Coordinate Gene Expression of the α3, α4, and α5 Chains of Collagen Type IV EVIDENCE FROM A CANINE MODEL OF X-LINKED NEPHRITIS WITH A COL4A5 GENE MUTATION

Canine X-linked hereditary nephritis is an animal model for human X-linked hereditary nephritis with a premature stop codon in the α5(IV) gene of collagen type IV. We used this model to examine the other α(IV) chains at the mRNA and protein level in the kidney, since in human X-linked hereditary nephritis, the α3(IV) and α4(IV) chains are often absent from the glomerular basement membrane, although both are encoded by autosomal genes. cDNA probes for the α1(IV)-α6(IV) chains were generated from normal dog kidney using the polymerase chain reaction. Sequences were ≥88% identical at the DNA level and ≥92% identical at the protein level to the respective human α(IV) chains. By Northern analysis, transcripts for the α1(IV), α2(IV), and α6(IV) chains were detected at comparable levels in both normal and affected male dog kidney RNA. As previously shown, the transcript for the α5(IV) chain was reduced to ~10% of normal. Unexpectedly, the α3(IV) and α4(IV) transcripts were both decreased ≥77% in affected male dog kidney, suggesting a mechanism coordinating the expression of these three basement membrane components. The NC1 domain of collagen type IV isolated from normal dog glomeruli was positive for the α3(IV), α4(IV), and α5(IV) chains by Western blotting. In contrast, in the NC1 domain isolated from affected dog glomeruli, these three chains were not detectable, except for a trace of α3(IV) dimer. In X-linked hereditary nephritis, the absence of the α3(IV) and α4(IV) chains from glomerular basement membrane may reflect factors acting at the transcriptional and/or translational level in addition to the protein assembly level.

Hereditary nephritis (HN) 1 refers to a group of genetic disorders of collagen type IV with glomerular disease that often progresses to renal failure and sometimes extrarenal disease such as high tone sensorineural hearing loss and anterior len-ticonus (1,2). About 80% of affected families show X-linked inheritance; the remainder are autosomal dominant or recessive (3)(4)(5). Patients present from childhood to early adulthood, usually with hematuria. Most male patients develop terminal renal failure by the end of the third decade, while female patients range from no renal dysfunction to terminal renal failure. The most characteristic morphologic finding in HN is multilaminar splitting of glomerular basement membranes (GBM) as seen by electron microscopy (6). Type IV collagen, the major component of basement membranes, has been linked to the pathogenesis of both X-linked and autosomal HN.
Type IV collagen has recently emerged as a family of triple helical isoforms consisting of six genetically distinct chains, designated ␣1(IV)-␣6(IV) (7). The entire coding sequences for all six human chains have now been determined (8 -14). Their primary structures are similar; each is characterized by a ϳ25residue noncollagenous sequence at the amino terminus, a long collagenous domain of ϳ1400 residues of Gly-X-Y repeats that forms the triple helix together with two other ␣ chains, and a ϳ230-residue noncollagenous (NC1) domain at the carboxyl terminus. The existence of six ␣-chains allows for 56 different kinds (isoforms) of triple helical molecules, which differ in type and stoichiometry of chains. Evidence has been obtained for heterotrimers that have chain compositions of (␣1) 2 ␣2 and (␣3) 2 ␣4 and homotrimers of (␣1) 3 and (␣3) 3 (7). Isoforms containing the ␣5 or ␣6 chains have not yet been described.
The genes that code for the six human ␣(IV) chains have a unique arrangement in that they are located pairwise in a head-to-head fashion on chromosome 13 (COL4A1 and COL4A2) (15), chromosome 2 (COL4A3 and COL4A4) (16), and the X chromosome (COL4A5 and COL4A6) (13,17,18). Over 50 mutations have been found in the COL4A5 gene in families with X-linked HN (Alport syndrome (reviewed in Ref. 19), and mutations have been found in the COL4A3 and COL4A4 genes in patients with autosomal recessive HN (5,20). Although the specific effects of each mutation are not well understood at the molecular level, the end result of all of them is predicted to be an abnormal collagen type IV molecule. This presumably leads to formation of abnormal GBM and progressive renal disease.
