Molecular underpinnings of integrin binding to collagen-mimetic peptides containing vascular Ehlers–Danlos syndrome–associated substitutions

Collagens carry out critical extracellular matrix (ECM) functions by interacting with numerous cell receptors and ECM components. Single glycine substitutions in collagen III, which predominates in vascular walls, result in vascular Ehlers–Danlos syndrome (vEDS), leading to arterial, uterine, and intestinal rupture and an average life expectancy of <50 years. Collagen interactions with integrin α2β1 are vital for platelet adhesion and activation; however, how these interactions are impacted by vEDS-associated mutations and by specific amino acid substitutions is unclear. Here, we designed collagen-mimetic peptides (CMPs) with previously reported Gly → Xaa (Xaa = Ala, Arg, or Val) vEDS substitutions within a high-affinity integrin α2β1-binding motif, GROGER. We used these peptides to investigate, at atomic-level resolution, how these amino acid substitutions affect the collagen III–integrin α2β1 interaction. Using a multitiered approach combining biological adhesion assays, CD, NMR, and molecular dynamics (MD) simulations, we found that these substitutions differentially impede human mesenchymal stem cell spreading and integrin α2–inserted (α2I) domain binding to the CMPs and were associated with triple-helix destabilization. Although an Ala substitution locally destabilized hydrogen bonding and enhanced mobility, it did not significantly reduce the CMP–integrin interactions. MD simulations suggested that bulkier Gly → Xaa substitutions differentially disrupt the CMP–α2I interaction. The Gly → Arg substitution destabilized CMP–α2I side-chain interactions, and the Gly → Val change broke the essential Mg2+ coordination. The relationship between the loss of functional binding and the type of vEDS substitution provides a foundation for developing potential therapies for managing collagen disorders.

vEDS substitution provides a foundation for developing potential therapies for managing collagen disorders. 4 is life-threatening and results from abnormal synthesis of, or pathogenic mutations to, collagen III in blood vessel walls and distensible organ linings. This leads to arterial aneurysms; rupture to arterial, uterine, and intestinal walls; and thin, translucent skin (1,2). vEDS is one of several debilitating genetic collagen diseases primarily caused by single Gly 3 Xaa mutations in the triplehelical (Gly-Xaa-XaaЈ) n repeating domain of fibrillar collagens that include osteogenesis imperfecta (OI), achondrogenesis type II, spondyloepiphyseal dysplasia syndrome, and Stickler syndrome. Because glycines stabilize the interior of collagen triple helices through an intricate hydrogen-bonding network (3)(4)(5) in the canonical (GXXЈ) n repeating domain, mutation of conserved glycines in fibrillar collagens may cause deficiencies in collagen structure, assembly, or production (6 -8).

