A Tripeptide Deletion in the Triple-helical Domain of the Prod(1) Chain of Type I Procollagen in a Patient with Lethal Osteogenesis Imperfecta Does Not Alter Cleavage of the Molecule by N-Proteinase’

Dermal fibroblasts from a fetus with perinatal lethal osteogenesis imperfecta synthesized normal and abnor- mal type I procollagen molecules. The abnormal molecules contained one or two proal(1) chains in which glycine, alanine, and hydroxyproline at positions 874, 876, and 876 in the triple-helical region were deleted as the result of a %base pair genomic deletion. Molecules that contained abnormal chains were overmodified from the site of the deletion toward the amino-terminal region of the molecule. Secretion of the ov- ermodified molecules was impaired. The thermal stability of molecules containing abnormal chains was lower than that of normally modified molecules. After cleavage of molecules with vertebrate collagenase, the temperature of thermal denaturation of the overmodified A fragments was greater than that of the fragments from the normal molecules. The rates of cleavage of the normal and the abnormal molecules by N-pro-teinase were indistinguishable. Our findings suggest that the tripeptide deletion introduces a shift in the phase of the chains in the triple helix. This structural change is propagated from the site of the deletion to- 5' CAGGTACAGGGAACTGGAGCCCAGC 3') located at the 5' end of intron 44 were used to amplify a region of genomic DNA from the 01 cell strain and from a control cell strain that spanned exon 44. The resultant 225-bp product was digested with the restriction endonuclease Bad, and the digestion products were analyzed on a 6% PAGE gel.

ase, procollagen N-proteinase (EC 3.4.24.14), the enzyme that removes the amino propeptides of type I procollagen in vivo; PAGE, polyacrylamide gel electrophoresis; pccollagen, intermediate in the processing of normal type I procollagen to type I collagen containing the carboxyl-terminal propeptides but not the amino-terminal propeptides; pCa-chains, chains of pccollagen; bp, base pair; CB-peptides, cyanogen bromide peptides. tively) of type I collagen have been identified as the molecular cause of different forms of osteogenesis imperfecta (01) (1,2). Mutations causing the perinatal lethal form of 01 (01 type 11) include genomic deletions (3, 4), an insertion in COLlAl (5), splicing errors (6,7), and point mutations that lead to the substitution of single glycine residues within the Gly-X-Y repeat unit characteristic of the collagen triple helix (1, 2). Most of the mutations interfere with the normal assembly of a stable triple helix, delay secretion, and cause increased posttranslational modification (lysyl and prolyl hydroxylation and hydroxylysyl glycosylation) of all chains amino-terminal to the site of the alteration in the molecule (7-9).
An additional feature of molecules that contain chains with deletions or amino acid substitutions for glycine is that they are resistant or partially resistant to cleavage by N-proteinase (10,ll). Type I procollagen N-proteinase removes the aminopropeptide domain from the type I procollagen molecule during the normal processing of procollagen to collagen (12,13). Previous studies have demonstrated that N-proteinase will not cleave heat-denatured procollagen and is a substrate conformation-dependent proteinase (14). To explain how an amino acid substitution in the triple-helical region of the type I procollagen molecule results in slower processing by N - proteinase, it has been proposed that a conformational change, such as a tripeptide shift in the phase of the chains in the helix (15), is propagated from the site of the substitution toward the amino terminus and thereby disrupts the Nproteinase cleavage site (10, 11).
We have characterized a mutation in a fetus with 01 type I1 that results in the deletion of a tripeptide: glycine, alanine, and proline (normally hydroxylated to hydroxyproline and referred to in that manner subsequently) at positions 874, 875, and 876 of the triple helix of the proal(1) chain (residue 1 is the first glycine of the major triple helix). We have shown that an altered structure is propagated from the site of this deletion toward the amino-terminal region of abnormal molecules but that the rates of cleavage of the normal and abnormal molecules by N-proteinase are indistinguishable. These findings demonstrate that a tripeptide shift in the phase of the chains in abnormal molecules does not alter the N-proteinase cleavage site in such a way as to cause a decrease in the rate at which the enzyme cleaves the amino-terminal propeptide from those molecules.

