Type V Collagen Controls the Initiation of Collagen Fibril Assembly*

Vertebrate collagen fibrils are heterotypically composed of a quantitatively major and minor fibril collagen. In non-cartilaginous tissues, type I collagen accounts for the majority of the collagen mass, and collagen type V, the functions of which are poorly understood, is a minor component. Type V collagen has been implicated in the regulation of fibril diameter, and we reported recently preliminary evidence that type V collagen is required for collagen fibril nucleation (Wenstrup, R. J., Florer, J. B., Cole, W. G., Willing, M. C., and Birk, D. E. (2004) J. Cell. Biochem. 92, 113–124). The purpose of this study was to define the roles of type V collagen in the regulation of collagen fibrillogenesis and matrix assembly. Mouse embryos completely deficient in pro-α1(V) chains were created by homologous recombination. The col5a1–/– animals die in early embryogenesis, at approximately embryonic day 10. The type V collagen-deficient mice demonstrate a virtual lack of collagen fibril formation. In contrast, the col5a1+/– animals are viable. The reduced type V collagen content is associated with a 50% reduction in fibril number and dermal collagen content. In addition, relatively normal, cylindrical fibrils are assembled with a second population of large, structurally abnormal collagen fibrils. The structural properties of the abnormal matrix are decreased relative to the wild type control animals. These data indicate a central role for the evolutionary, ancient type V collagen in the regulation of fibrillogenesis. The complete dependence of fibril formation on type V collagen is indicative of the critical role of the latter in early fibril initiation. In addition, this fibril collagen is important in the determination of fibril structure and matrix organization.

Type V collagen is a member of the fibril subclass of collagens, which have in common a triple helical domain composed of an uninterrupted series of Gly-X-Y triplets. Type V collagen is a quantitatively minor component of predominantly type I collagen fibrils in most non-cartilaginous tissues. Several isoforms of type V collagen exist, which differ in the type and ratio of constituent chains, including heterotypic molecules containing type XI collagen chains. The most abundant and widely distrib-uted isoform is ␣1(V) 2 ␣2(V), which forms heterotypic fibrils with type I collagen (1). The role of type V collagen in the organization and biological properties of collagenous extracellular matrix is poorly understood. Observations of an inverse correlation between type V collagen:type I collagen ratios and collagen fibril diameter in in vitro fibril assembly experiments (2), cell cultures (3,4), and in various tissues (5) have led to the hypothesis that type V collagen serves as a negative regulator of collagen fibril diameter (3)(4)(5). That function may be mediated by retention of the non-collagenous amino-terminal propeptide after type V collagen molecules are incorporated into fibrils (2, 6 -9). This non-collagenous domain projects outward through the gap between adjacent type I collagen molecules, leaving major portions present on the fibril surface (1,3,7) where they may limit lateral growth of the fibril by steric hindrance and charge interactions (5,11).
A novel role for type V collagen was suggested recently by the results of studies (4) of cultured dermal fibroblasts from patients with the heritable connective tissue disorder, Ehlers-Danlos syndrome. The fibroblasts had mutations that caused haploinsufficiency of COL5A1, which encodes pro-␣1(V) chains. In those cultures, the total incorporation of collagen into collagen fibrils of the cell layer was reduced by half and was associated with a proportional decrease in fibril number. Because type V collagen comprises less than 5% of dermal collagen, the observed decrease in total collagen fibril formation was more than 1 order of magnitude greater than the expected reduction in collagen mass caused by the loss of contribution from one COL5A1 allele. The corresponding decrease in fibril number indicated that type V collagen may control the utilization of type I collagen during collagen fibril initiation in some tissues.
To further investigate the function of type V collagen in collagen fibril formation, in vivo studies were performed on mice in which col5a1 was inactivated by homologous recombination. The complete absence of a functioning col5a1 gene resulted in lethality at embryonic day 10 (E10), 1 with evidence of cardiovascular failure at the time of fetal demise. Animals that are haploinsufficient for col5a1 manifest many of the clinical, biomechanical, morphologic, and biochemical features of the Ehlers-Danlos syndrome, classic type.