Vascular Ehlers-Danlos syndrome (vEDS)
The specific triple-helical conformation of collagen is important for recognition by its numerous binding partners in the extracellular matrix (ECM). Interactions of collagen III with cellular receptors are critical for cell homeostasis, wound healing, and platelet adhesion and activation (9 -12). In this study, we focus on the collagen III interaction with the endothelial and platelet receptor integrin ␣ 2 ␤ 1 , which plays a pivotal role in cell-ECM adhesion and firm platelet arrest, adhering to exposed subendothelial ECM upon vascular injury (13)(14)(15)(16). The collagen-binding integrins interact with native collagens via their inserted (I) domain located in the ␣ subunit (␣I) of the integrin headpiece (17)(18)(19)(20)(21). Isolated, ␣I retains the specificity and affinity of the parent integrins for collagen and has been used as a model of integrin in biological and structural studies (22)(23)(24)(25)(26). Integrins recognize a specific binding motif in the collagen sequence, GXXЈGEXЉ (in which X is any amino acid, XЈ is usually Hyp, and XЉ is frequently Arg or Asn), and the binding is metal-mediated, in which the Glu of collagen is coordinated to This work was supported in part by National Institutes of Health Grant GM 45302 (to J. B.) and American Heart Association Postdoctoral Fellowship 17POST33410326 (to C. L. H.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S6. 1  cro ARTICLE a divalent metal cation in the metal ion-dependent adhesion site (MIDAS) of the ␣I (27,28). The collagen III sequence GROGER has been determined to be a high-affinity integrinbinding site (29) and is in a region that is especially overrepresented in reported cases of Gly 3 Xaa mutations (Fig. 1A). Here, we investigated how Gly 3 Xaa mutations within this domain impact integrin ␣ 2 ␤ 1 binding and the molecular basis of any functional interference.
Previous studies have shown that substitution of these integral Gly residues causes structural deformities to the triple helix, and the extent of triple-helix destabilization depends on the amino acid substituted (30,31) and the surrounding sequence context (32)(33)(34)(35)(36). Recently, elegant studies have investigated the impact of Gly 3 Xaa mutations near integrinbinding sites in collagen I (31, 37) using recombinant bacterial collagen systems, which can incorporate long spans of collagenlike sequence and can probe selective mutation sites. However, these models lack hydroxyprolines that are important for triple-helix folding and are unable to model the heterotrimeric collagen I triple helix. Here, we use collagen-mimetic peptides (CMPs) that incorporate a native collagen III sequence. These are good physiological models of collagen III fragments given its homotrimeric nature, and incorporation of native hydroxyprolines allows us to more closely model effects of Gly 3 Xaa mutations on triple-helix structure and dynamics.
Structural consequences of Gly substitutions in collagen triple helices have been extensively studied in the context of OI mutations to collagen I (35,(38)(39)(40)(41)(42). OI is an autosomal dominant disorder, resulting from mutations to COL1A1 or COL1A2 genes. Because collagen I is a heterotrimer of ␣1(I) and ␣2(I) ␣-chains, only one or two mutant ␣-chains will be incorporated into the collagen I triple helix. In the homotrimeric collagen III, vEDS mutations may result in incorporation of up to three mutant ␣-chains. For OI, clinical phenotypes have been classified based on severity, and larger Gly substitutions more frequently result in lethal OI (43)(44)(45). No classification system has been established for vEDS. However, a clinical study found that survival of individuals with vEDS Gly 3 Xaa mutations was correlated with the substituting amino acid identity, with the trend Ser Ͼ Arg/Asp Ͼ Val (46). That is, those individuals with large hydrophobic Val and charged Arg or Asp substitutions exhibited shorter life spans than those with small Ser substitutions (46). This suggests that the structural impact of different amino acids on the collagen III triple helix may be relevant to disease outcome.
Three amino acid substitutions at Gly 240 , in the integrinbinding GROGER motif, have been associated with vEDS: Ala, Arg, and Val (https://eds.gene.le.ac.uk) 5 (97,98). In this work, we employed CMPs incorporating these naturally occurring vEDS substitutions within the GROGER ␣ 2 I-binding site. We probed how the specific amino acid substitutions impact functional cell spreading and ␣ 2 I adhesion and elucidated the underlying atomic-level structural and dynamic perturbations to the collagen triple helix and the CMP-␣ 2 I interaction interface. We found that although the Ala mutation locally disrupts the canonical triple-helix hydrogen bonding and enhances the backbone dynamics, it is not sufficient to significantly reduce ␣ 2 I interactions, in the context of recombinant protein or on human mesenchymal stem cells (hMSCs). However, larger substitutions, Arg and Val, do impede ␣ 2 I interactions through different mechanisms involving breakage of side-chain interactions or Mg 2ϩ coordination, as suggested by MD simulations. Novel insight into this sequencestructure/dynamics-function relationship sets the foundation for new drug therapy techniques to combat collagen disorders that lead to compromised collagen interactions.
Gly 3 Xaa mutations within the collagenous domain are not evenly distributed between amino acid types but rather are highly skewed toward charged amino acids, which make up 61% of Gly 3 Xaa mutations of all cases reported (Fig. 1C). To determine the impact of reported vEDS mutations of different amino acid types on triple-helix conformation and function, we pursued a sequence-structure/dynamics-function study of CMPs that contain an integrin ␣ 2 ␤ 1 -binding domain in the context of the collagen III sequence.

Gly 3 Xaa vEDS mutations reduce binding of the integrin ␣ 2 I domain and hinder cell spreading
The collagen III sequence 237 GROGER 242 was identified to be an ␣ 2 I-binding site by rotary shadowing and electron 5 Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site.