MATERIALS AND METHODS
Clinical History-The infant was the first child born to a 36-yearold mother and 32-year-old father. Neither parent, none of the mother's four older children, nor any of the father's four older children (all with previous spouses) had evidence of osteogenesis imperfecta. This pregnancy was unremarkable by history, and fetal activity was 25529 Collagen ProalU,, Tripeptide Deletion Does Not Alter Cleavage felt to be normal. There was a question of polyhydramnios detected by ultrasound. Because of frank breech presentation at the onset of labor at 32 weeks gestation, delivery was by cesarean section. The infant weighed 1,500 g, was 33 cm long, and had a head circumference of 31 cm. Apgar scores at 1 and 5 min were 1 and 1, respectively, and the infant died of respiratory failure shortly after delivery. The bead was soft with no palpable cranium, the chest was small, and all four extremities were short and bowed. Radiographs were consistent with the diagnosis of the perinatal lethal form of osteogenesis imperfecta (01 type 11). Chromosomes were normal male. Preparation and Electrophoretic Analysis of Procollngens and Collagens-Fibroblast cell strains were established from explants of skin from the fetus with 01 type I1 . Cells from unrelated healthy subjects served as controls. Cell cultures were maintained under standard conditions in Dulbecco-Vogt modified Eagle's medium (Irvine Scientific) as described previously (16). Labeling of proteins with 2,3,4,5-[3H]proline (101 Cilmmol, Amersham Corp.), harvesting of the medium and cell layer proteins, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), cleavage of proteins with cyanogen bromide (CNBr) in gels, and two-dimensional analysis of the resultant peptides of type I collagen were carried out as described previously (8,16). Vertebrate collagenase fragments of the d ( I ) and ot2(I) chains were prepared from pepsin-digested [3H]proline-labeled procollagens by digestion with fibroblast collagenase for 16 h at 20 'C (17). The collagenase was provided by Dr. Eugene Bauer, Stanford University. Vertebrate collagenase cleaves type I collagen between amino acids 775 and 776 of the triple helix, generating the aminoterminal A fragment and the carboxyl-terminal B fragment.
Preparation and Purification of Pr~collagen-~~C-Labeled type I procollagen was purified from the culture medium of normal and proband dermal fibroblasts (passages 7-10) using the methods described previously (18). In brief, fibroblasts were grown to confluence and incubated in Dulbecco's modified Eagle's medium supplemented with 1 pCi/ml of a mixture of uniformly labeled [14C]-~-amino acids, 25 pg/mI ascorbic acid, and no serum. Medium proteins were precipitated by ammonium sulfate, and the type I procollagen was chromatographed on two consecutive columns of DEAE-cellulose (19,20). The procollagen was concentrated by ultrafiltration and stored at -20 "C in storage buffer consisting of 0.1 M Tris buffer (pH 7.4 at 20 "C) containing 0.4 M NaCl and 0.01% NaN3. Procollagen concentration was determined by a colorimetric hydroxyproline procedure (21), assuming 10.1% hydroxyproline by weight procollagen (19). The procollagens from control and the 01 cells had specific radioactivities of 1,500 and 1,350 cpmlpg, respectively.
Rotary Shadowing Electron Microscopy-Type I procollagen purified from the medium of control and 0 1 cells was prepared for rotary shadowing electron microscopy using the "mica sandwich" method described previously (22). In brief, a I-cm2-piece of mica was cleaved, 5 pl of procollagen was added to one surface, the other mica surface was opposed thereby forming a thin film of the sample. The sample was frozen by immersion of the mica sandwich in liquid nitrogen, the mica surfaces were separated, and the sample (on two mica surfaces) was dried in vacuo. The mica squares were rotated, and the procollagen molecules were shadowed with platinum-tungsten particles using a measured angle of 6". Carbon was evaporated onto the molecules, the replicas were floated onto clean water, and samples were picked up on 400-mesh copper grids for subsequent examination using a JEOL 1200EX transmission electron microscope operated at an accelerating voltage of 80 kV. Images of the replicas were recorded on Ilford EM film.
