Mutations Near Amino End of α1(I) Collagen Cause Combined Osteogenesis Imperfecta/Ehlers-Danlos Syndrome by Interference with N-propeptide Processing*

Patients with OI/EDS form a distinct subset of osteogenesis imperfecta (OI) patients. In addition to skeletal fragility, they have characteristics of Ehlers-Danlos syndrome (EDS). We identified 7 children with types III or IV OI, plus severe large and small joint laxity and early progressive scoliosis. In each child with OI/EDS, we identified a mutation in the first 90 residues of the helical region of α1(I) collagen. These mutations prevent or delay removal of the procollagen N-propeptide by purified N-proteinase (ADAMTS-2) in vitro and in pericellular assays. The mutant pN-collagen which results is efficiently incorporated into matrix by cultured fibroblasts and osteoblasts and is prominently present in newly incorporated and immaturely cross-linked collagen. Dermal collagen fibrils have significantly reduced cross-sectional diameters, corroborating incorporation of pN-collagen into fibrils in vivo. Differential scanning calorimetry revealed that these mutant collagens are less stable than the corresponding procollagens, which is not seen with other type I collagen helical mutations. These mutations disrupt a distinct folding region of high thermal stability in the first 90 residues at the amino end of type I collagen and alter the secondary structure of the adjacent N-proteinase cleavage site. Thus, these OI/EDS collagen mutations are directly responsible for the bone fragility of OI and indirectly responsible for EDS symptoms, by interference with N-propeptide removal.

Osteogenesis imperfecta (OI) 1 is a genetic disorder of connective tissue characterized by bone fragility, growth deficiency, and blue sclerae (1,2). Defects in type I collagen are well known to cause the full clinical range of OI (3,4). Haploinsufficiency for type I collagen, caused by a null ␣1(I) allele, results in a very mild clinical phenotype (5). Collagen structural defects, which are usually glycine substitutions or exon skipping defects, have a dominant negative mechanism. They result in a phenotype that ranges from lethal to moderately severe depending on the chain in which the mutation occurs, its location in the chain, and the specific amino acid substituted (6,7). The great majority of mutations causing OI occur in the helical regions of either pro-␣1(I) or pro-␣2(I). Less than 5% of collagen structural mutations occur in the C-propeptides of the two chains; these mutations cause lethal to moderate OI by delaying chain association into heterotrimer (8).
Ehlers-Danlos VII A and B are also caused by mutations in type I collagen (9). These mutations have a well defined location and mechanism of action (10 -22). All EDS VII mutations involve a complete or partial loss of exon 6 sequences from either ␣ chain, with EDS VIIA due to mutations in pro-␣1(I) and EDS VIIB due to similar mutations in pro-␣2(I). Because exon 6 contains both the N-proteinase cleavage site and the interhelix cross-linking lysine, the N-propeptide cannot be cleaved from EDS VII mutant procollagen and cross-linking is defective. Patients with N-propeptide retention have severe generalized joint hypermobility and congenital bilateral hip dislocation. Many have scoliosis because of ligamentous laxity, dislocations of other joints and mild osteopenia, with a few fractures. In those patients in whom dermal fibrils were examined, cross-sectional diameters were normal or decreased and borders were irregular (13-15, 22, 23).
A small number of mutations at the amino end of the helical region of ␣2(I) collagen have been noted to cause a combination of EDS and OI symptoms. These mutations include two cases with exon 9 skipping, a case with exon 11 skipping, and a large duplication of E12-32 (24 -28). The symptoms of EDS predominate in the clinical presentation of these patients. All have bilateral hip dislocations and marked laxity of large joints. Two had shoulder joint dislocations and hernias (24,26) and one had scoliosis (27). In the case with duplication of E12-32, the dermal fibril diameter was 2 ⁄3 that of a matched control (24). OI symptoms in these cases were mild. All had blue sclerae, two had dentinogenesis imperfecta (24,27,28) and one had Wormian bones (27,28). Only the child with duplication of E12-32 had a single tibial fracture.