Impact of vEDS mutations on collagen-integrin interactions
microscopy (EM) and was determined to be a high-affinity binding motif through solid-phase assays using synthetic CMPs (29). Previously, a CMP containing the native GROGER sequence was found to be biologically functional, as it supported cell adhesion to human lung fibroblast MRC-5 cells (29), which express ␣ 2 ␤ 1 and ␣ 1 ␤ 1 integrins (67). This binding site is in the N-terminal region of collagen III in which 27 consecutive glycines have at least one reported mutation that results in vEDS. We have designed a series of CMPs, referred to as T3-237, that are composed of 12 native residues of the collagen III sequence, including the high-affinity ␣ 2 I-binding motif, 237 GROGERGLOGPO 248 , flanked by (GPO) 4 at the N terminus and (GPO) 3 at the C terminus to promote triple-helix formation, GPC at the N and C termini for enhanced adhesion to microtiter plates (68), and GY at the C terminus to monitor peptide concentration by absorbance. We have incorporated the three naturally occurring mutations at Gly 240 : T3-237 G240A, G240V, and G240R. We expected that mutation of this binding motif may hinder the ability of integrin ␣ 2 ␤ 1 to interact with collagen III at this site. We investigated the adhesion and spreading behavior of hMSCs, which express this integrin receptor (69,70), on three of these synthesized CMPs, T3-237 WT, G240A, and G240V, as well as a positive control (collagen III) and negative controls (BSA and a hexapeptide, GROGER, which cannot form the necessary triple helix for cell binding) conditions. Cultures were followed for 1 day, with representative images captured at 4, 7, and 24 h after cell seeding. After 4 h, hMSCs cultured on collagen III exhibited attachment and cell morphology consistent with strong adhesion ( Fig. 2A). By comparison, cells attached to the negative controls were fewer in number and displayed a "pancake"shaped morphology that is typical of weak adhesion (Fig. 2, B and C). When grown on T3-237 WT and G240A, hMSCs showed cell spreading trends comparable with collagen III (Fig. 2, D and E), whereas the cell morphology on G240V was consistent with the negative controls (Fig. 2F). Similar obser-vations were made from images collected after 7 h in culture. Interestingly, by 24 h, cells on all conditions, including the negative controls, demonstrated attachment and morphology consistent with the collagen III condition, perhaps indicating that the hMSCs had deposited new matrix to foster cell adhesion.
To probe the impact of the collagen III mutations specifically on ␣ 2 I domain binding, we used in vitro solid-phase enzymelinked immunosorbent assays (ELISAs). The ␣ 2 I-collagen interaction depends upon coordination of a divalent metal cation with the Glu in the recognition domain (27,28). Therefore, binding assays were performed in the presence and absence of Mg 2ϩ cations to control for unspecific binding. Specificity for the GXXЈGEXЉ binding motif was also assessed by a negative control, (GPP) 10 , which does not contain an integrin-binding motif, and BSA, a globular protein with no known specific interaction for the integrin ␣ 2 I domain. Both negative controls show significantly reduced adhesion relative to T3-237 WT. The relative adhesion of the ␣ 2 I domain to each CMP was assessed by measuring absorbance at 450 nm. Integrin ␣ 2 I has the highest adhesion to T3-237 WT and binds less to the mutants, decreasing in the order Ala Ͼ Arg Ͼ Val (Figs. 3 and S2), trending with the decrease of triple-helical stability when substituted into the Gly position (71) and consistent with the trend of the cell adhesion assay.

T3-237 CMPs with naturally occurring vEDS mutations maintain the triple-helix conformation but have decreased thermal stability
We characterized the conformation and dynamics of the T3-237 CMPs to determine the molecular underpinnings of reduced ␣ 2 I adhesion with bulkier mutations. The triple-helix conformation is a prerequisite for interaction with integrin ␣ 2 ␤ 1 (72). At 4°C, each of the T3-237 variants have CD wavelength profiles characteristic of triple helices, that is a minimum negative molar residue ellipticity (MRE) near 198 nm and a maximum positive MRE near 225 nm, with little deviation in  (62). Integrin ␣ 2 ␤ 1 -binding sites, Gly 237 -Arg 242 , Gly 288 -Arg 293 , Gly 303 -Arg 308 , Gly 678 -Arg 683 , Gly 726 -Arg 731 , and Gly 987 -Arg 992 (28,29), are highlighted in orange. C, percentages of Gly 3 Xaa substituted amino acids of the total cases reported. Amino acids are colored based on type (negatively charged, red; positively charged, blue; hydrophobic, green; small, yellow).

Impact of vEDS mutations on collagen-integrin interactions
intensity (Fig. 4A). This indicates that the triple-helical composition between these variants is unchanged at this low temperature. The Gly 3 Xaa mutations do perturb the sensitivity of these triple helices to temperature, however. We found that the melting points of G240R and G240V are decreased by ϳ10°C relative to T3-237 WT (47.8 Ϯ 0.5-38.1 Ϯ 0.7°C) (Fig. 4B) by monitoring a CD melt.

