Collagen’s enigmatic, highly conserved N-glycan has an essential proteostatic function

Significance

Development and repair of connective tissues require the rapid biosynthesis of large quantities of collagen, the most abundant protein in the human body. The functional importance of collagen has motivated extensive efforts to uncover molecular mechanisms responsible for matching collagen production to physiological needs. While many roles for collagen chaperones and modifying enzymes are well established, the function of the highly conserved N-glycosylation site within collagen’s C-terminal globular domain has remained elusive for decades. By assaying N-glycan function under conditions of impaired collagen folding, we show that, although the N-glycan is dispensable under normal conditions, it is essential for collagen folding and secretion under conditions that challenge proteostasis. Such environments are commonly encountered during development, tissue repair, and disease.

Abstract

Intracellular procollagen folding begins at the protein’s C-terminal propeptide (C-Pro) domain, which initiates triple-helix assembly and defines the composition and chain register of fibrillar collagen trimers. The C-Pro domain is later proteolytically cleaved and excreted from the body, while the mature triple helix is incorporated into the extracellular matrix. The procollagen C-Pro domain possesses a single N-glycosylation site that is widely conserved in all the fibrillar procollagens across humans and diverse other species. Given that the C-Pro domain is removed once procollagen folding is complete, the N-glycan might be presumed to be important for folding. Surprisingly, however, there is no difference in the folding and secretion of N-glycosylated versus non-N-glycosylated collagen type-I, leaving the function of the N-glycan unclear. We hypothesized that the collagen N-glycan might have a context-dependent function, specifically, that it could be required to promote procollagen folding only when proteostasis is challenged. We show that removal of the N-glycan from misfolding-prone C-Pro domain variants does indeed cause serious procollagen and ER proteostasis defects. The N-glycan promotes folding and secretion of destabilized C-Pro variants by providing access to the ER’s lectin-based chaperone machinery. Finally, we show that the C-Pro N-glycan is actually critical for the folding and secretion of even wild-type procollagen under ER stress conditions. Such stress is commonly incurred during development, wound healing, and other processes in which collagen production plays a key role. Collectively, these results establish an essential, context-dependent function for procollagen’s previously enigmatic N-glycan, wherein the carbohydrate moiety buffers procollagen folding against proteostatic challenge.

As the molecular scaffold for animal life, the various types of collagen perform diverse structural and biological functions (1). Fibrillar collagens, the most abundant collagen class, are synthesized as procollagen strands consisting of N- and C-terminal propeptide (N-Pro and C-Pro, respectively) domains sandwiching a lengthy, continuous triple-helical domain (Fig. 1A) (2). Both the N-Pro and C-Pro domains are proteolytically cleaved upon fibrillar procollagen secretion, such that only the triple-helical domain is deposited in the mature extracellular matrix (3). The function of the N-Pro domain remains largely unclear. The C-Pro domain, in contrast, has a well-established role in templating triple-helix assembly, controlling the composition and stoichiometry of triple helices, and setting the register of the triple helix (4–8).

Fig. 1.

The N-glycosylation motif within the procollagen-I C-Pro domain is highly conserved. (A) Schematic of the proα1(I) sequence showing the N-glycosylation sequon located within the C-Pro domain. (B) Two views of a representative N-glycan (from Protein Data Bank [PDB] ID: 1S4P) modeled onto the structure of the C-Proα1(I) homotrimer (PDB ID: 5K31). (C) Alignment of the amino acid sequences for the C-Pro domain of all human fibrillar procollagens illustrates the strong conservation of the N-glycosylation sequon. (D) Alignment of C-Proα1(I) (or equivalent) amino acid sequences from diverse species again illustrates the near universal conservation of N-glycosylation within the C-Pro domain.

The procollagen-I (composed of proα1 and proα2 strands) C-Pro domain (∼30 kDa in its monomeric form) is globular, cysteine rich, calcium binding, and N-glycosylated on a single asparagine residue (Fig. 1A; N1365 in proα1(I) or N1267 in proα2(I)). N-Glycosylation of the C-Pro domain was first observed nearly 50 years ago (9, 10). A recent crystal structure of the human C-Proα1(I) homotrimer established that the assembled C-Pro domain adopts a structure analogous to a flower (7). Mapping a model N-glycan onto the crystal structure of the C-Proα1(I) homotrimer suggests that the N-glycan is positioned on the outer face of the β-sheets that form each “petal” and is likely directed out into the solvent (Fig. 1B). This N-glycosylation site is conserved across nearly all fibrillar procollagen types and domains of life (Fig. 1 C and D). Even in instances where the specific site is not conserved, an N-glycosylation sequon is still present at least once in the C-Pro domain.

