Mapping the Effect of Gly Mutations in Collagen on α2β1 Integrin Binding

Abstract

The replacement of one Gly in the essential repeating tripeptide sequence of the type I collagen triple helix results in the dominant hereditary bone disorder osteogenesis imperfecta. The mechanism leading to pathology likely involves misfolding and autophagy, although it has been hypothesized that some mutations interfere with known collagen interactions. Here, the effect of Gly replacements within and nearby the integrin binding GFPGER sequence was investigated using a recombinant bacterial collagen system. When a six-triplet human type I collagen sequence containing GFPGER was introduced into a bacterial collagen-like protein, this chimeric protein bound to integrin. Constructs with Gly to Ser substitutions within and nearby the inserted human sequence still formed a trypsin-resistant triple helix, suggesting a small local conformational perturbation. Gly to Ser mutations within the two Gly residues in the essential GFPGER sequence prevented integrin binding and cell attachment as predicted from molecular dynamics studies of the complex. Replacement of Gly residues C-terminal to GFPGER did not affect integrin binding. In contrast, Gly replacements N-terminal to the GFPGER sequence, up to four triplets away, decreased integrin binding and cell adhesion. This pattern suggests either an involvement of the triplets N-terminal to GFPGER in initial binding or a propagation of the perturbation of the triple helix C-terminal to a mutation site. The asymmetry in biological consequences relative to the mutation site may relate to the observed pattern of osteogenesis imperfecta mutations near the integrin binding site.

Keywords: binding; collagen; extracellular matrix; integrin; missense mutations; molecular dynamics; osteogenesis imperfecta; recombinant protein expression; triple helix.

FIGURE 1.

A, schematic diagram of recombinant bacterial collagen VCL-Int with the insertion of human α1(I) chain residues from Gly-496 to Pro-513 (red), including the integrin α2β1 binding sequence, GFPGER. Residue numbers are taken from the UniProt entry P02452, and numbering starts from the beginning of the triple helix region. Bacterial collagen residues (blue) are denoted by their position relative to the human sequence insertion G(−12)POG(−9)PRG(−6)EQG(−3)PQGARGERGFPGERGVQGPP. Underlined Gly residues are mutated to Ser individually. B, SDS-polyacrylamide gel showing all the recombinant proteins, including VCL-Int and the proteins with Gly replacements. The right lane contains the Novex Sharp protein standard (Invitrogen); collagen monomer chains run slower than expected as reported previously. C, MALDI-TOF spectrometry of VCL, VCL-Int, and a representative mutated protein (G505S). a.u., arbitrary units.

FIGURE 2.

Structural characterization of VCL-Int and a representative mutated construct (G505S). A, CD thermal transitions of VCL-Int and G505S showing triple helix unfolding around 37 °C. The increase in ellipticity after 37 °C is due to the unfolding of the α-helix in the V domain. B, differential scanning calorimetry analysis showing thermal transitions of 37.1 °C for VCL-Int and 35.3 °C for G505S. C, SDS-PAGE of VCL-Int and VCL-Int G505S after trypsin digestion for time t = 0, 2, and 15 min at 20 °C. Similar results were observed for G496S, G499S, G502S, G508S, and G511S (data not shown). A small initial drop in intensity is observed that could reflect rapid digestion of a small amount of impurity or unfolded collagen. There is little change in the intensity following this initial decrease, suggesting that the mutations did not lead to major unfolding.

FIGURE 3.

Solid-state analysis of recombinant integrin α2 I domain binding to native and denatured recombinant collagens that are immobilized on 96-well plates and incubated with recombinant I domain (10 μg/ml). Anti-GST antibody was used to detect the bound I domain. Proteins were immobilized in their native state (black columns), in the denatured state (dotted columns), and in the absence of the Mg²⁺ cation (white columns). The native VCL showed no binding, whereas native VCL-Int showed a high binding similar to native type I collagen. Native G505S had a low binding, whereas native G511S had a binding similar to VCL-Int. All proteins in their denatured states showed no binding as triple helicity is essential for integrin binding. The divalent cation (Mg²⁺) is essential for the MIDAS motif, and BSA was used as a background value. Statistical analysis was performed by paired t test. The binding of G505S was significantly different from the binding of VCL-Int (p ≤ 0.01). All experiments were in triplicate, and error bars represent the standard deviation.

