Regulation of the Paired Type IV Collagen GenesCOL4A5 and COL4A6

Tissue-specific expression patterns of the paired type IV collagen genes COL4A5 and COL4A6form the basis for organ involvement in X-linked Alport syndrome, a disorder in which these genes are mutated. We investigated the proximal promoter region of COL4A5 and COL4A6 using glomerular visceral epithelial cells, in which COL4A5 alone is transcribed; keratinocytes, in which the genes are co-transcribed; and additional model cell lines. By RNase protection assays, the intergenic region is 292 base pairs. Transcription start sites for two 5′ splice variants of COL4A6 are 1 kilobase apart. Transient transfections with reporter gene constructs revealed that the minimal promoters for COL4A5 and COL4A6 are within 100 base pairs of their respective transcription start sites and are functionally distinct. In further transfection, gel shift and footprinting assays, we defined a bidirectional positive regulatory element, which functions in several cell types, but not in glomerular visceral epithelial cells selectively transcribing COL4A5. The existence of separate promoters for COL4A5 andCOL4A6 permits fine control over their expression. Activation through the bidirectional element can bring about co-expression of the genes, exploiting their paired arrangement. Features of the proximal promoter region frame its roles in a hierarchy regulating type IV collagen gene expression.

Type IV collagen is a major constituent of basement membranes (1). Its lattice network is assembled from six ␣(IV) chains, ␣1(IV)-␣6(IV), encoded by the genes COL4A1 through COL4A6 (2). In humans, these genes are arranged as head-to-head pairs COL4A1-COL4A2, COL4A3-COL4A4, and COL4A5-COL4A6, transcribed from opposite strands of chromosomes 13, 2, and X respectively. The functional differentiation of the individual ␣(IV) chains is evidenced in genetic diseases. Mutations in COL4A5 give rise to Xlinked Alport syndrome, a disorder of renal glomerular basement membrane (GBM), 1 variably associated with hearing loss and ocular abnormalities (3). Deletions involving COL4A5 and COL4A6 occur in Alport syndrome associated with diffuse leiomyomatosis (4). Mutations in COL4A3 or COL4A4 have been implicated in autosomal recessive Alport syndrome (5) and benign familial hematuria (6).
The mechanisms accounting for the specialized distribution of the minor ␣(IV) collagen chains are poorly understood. Expression of the ␣3(IV), ␣4(IV), and ␣5(IV) chains is activated in the developing renal glomerulus, as basement membranes of distinct ultrastructure and type IV collagen subunit composition are elaborated by endothelial and visceral epithelial cells (10,12,(15)(16)(17)(18)(19). Developmental regulation of minor chain expression has been characterized in other tissues as well (20,21). Special considerations relating to expression of the minor chains arise in Alport syndrome. In this condition, pathogenic mutations in one of the minor chains generally lead to loss of the remaining chains (11,12,22). This effect can be achieved by global downregulation at the mRNA level, as demonstrated in a recent study of a canine model with a COL4A5 mutation (23), or by post-transcriptional mechanisms (19,24).
It is likely that the pairing of the type IV collagen genes facilitates their coordinated regulation. The regulation of COL4A1 and COL4A2, separated by 130 bp, is served by a shared bidirectional promoter, which is subject to the effects of remote cis-acting elements (25)(26)(27)(28)(29)(30)(31)(32)(33)(34). Transcriptional mechanisms regulating the minor chains have not been elucidated. In contrast to the COL4A1-COL4A2 promoter, which applies to obligately co-expressed isoforms, a framework for the minor chains must account for coordinate expression in several tissues, as well as the discordant expression pattern of the ␣5(IV) and ␣6(IV) chains in GBM.
Using established models for keratinocytes, in which ␣5(IV) and ␣6(IV) are co-expressed, and renal glomerular visceral epithelial (GVE) cells, in which ␣5(IV) is expressed selectively, we provide evidence that tissue-specific expression patterns of type IV collagen arise at the transcriptional level. By detailed investigation of the proximal promoter region, we arrive at an initial understanding of the mechanisms regulating COL4A5 and COL4A6 transcription. We demonstrate the existence of two promoters for COL4A6, which are functionally separable from the COL4A5 promoter, and report the identification of a positive regulatory element, which can effect coordinated regulation. Finally, on the basis of findings in these and other cell lines, we frame potential roles for the proximal promoter region, in a hierarchy required to explain cell-specific transcription of this gene pair.