The energetic cost and high conservation of C-Pro N-glycosylation (11) argues for an essential biological function. Outside the cell, N-glycans can mediate protein–protein interactions, stabilize and/or organize extracellular matrices, and provide protection against proteases and other enzymes (12). However, the N-glycan–containing C-Pro domain is cleaved from procollagen prior to extracellular matrix assembly, indicating that the N-glycan is unlikely to have a significant extracellular function.

N-Glycans can also have critical functions inside the cell, largely by providing access to the extensive lectin-based proteostasis network in the endoplasmic reticulum (ER) and thereby extrinsically enhancing protein folding and quality control (13–15). In particular, the lectin chaperones calreticulin and calnexin are known to interact with procollagen-I (16, 17). Knockdown of calreticulin can reduce collagen secretion by mouse embryonic fibroblasts, and calnexin appears to be involved in directing misassembled procollagen to ER-phagy (18, 19). N-Glycosylation can also intrinsically modulate protein folding and stability via specific interactions with local amino acids and/or via entropic effects (20–23).

Given that 1) the procollagen N-glycan is widely conserved; 2) the C-Pro domain is only present in nascent procollagen molecules, not mature fibrillar collagens; and 3) procollagen interacts with the lectin chaperones, it seems reasonable to assume that the N-glycan plays an essential role in folding the C-Pro domain and thereby templating collagen triple-helix assembly. It was quite surprising, then, when nearly 30 years ago a non-N-glycosylated variant of proα1(I) was shown to fold properly, maintain appropriate hydroxylation, and secrete with kinetics similar to the wild-type (WT), normally N-glycosylated protein (24).

We are left with a decades-old enigma. N-Glycosylation of fibrillar procollagen C-Pro domains is nearly universally conserved and therefore almost certainly functional, but cleavage of the C-Pro domain prior to deposition of collagen triple helices means that the N-glycan must not have a function within the extracellular matrix. Meanwhile, the N-glycan apparently does not meaningfully contribute to intracellular procollagen folding. What, then, is the reason for N-glycosylation of the most abundant protein in the human body?

We hypothesized that the function of the N-glycan might be apparent only in the right proteostatic context. In particular, we speculated that N-glycan–mediated access to the ER’s lectin-based chaperone network could be essential when procollagen folding is challenged, such as in the context of misfolding-prone C-Pro domain mutations and/or ER stress. To test this hypothesis, we generated a non-N-glycosylated procollagen variant via genetic substitution of the native acceptor asparagine with glutamine. We then expressed either N-glycosylated or non-N-glycosylated versions of osteogenesis imperfecta (OI)-causing, misfolding-prone proα1(I) variants in human cells and assayed the secretion of both the full-length protein and the isolated C-Pro domain. We observed that misfolding-prone C-Pro variants were highly sensitive to N-glycan removal. Non-N-glycosylated variants proved defective in folding, assembling, and secreting to the extracellular milieu, instead forming insoluble intracellular aggregates. Chemical induction of low levels of ER stress revealed a similarly essential role for the N-glycan in the folding and secretion of even WT proα1(I) under conditions of proteostatic challenge. Cumulatively, these observations address the N-glycan enigma, unveiling the context-dependent essentiality of procollagen’s conserved N-glycan to enable proper folding and trafficking of procollagen in challenging proteostasis environments.

Results

Absence of the Conserved N-Glycan Impairs Secretion of Misfolding-Prone Collagen Variants.

A number of mutations in the C-Pro domain of procollagen-I are known to promote misfolding, with many leading to the debilitating disease OI (25–29). For example, disruption of the disulfide bond network in the C-Pro domain by the C1299W amino acid substitution leads to overmodification of triple helices and delayed procollagen-I secretion, resulting in mild OI (30). We hypothesized that the procollagen-I N-glycan could be critical for assisting the intracellular folding of such misfolding-prone variants.

To test this hypothesis, we created replication-incompetent adenoviruses encoding hemagglutinin (HA) epitope-tagged WT proα1(I) or misfolding-prone C1299W proα1(I), with or without an N1365Q substitution to prevent N-glycosylation (Fig. 2A). The HA tag, located at the proα1(I) N terminus, allows for selective immunoprecipitation (IP) and quantification of the ectopically expressed HA-proα1(I) separate from endogenously expressed proα1(I). We previously showed that introduction of an HA epitope does not disrupt normal procollagen-I folding, modification, or secretion (16). We used these viruses to transduce primary fibroblasts that also natively express WT procollagen-I and then performed pulse-chase experiments to quantify secretion of the HA-tagged proα1(I) variants in this relevant cellular context. We found that removal of the N-glycosylation sequon from WT proα1(I) did not alter its secretion, fully consistent with previous data (24). In contrast, removal of the N-glycosylation sequon from the C1299W variant led to a significant and substantial secretion defect (Fig. 2B and SI Appendix, Fig. S1A). These results suggest that the N-glycan plays a unique role in C1299W proα1(I) folding, a role that is not critical for successful WT proα1(I) folding.