We describe here a group of seven patients with mutations in the ␣1(I) chain who have a distinct OI/EDS phenotype, in which the symptoms of OI are more prominent and the EDS less severe than in the OI/EDS mutations in ␣2(I). The ␣1(I) OI/EDS mutation cluster is located in a high stability folding region at the amino end of the type I collagen helix. These mutations interfere with removal of the N-propeptide, although the N-proteinase site in exon 6 is intact. Thus, OI/EDS and EDS VII are shown to have a common mechanism for EDS symptoms.

EXPERIMENTAL PROCEDURES
Cell Culture-Skin fibroblast cultures were established from dermal punch biopsies. Fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum and 2 mM glutamine in the presence of 5% CO 2 . Osteoblast primary cultures were established from surgical bone chips using the method of Robey and Termine (29). In brief, osteoblasts were released from bone chips by digesting for 2 h at 37°C with 0.3 units/ml collagenase P in serum-free medium, and grown in 45% low-calcium DMEM, 45% low-calcium Ham's F-12 Kinase medium (Biofluids, Rockville, MD), 25 g/ml ascorbate, and 10% fetal bovine serum in the presence of 8% CO 2 .
Steady State Collagen Synthesis-To label procollagens, confluent fibroblast cultures of probands and control cells (ATCC 2127, American Type Culture Collection, Manassas, VA) were incubated for 2 h in serum-free medium containing 50 g/ml ascorbic acid, followed by incubation with 260 mCi/ml of 3.96 TBq/mmol L- [2,3,4, H]proline in serum-free medium for 16 h. Procollagens were harvested from media and cell layer and precipitated with ammonium sulfate; collagens were prepared by pepsin digestion (50 g/ml) of procollagen samples, as previously described (30).
Mutation Identification-Total RNA was isolated from cultured fibroblasts of patient and control cell lines using TriReagent (Molecular Research Center, Cincinnati, OH) according to the manufacturer's directions (31). The region of the ␣1(I) collagen mRNA corresponding to exons 5-12 was amplified by reverse transcription-polymerase chain reaction (RT-PCR) (32). Total RNA (1 g) was reverse transcribed with 50 units of murine leukemia virus RT (Applied Biosystems, Foster City, CA) using an antisense primer complementary to nucleotides 813-842 of the cDNA sequence (GenBank TM AF017178) in exon 12 (5Ј-CCAG-CAGGACCAGCATCTCCCTTGGCACCA-3Ј). The RT reaction was used as a template for PCR with a sense primer corresponding to nucleotides 377-406 of cDNA sequence and located in exon 5 (5Ј-CTGGCCGAGAT-GGCATCCCTGGACAGCCTG-3Ј). PCR used 0.1 mM dNTP, 2.5 units of Amplitaq, and 1ϫ PCR Buffer II (Applied Biosystems). PCR cycling conditions were as follows: 94°C for 5 min; then 35 cycles of 1 min at 94°C, 1 min at 65°C, and 1.5 min at 72°C; and finally 7 min at 72°C. RT-PCR products were sequenced directly on a Beckman Coulter CEQ2000 DNA Sequencer (Beckman Fullerton, CA) according to the manufacturer's protocol.
Pericellular Processing-Processing of procollagens secreted by fibroblasts was examined by labeling confluent cells from probands and control with 260 mCi/ml of 3.96 TBq/mmol [ 3 H]proline for 24 h and then replacing the media with DMEM containing 2 mM non-radioactive proline and 10% fetal bovine serum. Media from independent wells were harvested at 24-h intervals over a 5-day period as previously described (25). Media procollagen samples from fibroblasts were precipitated with ammonium sulfate and electrophoresed on 6% polyacrylamide-urea-SDS gels.
