Multiexon deletions in the type I collagen COL1A2 gene in osteogenesis imperfecta type IB. Molecules containing the shortened alpha2(I) chains show differential incorporation into the bone and skin extracellular matrix.

Osteogenesis imperfecta (OI) type IB is a rare subset of the mildest form of OI, clinically characterized by moderate bone fragility, blue sclera, and dentinogenesis imperfecta. Cultured skin fibroblasts from two unrelated individuals (OI-197 and OI-165) with the typical features of OI type IB produced shortened alpha2(I) chains. Reverse transcription-polymerase chain reaction of the alpha2(I)-cDNA revealed deletions in the triple helical domain of 5 exons (exons 7-11) in OI-197, and 8 exons (exons 10-17) in OI-165. This exon skipping was caused by genomic deletions in one allele of COL1A2 with the breakpoints located in introns 6 and 11 in OI-197, and introns 9 and 17 in OI-165. The secretion and deposition of the mutant collagen into the matrix was measured in vitro in cultures of skin fibroblasts and bone osteoblasts, grown in the presence of ascorbic acid to induce collagen matrix formation and maturation, as well as in collagen extracts from skin and bone. The secretion of mutant collagen was impaired and long term cultures of fibroblasts showed that the mutant collagen was not incorporated into the mature collagenous matrix produced in vitro by skin fibroblasts from both patients. Likewise, the shortened alpha2(I) chain was not demonstrable in skin extracts. In contrast, bone extracts from OI-197 showed the presence of the mutant collagen. This incorporation of the abnormal collagen into the mature matrix was also demonstrated in long term cultures of the patient's osteoblasts. The deposition of the mutant collagen by bone osteoblasts but not by skin fibroblasts demonstrates a tissue specificity in the incorporation of mutant collagen into the matrix which may explain the primary involvement of bone and not skin in these patients.

Osteogenesis imperfecta (OI) type IB is a rare subset of the mildest form of OI, clinically characterized by moderate bone fragility, blue sclera, and dentinogenesis imperfecta. Cultured skin fibroblasts from two unrelated individuals (OI-197 and OI-165) with the typical features of OI type IB produced shortened ␣2(I) chains. Reverse transcription-polymerase chain reaction of the ␣2(I)-cDNA revealed deletions in the triple helical domain of 5 exons (exons 7-11) in OI-197, and 8 exons (exons 10 -17) in OI-165. This exon skipping was caused by genomic deletions in one allele of COL1A2 with the breakpoints located in introns 6 and 11 in OI-197, and introns 9 and 17 in OI-165. The secretion and deposition of the mutant collagen into the matrix was measured in vitro in cultures of skin fibroblasts and bone osteoblasts, grown in the presence of ascorbic acid to induce collagen matrix formation and maturation, as well as in collagen extracts from skin and bone. The secretion of mutant collagen was impaired and long term cultures of fibroblasts showed that the mutant collagen was not incorporated into the mature collagenous matrix produced in vitro by skin fibroblasts from both patients. Likewise, the shortened ␣2(I) chain was not demonstrable in skin extracts. In contrast, bone extracts from OI-197 showed the presence of the mutant collagen. This incorporation of the abnormal collagen into the mature matrix was also demonstrated in long term cultures of the patient's osteoblasts. The deposition of the mutant collagen by bone osteoblasts but not by skin fibroblasts demonstrates a tissue specificity in the incorporation of mutant collagen into the matrix which may explain the primary involvement of bone and not skin in these patients.
Osteogenesis imperfecta (OI) 1 is a brittle bone disease that varies in severity from perinatal lethal to mild forms. In spite of the clinical variability, mutations in the genes for the pro-␣1(I) chains (COL1A1) and pro-␣2(I) chains (COL1A2) of type I collagen have been defined as the basis of the disease in more than 90% of cases studied to date (for reviews see Refs. [1][2][3]. The most common and mildest form of OI, OI type I (OI-I), is characterized by blue scleral hue, bone fragility with minimal deformities, and autosomal dominant inheritance (4). Dentinogenesis imperfecta occurs in some patients and this is used to subclassify patients into OI-IA (no dentinogenesis imperfecta) and OI-IB (dentinogenesis imperfecta present) (4,5).