vEDS mutations perturb conformation and dynamics of the triple helix local to the mutation site
To understand how this decreased stability impacts the ability for the ␣ 2 I domain to interact with its binding domain, we probed for residue-specific perturbations to conformation and dynamics by integrating NMR with MD simulations. Three CMPs, T3-237 WT, G240A, and G240V, were specifically 15 Nlabeled at Gly 16 , Xaa 19 (mutation site), and Gly 28 (Fig. 5A). The isotopic labels on Gly 16 and Xaa 19 allow us to probe perturbation of local structure and dynamics within the binding site and near the mutation. Enrichment of Gly 28 allows us to probe perturbations within a GPO-rich segment that is expected to have a stable triple-helical structure distant from the mutation site. Within the triple helix, each of the three ␣-chains is staggered by one residue from the adjacent chain. This creates degeneracy between like residues in the ␣-chains, giving distinct cross-  The EDTA condition is a negative control for unspecific binding independent of the divalent metal cations. BSA was used to block any uncoated surfaces and as a control for unspecific binding to the wells. (GPP) 10 is used as a control for unspecific binding of ␣ 2 I to collagen triple-helical segments that do not incorporate an integrin-binding motif. Experiments were performed in triplicate on one plate with 10 g/ml indicated CMP/well and 10 g/ml ␣ 2 I/well. Error bars indicate the standard deviation between the triplicate repeats. Statistical analysis was performed with an unpaired t test relative to WT binding in GraphPad Prism. *, p Յ 0.05; **, p Յ 0.01. The peptide sequence is given at the top. The integrin-binding motif is in bold, and native collagen III residues are underlined.

Impact of vEDS mutations on collagen-integrin interactions
peaks for isotopically labeled residues in each of the three chains in 1 H-15 N heteronuclear single quantum correlation (HSQC) spectra (Fig. 5A). A population of monomer species is also present for each residue. Because no adjacent residues were isotopically labeled, we could not unequivocally assign crosspeaks to specific ␣-chains. Gly 28 residues in all three ␣-chains of a GPO-rich segment have identical chemical environments and thus have a single overlapping triple-helical cross-peak (Fig.  5A). By comparison of 1 H-15 N HSQC spectra, Gly 16 has vastly different 1 H and 15 N chemical shifts between the three variants (Figs. 5A and S3), indicating that mutation of the Gly 19 residue perturbs the conformation and/or chemical environment of the Gly 16 backbone amide three residues N-terminal.
The structural integrity of the triple helix is provided by a critical hydrogen-bond network formed between the amide proton of each Gly and the carbonyl oxygen of the Xaa residue in the adjacent chain. As shown in Fig. 5B, in T3-237 WT, Gly 16 , Gly 19 , and Gly 28 all have amide chemical shift perturbations less than 4.5 ppb/°C (dashed line) upon increasing temperature, indicating that their associated hydrogen-bond network is not significantly modulated by temperature (73) up to 40°C. Within mutant peptides T3-237 G240A and G240V, Gly 16 and Gly 28 remain hydrogen-bonded as shown by their only minor chemical shift perturbations with temperature. However, each of the triple-helix ␣-chains of the mutated residues, Ala 19 and Val 19 , show considerable temperature-dependent chemical shift changes. This indicates that even the small mutation of Gly 3 Ala abolishes the local hydrogen-bond network. Regardless of the small (Ala) or larger hydrophobic (Val) mutation, however, the disruption to the conformation is only local to the mutation site, as Gly 16 , only three residues N-terminal to the mutation site, maintains its hydrogen-bonding capacity.
The backbone dynamics of the 15 N-labeled residues within the triple helices of each CMP were probed by heteronuclear { 1 H}-15 N NOEs. Lower NOEs indicate increased mobility on the fast, ps-ns timescale. In the WT CMP, { 1 H}-15 N NOEs at 15°C were similar for all triple-helical isotopically enriched glycines (Fig. 5C). Upon mutation to G240A, the small Gly 3 Ala mutation leads to a significantly more mobile Ala 19 backbone in only one ␣-chain. However, a larger, hydrophobic Gly 3 Val mutation increases the flexibility of all three ␣-chains dramatically, leading to { 1 H}-15 N NOE values near zero. In both variants, the increase in backbone flexibility is only local to the mutation site, as { 1 H}-15 N NOE values for triple-helical Gly 16 and Gly 28 are not decreased. In the case of G240V, one Gly 16 actually has an increased NOE value, indicating a rigidification of the ␣-chain backbone in this position.

