RNA and type I collagen were analyzed from cultured skin fibroblasts of a Beagle puppy with fractures consistent with type III osteogenesis imperfecta (OI). In a nonisotopic RNAse cleavage assay (NIRCA), the proband's RNA had a unique cleavage pattern in the region of COL1A2 encoding the C-propeptide. DNA sequence analyses identified a mutation in which nucleotides 3991-3994 (“CTAG”) were replaced with “TGTCATTGG.” The first seven bases of the inserted sequence were identical to nucleotides 4002-4008 of the normal canine COL1A2 sequence. The resulting frameshift changed 30 amino acids and introduced a premature stop codon. Reverse-transcription polymerase chain reaction (RT-PCR) with primers flanking the mutation site amplified two complementary DNA (cDNA) fragments for the proband and a single product for the control. Restriction enzyme digestions also were consistent with a heterozygous mutation in the proband. Type I procollagen labeled with [3H]proline was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Increased density of pC-α2(I) suggested comigration with the similarly sized pro-α2(I) derived from the mutant allele. Furthermore, α-chains were overhydroxylated and the ratio of α1(I):α2(I) was 3.2:1, consistent with the presence of α1(I) homotrimers. Analyses of COL1A2 and type I collagen were both consistent with the described heterozygous mutation affecting the pro-α2(I) C-propeptide and confirmed a diagnosis of OI.
MOST CASES of osteogenesis imperfecta (OI) in humans are caused by a mutation in COL1A1 or COL1A2, the two genes encoding type I collagen. The most common type of mutation in OI is the substitution of a glycine in the triple helical portion of the affected α(I)-chain. One human collagen mutation database(1, 2) recently recorded 108 glycine substitutions in α1(I) and 52 glycine substitutions in α2(I). Frameshift mutations have been reported much less frequently, with 20 in COL1A1 and only 1 in COL1A2. Most of the COL1A1 frameshifts involved the insertion or deletion of a single nucleotide within the region coding for the triple helix, resulting in a premature stop codon and a truncated pro-α1(I)-chain. In contrast, the single human COL1A2 frameshift was caused by the deletion of four nucleotides in the C-propeptide with no change in pro-α2(I)-chain length.(3)
In a normal cell, two pro-α1(I)- and one pro-α2(I)-chains associate at their C-propeptide ends and wind into a triple helix with globular N- and C-propeptides. In the extracellular space, the propeptides are cleaved by N- and C-proteinases to produce pN- and pC-collagen molecules. Cleavage by both proteinases produces mature collagen consisting of two α1(I)- and one α2(I)-chains. Because the C-propeptides play an integral role in triple helix formation, mutations altering their sequence can lead to disruptions of type I collagen structure, which are evident with protein analysis.(4–9)
There are few reports of spontaneous mutations in a COL1 gene causing OI in nonhuman species. The first of these was a mouse (oim) with a frameshift mutation in the region of COL1A2 coding for the C-propeptide of pro-α2(I).(10, 11) Homozygous oim/oim and some oim/+ mice have a severe phenotype analogous to human OI type III.(12) A recent report described abnormalities in type I collagen in three dogs with brittle bone disease, and COL1 mutations have since been identified in two of these animals.(13, 14) One of these probands (dog CU3) is the subject of this study.
Dog CU3 was a 12-week-old female Beagle puppy with a 10-week history of multiple fractures affecting almost every long bone.(13) Type I collagen harvested from cultured skin fibroblasts was overhydroxylated.(13) It was hypothesized that a spontaneous COL1A1 or COL1A2 mutation was the cause of the clinical and protein abnormalities observed. The objectives of this study were to determine if a COL1 mutation was present and to characterize the consequences of the mutation on the structure and function of the protein. A frameshift deletion-insertion mutation was identified in the region of COL1A2 complementary DNA(cDNA) coding for the pro-α2(I) C-propeptide. Cultured fibroblasts from the proband produced [α1(I)]3-homotrimeric and [α1(I)]2α2(I)-heterotrimeric type I collagen.