Matrix Deposition-Proband and control fibroblasts and osteoblasts were grown to confluence and stimulated every other day for 11 days (fibroblasts) or 9 days (osteoblasts) with fresh DMEM containing 10% fetal bovine serum and 100 g/ml ascorbic acid. Cultures were then incubated for 24 h with 260 mCi/ml of [ 3 H]proline in serum-free medium. Medium was collected and procollagens were precipitated with ammonium sulfate. Matrix collagens were serially extracted at 4°C as previously described (34). In brief, newly synthesized collagens were extracted for 24 h with neutral salt (50 mM Tris-HCl, pH 7.5, containing 0.15 M NaCl, 5 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, 10 mM benzamidine, and 10 mM N-ethylmaleimide), separated from matrix by centrifugation, and precipitated with 2 M NaCl. Collagens with acidlabile cross-links were extracted from the matrix for 24 h with 0.5 M acetic acid and precipitated with 2 M NaCl. Collagens with mature cross-links were extracted by pepsin digestion (0.1 mg/ml) for 24 h and precipitated with 2 M NaCl. All fractions were electrophoresed on 6% polyacrylamide-urea-SDS gels.
Transmission Electron Microscopy of Proband Dermal Fibrils-A dermal punch biopsy was obtained from each proband and from a control matched for age and race. The samples were fixed in 2.5% glutaraldehyde and then treated with 1% osmium tetroxide followed by en bloc staining with 2% uranyl acetate. After dehydration, the tissue was infiltrated with Spurr's plastic resin. 600 -800-Å sections were obtained with an AO Reichert Ultracut ultramicrotome mounted on copper grids and stained with lead citrate. The stained grids were examined in a Zeiss EM10 CA transmission electron microscope and representative areas were photographed (JFE Enterprises, College Park, MD).
Preparation of Proband Secreted Procollagens-Proband and control fibroblasts were grown to confluence at 37°C. Culture medium was removed and fresh serum-free DMEM supplemented with 2 mM glutamine and 50 g/ml ascorbate was added to the cell cultures. For procollagens used in calorimetry and N-propeptide processing studies, medium was harvested at 24-h intervals for 2 days and fresh medium containing 50 g/ml ascorbate was replenished daily. Medium was buffered with 100 mM Tris-HCl, pH 7.4, and cooled to 4°C. Protease inhibitors were added to obtain the following final concentrations: 250 mM EDTA, 0.2% NaN 3 , 1 mM phenylmethylsulfonylfluoride, 5 mM benzamidine, and 10 mM N-ethylmaleimide. Procollagen was precipitated by gradual addition of ammonium sulfate to a final concentration of 176 mg/ml and incubation at 4°C overnight, followed by centrifugation at 12,000 ϫ g for 2 h.
Differential Scanning Calorimetry-DSC scans from 10 to 50°C were performed at 0.125 and 1°C/min heating rates in a Nano II DSC instrument (Calorimetry Sciences Corporation, Lindon, UT). To prevent fibrillogenesis of collagen, 0.1-0.3 mg/ml protein solutions in 0.2 M sodium phosphate, 0.5 M glycerol, pH 7.4, were used. The denaturation temperature (T m ) in phosphate/glycerol buffers depends linearly on the concentration of all buffer components and the corresponding proportionality coefficients do not depend on the scanning rate, allowing extrapolation of T m to physiological conditions (35).
N-proteinase Cleavage-Ammonium sulfate procollagen precipitates were redissolved in 0.1 M sodium carbonate, 0.5 M NaCl, pH 9.3, labeled by covalent attachment of Cy2 and Cy5 fluorescent dyes (Amersham Biosciences, Piscataway, NJ) and transferred into 50 mM Tris, 0.5 M NaCl, 4 mM CaCl 2 , 0.5 mM phenylmethylsulfonylfluoride, 2.5 mM Nethylmaleimide, 0.02% Brij 35, pH 8, as described (36). Procollagen concentration was measured by Sircol assay (Biocolor Ltd., Belfast, Northern Ireland) and adjusted to 0.1 mg/ml. Binary mixtures of procollagens, one labeled by Cy5 and one labeled by Cy2, were prepared so that two different procollagens could be co-processed by an enzyme in the same sample tube and, therefore, under completely identical conditions. Procollagen N-proteinase (ADAMTS-2) was purified from fetal calf skin as described (37) and stored in 5-l aliquots in 50 mM Tris, 0.5 M NaCl, 2 mM CaCl 2 , 0.02% Brij 35, pH 7.6, at Ϫ80°C. Enzyme was added to each collagen mixture on ice for an enzyme:substrate ratio of 15 units of enzyme/mg of substrate. Reactions were incubated at 34°C to avoid partial denaturation of low-stability procollagens containing structural defects. Sample aliquots were collected at the indicated times after the start of the reaction, mixed with lithium dodecyl sulfate gel sample buffer (Invitrogen, Carlsbad, CA) with added EDTA (to stop enzymatic cleavage) and dithiothreitol. The samples were denatured and analyzed by gel electrophoresis on pre-cast 6% Tris/glycine minigels (Invitrogen, Carlsbad, CA). The gels were scanned on an FLA3000 fluorescence scanner (FUJI Medical Systems, Stamford, CT). Intensity profiles for each lane were extracted using ScienceLab software supplied with the scanner. Quantitative analysis of band intensities and deconvolution of overlapping bands were performed using PeakFit software (Systat, Point Richmond, CA).