Biochemically, patients with OI-I commonly show reduced production of structurally normal type I collagen as a result of a COL1A1 "null" allele caused by structural mutations that prevent procollagen assembly (6) or more commonly, by mutations that introduce premature stop codons, producing either unstable mRNA or the synthesis of truncated unstable collagen (7). Structural mutations within the triple helical domain of type I collagen usually cause more severe OI phenotypes; however, exon-skipping (7,8) and glycine substitution mutations (9 -16) have been defined in OI-I, but these are clustered toward the amino-terminal end of the triple helix domain, presumably reducing their impact on helix propagation and structure (10). Furthermore, in contrast to the structurally abnormal collagens, which are incorporated into the extracellular matrix in severe forms of OI (17,18), the abnormal collagen chains in OI-I may be excluded from matrix formation (12) producing a milder phenotype. Although COL1A1 is the predominant disease locus, OI-I can also result from COL1A2 exon-skipping mutations, which also clustered toward the amino-terminal end of the triple helix (19 -22), and in a single reported case, from a glycine substitution mutation in COL1A2 (23).
It is unclear whether the OI-IA and OI-IB phenotype results from distinct types of collagen mutations. This is due, in part, to the paucity of detailed clinical information provided in most publications on patients with defined COL1A1 or COL1A2 mutations, which does not allow classification of the patients as OI-IA and OI-IB with any degree of certainty. However, two of the probands with COL1A2 mutations displayed dentinogenesis imperfecta and could be classified as OI-IB (21,22) and detailed linkage analyses by Sykes et al. (24) further suggest that COL1A2 mutations may be a major cause of OI-IB.
While the definition of the spectrum of mutations causing OI remains an important goal, the major research challenge is the definition of the molecular mechanisms by which collagen structural mutations affect the complex organization and homeostasis of the extracellular matrix. In the present report, we describe the molecular defects in two probands and one affected parent with OI type IB. In contrast to the mild clinical phenotype, they had multi-exon deletions of the COL1A2 gene. The effects of these mutations on collagen synthesis, secretion, and matrix deposition were studied using skin, bone, and long term cultures of fibroblasts and osteoblasts. These studies demonstrate that skin and bone cells respond differently to the production of the mutant collagens, displaying differential incorporation of the mutant collagen into the extracellular matrix. These differences in the metabolism of the mutant collagen in bone and skin may account for the presence of bone fragility and the absence of clinical abnormalities of the skin.

EXPERIMENTAL PROCEDURES
Clinical Summary-A diagnosis of OI type IB was made for both patients based on persistence of gray-blue scleral color in association with dentinogenesis imperfecta and bone fragility (4).
OI-165-This 11-year-old boy with OI, whose birth length was on the 10th percentile, was diagnosed at birth because of leg bowing. A skeletal survey showed healing fractures in the leg, arm, and ribs. He had approximately 40 fractures including a malunion of a left femoral fracture at 2 months that resulted in a shortening with an 8.5-cm leg length discrepancy. His sclera were moderately dark blue-gray, grade 4/8 (25). Dentinogenesis imperfecta was present in primary teeth, although his permanent teeth were not noticeably affected. Head circumference was on the 98th percentile and height less than the 3rd percentile. He had marked hypermobility of distal and middle interphalangeal joints and moderate hypermobility of the left knee. Apart from the leg-length discrepancy and mild anterior bowing of both legs, there were no other deformities of the long bones. He had a mild postural scoliosis due to his leg length discrepancy. Skeletal radiographs showed generalized osteopenia but relatively wide cortices at the mid-shaft of long bones (Fig. 1a). Skull x-rays showed multiple Wormian bones. A CT scan of the cranio-cervical junction showed basilar impression. While neither parent was affected, heterozygosity for the mutation in the proband indicated that this was a new dominant mutation.
OI-197-This family included an affected father and daughter. The proband was a female aged 13 years. Her birth length was on the 3rd percentile. She had her first fracture aged 13 months and subsequently had more than 120 fractures. There were no known fractures of femora or vertebrae. Her sclera were intensely blue-gray, grade 5/8. Dentinogenesis imperfecta was present in both primary and permanent teeth. Head circumference was on the 98th percentile and height less than the 3rd percentile. Her skull showed temporal bulging and occipital bossing. The right arm had a fixed flexion deformity due to an old fracture. There was no structural deformity in the lower limbs apart from right genu valgum. She had marked hypermobility of proximal and distal interphalangeal joints of the fingers and in metarso-phalangeal joints. Hearing was normal. Skeletal radiographs showed generalized osteopenia in the long bones (Fig. 1b) and spine with normal height of most vertebrae (Fig. 1c). Skull x-ray showed multiple Wormian bones without evidence of basilar impression. The 51-year-old father had intensely blue-gray sclerae, arcus corneae, height less than the 3rd percentile, head circumference on the 98th percentile, and hearing impairment. His teeth had prematurely worn and were extracted at 15 years of age. He had no fractures but suffered back and knee pain associated with osteoporosis. X-rays of his spine and long bones showed mild osteopenia and widened disc spaces.