Assay of Thermal Stability-Thermal denaturation temperatures of normal and abnormal collagens and vertebrate collagenase digested collagens were determined using minor modifications of an enzymatic assay described previously (16,23).
Procollagen N-Proteinase-The enzyme was purified from the leg tendons of 250 dozen 17-day chick embryos as described previously (13). The preparation had 800 units/ml of activity where 1 unit is the amount required to cleave 1 pg of type I procollagen per h at 34 "C in a reaction system containing procollagen at a concentration of 10 pg/ ml.
Analysis of N-Proteinase Cleavage-Purified type I procollagen (IO pg/ml) was incubated with N-proteinase (10 units/ml) in 0.05 M Tris-HC1 (pH 7.4 at 20 "C) containing 0.15 M NaCl and 0.01% NaN3 at 34 "C. At specified times during incubation aliquots were removed and added to a one-quarter volume of 5 X SDS sample buffer, and the sample was heated at 100 "C for 3 min. The proteins were separated by SDS-PAGE, using 5% separating and 3.5% stacking gels (24), and visualized by fluorography. Fluorograms were prepared by equilibrating gels in 20% diphenyloxazole in ethanoic acid and exposing dried gels to preflashed Kodak X-Omat-AR film at -70 "C. The normally migrating procY(1) chains and the slowly migrating p r o d ) chains in the 01 sample were not resolved sufficiently by laser densitometry to permit analysis of the cleavage of normal molecules and abnormal molecules separately. To quantitate cleavage by N-proteinase, fluorograms (in triplicate) were scanned, and the areas under the peaks corresponding to the sum of the proal chains and the sum of the pCa chains were measured. The percent of pccollagen formed was calculated from (100 X pCd(I))/(prod(I) + pCal(I)), after correction for molecular mass.
DNA Sequence Determination-Total RNA was prepared from dermal fibroblasts (25)(26)(27). The preparation of cDNA from the al(1) mRNA and the amplification of the region encoding the al(I)CB6 (amino acid residues 823-1014 of the triple helix plus the 28 residues of the carboxyl-terminal telopeptide) were performed using oligonucleotide primers and methods that have been described previously C-IA2-6 (sequence: 5' AGAGTCGACAGTAGCATCAACTTCA- (28). Using the same approach, a SalI-tailed oligonucleotide primer, TAGT 3'), was used to prepare cDNA from the cu2(I) mRNA. The cDNA synthesized from the a2(I) mRNA in the domain encoding amino acids 821-1014 of the triple helix plus 26 residues of the carboxyl-terminal telopeptide was amplified using primer C-1A2-6 ATTCTTGGCATTGCCGGCCCTCCT 3'). The amplified cDNA and an EcoRI-tailed primer N-1A2-6 (sequence: 5' CTGGAproducts were separated on a 1% LMP gel (Bethesda Research Laboratories), and the appropriate fragments were excised and purified (29). The 712-bp amplified cul(1) cDNA fragment and the 741-bp a2(I) cDNA fragment were digested with EcoRI and SalI, and 30 ng was ligated to 100 ng of similarly cleaved M13mp19. Single-stranded DNA was prepared from clones containing inserts of the correct size (30) and the DNA was sequenced by the dideoxy chain termination method (31) and T7 polymerase (Sequenase, U. States Biochemical Corp.). Primer 6-2R (sequence: 5' TGACCACAGCCTTGTCTGC-TGCTTC 3') located in the 5' region of intron 43 (3' to exon 43) and primer 6-3.2 (sequence: 5' CAGGTACAGGGAACTGGAGCCCAGC 3') located at the 5' end of intron 44 were used to amplify a region of genomic DNA from the 01 cell strain and from a control cell strain that spanned exon 44. The resultant 225-bp product was digested with the restriction endonuclease B a d , and the digestion products were analyzed on a 6% PAGE gel.