How COL4A3, COL4A4, and COL4A5 gene mutations alter the structure of type IV collagen remains undefined. In normal GBM, this collagen is composed of five ␣(IV) chains (␣1-␣5) with ␣1 and ␣2 chains forming one kind of triple helical isoform, ␣3 and ␣4 chains another kind, and ␣5 chains an undefined kind. In the abnormal Alport GBM, there are several lines of evidence suggesting that COL4A5 gene mutations cause defective assembly of the ␣3 chain (reviewed in Ref. 21), and in one study defective assembly of the ␣4 chain was re-ported (22). This conundrum leads to the hypothesis that a mechanism exists for the synthesis of normal GBM that links the assembly, either at the mRNA or protein level, of triple helical molecules containing the ␣5(IV) chain with triple helical molecules containing the ␣3(IV) chain.
The purpose of the present study was to determine the influence of a COL4A5 mutation on the expression of ␣(IV) chains at the mRNA and protein levels. This was accomplished using a unique family of Samoyed dogs with an X-linked form of HN, which closely resembles human X-linked HN at the clinical, genetic, morphologic, and immunohistochemical levels (23)(24)(25)(26)(27) and which is caused by a single base substitution in the COL4A5 gene that results in a premature stop codon (28). The findings indicate the existence of a mechanism coordinating the expression of the ␣3(IV), ␣4(IV), and ␣5(IV) chains.

MATERIALS AND METHODS
Preparation of RNA from Dog Kidney-Normal and affected dogs (paired littermates born to two different carrier females) were sacrificed at 4 months of age, and samples of kidney tissue were snap-frozen in liquid nitrogen. Total RNA was prepared from 1 g of kidney tissue by lysis in guanidinium isothiocyanate followed by centrifugation over 5.7 M CsCl, as described previously (28).
cDNA Synthesis and Amplification by the Polymerase Chain Reaction (PCR)-Synthesis of first strand cDNA was carried out using normal dog RNA according to Sambrook et al. (29). The canine ␣2(IV), ␣3(IV), and ␣6(IV) cDNAs for the NC1 domains were obtained using nested PCR reactions. Pairs of synthetic oligonucleotide primers for the respective cDNAs were constructed based on the sequence of the human ␣2(IV) cDNA (9), the sequence of the bovine ␣3(IV) cDNA (30), and the sequence of the human ␣6(IV) cDNA (14).
The sequences for the ␣2(IV) primers used were as follows.
External Sense The sequences for the ␣3(IV) primers used were as follows.

External
Sense The sequences for the ␣6(IV) primers used were as follows.

External
Sense The first round of PCR was performed using l l of the 20 l of first strand cDNA prepared above as template and 100 ng of both external primers in Perkin-Elmer PCR buffer containing 2.5 mM MgCl 2 to a total volume of 100 l. Samples were denatured at 95°C for 5 min and cooled to 80°C and 0.5 l of Taq polymerase (2.5 units) added. 35 cycles were carried out in a Perkin-Elmer DNA thermal cycler. Each cycle consisted of denaturation at 98°C for 30 s and annealing and extension at 65°C for 2 min. For the second round of PCR, a 2-l aliquot of the first PCR reaction mixture was amplified by the same method, except using the internal primers.
The above strategy was unsuccessful for the canine ␣1(IV) and ␣4(IV) cDNAs, and instead 3Ј-rapid amplification of cDNA ends (3Ј-RACE) (31) was used followed by a nested PCR reaction. In this method, a 19-mer oligonucleotide "adapter" sequence coupled to a (dT) 17 -oligomer was used as the antisense primer for the reverse transcriptase reaction. For the first (external) round of PCR, the sense primer was based on human ␣1(IV) or bovine ␣4(IV) sequence, and an antisense primer was the "adapter." For the second (internal) round of PCR, both primers were based on human ␣1(IV) cDNA (8) or bovine ␣4(IV) cDNA (16) sequences.