Cell Culture-Samples of skin and bone from patients and agematched controls were obtained during routine surgery with informed consent and approval of the Ethics Committee of this hospital. Skin fibroblasts were established from biopsies and grown as described previously (26,27). Osteoblasts cultures (OI-197) were established from mechanically cleaned bone chips. Soft tissues and surface cells were removed by digestion with 2 mg of bacterial collagenase (Worthington CLS2)/ml of Ham's F-12 medium (Flow Laboratories) containing 5% (v/v) fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin, for 2 h. The bone chips were then placed in a Petri dish with DMEM containing 10% (v/v) fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Osteoblasts that grew out from the bone chips were subcultured and grown in DMEM containing 10% (v/v) fetal calf serum. The phenotype of the osteoblasts was verified by measurement of alkaline phosphatase activity.
Collagen Biosynthetic Labeling-After culture for 3 days in DMEM containing 10% (v/v) fetal calf serum and 0.25 mM ascorbate (Sigma), confluent cells were labeled with 10 Ci/ml L-[2,3-3 H]proline (44.5 Ci/ mmol, DuPont NEN) for 18 h in DMEM containing 10% (v/v) dialyzed fetal calf serum (26,27). The cell and medium fractions were harvested separately, and procollagens were precipitated with (NH 4 ) 2 SO 4 at 25% saturation and converted to collagen by limited pepsin digestion before electrophoresis (26,27). In some experiments the 0.25 mM sodium ascorbate in the labeling medium was replaced with 0.1 mM ␣,␣Јdipyridyl to prevent the post-translational hydroxylation of the procollagens. The cell fraction was collected, and the unhydroxylated procollagen was analyzed by electrophoresis after precipitation with (NH 4 ) 2 SO 4 at 25% saturation.
In Vitro Matrix Analysis-Matrix deposition was induced by ascorbate in long term culture of fibroblasts and osteoblasts (18,28). From confluence, the osteoblasts were grown for another 14 days, and fibroblasts for another 21 days, in DMEM medium containing 10% (v/v) fetal calf serum and 0.25 mM ascorbic acid. The medium was then replaced with 10 ml of DMEM medium containing 10% (v/v) dialyzed fetal calf serum, 0.25 mM sodium ascorbic acid, and 10 Ci/ml L-[2,3-3 H]proline. Following incubation for 18 h, the medium was collected and analyzed. The cell matrix was serially extracted at 4°C with a neutral salt buffer (50 mM Tris/HCl, pH 7.5 containing 0.15 M NaCl, 5 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, and 10 mM N-ethylmaleimide) to extract newly synthesized collagens, 0.5 M acetic acid to extract collagens with acid labile cross-links, and limited pepsin digestion (0.1 mg/ml pepsin in 0.5 M acetic acid) to extract mature cross-linked collagens. The extracts were collected by centrifugation, and portions of the neutral salt buffer and 0.5 M acetic acid extracts were subjected to limited pepsin digestion before electrophoresis.
In Vivo Matrix Analysis-Cleaned trabecular bone fragments were decalcified with 50 mM Tris/HCl, 0.2 M EDTA, pH 7.5, for 5 days at 4°C and washed with cold deionized water. The epidermis and fat tissues were mechanically removed from skin biopsies, and both the decalcified bone fragments and skin were then extracted for 18 h with cold chloroform:methanol (2:1) to remove any residual fat, and then dried under vacuum and weighed. The tissues were rehydrated in neutral salt buffer, freeze-milled, and extracted to remove the non-collagenous proteins. The pellets were extracted with 0.5 M acetic acid and subjected to limited pepsin digestion at an enzyme substrate ratio of 1:10 (27). The pepsin-solubilized collagens were lyophilized and analyzed by electrophoresis.