Phenotype of OI/EDS Patient Group-
The seven probands in this study have a distinct OI/EDS phenotype (Table I). All seven first came to medical attention for symptoms of osteogenesis imperfecta. All have types III or IV OI, with multiple fractures of long bones; the children with G25V, G76E, and G88E had bone deformity sufficient to require osteotomy procedures. Their L1-L4 DEXA z-scores range from Ϫ3.0 to Ϫ5.2. All have the significant short stature of OI and a height age that ranges from 20 to 80% of the mean height for their chronological age. Also characteristic of the relatively shorter lower extremities in OI, most have arm span significantly greater than length. All probands have strikingly blue sclerae.
The symptoms of EDS are notably more severe than the mild to moderate joint hyperextensibility frequently found in OI. In addition to significant hyperextensibility of large and small joints ( Fig. 1), these probands have laxity of paraspinal ligaments. This results in early scoliosis without vertebral compressions. The scoliosis is rapidly progressive and unresponsive to bracing. Spinal fixation has been required by the mid-teenage years.
Collagen Biochemistry and Mutation Detection-Because the probands have clinically significant osteogenesis imperfecta, we examined the type I collagen synthesized by their dermal fibroblasts electrophoretically on SDS-urea-PAGE. Proband steady state media and cell layer collagen did not have delayed electrophoretic migration, as would be expected with the well known overmodification of type I collagen chains frequently seen in OI ( Fig. 2A). Proband 1 had a leading edge in both media and cell layer ␣2(I) bands and proband 2 had a leading edge in the ␣2(I) band isolated from the cell layer.
The normal collagen electrophoresis results prompted us to screen these patients for collagen mutations at the amino end of the helical region of either ␣1(I) or ␣2(I), which would not be expected to cause overmodification. Direct sequencing of RT-PCR products spanning exons 5-12 of both ␣ chains revealed that all 7 probands had mutations located in exons 7-11 of COL1A1 (Fig. 2B), causing structural abnormalities in the 90 residues at the amino end of the ␣1(I) protein chain. Proband 1 has a mutation (IVS7 ϩ 4 A Ͼ T; g.3756AϾT) that causes skipping of exon 7 from the mutant transcript. The leading edge seen after pepsin digestion on SDS-PAGE is most likely the consequence of normal length ␣2(I) chain looping out of helices containing a shorter mutant ␣1(I) chain(s). The remaining 6 probands are heterozygous for glycine substitution mu-  tations, located at Gly13, Gly25 (2 patients), Gly34, Gly76, and Gly88, respectively. The G13D substitution presumably disrupts the collagen helix sufficiently to expose the amino end of the ␣2(I) chain to pepsin digestion, generating the leading edge seen on the ␣2(I) band. Only RT-PCR products of the expected size were obtained from both ␣1(I) and ␣2(I) transcripts in the G13D case; sequencing ruled out a second mutation or use of a cryptic splice signal. Mutations Interfere with in Vitro and Pericelluar NH 2 -terminal Processing-The mutations causing OI/EDS interfere with in vitro processing of the N-propeptide of proband procollagen by purified N-proteinase (Fig. 3), although the sequence of the N-proteinase cleavage site in exon 6 is intact in all cases. Because of random association of pro-␣ chains during procollagen assembly, heterozygous patients generate 25, 50, and 25%, respectively, of procollagen molecules containing two, one, and no mutant pro-␣1(I) chains. From probands heterozygous for mutations in exon 7 (⌬E7 and G13D), only about 25% of pro-␣1(I) chains were processed in vitro. Apparently, these mutations prevent processing of the N-propeptide from all helices that contain one or two copies of the mutant pro-␣1(I) chain, and only the 25% of helices with two normal pro-␣1(I) chains could be processed. The procollagens with mutations in exons 8 -11 (G25V, G34R, and G76E) underwent 70 -88% cleavage, suggesting that helices with one or two normal pro-␣1(I) chains were processed, whereas the 25% of helices containing two mutant chains could not be cleaved. The procollagen containing the G88E substitution was completely cleaved, but at a reduced rate compared with normal procollagen. Mutations located more distal to the cleavage site, at G121D and G136R, had complete in vitro cleavage with normal kinetics.