Fig. 2.

Procollagen’s conserved N-glycan is required for folding, assembly, and secretion of misfolding-prone procollagen variants. (A) Substitution of the N-glycan acceptor residue, N1365, with a glutamine prevents proα1(I) from being N-glycosylated. (B) Pulse-chase experiments reveal that removal of the N-glycan from WT and C1299W C-Proα1(I) significantly and substantially reduces secretion of the latter (p_{80 min} = 8 × 10⁻⁵, p_{160 min} = 0.001) but not the former (p_{80 min.} = 0.593, p_{160 min.} = 0.512). Representative autoradiographs are shown. (C) Pulse-chase experiments conducted with C-Proα1(I) expressed alone recapitulate the secretion patterns exhibited by full-length proα1(I). Removal of the N-glycan significantly reduces secretion of C1299W C-Proα1(I) (p_C1299W = 0.003) but not WT C-Proα1(I) (p_WT = 0.113). This effect is consistently observed for other misfolding-prone C-Proα1(I) variants (p_G1272V = 0.005, p_W1275R = 0.001, p_A1286D = 2 × 10⁻⁴) but not for the assembly-deficient D1277H C-Proα1(I) variant (p_D1277H = 0.185). Representative autoradiographs are shown. **p ≤ 0.01, ***p ≤ 0.001, n.s., not significant.

We next sought to ascertain whether the secretion defect observed upon removal of the N-glycan from C1299W was generalizable to other misfolding-prone C-Pro variants. Procollagen genes are large, highly repetitive, guanine/cytosine rich, and not readily amenable to genetic manipulation (17). Moreover, mechanistic analyses are challenging in the context of the full-length protein. Hence, for further studies, we used a more tractable system in which the C-Proα1(I) domain was expressed alone, in the absence of the N-Pro and triple-helical domains. We previously showed that expression of the C-Pro domain alone recapitulates the behavior and proteostasis defects associated with various C-Pro domain mutations in procollagen-I (31). In addition, both we and others have shown that the C-Proα1(I) domain folds and assembles correctly in human embryonic kidney (HEK293) cells and is robustly secreted into the media (6, 7, 32). We confirmed that WT C-Proα1(I) expressed alone was N-glycosylated, and that the N1365Q substitution eliminated N-glycosylation, based on the equivalent shifts in electrophoretic mobility caused by PNGase F digestion and the N1365Q substitution (SI Appendix, Fig. S1B).

We generated HA-tagged constructs encoding WT C-Proα1(I), C1299W C-Proα1(I), and four additional OI-causing C-Proα1(I) variants—G1272V, W1275R, A1286D, and D1277H (26, 27)—all with and without the N1365Q substitution. Pulse-chase experiments revealed that, consistent with our full-length proα1(I) data (Fig. 2B), removal of the N-glycosylation sequon from WT C-Proα1(I) had no impact on its secretion (Fig. 2C and SI Appendix, Fig. S1C). In contrast, removal of the N-glycosylation sequon from misfolding-prone C1299W C-Proα1(I) resulted in a significant and substantial secretion defect. Three of the four additional OI-causing C-Proα1(I) variants (G1272V, W1275R, and A1286D) also exhibited significant and substantial decreases in secretion upon removal of the N-glycan (Fig. 2C and SI Appendix, Fig. S1C), consistent with the notion that the N-glycan is somehow critical for ensuring the successful folding of misfolding-prone procollagen variants.

Notably, removal of the N-glycan from the OI-causing D1277H C-Proα1(I) variant did not significantly alter its secretion (Fig. 2C and SI Appendix, Fig. S1C). Unlike C1299W, G1272V, W1275R, and A1286D C-Proα1(I), which are all misfolding-prone but nonetheless capable of forming properly assembled trimers, D1277H C-Proα1(I) is known to be assembly-defective. The D1277H substitution permanently disrupts the Ca²⁺-binding site in the C-Pro domain that is critical for proper C-Pro assembly (6, 7). Immunoblots of secreted C-Proα1(I) variants confirmed that all the variants except for D1277H assembled into the expected disulfide-linked trimers (SI Appendix, Fig. S1D). D1277H C-Proα1(I) largely failed to assemble properly, instead forming a complex mixture of monomers, disulfide-linked dimers and trimers, and larger disulfide-linked aggregates (SI Appendix, Fig. S1D). Compellingly, these results suggest that, while the presence of the N-glycan substantially assists misfolding-prone C-Proα1(I) variants in navigating challenging protein-folding landscapes, it does not greatly promote the secretion of irrevocably misassembling structures, such as those induced by the D1277H substitution in C-Proα1(I).