SDS-Polyacrylamide Gel Electrophoresis-Collagen chains were resolved on a 5% (w/v) polyacrylamide separating gel containing 2 M urea and a 3.5% (w/v) stacking gel. The sample preparation, electrophoresis conditions, Coomassie Brilliant Blue staining, and fluorography of ra- dioactive gels are described elsewhere (26,27). Coomassie-stained collagen bands were quantified by densitometry (Bio-Rad GS-670 densitometer) by comparison to standard collagen samples loaded on to each gel. The radioactivity in each collagen band was determined by excision and scintillation counting (29).
RT-PCR and cDNA Sequencing-Total cytoplasmic RNA was purified from skin fibroblast cultures (30), and first-strand cDNA was synthesized using a RT-PCR kit (Perkin-Elmer) using an ␣2(I) specific primer COL1 (Table I). cDNA corresponding to exons 6 -30 of the COL1A2 gene was amplified with COL23 and COL1 primers ( Table I). The PCR products were purified by electrophoresis on a 0.8% (w/v) agarose gel and recovered by electroelution, phosphorylated with T4 polynucleotide kinase and cloned into the SmaI site of M13mp18 vector (Amersham Corp.). Single-stranded DNA preparations from the individual clones were sequenced using a Sequenase kit (U. S. Biochemical Corp.) (31).
Southern Blot Analysis-Genomic DNA was prepared from confluent skin fibroblasts (32) established from OI-165 and OI-197, and from unrelated healthy individuals as controls. Whole blood genomic DNA was also prepared from the father of OI-197 (33). A 1341-bp PstI/NcoI fragment encompassing exons 9 -30 of the COL1A2 gene was isolated from a cDNA clone (34) and used as a probe for Southern blot analysis. Approximately 30 ng of the probe was labeled with [ 32 P]dCTP (3000 Ci/mmol, Amersham Corp.) to high specific activity using a random primer labeling kit (Boehringer Mannheim). EcoRI-digested fragments of genomic DNA were separated on a 0.8% (w/v) agarose gel and transferred onto a nylon Hybond Nϩ membrane (Amersham Corp.). Hybridization was carried out using standard procedures at 42°C in the presence of 50% (v/v) formamide. The filters were washed under stringent conditions at 50°C with 0.1 ϫ SSC (saline sodium citrate) and exposed to Kodak XAR-5 films at Ϫ70°C with an intensifying screen.
Amplification and Sequencing of Genomic DNA-The sequence and location of primers used for genomic PCR are listed in Table I. To obtain normal intronic sequences, 0.1 g of control DNA was amplified with primers COL32/COL38 for intron 9 and intron 10, COL31/COL35 for intron 11, COL34/COL39 for intron 17, and COL45/COL43 for downstream sequence of intron 6. To identify the breakpoints of the deletion in OI-165 and OI-197, primers located in exons 9 and 18 (COL32/ COL39) and in exons 6 and 12 (COL23/COL31), respectively, were used for PCR amplification. Amplifications were carried out in the presence of 1 unit of Taq extender (Stratagene) per unit of Taq polymerase (Boehringer Mannheim), and with extension times of 1 min/kb of expected amplification product. The products were purified by centrifugation in Centricon 100 columns (Amicon). Approximately 33 ng/kb of the purified products were used for direct sequencing using a cyclesequencing kit (U. S. Biochemical Corp.)

RESULTS
Collagen Analysis-The pepsin-resistant collagen produced by short term culture of OI-165 and OI-197 fibroblasts contained normal ␣1(I) and ␣2(I) chains as well as abnormal, faster migrating ␣1(I) and ␣2(I) bands. They were designated ␣1(I)* and ␣2(I)* in OI-197 (Fig. 2a, lanes 2 and 5) and ␣1(I)** and ␣2(I)** in OI-165 (Fig. 2a, lanes 1 and 4). Mutant and normal collagen secretion was determined by the quantitation of the distribution of radioactivity in the pepsin-resistant normal and mutant ␣2(I) chains in the cell layer and medium fraction after the 18-h labeling period. In OI-165 fibroblasts, 13% of the radiolabeled mutant ␣2(I)** chains and 95% of the normal ␣2(I) chain were secreted from the cell layer into the medium. In OI-197 the percentages were 71% and 89%, respectively. The secretion of the normal chains did not differ significantly from the control values of 92 Ϯ 3% (n ϭ 10).
