SEC CRISPR repair plasmid as described above. The pCFJ352 plasmid (Table S4) containing the sgRNA was used to direct Cas9 cleavage near this region for insertion.

The emb-9 overexpression construct emb-9p::emb-9::mRuby2::emb-9 3′utr (pQD03) was built by Gibson assembling by ligating the following fragments into pBluescript SK(-): the ∼3.8-kb emb-9 upstream regulatory sequence, the full-length emb-9 gene, an ∼500-bp emb-9 3′utr (all amplified from N2 genomic DNA sequence), and mRuby2 amplified from the emb-9::mRuby2(C) CRISPR repair plasmid (using primers listed in Table S3). To create the emb-9p::emb-9::mNG::mRuby2::emb-9 3′utr (pQD04) construct, the mNG sequence amplified from the emb-9::mNG (C) CRISPR repair plasmid was inserted into pQD03 using the Gibson assembly.

Mixtures of the sgRNA plasmid (5–10 ng/μl), the SEC repair template plasmid (100 ng/μl), and co-injection marker plasmids (pCFJ90 [myo-2p::mCherry] and pCFJ104 [myo-3p::mCherry], 5 ng/μl each) (Table S4) were injected into the gonads of young adult N2 hermaphrodites as previously described (Dickinson and Goldstein, 2016). mNG knock-in strains harboring glycine mutation alleles for emb-9 and let-2 were generated using the respective C-site mNG CRISPR repair plasmid and C-site sgRNA sequence (described in Endogenous knock-in constructs). Translational reporters for emb-9 and let-2 were generated using the respective C-site sgRNA sequence (described in Endogenous knock-in constructs). The let-2 transcriptional reporter was generated using the sgRNA sequence in the pCFJ352 plasmid (Table S4) that targets the ttTi4348 transposon insertion site on chromosome I. Following injection, animals were singled onto OP50 plates and allowed to lay eggs at 20°C for 2–3 days before treating with 500 μl of hygromycin (2 mg/ml). After 3–4 days at 20°C posthygromycin treatment, surviving candidate knock-in animals were identified by the dominant roller phenotype (sqt-1(e1350)) and an absence of co-injection marker fluorescence. To excise the SEC from homozygous-viable lines, 4–8 L3/L4 animals were heat-shocked in a 34°C water bath for 4 h and then left to grow at 25°C for 3–4 days. Progeny lacking the roller phenotype were singled, and edited lines were identified by visualizing the fluorophore at the BM and via PCR genotyping DNA using the primers listed in Table S3.

Transgenic worms expressing emb-9p::emb-9::mRuby2::emb-9 3′utr or emb-9p::emb-9::mNG::mRuby2::emb-9 3′utr were created by injecting pQD03 or pQD04 (200 ng/μl) with the plasmid pRF4 containing rol-6(su1006) that causes a dominant “roller” phenotype (50 ng/μl), EcoRI-digested salmon sperm DNA (30 ng/μl), and the empty pBluescript II vector (40 ng/μl) into the gonads of young adult LET-2::mNG or EMB-9::mNG hermaphrodites, respectively (Yochem and Herman, 2003). The pQD03 and pQD04 were expressed as extrachromosomal arrays, which often have mosaic somatic expression (Yochem and Herman, 2003). mRuby2 fluorescence was used to confirm transgene expression and BM localization of the overexpressed protein. The dominant roller phenotype was used to track transgene propagation. Young adult rollers (48 h after hatch) were imaged for the experiments.

RNAi experiments were performed using the feeding method (Timmons and Fire, 1998), and the emb-9 RNAi construct was obtained from the Ahringer library (Fraser et al., 2000; Kamath et al., 2003). RNAi bacterial cultures were grown in selective media at 37°C for 18 h and then for one additional hour after adding IPTG (1 mM final concentration) (Table S4) to induce double-stranded RNA expression. Cultures were then used to seed NGM plates that had been treated with a 1:1 mixture of 1 M IPTG and 100 mg/ml ampicillin (18 μl per plate) (Table S4). Plates were allowed to dry overnight after seeding. Synchronized L1 larvae were placed on the seeded plates and allowed to grow for 72 h at 20°C. The L4440 empty vector was used as a control (Timmons and Fire, 1998).