The sequences for the ␣1(IV) primers were as follows.

External
Sense The first round of PCR was carried out as described in the 3Ј-rapid amplification of cDNA ends protocol, using the authors' 10 ϫ PCR buffer, except that 100 mM ␤-mercaptoethanol and 67 M EDTA were added. The second round of PCR was carried out in the same buffer for 40 cycles, and the internal primer set. Each cycle consisted of denaturation at 94°C for 1 min, annealing at 55°C for 1.5 min, and extension at 72°C for 2 min.
Southern Blot Analysis-Genomic DNA from Chinese hamster and a human-Chinese hamster somatic cell hybrid with single human 2, 13, or X as the only human chromosomes were purchased from Coriell (Camden, NJ). Genomic DNA samples from a normal human male and female were prepared from peripheral blood as described previously (28). Four g of each sample were cleaved with EcoRI for the ␣1(IV) and ␣3(IV) genes, PstI for the ␣2(IV), ␣4(IV), and ␣6(IV) genes. Southern analysis was performed as described previously (28) using the NC1 domain cDNAs obtained above by PCR as probes for the ␣1(IV)-␣4(IV) and ␣6(IV) genes. The probes were labeled with 32 P by random primer synthesis (32).
Nucleotide Sequencing-The amplified PCR products were purified by Magic PCR Preps DNA Purification System (Promega, Madison, WI) according to the manufacturer's instructions. The sequences of the cDNAs encoding the NC1 domains of the various ␣(IV) chains were obtained by subcloning the PCR products into the TA cloning PCR vector (Invitrogen, San Diego, CA) and then using M13 universal primer and "Sequenase" (U.S. Biochemical Corp.), according to the manufacturer's instructions.
Northern Blot Analysis and Densitometry-Ten-g samples of total RNA prepared from normal and affected male dogs were separated by electrophoresis and blotted as described previously (28). The probes were the PCR products encoding the NC1 domain of the ␣1(IV)-␣4(IV) and ␣6(IV) cDNAs labeled with [ 32 P]dCTP as described above, as well as the PCR product encoding the NC1 domain of the ␣5(IV) cDNA (28) and actin. Each gel was run in duplicate. Hybridization signals from the autoradiographs were quantitated using a Molecular Dynamics computing 300A densitometer. The signals from the ␣(IV) transcripts were quantitated relative to the signal from the actin transcript in order to control for variations in gel loading. All measurements were taken at two different exposure times.
Preparation of NC1 Domain from Dog Kidney-Glomeruli from the kidneys of normal and affected dogs at 4 months of age were isolated by graded sieving as described previously (33). Equal amounts of normal and affected glomeruli (1 mg) were digested with bacterial collagenase for 24 h at 37°C (34). The collagenase-solubilized fraction was recovered after centrifugation and collection of the supernatant. The protein in the supernatant was quantitated using the spectroscopic absorbance at 280 nm.
Western Blotting-One-dimensional SDS-PAGE analysis was performed to confirm equal concentrations of NC1 domain in the normal and affected collagenase-solubilized GBM. Western blotting was performed as described previously (34). The human antibodies used included serum from a patient with Goodpasture syndrome (diluted 1:100) and serum from a patient with HN who received a renal transplant and subsequently developed anti-GBM antibodies (diluted 1:1000). Both of these sera have been previously characterized to be directed against the ␣3(IV) chain of collagen type IV (35). The rabbit antibodies used were directed against the ␣3(IV), ␣4(IV), and ␣5(IV) chains of collagen type IV as described previously (36) (diluted 1:200).
Immunofluorescence-Immunofluorescence was performed as previously reported (26) using the two human sera described above. Sections were pretreated with 6 M urea, pH 3.5. The serum from a patient with Goodpasture syndrome was used at a dilution of 1:10 and the serum from the HN patient was used at a dilution of 1:20.