Abnormality in al(I)CBG of Type Z Procollagen-Cells from
the infant with 01 produced normal molecules and molecules that contained (Y chains that were delayed in electrophoretic mobility. The chains in the abnormal molecules were preferentially retained by the cells (Fig. L4). The altered electrophoretic mobility of the chains synthesized by the cells from the fetus was the result of overmodification of chains of type I procollagen along their full-length (Fig. 1B).

A 9-Base Pair Genomic Deletion Leads to an In-frame Deletion of a Tripeptide-The
cDNA clones encoding the al(1)CBG region were either of normal sequence or contained a 9-bp deletion (either 5' GGTGCCCCT 3' or 5' GTGCCCCTG 3') which would result in the deletion of amino acid residues GlyM7, Alas75, and HypS7'j of the proal(1) triplehelical domain (Fig. 2). This deletion left the frame of the remaining coding sequence intact. To confirm that the deletion did not arise during the preparation of the cDNA or during the amplification process, genomic DNA was isolated from the 01 cell strain and from a control cell strain, the region suspected to contain the mutation was then amplified and cleaved by the restriction enzyme BanI (the recognition site for this enzyme would be eliminated by the presence of the mutation). The uncut amplified DNA from the fetus migrated as a doublet on PAGE gels, and only the upper band was cut by BanI, whereas the DNA fragment amplified from the control DNA cut to completion (results not shown). Parental tissues were not available for study so it was not possible to prove conclusively that the deletion had occurred de novo in the infants genomic DNA and was responsible for the lethal phenotype. For this reason, nine cDNA clones encoding the region of the pron2(I) chain corresponding to t h e CR6 region of the pronl(1) were sequenced. No sequence alterations were found in these clones (results not shown).
T h e lack of an additional sequence change in the COLIAP region of the gene corresponding to the CRfi region of protrl(1) confirmed that the observed overmodification resulted from the Gly-Ala-Hyp tripeptide deletion.
Rotary Shadowing Electron Microscopy-The t-ype I procollagens purified from the medium of the 01 and normal cells in culture were examined by rotary shadowing electron microscopy. Analysis of 200 molecules from each sample showed that the 01 molecules were indistinguishable from those in the control samples in that they were thread-like, approximately 300 nm in length, and did not have kinks as has been observed in type I procollagen molecules containing suhstitution of cysteine for glycine residues (15,45).

The Melting Temperature of the Overmodifird Collagpn I s Decreased, whereas the Overmodified Co1lagana.w A Fragments
Have an Increased Thermal Stability-The overmodified type I collagen molecules isolated from the medium and cell lavers of the 01 cell strain had a T,,, of 41 "C, whereas normal molecules synthesized by control cells and by the 01 cells had a T,,, of 42 "C (Fig. 3). Following cleavage with fibroblast collagenase of the collagens isolated from the medium and cell layer of the 01 fibroblasts, the overmodified collagenase A fragments melted at 38 "C whereas the normally modified A fragments and the A fragments from the control molecules melted at 36 "C (Fig. 4). The rate of cleavage by collagenase of the abnormal molecules was not slower than that of the normal molecules.
The vertebrate collagenase A fragments of t.ype 111 collagen migrate in gels to approximately the same position as overmodified a l ( I ) chains and have a thermal stability slightly above that of normally modified t r l ( l ) chains (32). T o confirm that we were observing the thermal denaturation of the normally modified and the overmodified A fragments of type I collagen, the thermal stability assay was repeated using a preparation of purified t-ype I procollagen that lacked type I11 procollagen. The thermal denaturation of this material was as obtained for the unpurified material (results not shown).