Two additional features of the immunoblots (SI Appendix, Fig. S1 D and E) are noteworthy. First, the secretion defect observed in pulse-chase experiments for the non-N-glycosylated, misfolding-prone C-Proα1(I) variants was even more striking by immunoblotting of media, with virtually no protein detected in media immunoblots (SI Appendix, Fig. S1D). We speculated that this observation may be attributable to long-term ER stress caused by production of non-N-glycosylated, misfolding-prone C-Proα1(I) variants reducing protein synthesis over time and/or extensive aggregation of those same variants (33). Consistent with this notion, whereas the non-N-glycosylated, misfolding-prone C-Proα1(I) variants were readily observed in pulse-chase experiments, they were barely detectable by immunoblotting at steady state of the soluble intracellular fraction (SI Appendix, Fig. S1E).

Non-N-Glycosylated, Misfolding-Prone C-Proα1(I) Variants Induce ER Stress and Are Retained as Intracellular Aggregates.

We hypothesized that decreased secretion upon removal of the N-glycan could be attributed to exacerbated misfolding of C-Proα1(I) variants already rendered misfolding-prone by disease-causing mutations. These severely misfolding non-N-glycosylated variants may cause ER stress, leading to induction of the unfolded protein response (UPR). To test this hypothesis, we used qPCR to assess messenger RNA (mRNA) expression levels of UPR target genes (34). We observed that expression of the non-N-glycosylated, misfolding-prone C-Proα1(I) variants induced a substantial increase in mRNA levels of ER stress-responsive transcripts, including those encoding binding immunoglobulin protein (BiP), hypoxia up-regulated protein 1 (HYOU1), and endoplasmin (Grp94) (Fig. 3A and SI Appendix, Fig. S2A). Notably, none of the misfolding-prone variants induced the UPR when the N-glycan was present. Thus, the N-glycan appears to be essential to mitigate ER stress–induced misfolding caused by these mutations within C-Proα1(I).

Fig. 3.

Removal of the N-glycan from misfolding-prone but not WT C-Proα1(I) induces ER stress and intracellular aggregation. (A) mRNA expression levels of the UPR target genes HSPA5, HYOU1, and HSP90B1 show the induction of ER stress upon expression of non-N-glycosylated, misfolding-prone C-Proα1(I) variants (SI Appendix, Fig. S2A). Tg treatment was used as a positive control for ER stress induction. (B) Extended pulse-chase experiments demonstrate the sustained secretion defect exhibited for non-N-glycosylated C1299W C-Proα1(I) and corresponding intracellular retention. Treatment with the proteasome inhibitor MG-132 did not rescue either secreted or detectable intracellular protein levels. Representative autoradiographs are shown. (C) Representative confocal microscopy images showed the formation of intracellular aggregates in cells expressing non-N-glycosylated C1299W C-Proα1(I).

To evaluate whether the non-N-glycosylated, misfolding-prone C-Proα1(I) variants were being targeted to degradation, we treated cells expressing N-glycosylated or non-N-glycosylated WT or C1299W C-Proα1(I) with either MG-132, a proteasome inhibitor, or Bafilomycin A1, an autophagy inhibitor. Neither inhibitor treatment rescued either secreted or intracellular steady-state levels of the non-N-glycosylated C1299W C-Proα1(I) (SI Appendix, Fig. S2 B and C).

We next performed a more comprehensive pulse-chase experiment to monitor the intracellular retention, secretion, and degradation of N-glycosylated versus non-N-glycosylated WT and C1299W C-Proα1(I). We found that secretion of the non-N-glycosylated C1299W C-Proα1(I) plateaued at ∼50% after 80 min. Even after 160 min, ∼40% of the non-N-glycosylated C1299W variant was still retained intracellularly (Fig. 3B). These findings were recapitulated in a much longer 8-h chase: non-N-glycosylated WT C-Proα1(I) was exclusively present in the media after 8 h (SI Appendix, Fig. S2D), whereas large quantities of non-N-glycosylated C1299W C-Proα1(I) were still retained in cells. Treatment with MG-132 in the pulse-chase experiment had no substantial effect on secretion and only very modestly increased intracellular protein levels (Fig. 3B). Together, these results indicate that the reduced secretion of C1299W C-Proα1(I) following removal of the N-glycan can largely be attributed to sustained intracellular retention rather than degradation.

