A cysteine-based molecular code informs collagen C-propeptide assembly

Abstract

Fundamental questions regarding collagen biosynthesis, especially with respect to the molecular origins of homotrimeric versus heterotrimeric assembly, remain unanswered. Here, we demonstrate that the presence or absence of a single cysteine in type-I collagen's C-propeptide domain is a key factor governing the ability of a given collagen polypeptide to stably homotrimerize. We also identify a critical role for Ca²⁺ in non-covalent collagen C-propeptide trimerization, thereby priming the protein for disulfide-mediated covalent immortalization. The resulting cysteine-based code for stable assembly provides a molecular model that can be used to predict, a priori, the identity of not just collagen homotrimers, but also naturally occurring 2:1 and 1:1:1 heterotrimers. Moreover, the code applies across all of the sequence-diverse fibrillar collagens. These results provide new insight into how evolution leverages disulfide networks to fine-tune protein assembly, and will inform the ongoing development of designer proteins that assemble into specific oligomeric forms.

Fig. 1

C-Pro domain-mediated assembly of collagen type-I. a Schematic representation of collagen-I assembly. Two strands of Colα1(I) and one strand of Colα2(I) typically assemble into heterotrimers, a process that is initiated by the respective C-Pro domains. Colα1(I) is also known to form homotrimers, whereas Colα2(I) does not homotrimerize and only forms heterotrimers. The crystal structure of a homotrimeric C-Pro domain is used here for illustration purposes (PDBID 5K31), with each collagen-I C-Pro domain differentially colored to demonstrate possible assembly schematics. b Alignment of the Colα1(I) and Colα2(I) C-Pro domains highlights high sequence similarity. The cysteine network is numbered from C1 to C8 in the N-terminal to C-terminal direction, with each cysteine residue colored in red

Fig. 2

Phylogenetic analysis of the cysteine network of the C-Pro domain of collagen-I. a Crystal structure of C-Proα1(I) (top; PDBID 5K31) oriented to display the interstrand disulfide bonds (shown in stick mode) between the C2 and C3 residues (numbered as in Fig. 1b) of neighboring polypeptides that covalently stitch the identical subunits together. Each subunit is colored in a different shade of green to facilitate visualization. b Cladogram illustrating organismal evolution of animals. Each clade of the tree is listed at each branch point, with the organisms likely to have heterotrimeric collagen-I highlighted by the yellow box. c Alignment of ancient collagen-I C-Pro sequences, beginning with the earliest likely emergence of heterotrimeric collagen-I (chordates). Conserved cysteine residues are highlighted in yellow, with the cysteine number indicated at the top of the alignment. Sites where C2 (or C3) are not conserved are bolded. Conserved amino acids involved in the Ca²⁺ coordination network are marked with a “•” below the alignment (see also Fig. 4 for a representative image of the Ca²⁺ coordination). Collagen C-Pro domains were aligned using Clustal Omega

Fig. 3

The presence or absence of a single cysteine residue defines the capacity of collagen-I C-Pro domains to form stable, disulfide-linked homotrimers. a Immunoblot analysis of individually expressed wild-type, HA-tagged C-Proα1(I) (red) and wild-type, FLAG-tagged C-Proα2(I) (green) proteins under non-reducing and reducing conditions showing that wild-type C-Proα1(I) forms a disulfide-linked homotrimer, whereas wild-type C-Proα2(I) does not, recapitulating the known disulfide-dependent assembly patterns of full-length Colα1(I) and Colα2(I). Co-expression of wild-type C-Proα1(I) and wild-type C-Proα2(I) rescues wild-type C-Proα2(I) into a heterotrimer, as shown by the yellow color indicating red and green overlap. b Immunoblot analysis of individually expressed HA-tagged Cys1265Ser (C2S) C-Proα1(I) (red) and FLAG-tagged Ser1169Cys (S2C) C-Proα2(I) (green) proteins under non-reducing and reducing conditions showing that the serine variant of C-Proα1(I) is no longer able to form a disulfide-linked homotrimer. In contrast, the cysteine variant of C-Proα2(I) is able to form disulfide-linked homotrimers. Co-expression of C2S C-Proα1(I) and S2C C-Proα2(I) rescues C2S C-Proα1(I) into a heterotrimer, as shown by the yellow color indicating red and green overlap