Reporter Gene Constructs and Transient Transfections-Parent constructs p␣5E1, p␣6E1a, and p␣6E1b contain genomic restriction fragments upstream from a luciferase reporter gene, in the vector pGL3-Basic (Promega), as shown in Fig. 5. Deletion constructs were produced by excision of appropriate restriction fragments, with subsequent bluntending if necessary, and religation (Fig. 5).
For transient transfections, cell monolayers grown to 60 -90% confluence on 60-mm dishes were incubated for 6 h with 2.5 g of the luciferase test construct and 1.5 g of the normalizing vector pSV-␤ galactosidase (Promega), in 1.5 ml of Opti-MemI medium (Life Technologies) containing 10 g/ml Polybrene (American Bioanalytical). Cells were exposed to 30% dimethyl sulfoxide for 3 min, restored to complete medium and harvested 32 to 34 h thereafter in Reporter Lysis Buffer (Promega). Luciferase activities were determined using a Monolight 2001 luminometer (Analytical Luminescence Laboratory Inc.) with the Luciferase Assay System (Promega). They were normalized for transfection efficiency using ␤-galactosidase activities, determined by the ␤-Galactosidase Enzyme Assay System (Promega). Luciferase activities are expressed relative to paired promoterless controls, and represent mean Ϯ S.E. for at least three transfections.
Nuclear Extracts-Nuclear extracts were prepared by standard protocols, with minor modifications (39). Ten 175-cm 2 flasks with nearly confluent monolayers were used for each preparation, and all steps were carried out at 4°C. Cells were washed twice and scraped into phosphate-buffered saline, collected, centrifuged at 1000 ϫ g for 15 min, consolidated, and centrifuged again, yielding packed cell volumes of ϳ0.5 ml. Following resuspension in 2.5 ml of Buffer A (10 mM KCl, 10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride), cells were placed on ice for 10 min, then pelleted. They were resuspended in 1 ml of Buffer A, Dounce homogenized (A type pestle), then centrifuged at 1000 ϫ g for 15 min. The supernatant was removed, and the pellet was centrifuged at 30,000 ϫ g for an additional 10 min, yielding a crude nuclear fraction. This was resuspended in 0.5 ml of Buffer C (420 mM NaCl, 20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 0.2 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 20% glycerol), Dounce homogenized (B type pestle), then extracted in Buffer C for 30 min. Following centrifugation at 30,000 ϫ g for 30 min, the supernatant was collected, and dialyzed overnight against a buffer containing 100 mM KCl, 10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, and 20% glycerol. Extracts were stored in aliquots at Ϫ80°C, and protein concentrations were estimated using the Bio-Rad protein assay kit.
Gel Shift and Chemical Footprinting Assays-Probes for gel shift and chemical footprinting assays were prepared from 73-bp MslI-FokI fragments (nt 728 -800, GenBank TM /EBI accession D28116), which were isolated from p␣6E1a or mutant p␣6E1a(MT2) by multiple-step restriction digest and gel purification. Fragments were labeled at one end by fill-in of a CCCC overhang, using Klenow fragment (New England BioLabs) in the presence of [␣-32 P]dGTP, followed by chase with excess cold dGTP. Labeled fragments were isolated from 9% non-denaturing polyacrylamide gels. Unlabeled fragments were used in competition assays.
For gel shift assays, 2,000 cpm of labeled probe was added to 8 g of nuclear extract, 5 g of poly(dI⅐dC), and 3 g of bovine serum albumin in 20 l of gel shift buffer containing 50 mM KCl, 20 mM HEPES, pH 7.9, 5 mM MgCl 2 , 1 mM dithiothreitol, and 20% glycerol. Mixtures were incubated at 4°C for 30 min, then electrophoresed on 8% non-denaturing polyacrylamide gels in 1ϫ TGE buffer (25 mM Tris, 190 mM glycine, 1 mM EDTA).