RESULTS
Experimental Approach-In order to determine the influence of a COL4A5 mutation on the expression of ␣(IV) chains at the mRNA and protein levels, we used the Samoyed dog model of HN in which there is a premature stop codon in the COL4A5 gene and almost no ␣5(IV) mRNA (28). In order to address the expression of the various ␣(IV) chains at the mRNA level, it was necessary to clone the cDNA for the NC1 domain for each chain. Each of these cDNAs was then characterized at two levels: 1) the degree of cross-hybridization to the other ␣(IV) genes using rodent-human hybrid preparations, and 2) the nucleotide and derived amino acid sequence. After establishing that each cDNA was derived from its respective ␣(IV) gene, these cDNAs were used as probes to measure the message levels for the ␣1(IV)-␣6(IV) chains, comparing normal and affected dogs. Based on these results, experiments at the protein level were designed to compare the NC1 domain prepared from normal and affected dog glomeruli. Initially, two patient sera were used, one from a patient with Goodpasture syndrome, the other from a HN patient who had undergone a renal transplant, both known to be directed against the ␣3(IV) chain. These results were then refined by using antibodies specific for each of the ␣3(IV), ␣4(IV), and ␣5(IV) chains.
Chromosome Localization of the Amplified Canine cDNAs-To establish that each of the amplified cDNAs was encoded by its own ␣(IV) genes rather than some other collagen gene, we examined hybridization of these cDNAs to genomic DNA samples by Southern blot analysis. The cDNAs encoding the canine ␣1(IV) and ␣2(IV) NC1 domains hybridized to total genomic DNA from normal human male (Fig. 1, lane 1) and female (lane 2), and to total genomic DNA from Chinese hamster (lane 4). In the genomic DNA sample of the human-Chinese hamster somatic cell hybrid with chromosome 13 as the only human chromosome (lane 3), all the human and hamster bands were present (Fig. 1, a and b). In similar experiments, the cDNAs encoding the canine ␣3(IV) and ␣4(IV) NC1 domains hybridized to human chromosome 2 (Fig. 1, c and d) and the cDNA encoding the canine ␣6(IV) NC1 domain hybridized to human chromosome X (Fig. 1e). No cross-hybridization to other collagen genes was seen with any of the canine ␣(IV) cDNAs.
Comparison  Fig. 2. The sequence for canine ␣1(IV) chain included 6 residues of the collagenous region and 219 residues of the NC1 domain, lacking 10 residues at the C terminus, assuming a length of 229 residues as in the human ␣1(IV) NC1 domain (8).
The sequence for canine ␣2(IV) chain included 202 residues of the NC1 domain, lacking 9 residues at the N terminus and 16 residues at the C terminus, assuming a length of 227 residues as in the human ␣2(IV) NC1 domain (9). The sequence for canine ␣3(IV) chain included 210 residues of the NC1 domain, lacking 11 residues at the N terminus and 12 residues at the C terminus, assuming a length of 233 residues as in the bovine ␣3(IV) NC1 domain (30) or 10 residues at the N terminus, assuming a length of 232 residues as in the human ␣3(IV) NC1 domain (37). The sequence for canine ␣4(IV) chain consisted of 208 residues of the NC1 domain, lacking 11 residues at the N terminus and 12 residues at the C terminus, assuming a length of 231 residues as in the bovine and human ␣4(IV) NC1 domains (16,38). The sequence for canine ␣6(IV) chain included 205 residues of the NC1 domain, lacking 7 residues at the N terminus and 16 residues at the C terminus, assuming a length of 228 residues as in the human ␣6(IV) NC1 domain (14). The percentage of identity between corresponding canine and human sequences is shown in Table I. Of note, the positions of the cysteine residues are conserved between all canine ␣(IV) chains and between each canine and human ␣(IV) chain. The percentage of identity at the amino acid level between the canine ␣1(IV)-␣6(IV) NC1 domains is shown in Table II.