We next used confocal microscopy to visualize the status of the intracellularly retained protein. We costained the HA-tagged C-Proα1(I) variants with antibodies against the ER marker protein disulfide isomerase (PDI) or the Golgi marker GM130. We observed diffuse staining of WT C-Proα1(I), both with and without the N-glycan, that was primarily localized to the Golgi (Fig. 3C), consistent with previous observations of well-behaved, robustly secreted WT collagen (18, 31, 35). N-Glycosylated C1299W C-Proα1(I) stained diffusely within both the Golgi and the ER, consistent with a longer ER residence time and the excessive posttranslational modification previously observed with this variant (30), but also indicating that the protein was successfully transiting the secretory pathway. In contrast, the non-N-glycosylated C1299W C-Proα1(I) exhibited punctate staining, consistent with the formation of extensive intracellular aggregates.

N-Glycosylation Promotes the Interaction of Misfolding-Prone C-Proα1(I) with the Lectin Chaperones.

Formation of intracellular aggregates, coupled with induction of ER stress, strongly suggests that misfolding of these C-Proα1(I) variants is badly exacerbated by the absence of the N-glycan. We hypothesized that such misfolding could stem from a failure to effectively engage the ER’s lectin-based chaperone machinery—the calnexin/calreticulin cycle (Fig. 4A). Nascent N-glycoproteins carry a 14-residue precursor oligosaccharide, which is processed in the ER by cleavage of the two terminal glucose residues by glucosidases (GS)-I and -II. Calnexin and calreticulin bind the resultant 12-residue sugar on N-glycoproteins and, together with ERp57, ERp29, and CypB, assist client protein folding. Upon release from calnexin or calreticulin, the final glucose residue is cleaved by GS-II, and the folded glycoprotein is then transported to the Golgi for secretion. Misfolded proteins can either reengage with calnexin and calreticulin after readdition of the final glucose residue by UDP-glucose:glycoprotein glucosyltransferase 1 (UGGT1) or be targeted to quality control (15, 36, 37).

Fig. 4.

Procollagen’s N-glycan addresses proteostasis defects by endowing access to the ER’s lectin-based chaperoning system. (A) Schematic of the calnexin/calreticulin cycle and related cochaperones or enzymes. (B) IP of N-glycosylated and non-N-glycosylated WT and C1299W C-Proα1(I) revealed that the misfolding-prone C1299W C-Proα1(I) interacted much more extensively with the lectin-based chaperone system than WT C-Proα1(I) did. These interactions required the presence of an N-glycan on C-Proα1(I). (C) Treating cells with CST, which blocks N-glycan–mediated access to the lectin chaperones, impaired secretion of N-glycosylated C1299W, chemically phenocopying the secretion defect caused by genetic removal of the N-glycan. This effect was observed both by immunoblotting of the media (Left) and by pulse chase (Right, p = 0.001). Representative autoradiographs are provided in SI Appendix, Fig. S3A. (D) A nonnative N-glycosylation sequon can be introduced at C-Proα1(I) residue 1292 via the substitution E1294T. (E) Secretion of the otherwise non-N-glycosylated WT C-Proα1(I) was not significantly affected by introduction of the nonnative N-glycosylation sequon, while secretion of the otherwise non-N-glycosylated C1299W C-Proα1(I) was restored to the level of the natively N-glycosylated C1299W variant. This effect was observed both by immunoblotting (Top) and by pulse chase (Bottom, p = 0.003). Representative autoradiographs are provided in SI Appendix, Fig. S3B. Pulse-chase data in Fig. 3E and SI Appendix, Fig. S3B for WT + native N-glycan, WT Δ N-glycan, C1299W + native N-glycan, and C1299W Δ N-glycan are reproduced from Fig. 2C. **p ≤ 0.01, ***p ≤ 0.001, n.s., not significant.

To evaluate whether and how C-Proα1(I) interacts with the lectin-based chaperone system, we immunoprecipitated HA-tagged WT and C1299W C-Proα1(I) from cell lysates and probed for calnexin, calreticulin, and other related cochaperones or enzymes (Fig. 4B). We observed that all of the interactors we probed for interacted strongly with the N-glycosylated, misfolding-prone C1299W C-Proα1(I). In contrast, they only minimally or undetectably engaged WT C-Proα1(I). Removal of the N-glycan completely abolished or strongly reduced interactions between C1299W C-Proα1(I) and components of the lectin-based chaperone system, suggesting that the N-glycan is critical for the engagement of misfolding C-Proα1(I) by this system.