We found that coupled gel shift/footprinting assays were required to delineate DNA elements interacting with proteins in nuclear extracts, because of the low occupancy of these sites. We used a previously described copper phenanthroline footprinting technique, with minor modifications (40). Specialized reagents were purchased from Aldrich, and all steps were carried out at 4°C. Scaled-up gel shift assays were carried out using 100,000 cpm of labeled probe, 40 g of nuclear extract from SCC-25 cells, 15 g of poly(dI⅐dC), and 15 g of bovine serum albumin in 100 l of gel shift buffer. After electrophoresis, gels were immersed in 400 ml of 10 mM Tris-HCl, pH 8.0, followed by addition of 40 ml of DNA cleavage reagent (36 ml of distilled water, added to 2 ml of 9 mM CuSO 4, and 2 ml of 40 mM 1,10-phenanthroline monohydrate in ethanol). The cleavage reaction was initiated by addition of 40 ml of 58 mM 3-mercaptopropionic acid, and terminated at 3-90 min by addition of 40 ml of 30 mM 2,9-dimethyl-1,10-phenanthroline. Bands corresponding to unbound DNA and the major bandshift S 1 (see "Results") were visualized by autoradiography, then excised. DNA was eluted overnight at 37°C in 0.5 M ammonium acetate, 10 mM Mg acetate, 1 mM EDTA, 0.1% SDS and isolated by sequential phenol/chloroform extraction, chloroform extraction, and ethanol precipitation, with 10 g of glycogen as a carrier. Pellets were rinsed twice in 85% ethanol and electrophoresed on 10% sequencing gels. Digestion patterns were visualized by autoradiography and analyzed by scanning densitometry.

Characterization of Model Systems-COL4A5 and COL4A6
expression was analyzed in several cell lines by RNase protection assay. We report results for GVE cells, in which COL4A5 alone is expressed; SCC-25 keratinocytes, in which both genes are expressed; and SK-N-SH neuroblastoma cells, in which neither gene is expressed (Fig. 1). To ascertain whether transcription accounts for differences among these mRNA expression patterns, we amplified nascent COL4A5 and COL4A6 transcripts, using intron-directed primers. In this surrogate for the nuclear runoff assay (37,41), detection of nascent transcripts corresponded to steady-state mRNA levels (Fig. 2), indicating that transcription, rather than mRNA processing or degradation, is the major determinant of steady-state mRNA levels. The mRNA expression profiles in GVE and SCC-25 cells provide evidence that the type IV collagen composition of corresponding renal glomerular and epidermal basement membranes arises at the transcriptional level.
Structure of the Proximal Promoter Region-Transcription start sites were mapped by RNase protection assays, using RNA from SCC-25 and GVE cells (Fig. 3). Start sites for transcripts of COL4A5 and 5Ј splice variants A and B of COL4A6 are upstream relative to earlier determinations by cDNA sequencing (4,42). The intergenic region is 292 bp. By inspection, the 5Ј-flanking regions of exons E1 of COL4A5 and E1a of COL4A6 lack classical TATA boxes but lie downstream from AT-rich regions, which could confer specificity to sites of transcriptional initiation (Fig. 4). Additional features of the intergenic region include a tandem array of CCAAT boxes and a central CTC box, a functional motif, which is conserved in the intergenic region separating the COL4A1 and COL4A2 genes (28,31). Transcript B of COL4A6 is associated with at least two closely situated start sites. The 5Ј-flanking region of exon E1b also lacks a classical TATA box, but is associated at the upstream start site with a sequence that is identical to a tran-scriptional initiator (Inr) at 7 of 8 nucleotides (43).
We undertook a detailed analysis of the proximal promoter region, as this is where cell-specific regulatory elements are expected to exert their effects. In transient transfections with luciferase reporter gene constructs, we sought to characterize the minimal COL4A5 and COL4A6 promoters and to identify important proximal regulatory elements.
A parent construct p␣5E1 reporting in the COL4A5 orientation gave luciferase activity 17-and 47-fold higher than the promoterless vector pGL3-Basic, in SCC-25 and SK-N-SH cells, respectively (Fig. 5). This finding localizes the minimal COL4A5 promoter to the intergenic region. Inasmuch as the two cell types differ in COL4A5 expression ( Figs. 1 and 2), overlying mechanisms are implicated in regulation of the endogenous promoter. Transfections with construct p␣6E1a gave luciferase activity 17-and 40-fold higher than the promoterless control, in SCC-25 and SK-N-SH cells (Fig. 5). These activities place the minimal COL4A6 promoter in the intergenic region as well and implicate additional mechanisms in repressing COL4A6, in SK-N-SH cells. In GVE cells, luciferase activity was only 5-to 6-fold over promoterless controls, in both the COL4A5 and COL4A6 orientations (Fig. 5). The lack of orientation specificity for these constructs suggests that distal elements are involved in the selective activation of COL4A5 (see "Discussion").