Confocal images (excluding photoconversion experiments) were acquired at 20°C on Zeiss Axio Imager A1 microscopes, fitted with either a Yokogawa CSU-10 or a Yokogawa CSU-W1 spinning disk, and either an ORCA-Fusion CMOS or ORCA-Quest qCMOS camera using 488-, 505-, and 561-nm excitation lasers. All microscopes were controlled with μManager software v1.4.23 or v2.0.1 (Edelstein et al., 2010). Zeiss 40×, 63×, and 100× Plan Apochromat (1.4 numerical aperture) oil-immersion objectives were used. Animals were mounted on 5% noble agar pads with 0.002 M sodium azide (Table S4) for L1s and 0.01 M sodium azide for all other stages (Kelley et al., 2017).

For fluorescence recovery after photobleaching (FRAP) experiments, photobleaching was performed using an iLas² targeted laser system from BioVision equipped with an Omicron Lux 60 mW 405 nm continuous wave laser and controlled with MetaMorph version 7.10.4.407 software (Keeley et al., 2020). Laser power and duration of photobleach were consistent for all animals within an experiment. Imaging intervals and duration allowed for fluorescence recovery were determined empirically to capture quantifiable recovery. For all FRAP experiments, worms were immobilized on 5% agar pads with 0.002 M sodium azide. The half of the pharyngeal terminal bulb opposite to the L1 gonad was photobleached. Animals were then rescued from the slide and washed with M9 to dilute the sodium azide treatment. Recovery was assessed using the same imaging settings after allowing the worms to develop normally for 5 h. Animals rescued after posterior pharynx photobleaching all developed normally and had no pharyngeal morphological defects (n = 30/30 animals).

Photoconversion experiments were performed at 20°C on an inverted Zeiss 880 LSM confocal mounted to a Zeiss Axio Observer Z1 microscope using a Zeiss 63× Plan Apochromat (1.4 numerical aperture) oil-immersion objective and a pinhole size of 57 μm. The photoconvertible mMaple and mEos2 fluorophores were photoconverted with a 405-nm laser, and photoconversion settings were the same between samples. Unconverted mMaple and mEos2 were imaged using a 488-nm laser and converted mMaple and mEos2 with a 561-nm laser. For all optical highlighting experiments, worms were immobilized on 5% agar pads with 0.002 M sodium azide. An ∼350-μm² section of the body wall muscle BM EMB-9::mEOS2 or LET-2::mMaple between the pharyngeal bulbs was photoconverted. Animals were then rescued from the slide and washed with M9 to dilute the sodium azide. Recovery was assessed using the same imaging settings after allowing the worms to develop normally for 5 h. Animals rescued all developed normally and had no morphological defects (n = 20/20 animals).

Single z-slice confocal images were quantified using FIJI (2.0) (Schindelin et al., 2012). BM fluorescence intensity was measured by drawing 3 pixel-wide lines along the BM of interest outlined below and recording the mean fluorescence value. Pharyngeal BM measurements were recorded from both sides of the terminal bulb and averaged. Body wall muscle BM measurements were recorded on both sides of the head region flanking the pharynx and averaged. The spermathecal BM was measured on the side proximal to unfertilized eggs. Measurements were recorded on both dorsal and ventral sides of the organ (in regions where it was not folded in on itself) and averaged together. DTC BM measurements were taken around the entire cell. To account for autofluorescence, background measurements were acquired by drawing a line either next to the pharynx (for pharyngeal or body wall muscle BM measurements) or inside the gonad (DTC and spermathecal BM measurements), where no type IV collagen signal was apparent. mNG translational reporter measurements were taken by recording the mean fluorescence value in a single z-slice confocal image for the entire DTC, the entire spermatheca, or two body wall muscles flanking both sides of the pharynx (which were then averaged together). To account for autofluorescence, background measurements were acquired by drawing a line either next to the pharynx (for body wall muscle translation measurements) or inside the vulva (for DTC and spermatheca translation measurements), where no translation reporter signal was apparent.

In Fig. 4 B, Fig. 6 C, and Fig. 7 C, to directly compare LET-2/EMB-9 ratios between tissues, LET-2::mNG mean fluorescence intensity data points were normalized to the mean value of EMB-9::mNG within that tissue for each trial. Specifically, each data point in the boxplots was normalized with the following formula: (L/mean E), where “L” is a single mean fluorescence intensity value measurement of LET-2::mNG in a tissue BM, and “mean E″ is the mean value of all EMB-9::mNG measurements within the same tissue BM for the trial. For Fig. S4 A, Fig. S5 B, and Fig. S5 E, the EMB-9::mNG data points were normalized in a similar way to display the spread of data around the mean. Specifically, the formula used was (E/mean E), where “E” is a single measurement of EMB-9::mNG mean fluorescence in a tissue BM. For Fig. 5 B and Fig. S4 D, the same normalization methods were used, substituting measurements of let-2 translation reporter for L and emb-9 translation reporter for E.