We also examined the conversion of procollagen to collagen by the pericellular processing enzymes in a cell culture assay, by following the conversion of a pulse of [ 3 H]procollagen to collagen over 5 days (data not shown). The procollagen secreted by all seven of the OI/EDS probands showed delayed processing to collagen by the pericellular enzymes, consistent with the in vitro processing results. In comparison with processing of normal control procollagen, amino propeptide processing is substantially delayed in probands 1 and 2 (exon 7 mutations), delayed to a lesser extent in probands 3 and 5 (exon 8 mutations), and only modestly delayed in probands 6 and 7 (exon 11 mutations).
pN-collagen Deposited in Matrix in Fibroblast and Osteoblast Cultures-The pN-collagen present in the media of cul-tured fibroblasts and osteoblasts is incorporated into the matrix deposited by those cells (Fig. 4). In each proband, we see a substantial increase of pN-collagen in these fractions, as compared with control. For mutations in exon 7 (⌬E7 and G13D), the amount of pN-␣1(I) is equivalent to the amount of fully processed ␣1(I) chain. For mutations in exon 8 (G25V and G34R), pN-␣1(I) in fibroblast extracts is about 20 -30% of the amount of fully processed ␣1(I). Mutations further from the amino end of the collagen helix (G76E and G88E) have relatively less pN-␣1(I) in the fibroblast matrix. Two glycine substitution mutations beyond the first 90 residues of the chain, at G121D and G136R, do not have levels of pN-collagen greater than control. It is noteworthy that the proportion of pN-collagen is greater in the matrix deposited by cultured osteoblasts than fibroblasts in the four probands for whom osteoblasts were available (G13D, G25V, G76E, and G88E). For the G88E mutation, very little pN-collagen was detected in fibroblast matrix, but about one-third of total ␣1(I) was in pN form in osteoblast matrix. The matrix deposition assays also demonstrate that pN-collagen is efficiently cross-linked into matrix, rather than accumulating in the neutral salt extract.
Reduced Cross-section of Dermal Fibrils Corroborates pNcollagen Incorporation into Matrix in Vivo-In dermal fibrils from patients with EDS VII, the incorporation of pN-collagen into the fibril has been shown to result in the reduction of the fibril diameter (38), most likely by steric hindrance. Examination of dermal fibril diameter in our OI/EDS probands provides indirect evidence that pN-collagen is incorporated into OI/EDS matrix in vivo. The diameters of the dermal fibrils of all the OI/EDS probands are significantly smaller than those of agematched control fibrils (Fig. 5 and Table II). In all probands except one, the fibrils have about 2 ⁄3 the diameter of control fibrils. The fibrils with G76E mutant collagen are closer to normal diameter (Ϸ75%), but are still significantly smaller. Very few cauliflower forms were seen in OI/EDS fibrils. Fur- thermore, the borders of their fibrils were regular, unlike the irregular margins frequently seen in EDS VII.