To establish the origin of the faster migrating collagen chains, intact unhydroxylated procollagens were produced by incubation of OI-165 and OI-197 fibroblasts with ␣,␣Ј-dipyridyl. This unhydroxylated procollagen is retained within the cell in an unprocessed pro-␣-chain form, allowing a clearer interpretation of electrophoretic gels without the complexities introduced by the presence of procollagen processing intermediates. Electrophoretic analyses showed normal and shortened forms of the unhydroxylated pro-␣2(I) chains in OI-165 and OI-197, but only normal unhydroxylated pro-␣1(I) chains (Fig. 2a, lanes   7 and 8). The ratio of the normal and mutant unhydroxylated pro-␣2(I) chains were approximately 1:1 for OI-165 and 1:3 for OI-197. The ratios of the unhydroxylated pro-␣1(I) chains to the combined unhydroxylated normal and shortened forms of the pro-␣2(I) chains were 2.46 in OI-165 and 2.65 in OI-197. These values were similar to the control ratio of 2.38. From these results we concluded that each proband was heterozygous for a mutation of COL1A2 that yielded a shortened ␣2(I) chain. We also concluded that the shortened ␣1(I) chains observed following pepsin digestion were due to abnormal cleavage of the normal pro-␣1(I) chains resulting from the presence of shortened pro-␣2(I) chains in collagen heterotrimers.
In each proband, cleavage of pepsin-digested collagens with fibroblast collagenase yielded two forms of the NH 2 -terminal TC A and one form of the carboxyl-terminal TC B fragment of ␣2(I) chains (Fig. 2b, lanes 2 and 3). This observation localized the peptide deletion to the NH 2 -terminal three-quarter fragment of the triple-helical domain of the mutant ␣2(I) chains (see Fig. 3). The deletions were estimated to be approximately 15 kDa for OI-165 and 10 kDa for OI-197 by semi-log plot analysis using ␣2(I) collagen chains and CNBr-cleavage peptides as molecular weight standards.
Characterization of the COL1A2 Mutations-CNBr cleavage of the pepsin-digested collagen (see Fig. 3 for ␣2(I) peptide The collagens in the cell layer (lanes 1-3) and secreted into the medium (lanes 4 -6) were digested with pepsin and analyzed on a 5% (w/v) polyacrylamide gels unreduced. Unhydroxylated procollagens were also produced from fibroblast cultures in the presence of ␣,␣Ј-dipyridyl, and the procollagens in the cell layer were analyzed after reduction with 10 mM dithiothreitol (lanes 7-9). Shortened forms of the type I collagen chains were designated as ␣1(I)* and ␣2(I)* for OI-197 and ␣1(I)** and ␣2(I)** for OI-165. b, pepsin-digested collagens from the medium fractions were also subjected to fibroblast collagenase digestion, and the resultant TC A and TC B fragments were analyzed on a 7.5% (w/v) polyacrylamide gels.

COL1A2 Multiexon Deletions in Osteogenesis Imperfecta
locations) generated only normally migrating ␣2(I)CB3.5 peptides and reduced amounts of the ␣2(I)CB4 peptide in both patients (data not shown), further demonstrating that the deletions were most likely to be amino-terminal to the ␣2(I) CB3.5 peptide, and therefore within the amino-terminal 300 amino acids of the ␣2(I) helical domain. The corresponding region of cDNA was amplified with COL1 and COL23 (Table I).
In addition to the expected 1513-bp fragment, which was the only fragment amplified from control (Fig. 3, lane 2), shortened products of 1252 and 1055 bp were observed for OI-197 (Fig. 3,  lane 3) and OI-165 (Fig. 3, lane 4), respectively. Sequencing of these shortened fragments showed triple-helical in-frame deletions of exons 10 -17 in OI-165 and of exons 7-11 in OI-197 (Fig. 4). Five clones of each of the PCR products were sequenced, and no PCR errors were detected.
To identify the location of the multi-exon deletions, a Southern blot analysis was performed on genomic DNA. Hybridization of EcoRI-digested DNA to a cDNA probe specific for this region of the COL1A2 gene showed the expected 5.5-and 14-kb fragments from controls. Within the 14-kb fragment, there is a polymorphic EcoRI site (35) resulting in cleavage of this fragment into 10.5-and 3.5-kb fragments. The proband and the affected father of OI-197 were (Ϫ/Ϫ) for the EcoRI polymorphism. An additional fragment of approximately 11 kb was evident in OI-197 and her affected father, whereas an additional 8-kb band was observed in OI-165 (Fig. 5). These findings indicated genomic deletions of approximately 3 and 6 kb within the 14-kb EcoRI fragment of OI-197 and OI-165, respectively. OI-165 was (del/Ϫ) for the EcoRI polymorphism since the deletion encompasses the region of the polymorphic EcoRI site within the mutant allele. While the affected father of OI-197 showed a milder phenotype, he is unlikely to be a mosaic for the mutation since the mutant and normal bands are of equal intensity in both the father and the proband (Fig. 5).