Modeling ␣ 2 I interactions with vEDS Gly 3 Xaa variants
To understand how perturbations to the triple helix impact ␣ 2 I binding, we performed 500-ns all-atom MD simulations on each CMP variant as a free triple helix and in complex with ␣ 2 I. The MD simulations predict ns-timescale dynamics for all residues within the CMPs and ␣ 2 I domain, including those residues not probed by NMR. This provides information on the span of CMP residues around the substitution site that is impacted by dynamic perturbations due to a Gly 3 Xaa mutation. The diameter of the triple helix at a specific site is reflective of local unfolding. We therefore monitored how incorporation of the three mutations changed the diameter of the

Impact of vEDS mutations on collagen-integrin interactions
triple-helix backbone in the free state. Fig. 6A shows the average diameter of the triple helix around the integrin-binding site. The positions are numbered by the residue of the leading strand, and the diameters are measured from the circle encompassing three in-register C␣ atoms of the three ␣-chains at the indicated positions (74). The in-register residues for each position are indicated in the schematic in Fig. 6A. This places the mutations in the leading, middle, and trailing strands at positions 19, 20, and 21, respectively. We found that the triple helix is expanded, relative to the WT CMP, at each position that includes a Gly 3 Xaa mutation for all variants, and the effect asymmetrically propagates at least three positions C-terminal of the mutation sites, impacting functional domains. This asymmetric effect of Gly 3 Xaa mutations was found previously by Yigit et al. (37) in that Gly 3 Ser substitutions interfered with hydrogen bonding up to three triplets C-terminal of the mutation. In the T3-237 CMPs, we found that the diameter expansion extends two triplets in the C-terminal direction, from Xaa 19 to Hyp 24 of the leading chain (Fig. 6A). Val and Arg mutations have the greatest effect. Representative snapshots of the free triple-helix MD simulations show the diameter expansions for each CMP, with G240V and G240R having the greatest triple-helix distortion (Fig. 6B). Notably, the site of greatest expansion is at the position of Glu 20M , the metal-coordinating Glu for CMP-␣ 2 I binding. We would expect this distortion directly at the binding site to inhibit the ␣ 2 I interaction.
We probed for the impact of the mutations on complex formation with ␣ 2 I. We have confirmed that the proper Mg 2ϩ coordination and the van der Waals contacts, hydrogen bonds, and salt bridges that have been reported previously (29,75) are present and stable in the simulated complex formed with the WT CMP (Table 1). These interactions remain unchanged upon introduction of an Ala mutation in the Xaa 19 position, consistent with the retention of ␣ 2 I binding and hMSC spreading over the G240A CMP observed experimentally. However, in the case of an Arg mutation, several interactions become lost over the course of the simulation. The Mg 2ϩ -␣ 2 I Ser 155 coordination becomes unstable within 100 ns in complex with the G240R CMP (Fig. S4). The Mg 2ϩ coordination is fulfilled by ␣ 2 I Asp 151 , maintaining three coordinates with ␣ 2 I. However, only two direct interactions between the triple-helical CMP and ␣ 2 I remain, an Arg 21M backbone-His 258 side-chain hydrogen bond (Fig. S5) and an Arg 21M -Asp 219 salt bridge (Fig. S6). Other stabilizing hydrogen bonds and salt bridges formed between CMP and ␣ 2 I side chains are destabilized upon introduction of the G240R mutation, as given in Table 1 and shown in Figs. S5 and S6. Upon mutation to Val, ␣ 2 I loses Mg 2ϩ coordination at two residues, Ser 153 and Ser 155 , which is replaced by additional coordination to the CMP Glu 20M and a water molecule (Figs.  7C and S4). Although in the MD simulations of both Arg and

Table 1 Summary of CMP-␣ 2 I interactions observed in MD simulations of complexes formed with each variant
A gain of interaction relative to WT is shaded in green; a loss of interaction relative to WT is shaded in red.

Impact of vEDS mutations on collagen-integrin interactions
Val substitutions, the Mg 2ϩ coordination state is disturbed, the ␣ 2 I in the G240R complex maintains three coordinates; however, the ␣ 2 I in the G240V complex is able to stabilize just one Mg 2ϩ coordinate. This extreme loss in ␣ 2 I-Mg 2ϩ coordination in the case of the T3-237 G240V complex provides an explanation for the greatest reduction in ␣ 2 I adhesion to T3-237 G240V.