To compare the percent loss or gain of EMB-9::mNG or LET-2::mNG in RNAi or overexpression experiments, the data points in Fig. 6 B and Fig. 7 B were normalized by dividing each treatment mean fluorescence value measurement by the mean value of the control measurements within that trial. Specifically, the formula used was (T/mean C), where “T” is a single fluorescence intensity measurement from the treatment animal, and “mean C” is the mean value of all measurements from the control animals within a trial. For Fig. S5 A and Fig. S5 D, data points were normalized in a similar way to show the spread of control measurements around the mean. Specifically, the formula used was (C/mean C), where “C” is a single measurement from the control animals.

The number of EMB-9::mScarlet-I (IS1) and EMB-9::mCherry (IS2) fluorescent aggregates that localized within the ER (marked with the ER membrane protein ELO-1::mNG) in body wall muscles (Fig. 2 B and Fig. S2 B) was scored in a blinded manner—aggregates were first identified, and then, it was determined whether they were within the ER. A total of 10 aggregates were counted per animal, and 10 animals were analyzed for the EMB-9::mScarlet-I (IS1) and EMB-9::mCherry (IS2) strains.

Sum projections of the first four slices of the body wall muscle from confocal stacks of animals from Figs. 2, 6, 7, and 9 were generated. A manual threshold was applied to these images to create a binary mask of the aggregates. The aggregate area in the body wall muscle was estimated by drawing a polygon of area ∼4,000 pixels in the body wall muscle and measuring the raw integrated density. The aggregate area measured was then divided by the total area of the polygon to report the percent area occupied by aggregates in the body wall muscle.

For FRAP analysis, the mean fluorescence intensity values were generated by drawing an ∼0.3-μm-wide and 3.825-μm-long line through a clearly defined section of the midplane of a single z-slice of the pharyngeal terminal bulb. The background was calculated by drawing an equivalent line within the terminal bulb of the pharynx, where no type IV collagen is produced or localized. Mean fluorescence intensity was measured within the photobleached region and within an equivalent region on the unbleached side of the terminal bulb before bleach (prebleach), immediately after bleach (bleach), and 5 h after bleach (5 h postbleach). A bleach correction factor was used to account for general photobleaching during image acquisition across the duration of the experiment (Gianakas et al., 2023). In brief, the fluorescence intensity measurement in the unbleached region at the 5-h postbleach time point was divided by the value at the prebleach time point to obtain the bleach correction factor. Fluorescence intensity measurements at the prebleach and bleach time points in the photobleached region were then multiplied by the bleach correction factor to normalize these values to the 5-h postbleach time point. The background signal was then subtracted from bleach-corrected values and used in analysis. Recovery was calculated by dividing the measurement at the 5-h postbleach time point by that at the prebleach time point.

For each replicate, four L4 hermaphrodites were placed on an NGM plate seeded with 400 μl of OP50 E. coli. Plates were maintained at 20°C and checked twice daily for starvation. For every experiment, N2 wild-type animals were plated on the same day, using the same batch of plates. The time it took for experimental strains to starve was reported relative to the time it took for the corresponding N2 plates to starve.

Mixed populations of worms were washed off NGM plates with sterile water and placed into 1.5-ml Eppendorf tubes. Animals were then washed three times with sterile water and frozen in liquid nitrogen. Frozen animals were then ground to a powder under cryogenic conditions with a mortar and pestle and then resuspended in cooled RIPA buffer plus 1X SIGMAFAST Protease Inhibitor Cocktail (Table S4). Laemmli sample buffer and β-mercaptoethanol (Table S4) were then added to a final concentration of 1X and 4%, respectively. The final concentration of sample powder to buffer was 0.5 mg/ml. Samples were boiled for 5 min, and then, 15 μl of resuspended samples was loaded onto 4–15% gradient Mini-PROTEAN TGX Gels (Table S4) and run at 100 V. Wet transfers to nitrocellulose were carried out overnight in a cold room at 30 V for 16 h. Nitrocellulose membranes were blocked with 5% milk in TBST for 1 h, incubated with primary antibody against mNG in 1% milk and TBST overnight at 4°C (antibody dilution 1:1,000), and then treated with secondary peroxidase-conjugated antibody (Table S4) for 1 h at room temperature. Blots were treated with SuperSignal West Atto Ultimate Sensitivity Chemiluminescent Substrate (Table S4) for 5 min before imaging. Blots were imaged on an iBright 1500 imaging system (Invitrogen).