Northern Analysis of Canine ␣1(IV)-␣6(IV) mRNAs-The cDNAs for the NC1 domains for the ␣1(IV)-␣6(IV) chains obtained above were used as probes to determine the level of the messages for their respective genes, comparing normal and affected dog kidney. For the ␣1(IV) mRNA, transcripts were present at 6.8, 6.1, and 5.5 kb in both normal and affected dog kidney, at comparable levels when measured by densitometry (Fig. 3a). For the ␣2(IV) mRNA, a single transcript was present at 6.4 kb in both normal and affected dog kidney, at similar levels when measured by densitometry (Fig. 3b). The mRNAs are comparable in size with the values of 6.7 and 5.4 kb reported for the human ␣1(IV) mRNA (39) and 6.4 kb reported for the ␣2(IV) mRNA (40).
Using the canine ␣3(IV) NC1 domain cDNA as a probe, a single transcript of ϳ8 kb was identified in both normal and affected male dog kidney RNA (Fig. 3c). This size is the same as that reported for human ␣3(IV) mRNA (37). By densitometry, the abundance of the ␣3(IV) transcript in the affected dog was 14 -23% that seen in normal dog. When using the canine ␣4(IV) NC1 domain cDNA as a probe, a single transcript of ϳ10 kb was identified in both normal and affected male dog kidney RNA (Fig. 3d). This size is the same as that reported for human ␣4(IV) mRNA (38). The abundance of the ␣4(IV) transcript in the affected dog was 11-17% that seen in normal dog as determined by densitometry.
As reported previously (28), using the canine ␣5(IV) NC1 domain cDNA, a minor transcript at ϳ8.6 kb and a major transcript at ϳ6.7 kb were identified in both normal and affected male dog kidney RNA (Fig. 3e). As before, the abundance of both ␣5(IV) transcripts was decreased ϳ90% in the affected dog as determined by densitometry. The canine ␣6(IV) NC1   domain cDNA hybridized to transcripts of 7.3 and 6.6 kb in both normal and affected dog kidney (Fig. 3f). This value is comparable with the 6.8 -7.3-kb values reported for the human ␣6(IV) mRNA (13,14). By densitometry, the level of message in the affected dog was 74 -80% of normal.
Western Blotting of Canine NC1 Domain-The serum from the patient with Goodpasture syndrome and the serum from the HN patient who developed anti-GBM antibodies after renal transplantation gave similar results. Both sera stained a single band in the dimer and monomer regions of normal dog NC1 domain, while no staining was seen in the case of affected male dog NC1 domain (Fig. 4a). The control sera did not exhibit any specific binding. When chain-specific antibodies for the ␣3(IV), ␣4(IV), and ␣5(IV) chains were used, each chain was detected in normal dog NC1 domain as single bands in both the dimer and monomer band regions. In the case of the NC1 domain prepared from affected dog kidney, only a very faint band in the dimer region could be detected for the ␣3(IV) chain. No bands for the ␣4(IV) and ␣5(IV) chains were seen (Fig. 4b).
Immunofluorescence Studies-The serum from the patient with Goodpasture syndrome and the serum from the HN patient who developed anti-GBM antibodies after renal transplantation gave similar results. Linear staining of capillary loops of normal dog glomeruli was seen, but there was no staining in glomeruli of affected male dogs (Fig. 5). DISCUSSION The sequences of the human ␣1(IV), ␣2(IV), and ␣6(IV) chains and the bovine ␣3(IV) and ␣4(IV) cDNAs were used to obtain the equivalent canine cDNAs by PCR. The canine PCR products were established to be encoded by their respective canine genes in two ways. First, all PCR products hybridized to the appropriate human chromosome with no cross-hybridization. Second, the sequences of each canine ␣(IV) cDNA showed over 88% identity to the bovine and/or human NC1 domains at both the nucleotide and amino acid levels. The positions of all cysteine residues were conserved both between species and between canine ␣(IV) chains. In general, the amino acid sequences were more similar among the ␣1(IV), ␣3(IV), and ␣5(IV) chains (71-82% identity) and among the ␣2(IV), ␣4(IV), and ␣6(IV) chains (67-75% identity) than with the other ␣(IV) chains (53-66% identity). These results are in keeping with the concept that the ␣(IV) chains fall into two families based on sequence similarities, with the ␣1(IV), ␣3(IV), and ␣5(IV) chains in one family and the ␣2(IV), ␣4(IV), and ␣6(IV) chains in another (18). The sequences for the normal canine ␣3(IV) and ␣4(IV) chains will provide a base line of comparison for the situation where single amino acid substitutions are found in human families with autosomal recessive HN (5,20), in order to ascertain whether these changes are more likely pathogenic or simply polymorphisms.