Discussion
We have used an integrative approach, combining biological adhesion assays, CD, NMR spectroscopy, and MD simulations, to gain insight into the molecular underpinnings of the reduced ␣ 2 I-CMP adhesion in the presence of Gly 3 Xaa mutations in collagen III. We observed that substitution of a Gly in the integrin-binding site interferes with ␣ 2 I-CMP adhesion, dependent upon the identity of the Gly substitution, decreasing in the order WT Ͼ G240A Ͼ G240R Ͼ G240V, and T3-237 G240V substantially reduces hMSC spreading functionality relative to WT or G240A. Interestingly, in contrast to previous reports (31) using recombinant bacterial collagen models, a Gly 3 Ala mutation in the integrin-binding motif does not abolish ␣ 2 I binding, which may in part be due to the presence of hydroxyproline in our CMPs that stabilizes the ␣ 2 I-CMP interaction through hydrogen bonds with ␣ 2 I side chains. This highlights the importance of native hydroxyprolines in collagen interactions. The results indicate that, even within this short CMP, relative to the full-length collagen III, a single Gly 3 Xaa mutation does not abolish triple-helix formation, but the thermal stability of the triple helix formed is disrupted, depending upon the side chain of the substituted amino acid.
In concert with the insight gained from ELISA, CD, NMR, and MD, the cell adhesion trends of hMSCs, which express ␣ 2 ␤ 1 integrins on the cell surface, on CMPs could be predicted based on the substituted amino acid. The reduced cell adhesion of T3-237 G240V relative to WT and G240A can be explained by local backbone mobility and diameter expansion of the triple helix upon substitution of a bulky side chain. However, because the melting temperature of G240V is 10°C lower compared with WT and G240A, a more significant fraction of the G240V may be unfolded. Based on the CD analysis, at 37°C, ϳ70% of the ␣-chains for both WT and G240A are in the triple-helical conformation, whereas only ϳ50% of the ␣-chains for G240V have attained triple helix. This difference in the percentage of triple-helix structure presented by CMPs might present a reduced number of optimal integrin-binding sites to which the cells and recombinant ␣ 2 I can adhere. Hence, it is imperative to realize whether the mutation itself leads to reduced cell and ␣ 2 I adhesion or the variation in the number of triple-helical integrin-binding sites offered by the CMPs.
vEDS is one of several debilitating connective tissue disorders due to glycine mutations in fibrillar collagens (5,76,77). OI is another connective tissue disorder in which Gly 3 Xaa mutations in collagen I have been extensively studied (30 -36, 40, 42, 78 -84). The frequencies of substituted amino acids for OI are much different from those reported for vEDS. Over all OI mutations, Ser is the most substituted amino acid, accounting for ϳ41% of Gly 3 Xaa mutations (44). Small-residue mutations make up nearly 63% of all OI Gly 3 Xaa substitutions (44). Conversely, in vEDS, small substitutions make up only ϳ19% of Gly 3 Xaa mutations (Fig. 1C), with the most common substitutions being bulky Asp, Arg, Val, and Glu. Thus, Gly 3 Xaa mutations in the context of OI and vEDS cannot be treated as equal. It is not yet understood whether this is due to tissue specificity, the difference in sequence environment between the collagens, heteroversus homotrimeric nature of the triple helices, or other factors.
When substituted into the Gly position, the charged and bulky mutations reported in vEDS are the most triple helixdestabilizing amino acids (71), and Gly 3 Xaa substitutions of these amino acids in collagen I most often result in the lethal form of OI (43,44). Thus, it has been previously proposed that there is a correlation between OI phenotype and genetic mutation identity (44). In the case of vEDS, survival of a cohort of afflicted individuals was indeed found to trend with the identity of the substituted amino acid, in that those with Val, Arg, and Asp substitutions had the lowest survivability and that Ser substitutions were less severe (46). Our results show that substitution of the small amino acid Ala into a critical integrin ␣ 2 ␤ 1binding site only moderately disrupts ␣ 2 I adhesion to the Gly 3 Ala CMP, and hMSC spreading on the Gly 3 Ala CMP is minimally affected. Conversely, substitution of a larger Val mutation substantially reduced ␣ 2 I adhesion and hMSC spreading. This suggests that, despite the local structural and dynamic perturbations imparted on the triple helix by the Gly 3 Ala mutation, the CMP-␣ 2 I interaction maintains some plasticity wherein cellular functions are still able to occur. The ability for the small Gly 3 Ala mutation to maintain its functionality may be reflective of the underrepresentation of small mutations in vEDS, as milder phenotypes may go unreported. However, genotype-phenotype relationships will need to be further investigated for verification.
The collagen III-integrin ␣ 2 ␤ 1 interaction investigated here is critical for platelet adhesion and signaling and endothelial cell adhesion in blood vessel walls (Fig. 7, A and B). Disruption

Impact of vEDS mutations on collagen-integrin interactions
of the collagen III triple helix through even a single Gly 3 Xaa mutation may hinder both the structural integrity of collagen III-rich tissues such as blood vessel walls and distensible organs but also inhibit vital cellular interactions with the ECM. The severity of vEDS is attributed largely to the potential for aortic aneurysms and arterial ruptures, which may be exacerbated by reduced platelet activity. These novel atomic-level insights into how the identity of vEDS Gly 3 Xaa mutations in an integrinbinding site impact the interaction of ␣ 2 I with its recognition motif on collagen III provide a foundation for new drug therapy techniques to combat debilitating collagen disorders that compromise collagen interactions.