Chemically Abrogating Access to the Calnexin/Calreticulin Chaperoning System Phenocopies Genetic Removal of the N-Glycan.

The observations in Fig. 4B suggest 1) that the lectin chaperones specifically engage misfolding-prone C-Proα1(I) variants and 2) that productive interactions between C-Proα1(I) and these chaperones require the presence of the N-glycan. It follows that the secretion defect we observed for non-N-glycosylated, misfolding-prone C-Proα1(I) variants (Fig. 2) could be attributed to an inability of these proteins to fold in the absence of assistance from the calnexin- or calreticulin-based chaperone machinery.

To differentiate the effects of N-glycan–mediated access to the lectin chaperones from any intrinsic stabilization afforded by the N-glycan, we treated cells with castanospermine (CST) to chemically abrogate access to the calnexin/calreticulin cycle. CST inhibits GS-I and -II, preventing removal of the terminal glucose residues from the N-glycan and thereby ensuring that N-glycosylated proteins cannot access these chaperones via the normal sugar–lectin interaction (38–40). We observed that, upon treatment with CST, the formerly well-behaved N-glycosylated C1299W C-Proα1(I) variant now exhibited a significant secretion defect observable both by pulse-chase experiments and immunoblotting (Fig. 4C and SI Appendix, Fig. S3A), chemically phenocopying the same defect we had previously observed for genetically non-N-glycosylated C1299W C-Proα1(I) (Fig. 2C). Meanwhile, CST treatment did not affect the secretion of WT C-Proα1(I). Thus, the N-glycan is essential for the misfolding-prone C1299W variant specifically because it allows access to the calnexin/calreticulin cycle.

Introduction of a Nonnative N-Glycosylation Sequon Elsewhere in C-Proα1(I) Restores Secretion of Misfolding-Prone, Non-N-Glycosylated C-Proα1(I).

Our data indicate that the N-glycan allows misfolding-prone C-Proα1(I) variants to interact with the ER’s lectin-based chaperone system, thereby promoting proper folding, assembly, and secretion. We next asked whether this effect was specific to the N-glycan located at the native sequon or whether an N-glycan elsewhere within C-Proα1(I) could play a similar role in maintaining proper proteostasis. Notably, not all procollagens have an N-glycosylation sequon at the equivalent of the N1365 site in proα1(I). Several instead have a sequon at a different location in the C-Pro domain (Fig. 1 C and D).

We introduced a nonnative N-glycosylation sequon at residue 1292 in otherwise non-N-glycosylated WT and C1299W C-Proα1(I). We chose position 1292 because 1) it corresponds to the second most common site for N-glycosylation sequons in the fibrillar procollagens (Fig. 1C, see for example N1672 in proα1(V)), 2) it is already an asparagine in the WT C-Proα1(I) sequence, and 3) it is located in a solvent-exposed loop in the “base” region of the C-Pro domain, where an N-glycan likely would not sterically obstruct C-Proα1(I) folding (Fig. 4D). We used site-directed mutagenesis to introduce a E1294T substitution within the WT Asn-Met-Glu sequence at position 1292, thereby creating a nonnative N-glycosylation sequon of Asn-Met-Thr at position 1292 in C-Proα1(I) (Fig. 4D).

Based on electrophoretic mobility and PNGase F digests (SI Appendix, Fig. S1B), we determined that the nonnative sequon was successfully N-glycosylated when the E1294T/N1365Q variant was expressed in HEK293T cells. Both immunoblotting and pulse-chase experiments (Fig. 4E and SI Appendix, Fig. S3B) revealed that WT C-Proα1(I) with the nonnative sequon secreted at a similar level as both the native N-glycosylated and non-N-glycosylated WT C-Proα1(I) variants. Remarkably, introduction of the nonnative sequon in otherwise non-N-glycosylated C1299W C-Proα1(I) was sufficient to restore secretion to similar levels as natively N-glycosylated C1299W (Fig. 4E). This result is consistent with the notion that the key requirement for solving proteostasis defects in the C-Proα1(I) domain is access to the calnexin/calreticulin cycle via an N-glycan; the specific location of the N-glycan is flexible.

The N-Glycan Is Essential for Maintaining WT Procollagen Proteostasis in the Context of ER Stress.