Construct p␣6E1b containing the 5Ј-flanking region of E1b was used to assess whether transcription of COL4A6 is directed by two promoters. Reporter gene activity was 7-to 9-fold higher than the promotorless control in the three cell lines (Fig. 5), indicating the presence of a functional promoter, which is distinct from that upstream of E1a and thus a site of differential regulation for Transcript B. Given comparable promoter activities among the cell lines, additional regulatory mechanisms must apply, to explain differences in steady-state levels of Transcript B.
We refined localization of the minimal promoters for COL4A5 and COL4A6 by deletion analysis. In SCC-25 and SK-N-SH cells, deletion of nt 755-876 (GenBank TM /EBI accession D28116), in construct p␣5E1, caused 82 and 87% reductions in activity, respectively (Fig. 5), indicating the presence of a positive regulatory element in this region. Further deletion to nt 625 did not alter reporter gene activity appreciably. Construct p␣5E1(625), which contains 99 base pairs upstream from the transcription start site, gave luciferase activity significantly higher than promoterless controls, suggesting that the minimal promoter resides within this fragment. In GVE cells, serial deletions of p␣5E1 to nt 625 had no significant effects, with activity for p␣5E1(625) remaining significantly higher than control. These results consign the minimal promoter to the immediate 5Ј-flanking region, as in the other cell lines, but provide evidence against the function of a positive regulatory element. Deletion of the tandemly arranged CCAAT boxes or of the CTC box (see Fig. 4) had no significant effects on reporter gene activity.
By comparison to serial deletions of p␣5E1, those of p␣6E1a were associated with only modest (Ͻ55%) reductions in luciferase activity (Fig. 5). Plasmid p␣6E1a (755), containing 67 bp upstream of the transcription start site, gave luciferase activity significantly above promoterless controls. These data indicate the presence of a minimal promoter immediately upstream of exon E1a, which is thus functionally separable from the COL4A5 promoter, as it is from a second downstream COL4A6 promoter.
Taken together, our findings support a configuration of nonoverlapping minimal promoters for COL4A5 and COL4A6. In SCC-25 and SK-N-SH cells, the positive regulatory effects of an intergenic, potentially bidirectional, element are evident. This configuration has valuable attributes for regulation of the paired genes, although it is by itself insufficient to explain cell-specific expression patterns (see "Discussion").
A Positive Regulatory Element in the Intergenic Region-To investigate further the positive regulatory element between nt 755 and 876, we utilized constructs with directed substitutions in this region, in SCC-25 cells (Fig. 6). Four sites MT1 through MT4 were chosen to span the region, and a fifth, containing an AP-1 consensus site (MT5), was targeted directly. The MT2 mutation had a clear bidirectional effect, reducing luciferase activity 82% in the COL4A5 orientation, and 68% in the COL4A6 orientation, relative to the parent constructs. Among the remaining constructs, we observed appreciable changes relative to the parent constructs with MT4 only, which caused a 2-fold increase in luciferase activity, in the COL4A6 orientation. This mutation was associated with the recruitment of additional shifted bands in gel shift assays (data not shown), suggesting the introduction of binding sites for spurious transcriptional activation. Substitution of the AP-1 site had no significant effects, suggesting that this element is not required for basal transcription.
To investigate the interactions of nuclear factors with the bidirectional element, gel shift assays were carried out, using labeled probes containing the region of the MT2 mutation (nt 728 -800). Several band shifts were observed following incubation of wild-type probes with nuclear extracts from SCC-25 or GVE cells but not following incubation of MT2 mutant probes (Fig. 7). The banding patterns were complex and differed between the two cell types, suggesting multiple interactions between the DNA fragments and cell-specific nuclear factors. Most of the band shifts were abolished by competition with excess unlabeled wild-type fragment, but not with MT2 mutant fragments, supporting the specificity of the interactions, and the importance of the region affected by the MT2 mutation.