Fig. 4

The essential role of Ca²⁺ in templating non-covalent C-Pro assembly. a Crystal structure of C-Proα1(I) (PDBID 5K31), highlighting the presence of a Ca²⁺ ion at each subunit interface of the homotrimer. Each subunit is colored in a different shade of green with Ca²⁺ ions in blue. Inset: residues involved in binding Ca²⁺. In b–c, the slopes are directly proportional to the weight average molecular weight at each point. b Sedimentation equilibrium data showing that wild-type C-Proα1(I) is best fit as a single homotrimeric species in the absence of Ca²⁺ (blue) and in the presence of Ca²⁺ (red). The best fit is shown as a solid line. c Sedimentation equilibrium data showing that S2C C-Proα2(I) is best fit as a single homotrimeric species both in the absence of Ca²⁺ (blue) and in the presence of Ca²⁺ (red). The best fit is shown as a solid line in the corresponding color. In d–e, the slopes, which vary at each radial position, are directly proportional to the weight average molecular weight for the total protein concentration at that position. d Sedimentation equilibrium data for wild-type C-Proα2(I) in the absence of Ca²⁺ are best fit as a monomer–dimer equilibrium (primarily monomer; see Supplementary Fig. 4) with an association constant sufficiently small to be approximated as a single species of intermediate weight-average molecular weight, with no evidence of a trimeric species, as shown by the best fit in blue. In the presence of 0.5 mM Ca²⁺ (red), the data are best fit as a monomer–trimer equilibrium (primarily trimer; see Supplementary Fig. 4). e Sedimentation equilibrium data for C2S C-Proα1(I) in the absence of calcium are best fit as a monomer–dimer equilibrium (primarily monomer; see Supplementary Fig. 5) with an association constant sufficiently small to be approximated as a single species of intermediate weight-average molecular weight, with no evidence of trimeric species, as shown by the best fit in blue. In the presence of 0.5 mM Ca²⁺ (red), the data are best fit as a monomer–trimer equilibrium (primarily trimer; see Supplementary Fig. 5). Only every third data point is shown (b–e). f Size exclusion chromatography analysis of Asp1277His C-Proα1(I) in the presence and absence of Ca²⁺, as compared to wild-type C-Proα1(I), reveals the inability of this Ca²⁺-binding variant to assemble into a homotrimer, despite the presence of both C2 and C3 (left). Immunoblot analysis under non-reducing conditions show the predominant species is a dimer, with other oligomeric products also present (right)

Fig. 5

Co-expression analysis of heterotrimeric assembly products. This figure is supplemented with arrows from a to b and from b to c to indicate that the samples generated from each stage were then used in the subsequent analysis. Both wild-type and C2S C-Proα1(I) were tagged with HA. Both wild-type and S2C C-Proα2(I) were tagged with FLAG. All C-Proα1(I) signal was detected using an antibody raised against C-Proα1(I) (shown in red). All C-Proα2(I) signal was detected using an antibody raised against the FLAG epitope (shown in green). a Stage 1: Media harvested from cells transfected with the indicated combinations of C-Proα1(I) and C-Proα2(I) was immunoblotted for the presence of each construct on a reducing SDS-PAGE gel. b Stage 2: Immunoprecipitation of C-Proα2(I) from the Stage 1 media samples (a) under non-reducing, Ca²⁺-depleted conditions. Immunoprecipitated samples were immunoblotted for the presence of each construct on a reducing SDS-PAGE gel. c Stage 3: Final step in the purification of heterotrimers via immunoprecipitation of C-Proα1(I) from the eluent obtained in Stage 2 (b). Immunoprecipitated samples were immunoblotted for the presence of each construct on a reducing SDS-PAGE gel. d Quantification of the C-Proα1(I):C-Proα2(I) ratio of the purified heterotrimers obtained in Stage 3 (c). The bar chart shows the average ratio across four biological replicates. ***p value <0.001 as determined by a t test. Error bars indicate standard deviation from the mean for each sample. e Co-immunoprecipitation experiments on heterotrimers formed when wild-type C-Proα1(I) and S2C C-Proα2(I) were co-expressed at ratios of 1:1, 2:1, and 5:1. The conditioned media were first subjected to an HA immunoprecipitation to extract all C-Proα1(I)-containing species, and the supernatant (unbound fraction) was then subjected to a FLAG immunoprecipitation to purify any remaining S2C C-Proα2(I). Elutions from the FLAG immunoprecipitation were analyzed on a non-reducing, SDS-PAGE gel. n.s. = not significant