Collectively, our results demonstrate that procollagen N-glycosylation is required in order to preserve proper folding and secretion when misfolding-prone mutations occur in disease. They do not, however, establish a functional role for the N-glycan in WT procollagen. The demonstrated need for the N-glycan when procollagen misfolding is exacerbated does, however, suggest a hypothesis that we found quite compelling—specifically, the proper folding and secretion of even WT procollagen might be heavily reliant on the presence of the N-glycan in settings where general ER proteostasis is impaired. In this regard, it is noteworthy that collagen-producing tissues often experience physiological ER stress and chronic UPR activation owing to the high secretory output of extracellular matrix proteins that accompanies development and extracellular matrix remodeling processes (41–50).

To test this hypothesis, we first used immunoblotting to assess secretion of WT proα1(I) with or without the N-glycan in either the absence or presence of a low level of ER stress, induced by treatment with 750 nM thapsigargin (Tg), a Ca²⁺ homeostasis disruptor and well-known ER stress-causing UPR activator. ER stress induction was confirmed by a Tg-induced increase in the protein and mRNA transcript levels for Grp94, BiP, and other UPR-responsive genes (SI Appendix, Fig. S4 A and B). We found that secretion of both N-glycosylated and non-N-glycosylated proα1(I) was decreased by Tg treatment. However, non-N-glycosylated proα1(I) was much more sensitive to ER stress than N-glycosylated proα1(I). Only ∼15% of non-N-glycosylated proα1(I) was secreted under Tg treatment compared to vehicle (dimethyl sulfoxide, DMSO) treatment, whereas N-glycosylated proα1(I) was able to maintain nearly 50% of basal secretion (Fig. 5A). This observation was confirmed in pulse-chase experiments, in which non-N-glycosylated WT proα1(I) again proved more sensitive to induced ER stress than did the normally N-glycosylated proα1(I) (Fig. 5B and SI Appendix, Fig. S4C). Thus, the procollagen-I N-glycan protects against WT collagen misfolding when proteostasis is challenged by ER stress.

Fig. 5.

Procollagen’s N-glycan protects WT proα1(I) against misfolding in the context of ER stress. Primary fibroblasts were transduced with adenoviruses encoding WT, full-length proα1(I) with or without the N-glycosylation sequon and then treated with Tg (750 nM) to induce ER stress. (A) Immunoblotting of media indicated that secretion of both N-glycosylated and non-N-glycosylated proα1(I) was impaired by Tg treatment, with the non-N-glycosylated proα1(I) displaying a much larger defect (Top, p = 0.008). Quantification from three biological replicates is shown (Bottom) alongside a representative immunoblot. The up-regulation of BiP and Grp94 in cell lysates confirmed the UPR induction caused by Tg treatment (SI Appendix, Fig. S4B). (B) Pulse-chase analyses of N-glycosylated versus non-N-glycosylated proα1(I) likewise confirmed a significant difference in the fraction of proα1(I) secreted during Tg treatment relative to DMSO treatment at 80 min (p = 0.017). *p ≤ 0.05, **p ≤ 0.01, n.s., not significant.

Discussion

The high conservation of the N-glycosylation sequon in the C-Pro domain of the fibrillar procollagens has been noted for decades (9, 10), yet previous attempts to establish a function for the N-glycan have proved inconclusive. Lamandé and Bateman revealed that genetic removal of the N-glycan attachment site does not alter assembly, secretion, propeptide processing, or matrix deposition of WT collagen-I in normal in vitro model systems (24). Separately, the cleaved C-Pro domain has been proposed to function as an extracellular signal to regulate collagen synthesis (51–54) or to promote C-Pro uptake and degradation (55), suggesting that the N-glycan could theoretically have a role in C-Pro recognition by a receptor. However, the portion of the C-Pro domain that is important for these signaling processes does not include the N-glycan (52, 53).

Here, we demonstrate that the procollagen-I N-glycan has an essential intracellular role in maintaining procollagen folding and secretion under conditions of proteostatic challenge, such as those incurred during disease or ER stress. We first confirmed that the N-glycan is dispensable for folding and secretion of WT procollagen-I, fully consistent with previous results from Lamandé and Bateman (24). We then used OI-causing variants of procollagen-I, specifically those located within the C-Pro domain, as models for conditions under which procollagen folding is challenged. These variants exhibit delayed folding but are ultimately able to trimerize and pass cellular quality control checkpoints required for secretion (26, 30). We discovered that removal of the N-glycan from these misfolding-prone proα1(I) variants led to dramatically decreased secretion of both the full-length protein and the C-Pro domain when expressed alone. In contrast, secretion of the irrevocably assembly-defective D1277H variant does not benefit as significantly from the presence of the N-glycan. In addition, we observed activation of ER stress pathways and formation of intracellular aggregates upon expression of non-N-glycosylated, misfolding-prone C-Proα1(I) variants.