EXPERIMENTAL PROCEDURES
Generation of col5a1-deficient Mice-A col5a1 targeting vector was generated using PCR to amplify gene-specific 5Ј-and 3Ј-targeting arms from KG-1 ES cell DNA using Herculase (Stratagene, La Jolla, CA). PCR products were verified by sequence analysis. A 5Ј-targeting arm from a region that included part of exon 3 was designed utilizing the primers 5Ј-GATGAATTC AAGCTTCACGGTGGGCACAGAGACTGGA-* This work was supported by National Institutes of Health Grants AR47054 (to R. J. W.) and EY05129 (to D. E. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  TA-3Ј and 5Ј-CAGCCGAAGCTTGGATCCTGGATGCCCTGCTCATTG-TAAATG-3Ј. These primers include HindIII restriction sites to facilitate cloning and a BamHI restriction site that introduces a unique restriction polymorphism. A 1.8-kb targeting arm was amplified by PCR, digested with HindIII, and cloned into the HindIII site of pNTKV-KO (Stratagene). A 3Ј-targeting arm from a region that included part of exon 4 was designed utilizing the primers 5Ј-TAGCCCGAATTCTGGA-TGATGAAATATTTGAGGTAG-3Ј and 5Ј-GAGTCAGAATTCGCTTGG-ACGGTCATTAGCCGTCAG-3Ј. These primers include EcoRI restriction sites (underlined). A 4.4-kb targeting arm was amplified by PCR, digested with EcoRI, and subcloned into the EcoRI site of pNTKV-KO. The col5a1 targeting vector was linearized with NotI, and 20 g was electroporated into KG-1 ES cells using standard conditions (42). Positive ES cell clones were identified by Southern blot analysis using gene-specific 5Ј and 3Ј probes. ES cells targeted correctly were expanded and injected into C57BL/6 blastocysts. Germ line transmission was obtained by breeding chimeric animals to either C57-BL/6 or 129SVE. Wild type and recombined alleles in mouse genomic DNA were detected by PCR using the primers 5Ј-CTGTAGAGGTTTG-ATCTTAGGGCG-3Ј, 5Ј-CATCATAAACCATCTACTATCGGG-3Ј, and 5Ј-CTTCTATCGCCTTCTTGACGAGTT-3Ј.
Measurement of Total Collagen Deposited in the Skin-Skin was harvested from 12-week-old animals. Two 4-mm punch biopsies were taken, from the right and left of midline on the lower back of each animal. Samples were washed with phosphate-buffered saline and then suspended in 6 N HCl at 100°C for 18 h for hydrolysis. Colorimetric analysis of the hydroxyproline content of each skin sample was performed after acid hydrolysis (12). The conversion ratio of 0.12:1.0 was used to convert micrograms of hydroxyproline to total collagen (12).
In Situ Hybridization-Whole mount in situ hybridizations were performed as described previously (13). The sense and antisense probes extended from exon 51 to part of exon 65 of col5a1. This was accomplished either by using a PCR strategy in which the bacterial T7 promoter sequence was added to the 5Ј end of the antisense sequence for the transcription of an antisense RNA probe or by subcloning into a vector with the T7 promoter for the sense probe. The primer pairs for the probes were as follows: col5a1 antisense, 5Ј-GGTTTCCTGGAGAT-CCTGGC-3Ј and 5Ј-GATCCTAATACGACTCACTATAGGGAGGGCCT-TCAGCATCCAC-3Ј; col5a1 sense, 5Ј-GGTTTCCTGGAGATCCTGGC-3Ј and 5Ј-GAGGGCCTTCAGCATCCAC-3Ј.
Analysis of Cell Culture Proteins-Embryos dissected at day 11, 12, or 14 were rinsed in phosphate-buffered saline, disaggregated physically into Dulbecco's modified Eagle's complete medium with 10% fetal bovine serum, 100 units of penicillin, and 100 g/ml streptomycin, and cultured at 37°C, 95% humidity, and 5% CO 2 . For analysis of procollagens and collagens, 12,000 -25,000 cells/cm 2 were plated in complete medium plus 0.2 M ascorbate-2-P. For Western blot analysis, pro-collagens were collected from the tissue culture medium and separated in a 5% SDS-PAGE procedure as described previously (14). Polyclonal antiserum to a unique decapeptide from the NC3 domain encoded by exon 6 of col5a1 (15,16) was obtained from Invitrogen. The antiserum was tested against a recombinant COL5A1 expression construct (generously provided by D. S. Greenspan, University of Wisconsin) and showed no cross-reactivity to COL11A1 (data not shown). The membrane used was Zetabind TM (Cuno, Inc., Meriden, CT). After incubation with the primary antibody, the blots were treated with anti-rabbit IgG conjugated to alkaline phosphatase (Bio-Rad) and developed using reagents from Bio-Rad. To measure short term synthesis and secretion of types I and V collagen, media from cells incubated with 50 Ci of [2,3,4,5-3 H]proline (PerkinElmer Life Sciences) for 16 h was harvested and digested with pepsin as described previously (4). Collagen chains were separated by SDS-PAGE in 5% bisacrylamide gels under nonreducing conditions as described previously (4,17). Radioactive proteins were detected by autoradiofluorography using EnHance TM (PerkinElmer Life Sciences) as the fluorescing agent.