Fig. 6

A priori prediction of collagen trimerization propensities based on the C2/C3 pattern (C1 through C4 shown). a Known disulfide-bonding network for a homotrimeric collagen strand with C2 and C3 intact, as revealed by the crystal structures of C-Proα1(I) and C-Proα1(III) homotrimers^, . b Predicted disulfide-bonding network for a C-Pro domain lacking a cysteine residue in position 2 or 3. Although intrastrand disulfide bonds likely form, interstrand disulfide bonds cannot form in the absence of C2 or C3. c Predicted disulfide-bonding network for a collagen heterotrimer that requires 2:1 assembly. The accuracy of this predicted disulfide-bonding pattern in 2:1 heterotrimers is supported by previous work modeling the replacement of a single C-Proα1(I) subunit in the homotrimer crystal structure with a C-Proα2(I) monomer. Note the single free cysteine in the assembled protein owing to the presence of an odd number of cysteine residues. d Predicted disulfide-bonding network for a collagen heterotrimer that requires 1:1:1 assembly, illustrating the minimum requirements for stable 1:1:1 heterotrimer formation, which are one strand having cysteine residues at positions 2 and 3, one only at position 2, and one only at position 3. e Alignment of rainbow trout and zebrafish collagen-I C-Pro domains (C1–C4 only shown). The expected cysteine pattern for a 1:1:1 homotrimer in d is present in both instances where collagen-I has three α strands in nature and in human collagen type-V, which all form 1:1:1 heterotrimers. Note that in all images the potentially interstrand disulfide-forming cysteines (C2 and C3) are colored red, while the intrastrand disulfide-forming cysteines are colored blue

Fig. 7

The cysteine-based code for collagen C-Pro assembly is generalizable across the fibrillar collagens. a Alignment of the interstrand disulfide-bonding region of the human fibrillar collagen C-propeptides highlights that C-Pro domains known to homotrimerize contain all four conserved cysteine residues, whereas C-Pro domains known to only form heterotrimers lack a single cysteine at C2 or C3. The residue colored red in each protein sequence corresponds to the mutated residue analyzed in b, c. Amino acid numbering was derived from the corresponding full-length procollagens. b Immunoblot analyses of individually expressed wild-type fibrillar collagen C-Pro domains known to only form heterotrimers. Assembly was analyzed under non-reducing and reducing conditions. Wild-type variants (HA-tagged; red) all migrated as monomers, while all variants in which the missing cysteine residue was re-introduced (FLAG-tagged; green) gained the ability to homotrimerize in a disulfide-dependent manner. c Immunoblot analysis of individually expressed wild-type fibrillar collagen C-Pro domains known to have the capacity to homotrimerize. Assembly was analyzed under non-reducing and reducing conditions. Wild-type variants (HA-tagged; red) all migrated as disulfide-dependent homotrimers, while variants in which a single cysteine residue was mutated to serine (FLAG-tagged; green) lost the capacity to form disulfide-linked homotrimers. Note that C-Proα2(XI) was not secreted and was detected only in cell lysate. d Co-expression of the wild-type C-Proα1(III) domain with the C2S C-Proα1(III) demonstrates the wild-type protein’s ability to rescue the monomeric C2S variant into a disulfide-linked trimer, shown by the overlap of green and red signal (yellow)

Fig. 8

Model reaction coordinate diagram for collagen C-Pro assembly. In the presence of Ca²⁺, collagen C-Pro domains can all transiently trimerize (presumably via initial dimerization; not illustrated here), with the trimers in a dynamic equilibrium with the other states. The non-covalent trimeric assembly products serve as an intermediate to the final covalently immortalized trimers. The transient trimers are bracketed to indicate that their relative energies are not depicted in this plot. Covalent immortalization that is possible only when sufficient cysteine residues are present at positions 2 and 3 to form the requisite disulfides further stabilizes the trimer state, templating collagen triple-helix formation