Preparation of CMPs
All CMP variants were purchased from LifeTein LLC (Somerset, NJ) as purified peptides. CMPs were dissolved in assay buffer from lyophilized powder and equilibrated at 4°C overnight before use. For CD studies, CMPs were further purified using PD Midi-trap G10 desalting columns (GE Healthcare). Concentrations of peptides were determined by measuring the absorbance at 280 nm using a molar extinction coefficient of 1280 M Ϫ1 cm Ϫ1 , with the exception of the GROGER hexapeptide (GenScript), which does not have a Tyr residue. The concentrations of this peptide were determined by weight, considering the net peptide content and HPLC purity as provided by the synthesis company.

Cell culture
hMSCs were continuously cultured in complete MEM-␣ containing 10% fetal bovine serum, 1% penicillin-streptomycin, 1% L-glutamine, and 0.001% basic fibroblast growth factor. Growth medium was changed every alternate day. The cells were detached by TrypLE and split at a ratio of 1:3 in a T175 flask upon reaching 80% confluence. The cells were incubated at 37°C in a cell incubator with 95% air and 5% CO 2 .

Cell spreading assays
Cell spreading assays were performed as follows. 96-well solid white polystyrene plates (Corning, catalog number 3917) were coated with 100 l/well multiple concentrations of CMPs in 10 mM acetic acid overnight at 4°C. 50 g/ml collagen III in 10 mM acetic acid was used as a positive control. 75 g/ml GROGER hexapeptide in 10 mM acetic acid and 5% (w/v) BSA in 1ϫ sterile-filtered PBS were used as negative controls. The GROGER hexapeptide consists of the sequence 237 GROGER 242 in a single chain and does not form a triple-helical structure, leading to no cell adhesion or spreading. The next day, peptide solutions were aspirated, and the wells were then blocked with 200 l/well BSA solution (5% (w/v) BSA in 1ϫ sterile-filtered PBS) for 2 h at room temperature with the exception of the tissue culture treated wells used as positive controls. Meanwhile, near-confluent hMSCs were labeled with 3 M Cell-Tracker Green 5-chloromethylfluorescein diacetate (CMFDA) dye (Thermo Fisher Scientific, catalog number C7025) in MEM-␣ (serum-and thiol-free) according to the manufacturer's instructions. After the blocking step, the BSA solution was aspirated, and the wells were washed three times with 200 l/well PBS. After the washes, hMSCs were harvested from culture and reseeded onto the peptide-coated plates at a density of 5,000 cells/well in complete MEM-␣ and returned to the incubator. At 4, 7, and 24 h, plates were imaged using an Olympus IX81 inverted epifluorescence microscope (Olympus Scientific, Waltham, MA) with a Hamamatsu ORCA digital camera (Hamamatsu Photonics, Bridgewater, NJ).

Expression and purification of integrin ␣ 2 I
The ␣ 2 I domain used in these studies corresponds to residues 142-336 of the integrin ␣ 2 subunit. Integrin ␣ 2 I was recombinantly expressed in Escherichia coli BL21(DE3) cells by induction with 1 mM isopropyl 1-thio-␤-D-galactopyranoside overnight at 25°C. The cells were lysed using a 20% sucrose TES buffer. The ␣ 2 I domain was purified by Ni 2ϩ -nitrilotriacetic acid-agarose affinity chromatography (Qiagen) and buffer-exchanged with PD-10 desalting columns (GE Healthcare). Protein concentration was determined by measuring the absorbance at 280 nm using a molar extinction coefficient of 20,400 M Ϫ1 cm Ϫ1 .

ELISA
Adhesion of recombinant ␣ 2 I to T3-237 CMP variants was determined colorimetrically in solid-phase ELISAs. Immunolon 2HB 96-well plates (Thermo Fisher) were coated with 100 l of CMP (10 g/ml in 10 mM acetic acid) overnight at 4°C. The wells were then blocked with 200 l of 5% (w/v) BSA in PBS, pH 7.4, for 1 h at room temperature. From this point, for each ␣ 2 I-CMP adhesion, half of the wells were treated with washing and binding buffers that consisted of PBS, pH 7.4 ϩ 0.5% (w/v) BSA in the presence of 5 mM MgCl 2 or 5 mM EDTA. After three washes with 200 l of washing buffer, 100 l of ␣ 2 I (10 g/ml in binding buffer) was incubated with each CMP for 1 h at room temperature. After washing three times with 200 l of washing buffer, 100 l of mouse anti-␣ 2 I mAb (1:2000 (v/v) dilution; Thermo Fisher, catalog number MA5-16571) in binding buffer was incubated in the wells for 45 min at room temperature. After washing three times with 200 l of washing buffer, 100 l of goat horseradish peroxidase-conjugated antimouse IgG antibody (1:5000 (v/v) dilution; GenScript, catalog number A00160, lot number 17B001197) in binding buffer was incubated in each well for 30 min at room temperature. Following a final four washes, ␣ 2 I binding was detected using a 3,3Ј,5,5Ј-tetramethylbenzidine substrate kit (Pierce) as directed by the manufacturer's instructions. Absorbance was measured at 450 nm using a Tecan Infinite F50 plate reader equipped with Magellan software.