Samples were processed by chemical fractionation to enrich for BM proteins for mass spectrometry–based proteomics as described previously (Morais et al., 2023). Briefly, a mixed population of L1, L2, and L3 larvae were collected in a Tris lysis buffer (10 mM Tris, 150 mM NaCl, 1% Triton X-100, 25 mM EDTA added of EDTA-free Roche cOmplete Protease Inhibitor Cocktail) and homogenized to extract soluble proteins. After centrifugation at 14,000 × g, the protein supernatant was collected. This fraction is enriched for soluble cellular proteins. The remaining pellet was then resuspended in an alkaline detergent buffer (20 mM NH₄OH and 0.5% Triton X-100 in 0.1 M PBS) to disrupt cell–matrix bonds and centrifuged again, and the supernatant was removed. The final pellet was resuspended in TEAB/SDS lysis buffer (50 mM TEAB and 5% SDS in LC-MS grade water). This matrix-enriched fraction was sonicated with Covaris LE220+ Focused Ultrasonicator (Covaris). The cellular fraction was added to 2X TEAB/SDS lysis buffer, and then, both soluble and insoluble protein lysates were reduced with 5 mM dithiothreitol and further alkylated with 15 mM iodoacetamide and processed for overnight digestion at 37°C in S-Trap columns (ProtiFi) using Sequencing Grade Modified Trypsin (Catalog # V5111; Promega). The peptide mixtures were eluted from the columns, desalted, and dried to completeness by vacuum centrifugation for mass spectrometry.

Peptide samples were analyzed by LC-MS/MS with a Thermo Exploris 480 mass spectrometer (Thermo Fisher Scientific). The raw spectra data were analyzed using MaxQuant version 2.5.0.0 (Cox and Mann, 2008). MS/MS data were searched against a modified version of the UniProt C. elegans proteome (v. 2024_01; containing 26,706 reviewed and unreviewed entries, with the addition of the protein sequence for mRuby2 and mNG [downloaded from https://www.fpbase.org/]) using MaxQuant’s integrated search engine, Andromeda. Carbamidomethylation of cysteine was specified as a fixed modification, and oxidation of methionine, proline, and lysine, and N-terminal acetylation as variable modifications. Up to two missed tryptic cleavage sites were allowed. Mass tolerance settings for precursor peptide ions were 20 ppm for the first MS/MS search and 4.5 ppm for the main MS/MS search, with 0.5 Da for fragment ions. A target-decoy search strategy was employed to control false-positive identifications, maintaining a 1% false discovery rate (FDR) for both peptide-spectrum matches and protein groups. The match-between-runs algorithm was enabled to increase the number of identifications. Contaminants were identified from matches to the MaxQuant contaminants database. Data processing and analysis were performed using RStudio version 2023.9.1.494 (https://rstudio.com). Peptide data obtained with MaxQuant were processed using the QFeatures package version 1.12.0 (Gatto, 2024). Peptides from decoy and contaminant proteins, nonunique, and detected in only one sample were removed. Peptide intensities were log2-transformed and normalized through median centering. Peptide data were summarized to protein groups via robust summarization, and statistical comparisons were performed with multiple t tests with the Benjamini–Hochberg FDR correction using the msqrob2 package (Goeminne et al., 2016, 2020).

Statistical analysis was performed in GraphPad Prism 10. The distribution of data was assessed for normality using the Kolmogorov–Smirnov test. To compare fluorescence intensities or peptide levels between two populations, we used an unpaired t test for normally distributed data or a Mann–Whitney test for data that were not normally distributed. For Fig. 6, we used a t test with Welch’s correction because the variance between datasets was unequal. To compare fluorescence intensities between three or more populations, we used a one-way ANOVA followed by Tukey’s multiple comparisons test for normally distributed data with equal variance between groups or a Brown–Forsythe and Welch ANOVA test followed by a Dunnett’s test for normally distributed data with unequal variance between groups. For data that were not normally distributed, we used a Kruskal–Wallis test followed by Dunn’s test. Sample sizes were determined before each experiment. Sample sizes of at least 10 were chosen for each biological replicate of experiments in Fig. 2, A and B; Fig. 4 A; Fig. 5; and Fig. 8. Where protein ratios were calculated, three biological replicates of those experiments were carried out to ensure reproducibility of the results. Sex was not a factor in analyzing data because C. elegans primarily exist as hermaphrodites. Data points were only excluded if the animal moved during acquisition or developed vacuoles and died.