To delineate the boundaries of the bidirectional enhancer, a coupled gel shift/chemical footprinting assay was carried out, using the predominant shifted band S 1 (Fig. 7). The shifted probe was protected from cleavage at the site 5Ј-TAGC-3Ј (Fig.  8), a site of preferential cleavage for copper phenanthroline (44). Analysis by scanning densitometry indicated strong protection of the region 5Ј-TAGCGGATGGATCTCA-3Ј, with weaker protection of the adjacent region 5Ј-GAGGGG-3Ј. This site lacks known regulatory elements, by database search (TRANSFAC, release 3.5; Ref. 45). The MT2 mutation likely disrupts binding at the site secondarily. FIG. 4. Structure of the COL4A5 and COL4A6 promoter region. Transcription start sites, sequence bearing identity to a transcriptional initiator (Inr) at 7 of 8 nucleotides, and CCAAT, AP-1, Sp-1, and CTC consensus regulatory elements are indicated (see text).

DISCUSSION
Several divergently transcribed gene pairs have been described in higher eukaryotes. Within the type IV collagen gene family itself, pairing has been conserved through an evolutionary pattern of duplications, across mammalian species (46). This implies biological advantage, deriving perhaps from stoichiometric requirements for related gene products. Pairing has been conserved within other gene families as well, including the histone genes, which are arranged within duplicated clusters as oppositely transcribed pairs, which share common promoter elements (47,48). Additional examples such as that afforded by the functionally related genes GPAT-AIRC (49), whose products catalyze steps in the purine nucleotide biosynthetic pathway, and TAP1-LMP2 (50), whose products are central in antigen presentation, reinforce the notion that paired gene arrangements have been conserved in response to evolutionary selection pressures. For the type IV collagen genes specifically, regulatory schemes operating within this framework account for a highly specific subunit distribution, among tissues that vary considerably in their basement membrane composition and turnover requirements.
We have demarcated the promoter region of the genes COL4A5 and COL4A6 and defined a novel positive regulatory element situated in the intergenic region. Our results are consistent with a scheme in which COL4A5 and COL4A6 are associated with functionally distinct promoters, which can be regulated separately or coordinately. The existence of separate promoters for Transcripts A and B of COL4A6 in particular affords a basis for their differential expression in several tissues, including kidney, lung, skin, and placenta (42). If the chief benefit of type IV collagen gene pairing is to facilitate coordinated transcription, then the overlapping COL4A1-COL4A2 promoter, and the distinct COL4A5 and COL4A6 promoters, represent minor but important variations permitting finely regulated synthesis of their respective gene products. The positive element defined here may confer the capacity for co-expression of COL4A5 and COL4A6 genes in keratinocytes and other cell types (12,13).
The complexity of DNA binding in the region of the bidirectional element is evidenced by the appearance of multiple band shifts in our gel shift assays, each reflecting one of multiple DNA-protein interactions. Abolition of these band shifts by the MT2 mutation, with corresponding functional effects in SCC-25 cells, suggests that binding of this cis-acting element is a condition for bidirectional transcription. If the site is viewed as a nidus for the binding of transcription factor complexes, then differential effects in SCC-25 and GVE cells can reflect the activity of a transcriptional activator in SCC-25 cells, and/or the activity of a transcriptional repressor in GVE cells. Inasmuch as the footprints in our assays do not correspond to consensus regulatory elements, we cannot posit roles for specific transcription factors in binding the bidirectional element. Moreover, the activity of this element in nonexpressing SK-N-SH cells implicates higher-order regulation by sequencespecific transcriptional repressors operating through distal elements, or by general, i.e. chromatin-mediated, mechanisms of transcriptional repression (51).
We have also demonstrated selective transcription of COL4A5 in GVE cells, corresponding to the discordant expression pattern of the ␣5(IV) and ␣6(IV) chains in the renal GBM. On the basis of our promoter studies, distal elements are required to account for selective activation of the COL4A5 promoter. Nonetheless, features of the proximal promoter region may afford a partial explanation. In GVE cells, inactivity of the bidirectional element is evidenced functionally in transient transfections, and is corroborated by differential patterning in gel shift assays. This inactivity could be permissive for selective COL4A5 activation, ensuring absence of the ␣6(IV) chain in GBM, by uncoupling transcription of the two genes. In principle, selective transcription could also be accomplished through asymmetric effects of cis-acting elements on a shared promoter; however, this scheme, described for other co-expressed gene pairs including COL4A1-COL4A2 (31, 32), may not afford sufficient selectivity for COL4A5 and COL4A6.