Transmission Electron Microscopy-Mouse embryos at day 10 of development and subscapular skin from 12-week-old male mice were used in these experiments. The ectoderm and subjacent mesenchyme lateral to the neural tube were analyzed for col5a1 wild type and col5a1-deficient embryos. The subscapular dermis was analyzed for wild type and haploinsufficient postnatal animals. Tissues were prepared for transmission electron microscopy as described previously (18,19). Thick sections (1 m) were cut and stained with methylene blueazure blue for light microscopy and for selecting specific regions for further analysis. Tissue samples were stained with 2% aqueous uranyl acetate followed by 1% phosphotungstic acid, pH 3.2. Sections were examined and photographed at 75 kV using a Hitachi 7000 transmission electron microscope. The microscope was calibrated using a line grating. For measurement of collagen fibril density, the dermis was divided into four equal regions. Analysis was performed for both the superficial and deep dermis. The superficial dermis was defined as the region subjacent to the epidermis, and the deep dermis was the lower fourth. Both regions were photographed in the central portion. Micrographs were taken at a magnification of ϫ31,680. Calibrated micrographs from each region were chosen randomly in a masked manner from the different regions. The micrographs were digitized, and all diameters were measured within a 2.0-m mask. The mask was placed based on fibril orientation, that is, by cross-section and absence of cells. Diameters were measured along the minor axis of cross-sections using a RM Biometrics-Bioquant Image Analysis (Memphis, TN) system.
Statistical Methods-In the statistical analyses, it was assumed that the dermal fibril diameters in each measured field in a negative have a mixed distribution consisting of a dominant symmetric component contaminated with a relatively small percentage of outliers. To identify potential outliers in each measured field, a well established resistant outlier detection rule was applied (20). For the outlier-trimmed dermal fibril diameters, it was adequate to assume a symmetric but not Gaussian dominant component. Therefore, the median was used as a measure of central location for each field. The negative-specific medians of outlier-trimmed data were modeled in a linear mixed effects model (21) incorporating animal-to-animal and negative-to-negative variability. Analyses of the number of fibrils/unit area (including the identified outliers) also were modeled in a linear mixed effects model because the normal distribution assumptions were reasonable.

Targeted Disruption of col5a1 and Survivability of Offspring-
The region of col5a1 to be targeted was identified in the Celera data base of mouse genomic sequences. Two sequences flanking exons 3 and 4 were amplified from KG-1 ES cell DNA and cloned into a targeting vector containing a neomycin cassette (Fig. 1A). The construct was confirmed by sequence analysis. KG-1 ES cells were transfected with linearized target vector by electroporation. Southern blot analysis ( Fig. 1B) of G418-resistant ES cell clones identified one of 300 clones containing the homologously recombined gene that replaced most of exons 3 and 4 with the neomycin cassette. Chimeric males that were generated using this clone were mated to C57BL/6 or 129SVe females to obtain germ line transmission. Genetic analysis (Fig. 1C) of offspring of heterozygous matings was consistent with preterm lethality of col5a1Ϫ/Ϫ offspring, but embryos harvested between embryonic days 8.5 and 11 showed expected Mendelian ratios (Table I). col5a1Ϫ/Ϫ embryos were found dead at day 11 and were resorbed completely by day 12 (data not shown).
Targeted Recombination of col5a1 Prevents Gene Expression-Total RNA isolated from the tails of col5a1ϩ/ϩ and col5a1ϩ/Ϫ animals showed a 50% reduction of col5a1-derived mRNA compared with col5a2 and col1a1, as found by ribonuclease protection assay (data not shown). The ribonuclease protection assay probe was complementary to nucleotides encoding a portion of the C-propeptide, including exon 64, thus excluding the possibility that a transcript encoding a truncated peptide would have a dominant negative effect by co-assembling with wild type pro-␣1(V) or pro-␣2(V) chains. Type V collagen was not secreted by fibroblasts derived from col5a1Ϫ/Ϫ embryos. Immunoblot analysis demonstrated that pro-collagens precipitated from the medium of cultured fibroblasts lacked pro-␣1(V) chains that were detected on Western blots from wild type fibroblasts ( Fig. 2A). Detection using antibody to COL1A1 (LF41) showed similar amounts of type I collagen from both ϩ/ϩ and Ϫ/Ϫ animals (data not shown). Type V collagen was absent from the medium of fibroblasts that was labeled biosynthetically with [ 3 H]proline. Electrophoretic analysis of the labeled collagens from the medium of the col5a1Ϫ/Ϫ cells compared with the col5a1ϩ/ϩ cells also demonstrated that the ␣1(V) chains were absent (Fig. 2B).