Genomic Deletion Breakpoints-The deletion breakpoints were identified using genomic PCR with primers positioned in adjacent exons (Table I). The primer sets selected for genomic PCR would not have amplified any fragments from controls as the expected products were too large for the PCR protocol used. However, a 2.7-kb product was obtained using primer set COL23/COL31 from OI-197 and her affected father (Fig. 6b) and a 0.27-kb product was obtained using primer set COL32/ COL39 from OI-165 (Fig. 6a). Sequencing of the 0.27-kb fragment from OI-165 showed the deletion breakpoints to be in introns 9 and 17 (Fig. 6a). Sequencing from both the upstream and downstream ends of the 2.7-kb product from OI-197 showed sequences from intron 6 and intron 11, respectively (36). The deletion breakpoint in intron 6 was approximately 2250 bp from its upstream end and approximately 1150 bp from

FIG. 5. Southern blot analysis.
EcoRI-digested genomic DNA were resolved on a 0.8% (w/v) agarose gel, transferred onto a nylon Hybond Nϩ membrane, and hybridized to an ␣2(I) cDNA probe (see "Experimental Procedures" for details). wt (ϩ/ϩ) represents the wild type DNA homozygous for an EcoRI polymorphic site within the 14-kb fragment. wt (Ϫ/Ϫ) represents the homozygous wild type DNA, which lacks the EcoRI polymorphic site. The position of the EcoRI fragments, the polymorphic EcoRI site (arrow E) and the genomic deletions in the probands relative to the exon/intron organization of the COL1A2 gene are shown. The COL1A2 exons that hybridize to the labeled cDNA probe are indicated. its downstream end. The deletion breakpoint in intron 11 was 63 bp from its upstream end and 458 bp from its downstream end (Fig. 6b). The breakpoints for both OI-165 and OI-197 were determined precisely by comparison with normal amplified intron sequences using primers detailed under "Experimental Procedures." The location of all the primers used in genomic PCRs also shown in Fig. 6. The primer upstream to the breakpoint in intron 6 (primer 45) was designed based on sequence obtained from the mutant allele of OI-197 PCR-amplified with primers COL23 and COL31.

Deposition of the Mutant Collagen into the Matrix Formed in Vitro and in Vivo -
The deposition of an ascorbate-induced collagenous matrix was studied in fibroblasts (OI-165 and OI-197) and osteoblasts (OI-197) cultured for 21 and 14 days, respectively (Fig. 7) by serial extraction with neutral pH isotonic buffer (newly synthesized procollagens and collagens), acetic acid (newly cross-linked collagens) and pepsin (insoluble mature cross-linked collagens). The complete extraction of the labeled collagens from the in vitro matrices was confirmed by scintillation counting of the acid hydrolyzed pepsin residues. The collagenous matrices of skin and bone were also studied.
Collagen deposition by OI-197 osteoblasts after 14 days of culture was reduced to approximately 30% of the control in three separate experiments. The collagen content of extracts of OI-197 bone was also reduced to approximately 35% of control. OI-197 osteoblasts synthesized mutant ␣2(I)* chains. The secretion of mutant collagen was impaired (comparable to that demonstrated for OI-197 skin fibroblasts) as there was a greater proportion of mutant chains in the neutral salt extract, which contained mainly intracellular collagen, than in the medium. However, a significant proportion of the mutant collagen was deposited into the osteoblast-produced extracellular matrix in vitro (Fig. 7a), demonstrated by the presence of the mutant ␣2(I)* chain in the collagen extracted with 0.5 M acetic acid and, to a lesser extent, in the pepsin extract. This result was also consistent with analysis of OI-197 bone, which demonstrated the presence of mutant molecules in pepsin extracts of the mature bone matrix (Fig. 7a).