MATERIALS AND METHODS
Nonisotopic RNAse cleavage assay
Dermal fibroblast cultures were established from full-thickness skin biopsy specimens of normal and affected dogs as previously described.(13, 15) Total RNA was harvested from confluent cultures of these fibroblasts as previously described,(16) and reverse transcribed (RT) to cDNA using primer 3′AP (Gibco BRL, Rockville, MD, USA). The translated region of canine COL1A1 cDNA was amplified in two overlapping fragments using polymerase chain reaction (PCR) primer pair S1 (5′-AGCAGACGGAGTTTCTC-3′) and A2888 (5′-ACCGACTTCACCGGGACGTC-3′) and primer pair S2042 (5′-ATTCCAGGGTCTCCCT-3′) and 3′AUAP (Gibco BRL). The translated region of canine COL1A2 cDNA was amplified in two overlapping fragments using PCR primer pair S20 (5′-GGTTCAGCTAAGTTGGAGGTACT-3′) and A2999 (5′CCAATGTTGCCAGGGTAAC3′) and primer pair S2530 (5′-GGACCCTCTGGTATCACTG-3′) and 3′AUAP. Nested PCR was then performed using six overlapping primer pairs for canine COL1A1 cDNA and five overlapping primer pairs for canine COL1A2 cDNA,(14) with each primer containing a 5′ T7-RNA polymerase site. A nonisotopic RNase cleavage assay (NIRCA) was performed using a commercial kit (MutationScreener; Ambion, Inc., Austin, TX, USA). Briefly, the nested PCR products were transcribed to RNA with T7-RNA polymerase, hybridized to form RNA duplexes, cleaved at mismatched sites with RNAse, resolved by electrophoresis on a 2.5% high-resolution agarose gel, and visualized with ethidium bromide staining. In cases in which the NIRCA was positive (as indicated by the presence of RNAse cleavage products), the nested PCR fragment used for that assay was gel-purified and sequenced using automated fluorescent dideoxy chain termination (Cornell University DNA Services Laboratory, Ithaca, NY, USA).
After NIRCA and preliminary sequencing identified a mutation in the region of COL1A2 cDNA amplified by the primer pair T7-S3341 and T7-A4332, a 481-base pair (bp) PCR fragment containing the mutated region was amplified with sense primer S3750 (5′-GGCTCAACCTGAAAACATCC-3′) and an antisense primer A4231 (5′-TGAAACAGACTGGGCCAACG-3′). This (S3750 + A4231) fragment was cut from an agarose gel, purified on a spin column (QIAquick; Qiagen Inc., Chatsworth, CA, USA), blunt-ended with Klenow, and ligated into SmaI-digested pGEM-3Zf(+) (Promega, Madison, WI, USA). The recombinant plasmid was electroporated into DH10β Escherichia coli, which was then plated on LB-agar containing ampicillin. Colony lifts were made onto filter paper, which were then hybridized with a [32P]deoxycytidine triphosphate (dCTP)-labeled PCR fragment. Positive colonies were identified with autoradiography, and plasmid DNA was purified on an anion-exchange modified silica gel resin (Qiagen, Inc.). DNA sequence analysis was used to identify plasmids containing the normal allele product and plasmids containing the mutant allele product.
A 106-bp PCR fragment containing the site of the mutation was amplified with sense primer S3937 (5′-ATCACCTACCACTGCAAGAAC-3′) and antisense primer A4043 (5′-TCGGCAACCAGTTCAACATC-3′). This (S3937 + A4043) fragment was digested with EcoNI (New England BioLabs, Inc., Beverly, MA, USA), Bst4C I (SibEnzyme, Novosibirsk, Russia, via New England BioLabs, Inc.), or both according to manufacturers' directions, and analyzed with ethidium bromide staining after electrophoresis on a 7.5% polyacrylamide gel.