Cleavage of the Abnormal MolPculrs by N-ProtPinase-In experiments that examined the consequences of the deletion for cleavage of the procollagen by N-proteinase, purified t-ype I procollagen was incubated with partially purified N-proteinase in vitro under conditions previously shown to detect changes in activity of the enzyme toward the substrate (10, 45). The abnormal procollagen molecules containing one or  two abnormal pronl(1) chains were posttranslationally overmodified and migrated slowly in SDS gels. The difference in migration of the procu chains was not sufficient to allow laser densitometry of the normally migrating and slowly migrating chains separately, so the normally migrating and slowly migrating pron chains were analyzed together. The rates of cleavage of pronl(1) chains in control samples and of the normally migrating and slowly migrating pronl(1) chains in the 01 samples by N-proteinase were indistinguishable from each other using the methods described here (Fig. 5). At 50 units/ml of N-proteinase instead of the usual 10 units/ml of enzyme, the rates of cleavage of control procollagen and of normal and abnormal procollagens synthesized by the 01 cells were again indistinguishable from each other (inset in Fig.

DISCUSSION
Although considerable information is available about the spectrum of mutations that cause different forms of 01 and it is possible, to some extent, to predict the severity of the phenotype from the nature of the causative mutation, little is known about the ways in which mutations alter the structure and properties of the collagen molecule.
Most of what is known about the effects of mutations in collagen genes on the stability of the triple helix is the result of studies of substitutions of other amino acids for glycine, the invariant occupant of every 3rd residue position in the triple helix of each of the three chains. Substitutions by residues with bulkier side chains might be expected to introduce structural changes at the site of the substitution that delay helix propagation (from carboxyl to amino terminus) or t o modify the structure of the triple helix and, thereby, result in overmodification of the chains amino-terminal to the site of the substitution (because the modifying enzymes recognize the free chains or the abnormal structure). Two models have been proposed to explain how a substitution for glycine could delay helix formation. Traub and Steinmann (33) proposed that where a cysteine is substituted for a glycine, a bulge occurs in the triple helix at the site of the substitution to accommodate the bulkier side chain. Because the formation of a bulge would be energetically unfavorable, helix formation would be delayed. In the model proposed by Vogel et al. (15), a cysteine for glycine substitution at position 748 in the nl(1) chain produced a flexible kink at the site of the substitution and a tripeptide shift in the phase of one or both nl(1) chains with respect to n2(I). Again, the introduction of the kink would be energetically unfavorable and so would delay helix propagation and result in the observed overmodification. The molecules containing the Gly-Ala-Hyp tripeptide deletion at position 874-876 described here, like those containing amino acid substitutions, were overmodified from the site of the deletion to the amino terminus. The tripeptide deletion conserves the Gly-X-Y repeating structure of the triple helix and would not be expected to introduce a bulge or kink in the molecule. Indeed, rotary shadowing electron microscopy of purified procollagen did not detect a kink or a bulge in the molecules. Similarly, overmodification has been observed in molecules containing exon deletions (6, 7) and a genomic deletion in the n2(I) chain (4). Again, the GIy-X-Y repeat unit in those resultant abnormal molecules remained intact.
It is not obvious why molecules with deletions that conserve the GIy-X-Y repeat unit are overmodified, but it seems likely that specific interactions between the -X and -Y residues in adjacent chains are important in determining the rate of helix formation or the structure of the triple helix (4).
The thermal stability of the overmodified collagen containing the tripeptide deletion was reduced by 1 "C. The discrepancy between our findings and thos reported by Hawkins et al. (34), who identified a similar deletion in one of the adjacent Gly-Ala-Hyp tripeptides, is unclear but may reflect minor difference in the methods used to measure thermal stability.
The denaturation of type I collagen is a highly cooperative process, and no unfolding intermediates are observed (35,36). Microcalorimetry measurements of the triple helix over a wide range of temperatures shows that blocks of the triple helix "micro-unfold" in the predenaturational range of temperatures (36, 37). Denaturation of the triple helix is, therefore, envisaged as a cooperative unfolding of adjacent blocks of residues until, at the denaturation temperature, a critical length of the molecule is unfolded sufficiently to prevent refolding of the chains (38). One explanation for why the deletion of the Gly-Ala-Hyp tripeptide decreases the thermal stability of the triple helix is that the region of 874-876 is particularly stable. Gly-Ala-Hyp triplets occur in clusters along the d(1) chain (39). These triplets can form stable triple helices in collagen-like polypeptides and most likely contribute significantly to the stability of type I collagen (39,40). The Gly-Ala-Hyp tripeptide that is deleted is the third of three identical tripeptides extending from residues 868-876 and is preceded by a Gly-Pro-Hyp triplet. Perhaps, the lower T,,, of the molecules that contain the abnormal chains results, in part, from the disruption of a cooperative block containing these residues.