In the X-linked form of HN, there are several observations that indicate that the ␣3(IV) and ␣4(IV) chains are abnormal even though both are encoded by autosomal genes (reviewed in Ref. 21). It is unclear how a mutation in the ␣5(IV) gene accounts for the abnormalities in these other ␣(IV) chains, but implied is the existence of one or more mechanisms that link the incorporation of the ␣3(IV), ␣4(IV), and ␣5(IV) chains, and that could operate at the protein assembly and/or at the translational/transcriptional level (35). At the protein level, events at both triple helix formation and supramolecular assembly need to be considered. Should the ␣3(IV), ␣4(IV), and ␣5(IV) chains form heterotrimers, then an abnormal ␣5(IV) chain could lead to faulty heterotrimer assembly, with resultant absence of these chains in GBM of patients with X-linked HN. Should the ␣3(IV) and ␣4(IV) chains be in trimers distinct from those containing the ␣5(IV) chain, the latter may be necessary for incorporation of ␣3(IV)and ␣4(IV)-containing trimers during the supramolecular assembly of collagen type IV molecules. An abnormality of ␣5(IV)-containing trimers in X-linked HN could then lead to absence of the ␣3(IV)and ␣4(IV)-containing trimers from GBM. At the translational/transcriptional level, should the expression of the ␣3(IV), ␣4(IV), or ␣5(IV) chains be coordinated, then in X-linked HN the transcription or translation of the ␣3(IV) and ␣4(IV) chains might be impaired secondary to a mutation in the ␣5(IV) gene. The protein assembly and the translational/transcriptional mechanisms need not be mutually exclusive.
In canine X-linked HN, there is a premature stop codon in the gene encoding the ␣5(IV) chain, resulting in ϳ90% reduction in the level of ␣5(IV) mRNA in affected male dogs (28). This model provides a unique opportunity to investigate the effect of a selective absence of the ␣5(IV) chain on the ␣3(IV) and ␣4(IV) chains and relate our findings to the above proposed mechanisms. We found that the levels of mRNA for the ␣3(IV), ␣4(IV), and ␣5(IV) chains were all significantly reduced in affected dog kidney, suggesting that there is coordinated expression of the ␣3(IV), ␣4(IV), and ␣5(IV) genes in the kidney and that in canine X-linked HN, a reduction in the ␣5(IV) mRNA somehow results in reduced message for the ␣3(IV) and ␣4(IV) genes. The concept of co-regulation of the ␣3(IV), ␣4(IV), and ␣5(IV) genes in the kidney is supported by the observation that developing rat glomeruli lack the ␣3(IV), ␣4(IV), and ␣5(IV) chains in GBM until the capillary loop stage, at which time all three chains appear (41). Furthermore, in those basement membranes that contain the ␣3(IV) and ␣4(IV) chains, these two chains are always present in association with the ␣5(IV) chain (41)(42)(43)(44).