Fig. S1, related to Fig. 1, shows the endogenous locus of fluorophore insertion for all genome-edited strains. Flexible linker sequences and Cas9 guide RNA sequences are listed. The fluorophore aa length and aa sequence similarity between fluorophores are shown. Fig. S2, related to Fig. 2, shows intracellular aggregates of LET-2::mNG (C) and EMB-9::mNG (C) in the body wall muscles of C. elegans as they develop from L1 larvae to gravid adults. Also shown are the localization of EMB-9::mCherry (IS2) aggregates within the ER and FRAP analysis of LET-2::mNG (C), EMB-9::mNG (C), and EMB-9::mNG (IS1) at the pharyngeal BM. Fig. S3, related to Fig. 3, shows the abundance of mNG and mRuby2 proteins determined by LC-MS/MS in the larval cellular fractions of strains expressing the knock-in proteins Strain 1 (EMB-9::mRuby2 [C]; LET-2::mNG [C]) and Strain 2 (EMB-9-mNG [C]; LET-2::mRuby2 [C]). Also shown is the comparable abundance of the LET-2 and EMB-9 protein in the matrix fractions of both strains. Fig. S4, related to Figs. 4 and 5, shows the normalized LET-2::mNG (C) data with the normalized EMB-9::mNG (C) data and the normalized LET-2 translation reporter (let-2::P2A::PEST::mNG) data with the normalized EMB-9 translation reporter (emb-9::P2A::PEST::mNG) data for all tissues and trials. Also shown is a western blot confirming the stability of mNG associated with the LET-2::mNG (C) and EMB-9::mNG (C) fusion proteins and mean fluorescence intensity measurements of both LET-2::mNG (C) and EMB-9::mNG (C) at the pharyngeal BM of homozygotes and heterozygotes. Fig. S5, related to Figs. 6, 7, and 8, shows the normalized LET-2::mNG (C) data with the normalized EMB-9::mNG (C) data and the normalized treatment (RNAi or overexpression) data with the normalized control (L4440 empty vector RNAi or absence of overexpression) data for all trials. Also shown are intracellular aggregates of LET-2::mNG (C) and EMB-9::mNG (C) in the muscles of day 1 adult C. elegans after emb-9 RNAi and emb-9 overexpression. Additionally, sum projections from confocal z-stacks of L1 larvae carrying let-2 or emb-9 glycine mutant alleles grown at the restrictive temperature are shown. Table S1 reports the quantification of intramuscular aggregates. Table S2 outlines the strains used in this study. Table S3 shows the primer sequences used in this study. Table S4 lists the chemical and DNA reagents used.

Plasmids generated, C. elegans strains generated, and data are available upon request. Strains will be donated to the Caenorhabditis Genetics Center (CGC). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Perez-Riverol et al., 2022) with the dataset identifier PXD055575 and https://doi.org/10.6019/PXD055575. All other data are included in the manuscript and supplemental materials.

We thank S. Boudko for extensive discussions on type IV collagen biochemistry, R. Jayadev for valuable suggestions when conceptualizing the project, and members of the Sherwood laboratory for feedback on experimental analysis. We also thank G. Poulin for help maintaining and handling C. elegans strains used for mass spectrometry. Some strains were provided by the CGC funded by National Institutes of Health (NIH) Office of Research Infrastructure Programs (P40 OD010440).

This work was supported by NIH grants (R35GM118049, R21OD032430) and a Wellcome Trust Discovery Award (Ref: 227417/Z/23/Z) to D.R. Sherwood, the NIH Ruth L. Kirschstein Individual Postdoctoral Fellowship (F32AG077907) to W. Ramos-Lewis, and the Jane Coffin Childs Memorial Fund for Medical Research to A.W.J. Soh. R. Lennon was supported by the Wellcome Trust (Ref: 226804/Z/22/Z, Ref: 227417/Z/23/Z, Ref: 301803/Z/23/Z), the Kidney Research UK and the Stoneygate Trust (Alport Research Hub), and the National Institute for Health Research (NIHR-Manchester BRC).

Author contributions: S. Srinivasan: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, software, validation, visualization, and writing—original draft, review, and editing. W. Ramos-Lewis: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing. M.R.P.T. Morais: formal analysis, investigation, visualization, and writing—review and editing. Q. Chi: investigation and methodology. A.W.J. Soh: investigation. E. Williams: investigation and writing—review and editing. R. Lennon: resources, supervision, and writing—original draft, review, and editing. D.R. Sherwood: conceptualization, funding acquisition, methodology, project administration, supervision, and writing—original draft, review, and editing.