Calorimetry Demonstrates Disruption of a Distinct Folding Region-Differential scanning calorimetry provides insight into the structural basis of OI/EDS. Typical DSC tracings (thermograms) for procollagen and collagen are illustrated in Fig.  6A for the control protein and two classical type IV OI mutations (G121D and G136R), which do not result in N-propeptide retention or cause EDS symptoms. The thermograms of normal procollagen and collagen each have a single peak of protein denaturation. Mutations in the helical region of collagen may result in one or two additional DSC peaks with reduced stability, representing denaturation of molecules with one and two mutant chains, respectively, in addition to the normal peak for molecules with no mutant chains. However, regardless of their specific shape, thermograms for each procollagen and the corresponding collagen are usually identical, because the propeptides do not contribute to overall stability of normal and most mutant molecules.
In contrast, the DSC of the OI/EDS probands is distinctive, with procollagen and collagen tracings that differ from each other: (i) a single procollagen peak in ⌬E7 and G13D splits into normal and reduced stability collagen peaks; (ii) the reduced stability procollagen peaks in G25V and G34R shift to lower temperature in collagen; (iii) an additional low-temperature peak appears in the collagen of G76E; and (iv) the thermograms of G88E procollagen and collagen have a reciprocal change in the heights of the low and normal stability peaks.
This difference in thermal stability with and without Npropeptide suggests that mutations in the first 90 residues at the amino end of the collagen helix disrupt a distinct folding region that anchors the collagen triple helix (Fig. 7, A and B). The presence of the N-propeptide in these mutant molecules compensates for the loss of helical stability, resulting in pro-collagen that is more stable than collagen. The anchor region is bordered by a low-stability microunfolding region (Fig. 6B); mutations beyond the first 90 residues have normal DSC patterns, do not disrupt the N-anchor region, and do not affect N-propeptide cleavage. DISCUSSION The cohort of patients presented here delineates a 90-residue region at the amino end of the ␣1(I) collagen chain in which mutations cause a distinct OI/EDS phenotype. We demonstrated that mutations in this region interfere with processing of the N-propeptide of type I collagen even though the Nproteinase cleavage site is intact. This abnormality was detected in both pericellular assays and in vitro with purified proband procollagen and N-proteinase. In vitro, mutations closest to the amino end in exon 7 prevented N-propeptide cleavage from 3 ⁄4 of procollagen molecules, presumably those containing one or two mutant chains. Mutations further from the Nproteinase site (exons 8 -11) prevented N-propeptide cleavage from 1 ⁄4 to 1 ⁄10 of secreted procollagen, most likely only molecules containing two mutant chains. The most remote OI/EDS mutation at the end of exon 11 (G88E) delayed the kinetics of propeptide removal.
We demonstrated that the resulting pN-collagen is incorporated into matrix deposited by cultured fibroblasts and osteoblasts of the probands. The amount of pN-collagen seen in the matrix deposited by fibroblasts is consistent with the in vitro processing results; mutations with impaired processing of a fraction of the ␣1(I) chains have a greater proportion of pN-␣1(I) in matrix than do mutations with a simple delay in processing. Interestingly, osteoblast matrices always have more pN-collagen than the corresponding fibroblast matrix. This is most dramatically seen in the case of G88E, in which at least one-third of total osteoblast ␣1(I) is in the pN-form and very little pN-collagen is seen in the fibroblast matrix. This difference between fibroblast and osteoblast matrices may reflect the lower activity of N-proteinase in bone than skin tissue (39). It may also reflect the relatively greater modification of osteoblast than fibroblast collagen (40) and the slower cleavage of the conformation-dependent cleavage site that is known to occur with overmodified collagen even in the absence of a collagen structural mutation (41). We also present evidence strongly supportive of pN-collagen incorporation into fibrils in vivo. The dermal fibrils of all seven probands have a strikingly smaller diameter than those of age-matched controls. This is similar to the findings in fibrils of EDS VIIA and -B patients who are known to have pN-collagen in fibrils (13-15, 22, 23), although the ␣1(I) OI/EDS probands do not have fibrils with irregular borders.