Primary cultures of skin fibroblasts were grown in Dulbecco's modified Eagle's medium (DMEM; Gibco BRL, Rockville, MD, USA) containing 20% (vol/vol) fetal bovine serum (FBS), and postconfluent cultures were used in pulse chase experiments to analyze procollagen processing.(13, 15) As described previously, 2- to 4-day postconfluent cultures were stimulated for 2 days with ascorbate, pulsed for 1 h with [3H]proline, and chased for 0, 1, 2, 4, and 8 h with medium without label.(14) One dish (25 cm2) of cells was collected at each time point and the cell layer and culture medium were processed separately. The cell layer was washed with cold phosphate-buffered saline (PBS), homogenized in 320 μl of homogenizing buffer, and denatured with 2% sodium dodecyl sulfate (SDS) and Laemmli sample buffer at 100°C for 3 minutes. Protease inhibitors and ammonium sulfate were added to the culture medium and the proteins were precipitated with stirring overnight at 4°C. The precipitate was collected by centrifugation, solubilized with extraction buffer, and denatured with 2% SDS and Laemmli sample buffer as mentioned previously. To analyze the effects of hydroxylation of α-chains, selected cultures were preincubated for 1 h with or without α,α′-dipyridyl and labeled with [3H]proline (20 μCi/ml) for 4 h with or without α,α′-dipyridyl at a concentration of 0.3 mM.(15) Cell layer and medium proteins were collected as described previously and were digested with pepsin in 0.5 M acetic acid at pH 2.6 as described previously.(16) For analysis with SDS-polyacrylamide gel electrophoresis (PAGE), all samples were dialyzed for 24 h against fresh Laemmli sample buffer. Aliquots containing 1500 disintegrations per minute (dpm; medium protein) or 2500 dpm (cell layer protein) were reduced with 2% β-mercaptoethanol for 3 minutes at 80°C immediately before separation with 5.0% acrylamide (procollagen) or 5.5% acrylamide (collagen) gels.(14)
In all gels, unlabeled rat tail-tendon collagen (RTTC) was used as a control. The lane containing RTTC was cut from the SDS-PAGE, stained with Coomassie blue, realigned to the remainder of the gel containing the unstained labeled lanes, and dried. Autoradiography was then performed using the reassembled gel. Alignment of the autoradiograph with the dried gel allowed comparison of labeled bands with the bands containing RTTC chains.
NIRCA and mutation analysis
NIRCA was used to screen the COL1A1 and COL1A2 cDNAs of one normal dog and six dogs with OI. In dog CU3, a 12-week-old Beagle with severe OI, there was cleavage of the RNA duplex spanning nucleotides 3341-4332 of COL1A2 (Fig. 1). The cleavage products were calculated to be 670 bp and 356 bp. After the T7 extensions on the primers were taken into account, a mismatch in the RNA duplex was predicted at either nucleotide 3677 or 3991.
COL1A2 cDNA from the proband was amplified from nucleotide 3750-4231, and the resulting PCR products were cloned into pGEM-3Zf(+). Sequencing of the plasmid inserts confirmed the presence of two different cloned PCR products: one with the normal COL1A2 cDNA sequence and one with a frameshift mutation (Fig. 2). In the mutant allele, nucleotides 3991-3994 (CTGA) were replaced with the sequence TGTCATTGG. This deletion of four nucleotides and insertion of nine nucleotides resulted in a net gain of five nucleotides, causing a shift in the translational reading frame. Translation of the mutant allele predicted an abnormal amino acid sequence for 30 codons downstream of the mutation site, where a premature stop codon truncated the C-propeptide of pro-α2(I) by 51 amino acids (Fig. 2).
Restriction enzyme digests were used to further show the mutation. PCR amplification of cDNA from nucleotide 3937 to 4043 produced two bands in dog CU3: a 107-bp product from the normal allele and a 112-bp product from the mutant allele (Fig. 3). Restriction enzyme digestion of these fragments with EcoN I, which cuts the normal (but not the mutant) allele product at one site, and/or Bst4C I, which cuts the mutant (but not the normal) allele product at one site, was consistent with dog CU3 being heterozygous for the COL1A2 frameshift mutation identified by sequence analysis (Fig. 3).
In the 0- to 8-h pulse chase experiments, processing of procollagen to collagen proceeded at a normal rate in cultures of proband fibroblasts.(13) However, the density of pC-α2(I) was increased compared with a normal control (Fig. 4). The electrophoretic mobility of pepsin resistant α(I)-chains produced by dog CU3 cells was slowed (Fig. 5A). When hydroxylation was prevented by adding α,α′-dipyridyl to the fibroblast culture media, the pepsin-resistant α(I)-chains migrated normally, suggesting that overhydroxylation was the cause of the slower mobility (Fig. 5B). As measured by densitometry, the ratio of α1(I):α2(I) from dog CU3 was 3.2:1 instead of the normal 2:1 (Fig. 5A).
A unique heterozygous COLA2 frameshift mutation was identified in the cDNA of a canine variant of OI. The mutation involved deletion of four nucleotides (3991-3994) and their replacement with an insertion of nine nucleotides, the first seven of which were identical to a sequence 11 bp downstream of the deletion site. The gain of five nucleotides resulted in a frameshift in the region of COL1A2 coding for the C-propeptide of the pro-α2-chain. The new reading frame predicted 30 novel amino acids before introducing a stop codon that would truncate the mutant pro-α2(I)-chain by 51 amino acids.