In contrast, mutant ␣2(I) chains were not detected in the matrix of long term (21 days) skin fibroblast cultures from OI-197 and OI-165 (Fig. 7b). This finding was consistent in three separate experiments. The mutant ␣2(I) chains were produced by the fibroblasts but were largely restricted to the neutral salt extracts. In OI-165 fibroblasts, the labeled mutant chains were poorly secreted with only a trace amounts of the mutant molecules in the medium fraction, while more significant proportion of the mutant molecules were secreted from OI-197 fibroblasts (Fig. 7b). These findings were again consistent with the in vivo analysis of the collagen extracted from skin for OI-165 and OI-197, which showed a normal collagen composition with no detectable mutant ␣2(I)* (OI-197) or ␣2(I)** (OI-165) chains. As expected, the fibroblasts from both patients produced a matrix in vitro with a reduced collagen content (estimated from three separate experiments to be approximately about 45% for OI-197 and 29% for OI-165 of that produced by control cells) consistent with the reduced collagen content of skin extracts. DISCUSSION The two OI type IB patients were shown to be heterozygous for large multi-exon deletions in the COL1A2 gene for the ␣2(I) chain of type I collagen. In one patient (OI-197) the genomic deletion of 3 kb encompassed exons 7-11, and in the other patient (OI-165) the 6-kb deletion removed exons 10 -17. Both deletions yielded intron junctions that did not alter the splice donor and acceptor sites. The 5Ј and 3Ј exons were spliced normally, maintaining the translational reading frame and the repetitive Gly-X-Y triplet sequence. The deletions removed amino acids 145-296 2 and 94 -200 from the triple helical domain of OI-165 and OI-195, respectively.
Large deletions of the type I collagen genes are infrequent in OI, and the majority of mutations are point mutations, which result in helix-destabilizing glycine substitutions, single exonskipping mutations, premature stop codons; and frameshift mutations, which produced shortened or elongated dysfunctional collagens (1,3,14). There are two previously described large deletions, both of which lead to OI type II (perinatal lethal). In one of them, a deletion in COL1A1 extends from exon 23 to exon 25 (84 amino acids) (37)(38)(39); in the other, the deletion in COL1A2 extends from exon 34 to exon 40 (40). In these cases it was suggested that the deletions were produced by nonhomologous recombination. In the two patients described here, analysis of the sequences around the intron deletion breakpoints did not demonstrate the present of inverted sequences or direct repeats and there was no significant homology between the 5Ј and 3Ј deletion junctions. Furthermore, it was unlikely that Alu repetitive elements were involved in the deletion, since the Alu repeat elements of COL1A2 (41) are located well away from either of the deletion junctions. Thus, it is most likely that the deletions also arose through nonhomologous intron-mediated recombination.
The deletions produced pro-␣2(I) chains with large internal helical deletions, and the type I collagen molecules containing these shortened chains were secreted poorly. It should be noted that the calculation of secretion was based on the distribution of pepsin-stable collagens, and therefore did not take into account intracellular mutant collagen degradation or any altered pepsin stability of mutant ␣2(I) chain-containing collagen trimers. However, the presence of shortened ␣1(I) chains following pepsin digestion, and the relatively constant ratio of total ␣1(I):␣2(I) in combined cell and medium pools (1.98 for OI-165; 2.12 for OI-197; 2.12 for controls) suggests that the collagen triple helix carboxyl-terminal to the mutation is largely pepsin resistant. Increased pepsin sensitivity, if present, would probably be most evident in intracellular mutant collagens undergoing assembly and helix folding, and result in dramatically reduced intracellular collagen levels after pepsin digestion. Both this and intracellular degradation of mutant-containing collagen molecules would result in an artificial overestimate of mutant collagen secretion efficiency. These arguments suggest that the reduced collagen secretion measured in our experiments is probably an underestimate of the extent of the mutant collagen secretion defect.
The retarded secretion of structurally abnormal type I collagen is a common finding in OI and represents an important intracellular "quality control" mechanism (26,42). The retention of the mutant-containing collagen trimers within the cells results in increased intracellular breakdown, via endoplasmic reticulum-mediated and lysosomal degradative pathways (43), resulting in reduced collagen in the extracellular matrix. This deficiency was reflected in the marked reduction in the collagenous matrix deposited by fibroblasts (OI-165 and OI-197) and osteoblasts  grown in long term cultures in the presence of ascorbic acid. This collagen deficiency was confirmed in the skin and bone from OI-197 and skin from OI-165.
In OI-165, serial extraction of the in vitro extracellular collagen matrix formed by dermal fibroblasts with acetic acid and then pepsin to extract the progressively more cross-linked collagen, demonstrated that the mutant collagen was excluded from the mature in vitro matrix. Pepsin extracts of skin tissue samples confirmed the absence of the mutant collagen. Exclusion of collagen containing mutant pro-␣1(I) chains from a dermal fibroblast matrix formed in vitro in the presence of dextran sulfate has been previously reported (12). Unfortunately bone samples or osteoblast cultures were not available from OI-165, and we were unable to determine whether this mutation was included or excluded from the bone matrix.