CD spectroscopy
CD wavelength profiles and temperature scans of each CMP were acquired on an AVIV Model 400 CD spectrometer (AVIV Biomedical Inc.). Wavelength scans were obtained at 4°C from 260 to 190 nm, collecting points every 0.5 nm with a 1-nm bandwidth for 4 s, averaging three scans for each sample. Temperature scans were acquired by measuring MRE at 224 nm from 0 to 70°C with a 10-s averaging time and 1.5-nm bandwidth. Samples were equilibrated for 2 min at each temperature

Impact of vEDS mutations on collagen-integrin interactions
before acquiring. The melting temperature was determined by first normalizing the melting curves assuming a fully folded state at 0°C and a fully unfolded state at 70°C and then calculating the temperature at which 50% of the population is folded based on a linear fit of the central temperature-dependent decay in the melting curves using GraphPad Prism.
All NMR experiments were performed on a Bruker Avance III 600-MHz spectrometer equipped with a TXI probe. Amide proton temperature gradient experiments were acquired as 1 H-15 N HSQC (85,86) experiments at temperatures from 5 to 40°C in increments of 5°C. The samples were equilibrated for at least 1 h between temperature changes. Amide proton temperature gradients were calculated by linear fitting of the amide proton chemical shifts versus temperature. The amide proton gradient is taken as the slope of the line. Heteronuclear { 1 H}-15 N NOE experiments were performed at 15°C. All data were processed with NMRPipe (87) and analyzed in Sparky (88).

MD simulations
The initial model of the T3-237 WT CMP was built based on the GFOGER peptide from the crystal structure of the GFOGER-␣ 2 I complex (Protein Data Bank (PDB) code 1DZI) (75). The Phe residue in the GFOGER peptide was replaced by an Arg residue using PyMOL (Schrödinger, LLC). The extra GPO repeats, the terminal GPCs, and the C-terminal Tyr of T3-237 were built using the Triple-Helical Collagen Building Script (89). The N and C termini were capped with acetyl and NH 2 groups, respectively. The initial models for the mutants were generated by replacing Gly 19 of T3-237 with Ala, Val, and Arg residues using PyMOL.
The initial coordinates for the T3-237 CMP-␣ 2 I complex were taken from the X-ray structure (PDB code 1DZI) (75). The coordinates for T3-237 were generated by aligning the backbone to the X-ray structure of the GFOGER peptide. The Co 2ϩ ion was replaced by a Mg 2ϩ ion. All of the water molecules in the X-ray structure were retained.
The T3-237 CMP and CMP-␣ 2 I complex models were laid in a cubic box of TIP3P water molecules (90) with the box border at least 10 Å away from any atoms of the CMP or ␣ 2 I. Extra Cl Ϫ ions were added to neutralize the positive charges.
The protein was treated with the ff14SB force field (91). The simulations were performed with the CUDA version of the pmemd module of the AMBER 2018 package (92). Periodic boundary conditions were used, and electrostatic interactions were calculated by the particle mesh Ewald method (93,94), with the nonbonded cutoff set to 8 Å. The SHAKE algorithm (95) was applied to bonds involving hydrogens, and a 2-fs integration step was used. Pressure was held constant at 1 atm with a relaxation time of 2.0 ps. The temperature was held at 300 K with Langevin dynamics with a collision frequency of 2.0 ps Ϫ1 .
Prior to MD simulations, the systems were subjected to energy minimizations and equilibration. The minimization started with 1000 steps of steepest descent minimization followed by 4000 steps of conjugate gradient minimization with 10 kcal mol Ϫ1 Å Ϫ2 position restraints on the CMP and ␣ 2 I. The following minimization was carried out without any restraints. Then the system was heated from 0 to 300 K for 100 ps with position restraints of 10 kcal mol Ϫ1 Å Ϫ2 on the CMP and ␣ 2 I. The system was first equilibrated for 1 ns at a constant temperature of 300 K and pressure of 1 atm with position restraints of 2 kcal mol Ϫ1 Å Ϫ2 on the CMP and ␣ 2 I. The following equilibration was conducted without any restraints. The production runs for all of the models were 500 ns. The trajectories were analyzed using CPPTRAJ (96).