Human and canine pro-α2(I)-amino acid sequences normally have 100% identity for the 30 amino acids that were altered in the mutant allele product of the canine proband.(16) Such a high degree of conservation indicates that these specific residues are important for normal function of the pro-α2(I)-C-propeptide. Alvares et al.(17) concluded that interactions between the pro-α1(I)- and pro-α2(I)-C-propeptides were determined by amino acid sequence and the conformations of interacting domains. Premature truncation of pro-α2(I) in dog CU3 eliminated residues that impart important conformational characteristics to the C-propeptide. These losses included a proline residue that is unmatched in the corresponding pro-α1(I)-sequence,(17) and the two most 3′ cysteines, which are normally involved in the formation of separate intrachain disulfide bonds.(18) Using site-directed mutagenesis of cysteine residues in the C-propeptide of pro-α2(I), Doyle et al.(18) found that triple helix formation could occur if one intrachain disulfide bond was eliminated, but not if both were absent.
The frameshift mutation described here created two new cysteine codons in the pro-α2(I)-C-propeptide. Thus, the truncated pro-α2(I) from dog CU3 has a normal number of cysteine residues and still may be able to form intrachain disulfide bonds. However, the altered location of the new cysteine residues may compromise the ability of mutant pro-α2(I) to achieve the conformation required for triple helix formation.
In the mutant COL1A2 allele of dog CU3, the three cysteine residues involved in interchain disulfide bond formation were preserved, lending further support to findings by others that the cysteines involved in interchain disulfide bond formation are not sufficient to permit winding of the triple helix.(17–19)
Because of the extensive alterations in the C-propeptide of the mutant allele product, it is unlikely that the truncated canine pro-α2(I) was successfully incorporated into a triple helix. Alvares et al.(17) showed that deletion of the terminal 36 amino acids of C-pro-α2(I) completely prevented binding with C-pro-α1(I). Indeed, truncation of only the last 10 residues of pro-α2(I) prevented its incorporation into heterotrimeric procollagen molecules.(20)
When there are insufficient pro-α2(I)-chains available to associate with pro-α1(I)-chains, the pro-α1(I)-chains will interact with each other and form triple helical homotrimers.(3, 17, 21) The increased ratio of α1(I):α2(I) seen after SDS-PAGE analysis of pepsin-resistant collagen synthesized by proband fibroblasts is consistent with the production of both homotrimers of α1(I) and heterotrimers of [α1(I)]2α2(I). The α1(I) homotrimers were found in the two other examples of OI caused by a frameshift mutation affecting the pro-α2(I)-C-propeptide. The oim/oim mouse, in which the 48 C-terminal amino acids of pro-α2(I) are altered and one additional amino acid is added, has no α2(I)-chains in the extracellular matrix and produces only α1(I)-homotrimers.(10) Heterozygous oim/+ mice have both heterotrimeric and α1(I)-homotrimeric type I collagen in the same fibrils.(22) The oim/+ phenotype ranges from clinically normal to severely affected mice that are clinically indistinguishable from oim/oim individuals; the latter is comparable with the severe phenotype seen in this heterozygous Beagle dog.(12) The α1(I)-homotrimers were the only form of type I collagen in the matrix of a human OI variant in which the last 33 amino acids in pro-α2(I)-chains were changed by a homozygous 4-bp frameshift deletion near the 3′ end of COL1A2.(3) This individual had moderately severe OI, with multiple fractures and severely deformed limbs by 2 years of age.(23) The α1(I)-homotrimers were post-translationally overmodified, as seen in the canine proband of this report.(3)
An unusual finding in the electrophoretic migration of proteins from the canine proband was an increase in band density in the region of pC-α2(I), which was seen after SDS-PAGE of both the cell layer (data not shown) and the culture medium proteins (Fig. 4). This may be caused by delayed C-terminal processing of overhydroxylated procollagen molecules. Another explanation is that there may be comigration of truncated pro-α2(I) from the mutant allele with normal pC-α2(I), because these polypeptides differ in size by only 32 Da. Evidence against this argument is found in the Mov-13 mouse, in which pro-α1(I) is not synthesized, and pro-α2(I)-chains are degraded intracellularly, suggesting that stability of pro-α2(I) requires association with pro-α1(I).(24) However, Uitto et al.(25) showed that, when both pro-α1(I) and pro-α2(I) were present (as in dog CU3), nontriple helical, reducible, pro-γ-chains were secreted by tendon fibroblasts when cis-hydroxyproline incorporation was used to prevent triple helix formation. Because truncation of pro-α2(I) eliminated residues required to form a triple helical heterotrimer but preserved the C-propeptide cysteines involved in interchain disulfide bonds, secretion of nontriple helical disulfide-bonded heterotrimers containing truncated pro-α2(I) could explain the band density change in Fig. 4. Further support for this explanation is the absence of a band of the expected size for mutant pC-α2(I). Generation of this intermediate form could not occur if mutant pro-α2(I) is not incorporated into trimeric procollagen, because N-proteinase requires a triple helical substrate.(26) Further studies, such as cyanogen bromide digestion and amino acid sequencing, should characterize more definitively the nature of the increased density of the pC-α2(I)-band.