In human HN, mutations have been reported that result in a premature stop codon in the ␣3(IV) gene (45) or the ␣5(IV) gene (46 -48); however, message levels for the ␣3(IV), ␣4(IV), and ␣5(IV) chains have not been determined, likely the consequence of limited material from patient biopsies, at the same time underscoring the advantage of an animal model. In other collagen types, premature stop codons have been identified in COL2A1 gene in Stickler syndrome (arthro-ophthalmopathy) (49), but message levels have not been reported. In dystrophic epidermolysis bullosa, a homozygous mutation resulting in a premature stop codon in both alleles of the COL7A1 gene has been reported (50). In this case, there was a marked reduction in the message levels for the COL7A1 gene and no detectable ␣1(VII) chain. Since both collagen type II and type VII are homotrimers, the message levels of other collagen chains on different chromosomes is not relevant. In osteogenesis imperfecta type I, a premature stop codon in the mutant allele results in only about half of the normal amount of collagen type I being made. There is a marked reduction in the amount of message from the mutant COL1A1 allele, with normal amounts of mRNA produced from the normal COL1A1 allele and the two normal COL1A2 alleles (51,52).
The reductions we detected for the ␣3(IV), ␣4(IV), and ␣5(IV) chains at the message level were then investigated at the protein level, first by using spontaneously occurring antibodies in patient sera directed against the ␣3(IV) chain (35,36) to confirm that a similar situation exists in canine X-linked HN as seen in human X-linked HN, namely affected male dogs show absence of the ␣3(IV) chain both by Western blotting and by immunofluorescence. These results correlate with our previous immunofluorescence results in which other sera from patients with Goodpasture syndrome (25) and anti-GBM nephritis (26) failed to stain GBM of affected male dogs. We next refined these studies by performing Western blotting with chain-specific antibodies. The ␣3(IV), ␣4(IV), and ␣5(IV) chains were present in GBM of normal dogs but undetectable in affected male dogs other than a faint dimer band for the ␣3(IV) chain. Thus, a comparable situation exists with canine and human X-linked HN, namely a primary mutation in the ␣5(IV) gene is associated with virtual absence of the ␣3(IV) and ␣4(IV) chains as well as the ␣5(IV) chain.
The mechanisms that might lead to decreased message levels for the ␣3(IV), ␣4(IV), and ␣5(IV) chains in a coordinate fashion are currently not understood and could include factors acting at the gene (transcriptional) level or at the mRNA (translational) level. The ␣1(IV) and ␣2(IV) genes are in a head-to-head arrangement and share a bidirectional promoter located between the two genes and an enhancer element in intron 1 of the ␣1(IV) gene (40,53). Transcription factors and specific DNA motifs that increase transcriptional activity have been identified in both the promoter region (54 -56) and the enhancer region (56 -58). Recently, it has been shown that this common promoter region is not equally bidirectional and is more correctly viewed as two overlapping promoters that share common elements but are unidirectional and gene-specific (55,56). The ␣3(IV) and ␣4(IV) genes are both on chromosome 2 (16), and it is reasonable these two genes share regulatory elements. Should similar motifs be present in the promoter regions of the ␣3(IV), ␣4(IV), and ␣5(IV) genes (currently unknown), there would be a potential molecular basis for coordinate transcriptional control. Still, it would be difficult to explain how, through shared positive transcription factors, a premature stop codon in one ␣(IV) chain could then bring about decreased transcription of two other ␣(IV) chains. Instead, one would need to hypothesize increased levels of some negative or inhibitory transcription factor. For instance, such a factor (protein) might bind to regulatory elements of the ␣3(IV) and ␣4(IV) genes and to the ␣5(IV) mRNA in an equilibrium state. When the ␣5(IV) mRNA is being produced, the protein would tend to bind to this message instead of the ␣3(IV) and ␣4(IV) genes, thereby derepressing the transcription of these two genes. When the ␣5(IV) mRNA is virtually absent, as in the case of canine X-linked HN, the protein stays bound to the ␣3(IV) and ␣4(IV) genes, repressing their transcription.
It has also been shown that DNA demethylation of the 5Јflanking region of the ␣1(IV) gene is associated with increased transcription (59). The relevance of this finding to the other ␣(IV) genes remains to be determined. Recently, a group of proteins known as DEAD box proteins have been recognized that have RNA helicase activity and affect ribosome and spli- ceosome assembly, RNA splicing, and initiation of translation (60). These proteins are believed to regulate RNA synthesis, processing, stability, and degradation. Whether these play a role in canine X-linked HN is unknown.