A working model for the proximal promoter region is presented in Fig. 9. Its major features are separability of the minimal promoters for the two genes, and provision for the bidirectional element, which can act as a shared near-upstream positive regulator. In this model, as applied to GVE cells, inactivity of the bidirectional element is an important feature of selective COL4A5 transcription, which is mediated by distal enhancers (panel A). In other cell types, activity of the bidirectional element is determined by the activation state of the locus as a whole, and the presence of specific MT2-binding transcription factors (panels B, C). Distinctions between SCC-25 and SK-N-SH cells reside in the mechanisms by which the genes are "pre-activated," i.e. rendered activator-responsive, by favorable local chromatin structures and DNA methylation patterns FIG. 8. Footprinting of the bidirectional element. S 1 -bound and unbound probes were resolved by gel shift assays, subjected to chemical cleavage in situ, for the indicated periods, and isolated as described under "Experimental Procedures." Approximately equal amounts of the recovered products were analyzed on 10% sequencing gels. A strongly protected region is indicated by the filled box, with a weakly protected region indicated by the open box. The shaded sequence 5Ј-TAGC-3Ј is cleaved preferentially by copper phenanthroline (44). The site of the MT2 mutation is indicated.  (51,52). These mechanisms could involve keratinocyte-specific enhancers, and/or somatic silencing mechanisms in cells like those from the SK-N-SH line, derived from the central nervous system. Predictions from the working model can now be addressed in further studies, including comparisons among DNase I-hypersensitive sites in chromatin from these model cell lines, and analysis of promoter constructs in vivo, in transgenic mouse lines.
An interesting issue raised by this and earlier work on other paired genes is the mechanism by which a proximate bidirectional element coordinates transcription. One possibility is that elements bound at the MT2 site alternate among minimal promoters, contributing to assembly and/or activation of basal transcription factors at each of these sites discretely. Another possibility is that promoter modulation occurs simultaneously, even to the point of interlocking divergent transcriptional initiation events. Simultaneous interactions may provide a more compelling case for conservation of the paired gene arrangement, as has been suggested for other gene pairs (53).
The bidirectional element is the most potent cis-acting element in the intergenic region, in SCC-25 and SK-N-SH cells. By deletion analysis, we could not demonstrate roles for a tandem array of CCAAT boxes, or for a central CTC box. These elements may play important roles in directing COL4A5 and COL4A6 transcription under different experimental conditions, or in cellular response to external mediators, such as have been shown to affect the COL4A1-COL4A2 promoter (54). A putative regulatory mutation causing ectopic ␣6(IV) expression in the GBM is in close proximity to tandem CCAAT boxes (14), raising the possibility that the mutation facilitates transcriptional activation through these sites. It remains to be seen whether any of these promoter elements are conserved in the intergenic region separating the COL4A3 and COL4A4 genes, therein providing a possible means of coordinating the expression of the four minor chains. Such elements could be targets of altered regulation in cases of Alport syndrome in which decreased mRNA levels for all of the minor chains accompany mutation in a single type IV collagen gene (23). Whereas it is possible that this effect arises at the level of mRNA stability, as has been proposed (23), it is also possible that the presence of a mutation generates a signal feeding back on the transcription of related genes.
In summary, we have shown that the transcription of the COL4A5 and COL4A6 genes is directed by distinct promoters, which permit highly selective expression of their gene products. In keratinocytes, co-expression of ␣5(IV) and ␣6(IV) correlates with the activity of a bidirectionally active proximal promoter element, which is bound at a previously unrecognized site. This element is likely to be important in establishing tissue-specific patterns of ␣5(IV) and ␣6(IV) co-expression. Nonetheless, it is clear from the complexity of binding, as well as from the largely unexplained phenomenon of discordant expression in the GBM, that the proximal promoter region functions within a transcriptional hierarchy, which is largely responsible for the specialized distribution of type IV collagen.