Following cleavage with fibroblast collagenase of the overmodified collagens to remove the domain containing the deletion, the thermal stability of the overmodified A fragments was higher than that of the normally modified fragments. An increased thermal stability for overmodified A fragments has been observed in other 01 type I1 mutant cell strains with substitutions in the B fragment ofthe al(1) chain: al(1) Glya7 to Arg (28) and d ( I ) Glyea3 to Asp (41).2 Again, in these instances, the thermal stability of the intact collagen molecules was reduced. The thermal stability of the overmodified A fragments generated from molecules containing a substitution of serine for glycine at position 844 was, however, indistinguishable from the normally modified A fragments (42). Thus, it does not appear that overmodification, per se, is responsible for the increased melting temperatures observed for the A fragments in molecules with amino acid substitutions or with the tripeptide deletion. Further, there is some evidence suggesting that the thermal denaturation of mutant 0 1 collagen is not altered by the degree of lysyl hydroxylation and hydroxylysyl glycosylation (43), and in some instances overmodified, abnormal molecules have been shown to have normal melting temperatures (44). These findings suggest that in molecules containing the deletion, the decrease in the thermal stability observed for the intact chains was caused by the structural change in the region of the deletion. Following removal of the region containing the deletion, the A fragment has a novel cooperative block structure resulting from the new alignment of the chains in the triple helix. Similarly, substitution of glycine by bulky residues such as arginine or aspartic acid may introduce structural changes that are propagated from the site of the substitution toward the amino terminus, and the resultant conformations may be more stable. In contrast, substitution of glycine by residues with small side chains, such as serine, may not introduce major structural changes. It was surprising that the cleavage of the abnormal molecules by fibroblast collagenase was not markedly delayed; the implications of this observation for understanding the * B. J. Starman, unpublished observation. structure of the molecules is unclear, and further examination is under way.
These arguments imply that the native type I collagen molecule has built-in thermal instabilities. Such instabilities may be biologically advantageous during turnover of connective tissues. The initial registration of the chains at the carboxyl-terminal end dictates the conformation along the length of the molecule, but a pause in folding of the triple helix in the B fragment of the molecule could allow the three chains to associate in a conformation that is thermodynamically more favorable than the native molecule. The propagation of a conformational change from the site of an amino acid substitution toward the amino terminus has been proposed to explain how amino acid substitutions in the triple-helical region of the type I procollagen molecule could result in slower processing of the molecule by N-proteinase. In the model proposed by Vogel et al. (15), it was suggested that the tripeptide shift introduced to accommodate the cysteine substitution altered the putative hairpin structure formed by each of the three proa chains in the region containing the N-proteinase cleavage site and that the altered conformation was responsible for the slower cleavage by Nproteinase. However, we have shown that N-proteinase readily cleaves molecules containing a tripeptide deletion. This finding offers direct evidence that a tripeptide shift that is propagated toward the amino terminus does not alter the Npropeptide cleavage site in such a way as to reduce the rate of cleavage by the enzyme. Therefore, it seems likely that even if a tripeptide shift were introduced at the site of a substitution for glycine, it is not sufficient to explain the decreased rate of cleavage by N-proteinase. Lightfoot et al. (45) have excluded the possibility that the site of a substitution for glycine sequesters enzyme molecules and thereby acts as a competitive inhibitor. Instead other alterations in structure such as a change in chain order, change in helix pitch, or modifications of residues in or near the cleavage site might be invoked to explain the abnormal kinetics of cleavage of those molecules containing substitutions for glycine.