The NIRCA has successfully identified mutations in a number of different human genetic diseases.(27, 28) In addition to the canine COL1A2 frameshift mutation characterized in this report, the authors also have used NIRCA in the identification of a point mutation (Gly208Ala) in COL1A1 cDNA from a different case of canine OI (dog OI-CU1 in Fig. 1).(14) The advantages of the NIRCA over other mutation detection systems such as single-stranded conformational polymorphism include the ability to use ethidium bromide staining instead of radioactivity, the use of low-power electrophoresis, and the ability to detect mutations in fragments up to 1 kilobase (kb) in size.
The NIRCA did not detect mutations in COL1 cDNA from four other dogs in which clinical presentation and type I collagen analysis were consistent with OI. It is possible that the condition in these animals is analogous to type V OI in humans, which does not appear to be associated with COL1 mutations.(29) Another possibility is the presence of a heterozygous mutation that prevents expression of the mutant allele. The ability of the NIRCA to detect a mutation is dependent on the location of the mutation relative to the annealing sites of the primers used, as well as the surrounding sequence. Additional primer pairs designed to overlap those used in the current study may allow detection of mutations in the four affected dogs. Automated sequencing of genomic DNA is a more direct method for detecting mutations, but it can be costly for large genes like COL1A1 and COL1A2 and false negative results may still occur. It is likely that a combination of mutation-detection strategies needs to be used before absence of a COL1 mutation can be assured in an individual with the clinical phenotype of OI.
The frameshift mutation discovered in this dog is unusual in that it involved the deletion of four nucleotides and their replacement with a nine-nucleotide sequence, the first seven bases of which were identical to the COL1A2 cDNA sequence found 11 bp downstream of the deletion site. This seven-nucleotide sequence is not repeated anywhere else in either the normal canine COL1A2 or COL1A1 cDNA sequences.(16, 30) It is tempting to speculate that the generation of this mutation involved unequal crossover because of aberrant alignment between the two sequential adenosines 5′ to the deletion site on one allele and the five adenosines 3′ to the deletion site on the other allele. However, the actual mechanism must be more complex to explain the origin of the final two guanines (nucleotides 8 and 9) of the inserted sequence.
This article describes the first COL1A2 mutation identified in canine OI. A canine COL1A1 mutation also has been identified,(14) and type I collagen abnormalities have been identified thus far in five additional dogs with clinical signs of OI.(13, 30) It is beneficial to study OI in multiple species, because identification of interspecies similarities in the disease solidifies our understanding of the underlying pathophysiology, while recognition of differences suggests areas where therapeutic manipulation of the disease may be possible. Dogs and cats are increasingly being used for comparative research of diseases shared with humans, especially for therapeutic studies.(31, 32) Biomechanically, the dog is a better size-matched model for the forces applied to the human skeleton (especially in childhood, when OI can be the most devastating) than is the mouse. Furthermore, the dog provides a heterogeneous population in which a variety of spontaneous COL1 mutations may occur, as is the case in the human condition. A canine model of OI offers a new perspective for the study of the effects of spontaneous COL1 mutations on the structure, organization, and function of type I collagen in both affected and unaffected tissues and a resource for investigating the effects of therapeutic modalities on OI.
The authors thank Dr. Da-Nian Gu and Ms. Barbara Hover for their contributions to this study. This work was supported by training grant T32 RR07059-01 (B.G.C.) and research grants AR08428 (B.G.C.) and AR20793 (R.R.M.) from the National Institute of Health and grants from The Iams Company, Lewisburg, OH.