The OI/EDS region has a structural basis. Analysis of collagen amino acid sequence (Fig. 6B) shows that the first 85 residues form a highly stable region that serves as the anchor for the amino end of the collagen triple helix. On its carboxyl end, it is bordered by a microunfolding region of 15 residues containing no proline or hydroxyproline and a chymotrypsinsensitive site (42). The DSC data presented here indicate that mutations in the first 90 amino acids disrupt the N-anchor region by destabilizing the NH 2 -terminal end of the triple helix and cause structural changes that extend into the propeptide. These structural changes are a distinct feature of OI/EDS. They are most striking for mutations closer to the extreme amino end of the helix and become more subtle as mutations approach the microunfolding region. We did not observe similar effects for G121D and G136R as well as several other glycine substitutions beyond the first 90 residues (data not shown).
FIG. 6. Structural basis of OI/EDS region. A, differential scanning calorimetry (DSC) thermograms of procollagen and collagen peaks on each tracing correspond to apparent denaturation temperatures. All three possible peaks for molecules with no, one, and two mutant ␣1(I) chains are clearly distinguishable in the G76E thermogram. In other thermograms two or all three peaks strongly overlap and become indistinguishable. B, amino acid sequence of the amino-terminal triple helical fragment of human collagen. Proline (P) and hydroxyproline (O) residues required for triple helical stability are underlined and highlighted in green. Based on their content, the first 85 residues are expected to form a stable triple helix that anchors the amino end of the molecule. The 15-amino acid stretch on the carboxyl end of this N-anchor (highlighted in yellow) contains no proline or hydroxyproline and is expected to be highly flexible. It is homologous to a known microunfolding region in mouse and rat tail tendon collagens (42). Their processing, matrix deposition, and DSC were typical for type I collagen helical mutations causing OI. Apparently, the flexible microunfolding region prevents propagation of structural defects caused by such mutations into the N-anchor.
Thus, the mutations that are responsible for an OI/EDS phenotype play both direct and indirect roles. As collagen structural mutations, they are directly responsible for the OI symptoms of the probands. By interfering with N-propeptide FIG. 7. Model for mechanism of EDS symptoms in EDS VIIA and OI/EDS. a and b, OI/EDS mutations (small circles) alter the secondary structure of the N-anchor region at or near body temperature. The conformational change results in up to a 5°C decrease in collagen T m (Fig. 6A), most pronounced in collagen obtained by pepsin treatment in 0.5 M acetic acid at 4°C (pepsincollagen). The extent of the conformational change in procollagen is limited by the clamping action of the N-propeptide, which holds the three collagen chains together and reduces the effect of the mutation on T m . C, normal procollagen is processed by N-and C-proteinases at specific cleavage sites and self-assembles into fibrils. D, EDS VII procollagen has a mutation involving partial or full deletion of the N-proteinase cleavage site. The N-propeptide is not cleaved because the cleavage site is absent and pN-collagen results. Incorporation of pN-collagen restricts lateral growth of fibrils by steric hindrance, resulting in reduced diameter and irregular shape of fibril cross-sections and reduced fibril strength. E, in OI/EDS, the amino acid residues comprising the N-proteinase cleavage site are intact. Still, the conformational change in the adjacent N-anchor region affects the configuration of the N-proteinase cleavage site, the enzyme cannot cleave, and pN-collagen results. As in EDS VII, incorporation of pN-collagen leads to reduced fibril diameter and strength. More regular shape of fibril cross-sections in OI/EDS might be related to different conformation of the flexible arm connecting the triple helix with uncleaved N-propeptide. processing, they are indirectly responsible for the EDS symptoms by a mechanism similar to EDS VIIA and -B (Fig. 7, C-E). In EDS VIIA and -B, however, the retention of the N-propeptide is caused by a deletion of the N-proteinase cleavage site. In OI/EDS, the cleavage site is intact. Instead, resistance to cleavage must be based on unfolding of the high-stability region at the amino end of the helix extending into the propeptide. Cleavage of the N-proteinase site requires an intact helical trimer (43)(44)(45) and maintenance of the normal conformation of the site itself, with an 18-amino acid hairpin-loop preceding the short ␣ helix in which cleavage occurs (46). Although the presence of the N-propeptide provides the overall thermal stability detected by DSC, the N-proteinase cleavage site must be in an inaccessible configuration in the OI/EDS procollagens. There is experimental evidence supporting a normal procollagen configuration in which the N-propeptide is folded back over the end of the helical region (47)(48)(49). Our data supports unfolding of the N-proteinase cleavage site but does not shed new light on the normal configuration of the N-propeptide itself.