A control mechanism might also exist at the RNA or translational level. Collagen chains assemble into a triple helix within the cytoplasm in the rough endoplasmic reticulum (61). The mechanism directing the association of specific chains into one molecule are poorly understood. It is reasonable to predict that a process exists to bring together the chains rather than depending on individual chains coming together by chance to form a triple helix. This could be achieved by bringing together the appropriate chains after translation or at an earlier stage by associating the appropriate mRNAs. The latter mechanism would ensure that as the chains are translated, they are already in close association, facilitating the process of triple helix formation. Proteins have been identified that are involved in polypeptide transport across membranes, before these proteins are in their final conformation, and in the correct association of proteins that assemble into oligomers (62). Such proteins have been termed "molecular chaperones" and show extensive homology to each other; examples have been identified in bacteria, yeast, plants, and humans (63).
A molecular chaperone protein might exist to facilitate the assembly of appropriate ␣(IV) chains into a triple helix. This protein could bring together either mRNAs or nascent ␣(IV) chains. Can a "molecular chaperone" mechanism then be used to explain how a premature stop codon in the ␣5(IV) mRNA could bring about decreased message levels of the ␣3(IV) and ␣4(IV) chains? It has been observed that a premature stop codon often leads to decreased message levels rather than to a truncated protein as might be predicted (64,65). In osteogenesis imperfecta type I, the message from the mutant gene with the premature stop codon is present in the nucleus but not in the cytoplasm, implying the mutant mRNA is prohibited from leaving the nucleus (52). This is not felt to be the result of decreased transcription or decreased message stability, and it has been proposed that normal mRNA processing involves a translation-like process that is disturbed in the setting of a premature stop codon (64). Two models have been proposed, termed "translational translocation" and "nuclear scanning." The former supposes that splicing and nuclear transport of mRNA molecules is coupled to translation, such that protein synthesis pulls the RNA molecule out of the nucleus. If there is a stop codon, the pull ceases, and the mRNA is degraded in the nucleus. The nuclear scanning model postulates a mechanism to scan mRNA for a stop codon; if one is found, the mRNA is degraded rather than released from the nucleus for translation. Neither model is predicted to be 100% effective, since low levels of message are detected in cases of a premature stop codon (as is also the case in canine X-linked HN). Therefore, in canine X-linked HN, one can rationalize that the decreased levels of the ␣5(IV) mRNA are the result of premature degradation of this message due to the presence of a premature stop codon. Should the ␣3(IV) and ␣4(IV) mRNAs be associated with the ␣5(IV) mRNA through a molecular chaperone, these two mRNAs might be degraded at the same time. Under normal circumstances, mRNA stability depends on both stabilizing sequences as well as masking destabilizing sequences in the mRNA molecule (65). The premature degradation of the mutant ␣5(IV) mRNA may involve mechanisms that act on specific sequences in that mRNA and that could act on similar sequences in the ␣3(IV) and ␣4(IV) mRNAs, leading to their degradation as well. Additional studies will be necessary to examine these hypotheses.
It is still possible that some of the abnormalities of GBM in canine X-linked HN occur at the level of protein assembly of the ␣(IV) chains into collagen type IV molecules in the setting of an absent ␣5(IV) chain. There is biochemical evidence for heterotrimers that have chain compositions of (␣1) 2 ␣2 and (␣3) 2 ␣4 and homotrimers of (␣1) 3 and (␣3) 3 (7). The interaction of the ␣5(IV) chain with the other ␣(IV) chains has not yet been determined. Whatever the mechanism, a deficiency of the ␣3(IV) and ␣4(IV) chains may contribute as much to renal disease in X-linked HN as a primary mutation in the ␣5(IV) gene. If so, the X-linked and autosomal recessive forms of human HN would share a common mechanism toward producing renal disease involving abnormalities of the ␣3(IV) and/or ␣4(IV) chains.