FIGURE 4.

A, solid-state analysis of recombinant α2 I domain binding to all recombinant bacterial collagens containing Gly to Ser substitutions at an I domain concentration of 0. 4 μm. The two striped columns on the left are the negative control VCL with no binding and the wild-type VCL-Int with strong binding. Experiments were carried out in triplicate, and error bars represent the standard deviation. Statistical analysis was performed with the paired t test. Columns marked with an asterisk showed a statistically significant difference in binding compared with the VCL-Int control (p ≤ 0.01). B, dose response of I domain binding to recombinant collagens adsorbed onto 96-well plates. Bound protein was detected with antibodies and measured as absorbance at 450 nm. Red, collagen type I; green, VCL-Int; cyan, G508S; purple, G505S; orange, G(−9)S; blue, G(−15)S. Data were fit in GraphPad Prism® using a non-linear fit (one site, specific binding). Error bars are not included because only one set of experiments can be carried out on a given 96-well plate, and it is not possible to average values between different plates due to slight changes in conditions. All assays have been independently repeated at least three times on new plates to confirm that results are consistent.

FIGURE 5.

HT1080 cell adhesion assay shows the in vitro binding profile of recombinant proteins with and without mutations. Live imaging (shown on top) was performed after calcein-AM staining of bound cells. Scale bars, 50 μm. Quantitative analysis for the cell attachment was performed by ImageJ. Experiments were carried out in triplicate with error bars representing the standard deviation and statistical analysis performed with the paired t test. G505S had a statistically significant decrease in cell adhesion compared with VCL-Int (p ≤ 0.05).

FIGURE 6.

A, the triple helix structure of the wild-type collagen peptide generated with the Triple-Helical Collagen Building Script (THeBuScr) (47). Chains A, B, and C are colored in blue, green, and red, respectively. B, interchain backbone NH···CO interactions are illustrated. Different backbone NH···CO interactions are disrupted in the mutants depending on the location of the substitution. Backbone hydrogen bonds whose average N···O distances were ≥3.5 Å during the course of simulation are labeled with red dashed lines to indicate their disruption. The hydrogen bonding is shown only for chains A and B of the triple helix because these are the chains that interact with integrin (23).

FIGURE 7.

A, the key interactions that participate in the integrin-collagen binding and the residues involved in each interaction. The residues from integrin and chains A and B of collagen are colored in black, green, and blue, respectively. All these interactions are observed to be stable in the 100-ns MD simulation of integrin with WT collagen (indicated by + symbols). The van der Waals interaction of collagen Phe-503^B with integrin Gln-215 and Asn-154 is disrupted by the Gly to Ser mutations at position 502 (indicated by the − symbol). B and C, the simulation structures of integrin with WT collagen (B) and integrin with G502S mutant (C). Chains A, B, and C for collagens are colored in blue, green, and red, respectively. The Mg²⁺ ions are represented by magenta spheres. The substituted Ser residues are shown in ball-and-stick. The disrupted van der Waals interaction is illustrated.

FIGURE 8.

Top, known OI mutations in the integrin binding region as reported in the OI database (5, 6). The residue replacing Gly is shown below together with the Sillence classification of the phenotypic severity of the OI (OI II, perinatal lethal; OI III, severe; OI IV, moderate; OI I, mild) (7). Middle, the experimental effect of Gly to Ser mutations on integrin binding as reported here is indicated with + for normal binding and − for weak or no binding. Bottom, the molecular model of collagen-integrin binding along the triple helix shows the distance of N-terminal mutations from the well defined binding site.