Children with OI/EDS mutations in ␣1(I) have a distinct phenotype, with joint hyperlaxity and early progressive scoliosis, as well as type III and IV OI symptoms, including fractures, osteoporosis, short stature, and blue sclerae. Within the ␣1(I) OI/EDS patient group, those with exon 7 mutations have relatively more N-propeptide retention, more severe EDS, and mild to moderate OI, whereas those with exon 11 mutations have less N-propeptide retention, milder EDS, and more severe OI. It is interesting that the OI/EDS symptoms of these patients are different from the clinical presentation of mutations at the amino end of the ␣2(I) chain. The patients with mutations in ␣2(I) have a presentation that is similar to EDS VIIA and -B, with severe hyperextensibility of large and small joints and bilateral congenital hip dysplasia (24 -28). They have mild manifestations of OI, limited to osteoporosis and blue sclerae. The distinction in phenotype between EDS VII/␣2(I) OI/EDS and ␣1(I) OI/EDS suggests that there is a difference at the level of the collagen fibril or the matrix. It is important to note that the mutations of the two sets are of different types. The EDS VIIA/B and ␣2(I) OI/EDS cases all involve total or partial loss of an exon or a large duplication. This will result in a shift of register of the mutant chain with respect to the remaining two chains of the collagen trimer and may cause a similar secondary configuration at the junction of propeptide and helix. The ␣1(I) OI/EDS cases are almost all glycine substitutions that cause local unfolding but would not be expected to shift the register of the chains in a trimer with respect to each other.
Other factors that may contribute to the difference in phenotype are the location of the retained N-propeptides in the fibrils and their accessibility to cleavage by tissue proteases. Holmes and co-workers (38) used scanning transmission electron microscopy to show that N-propeptides in EDS VIIB fibrils were in a bent-back configuration and were located in both gap and overlap zones on the fibril surface. Co-polymers of collagen and pN-collagen maintain a circular cross-section with N-propeptides located in the overlap zone until N-propeptides reach a critical density on the fibril surface. In the EDS VII fibrils with irregular cross-sections, N-propeptides are equally distributed in gap and overlap zones. Furthermore, in the co-polymers containing both collagen and pN-collagen, Watson et al. (33) demonstrated that hieroglyph fibrils could be converted to near-circular cross-section by protease digestion after fibrils formed.
In EDS VIIA and -B and ␣2(I) OI/EDS, the occurrence of aggregate (cauliflower) and hieroglyph forms in fibrils is consistent with the persistence of N-propeptides on the fibril surface, perhaps in a bent-back conformation. In our ␣1(I) OI/EDS probands, in contrast, the fibril diameter is consistently very small, presumably because steric hindrance by partially unfolded N-propeptides has severely limited lateral growth. The retained N-propeptides may be located only in overlap zones. Alternatively the pN-collagen at the fibril surface may have been non-specifically cleaved by tissue proteases.
The ␣1(I) N-propeptides from ␣1(I) OI/EDS probands differ from those of EDS VIIA in both their primary structure, in that the 24 amino acid residues encoded by exon 6 are absent in EDS VIIA, and presumably in their secondary structure as well. These differences may account for the cleavage susceptibility of OI/EDS ␣1(I) N-propeptides after fibril formation. Finally, the extent of formation of cross-links by the lysine residue in exon 6 may contribute to the phenotype by altering resistance to mechanical shearing. Investigation of the structure and location of the retained N-propeptide in ␣1(I) versus ␣2(I) OI/EDS fibrils and its impact on tissue function will provide further insight into phenotypic mechanisms.