For OI-197 skin fibroblasts, bone cells and tissue samples were available, enabling us to compare the behavior of the mutant collagen in these tissues. The results confirm that the mutant pro-␣2(I) chains are expressed by both fibroblast and osteoblast cultures (44). Studies comparing matrix formation in vitro in long term fibroblast and osteoblast cultures provided some interesting findings. Collagen deposition into the matrix was dramatically reduced in both fibroblast and osteoblast matrices, consistent with the tissue collagen deficiencies and with previous in vitro experiments demonstrating the reduced collagen deposition by OI fibroblasts (18). A reduction in collagen production has also been demonstrated for OI osteoblasts (45), but in these studies matrix deposition was not assessed. However, while serial extraction demonstrated that the mutant collagen was excluded from the mature in vitro and in vivo skin matrix, when bone cell culture and bone tissues were examined a totally different pattern emerged. In extracts of the osteoblast in vitro matrix, the mutant collagen was incorporated into the mature cross-linked collagenous matrix. The presence of the mutant collagen in pepsin extracts of mature bone tissue directly demonstrated that not only is the mutant incorporated into the matrix, but the mutant collagen is not degraded and remains a stable component of the mature matrix.
The mechanism of how bone and skin matrices discriminate and differentially incorporate the OI-197 mutant collagen is not known. It is becoming increasingly clear that collagen fibrillogenesis and maturation is a complex multistep process involving heterotypic collagen associations and interactions with other matrix components such as decorin and fibromodulin (46,47). While the predominant collagen of both skin and bone is type I collagen, there are many differences in the composition of other matrix components between these tissues, including the ability to mineralize, which may influence the pattern of collagen deposition and maturation. Several detailed structural studies have determined that the packing of collagen into fibrils, and the resulting molecular arrangement of adjacent collagen molecules, is different in bone to that in soft tissues such as skin and tendon (48,49). This alternate packing is most noticeably reflected in the different patterns of intermolecular cross-links in soft and hard tissues (50,51). The inclusion of the mutant collagen in the bone, but not skin matrix, may reflect this difference in packing, suggesting that the molecular arrangement in bone may be more permissive for the incorporation of the mutant molecules.
Previous studies have demonstrated that some collagen helical mutations have a long range effect on procollagen structure at the N-proteinase cleavage site (52). The resulting altered conformation of this cleavage site reduces procollagen processing and results in an accumulation of pN-collagen. The retention of the type I procollagen N-propeptide on a proportion of the mutant collagen molecules in vivo would present a steric barrier to correct fibrillogenesis analogous to that seen in Ehlers-Danlos syndrome type VII (53,54). In these Ehlers-Danlos syndrome patients, cleavage site mutations or N-proteinase deficiency result in the processing defect, which manifests clinically in skin and joint laxity, although there are reports of Wormian bones and blue sclera (55) suggesting that the defects are not totally confined to the soft tissues. In both the OI type IB patients, the mutant collagens are excluded from the skin matrix, and thus there is no opportunity the expression of the Ehlers-Danlos syndrome type VII clinical phenotype. In OI-197 bone the mutant collagen, which is shortened by a multi-exon deletion in the helical domain, may also have reduced N-propeptide processing, and the inclusion of this structurally abnormal collagen pN-collagen into the bone matrix may contribute to the pathological effect of the mutations on the bone matrix.
These studies also highlight important issues in the extrapolation from biochemical data obtained in fibroblast culture to the definition of the molecular pathology of the mutant protein in bone, the primary affected tissue. Likewise, these studies also demonstrate the differential matrix incorporation of collagens with different mutations, since our previous studies have shown that collagen with a helical glycine substitution mutation is efficiently incorporated into the in vitro fibroblast matrix but is then selectively degraded (18). Thus, the effect of the mutation on the incorporation of collagen into a functional matrix depends on the mutation position and type, its effect on parameters such as the intracellular and extracellular stability of the mutant collagen, secretion, and presumably also on its ability to interact appropriately with the developing fibrillar matrix. In an attempt to address these questions, further experiments are under way involving the stable transfection of fibroblast and bone cells with mutant ␣1(I) collagen genes and the comparison of the biochemical phenotypes that result.