We hypothesized that the secretion defect observed for the non-N-glycosylated, misfolding-prone C-Proα1(I) variants could be attributed to either intrinsic destabilization upon loss of the N-glycan or exacerbated misfolding due to the inability of non-N-glycosylated variants to access lectin-based chaperones in the ER. IP experiments revealed that the misfolding-prone C1299W C-Proα1(I), but not WT, associated extensively with the ER’s lectin-based chaperone machinery in an N-glycan–dependent manner. Chemical inhibition of lectin chaperone association using CST in the context of N-glycosylated C1299W C-Proα1(I) fully phenocopied the secretion defect we observed for non-N-glycosylated C1299W C-Proα1(I). Taken together, these data indicate that misfolding-prone C-Proα1(I) variants depend heavily on N-glycan–mediated access to the lectin chaperones in order to fold, assemble, and pass quality control checkpoints required for secretion. We further found that the specific location of the N-glycan is not critical for this assistance, merely its presence.

The intracellular C-Proα1(I) aggregates we observed in the absence of the N-glycan were not readily cleared by either ER-associated degradation (ERAD) or autophagy, even though both pathways are reported to have roles in procollagen-I quality control (19, 56–58). Calnexin has notably been found to interact with FAM134B, an ER-phagy receptor, thereby targeting misfolding procollagen to an autophagic fate that can be inhibited by bafilomycin A1 (19). Such clearance would likely be slowed or eliminated by removal of the N-glycan, which reduces binding to calnexin. Separately, another recent report of autophagic procollagen clearance that is bafilomycin A1 insensitive suggests that noncanonical mechanisms may be involved (56). Regardless, clearance by any of these or other pathways is very slow for the variants studied here, requiring >8 h after synthesis of non-N-glycosylated, misfolding-prone C-Proα1(I).

Misfolding-prone procollagen-I variants are found only in a very small fraction of the human population, suggesting that the highly conserved N-glycan is unlikely to function only in this context. Indeed, we find that the N-glycan is also critical for proper folding and secretion of WT procollagen-I under conditions of mild ER stress. ER stress and UPR activation occur in various commonly encountered physiological contexts, including differentiation of collagen-producing cells such as osteoblasts, chondrocytes, and fibroblasts (42–45). Chronic activation of the UPR has additionally been observed in connection with processes that require synthesis of large amounts of matrix proteins, such as development (41, 49) and wound healing (46, 47, 50). The N-glycan thus appears to be crucial for maintaining WT procollagen proteostasis and secretion during these critical, yet commonplace, circumstances.

Our data point to the N-glycan as a key structural motif responsible for enabling procollagen folding under challenging proteostasis conditions, adding to the repertoire of critical functions for procollagen’s C-Pro domain. Under normal conditions, the highly conserved N-glycan on procollagen-I is dispensable because the protein can fold and secrete efficiently without relying on lectin chaperones. However, when protein folding is challenged, whether by disease-associated mutations or ER stress, the N-glycan becomes essential for preventing procollagen misfolding by providing access to the lectin chaperone network in the ER (Fig. 6). Absence of the N-glycan not only exacerbates the folding defect but also disrupts procollagen quality control, leading to extensive intracellular aggregation. Taken together, these observations highlight the remarkable efficacy of procollagen proteostasis, wherein multiple systems cooperate to safeguard the production of a critical protein under changing conditions. Moreover, they suggest the potential of modulating the levels and/or activities of the ER’s lectin chaperone network as a strategy to resolve dysregulated procollagen proteostasis in the collagenopathies (29). In conclusion, our results reveal a fundamental, context-dependent role for the highly conserved procollagen N-glycan, whose implications for cellular signaling, development, and collagen misfolding diseases provide ample subject for future work.

Fig. 6.

Model for the context-dependent essentiality of procollagen’s conserved N-glycan. Under normal conditions, procollagen folds and secretes properly without significant aid from N-glycan–mediated access to the lectin chaperones. When folding is challenged, whether by misfolding-prone mutations or physiological ER stress, the N-glycan is essential to allow procollagen to associate with lectin chaperones in the ER to promote proper folding, assembly, and secretion. The inability to access lectin chaperones leads to aggregation and impaired secretion.

Materials and Methods

Detailed information on sources for materials used and detailed protocols for experimental procedures can be found in SI Appendix. These include a description of cell lines and reagents used and sequence information for constructs employed. Procedures for sequence alignments, vector construction, adenovirus production, transfection and cell culture, immunoblotting, pulse-chase experiments, RNA harvesting and qPCR assays, immunohistochemistry, and IPs are provided.

Data Availability

All study data are included in the article and/or SI Appendix.