Phenotype of col5a1Ϫ/Ϫ Embryos-Before embryonic day 10.5, col5a1Ϫ/Ϫ embryos were indistinguishable from ϩ/ϩ and ϩ/Ϫ littermates. At the time of dissection, E10 embryos were approximately the same size; however, more blood was apparent in col5a1ϩ/ϩ and ϩ/Ϫ embryos (Fig. 3, A and B). Similarly, blood-filled vessels were evident in col5a1ϩ/ϩ and ϩ/Ϫ but not in col5a1Ϫ/Ϫ yolk sacs (Fig. 3, C and D). Pooled blood was observed frequently in the col5a1Ϫ/Ϫ embryos even before cessation of the rhythmic contractions of the heart, indicating that cardiovascular insufficiency was a factor in embryonic demise. Whole mount in situ hybridization experiments in wild type embryos demonstrated that col5a1 is expressed throughout the ectoderm, with higher levels of expression within the somites, condensing limb mesenchyme, and umbilical vessels at E11 (Fig. 4). In the yolk sac, the expression is intense in the umbilical vessels and in some smaller vessels as well (Fig. 4E). col5a1 gene expression appears to be a normal part of early connective tissue formation in vascular tissue and dermis, tissues that are associated with phenotypic abnormalities in humans with Ehlers-Danlos syndrome and COL5A1 haploinsufficiency (19,22). To examine the effects of type V collagen deficiency on the development of connective tissues, ultrastructural analysis of predermal mesenchyme of col5a1Ϫ/Ϫ and wild type embryos was performed.
Death of E10 Embryos Is Associated with a Lack of Mesenchymal Fibrils-The absence of type V collagen in the col5a1Ϫ/Ϫ mice was associated with a lack of collagen fibril formation in the mesenchyme. In wild type embryos, there were numerous small diameter fibrils located subectodermally (Fig. 5A) that were localized throughout the mesenchyme (Fig.  5C). In contrast, collagen fibrils were completely absent in the mesenchyme of col5a1Ϫ/Ϫ mice (Fig. 5D). A small number of very large diameter fibrils were observed in the embryos; however, the distribution of these abnormal fibrils was limited to   the ectodermal basement membrane zone at the stromal interface (Fig. 5B). Analysis of the very limited number of fibrils that were observed in the region adjacent to the ectodermal basement membrane demonstrated that the fibrils were much larger than fibrils found in the same region in ϩ/ϩ animals (Fig. 6). In contradistinction to what was observed in the normal embryos, all fibrils found in the ectodermal basement membrane region in the Ϫ/Ϫ animals were misshapen and had irregular borders (Fig. 6, compare A and B insets).
These data indicate a lack of normal fibril assembly in the mesenchyme of mice lacking type V collagen. These roles are implied by the observed lack of type V collagen in the initiation of fibril assembly and regulation of the initial assembly. These roles are associated with a lack of normal collagen fibrils, a severe reduction in number and distribution, and resulting embryonic lethality. To further address the role of type V collagen in the regulation of fibrillogenesis, the effects of reduced copy number were studied in the postnatal dermis.
The Dermis of Mature col5a1ϩ/Ϫ Mice Contains Abnormal Collagen Fibrils-The dermis of mature col5a1 haploinsufficient mice is characterized by large numbers of structurally aberrant collagen fibrils (Fig. 7). The mature wild type dermis is composed of a relatively homogeneous population of cylindrical fibrils at 12 weeks (Fig. 7, A and B). In contrast, the  C and D). The first is a population of cylindrical fibrils that is comparable with those seen in the wild type dermis but larger in diameter (small arrows). The second population is composed of very large, heterogeneous fibrils with very irregular fibril contours (*) in cross-section (C). In the longitudinal section (D), the fibril surfaces are irregular, and the diameter is inconsistent along the fibril length. Bar ϭ 300 nm. 12-week-old col5a1ϩ/Ϫ dermis contained cylindrical fibrils and a population of larger, abnormal fibrils with very irregular contours (Fig. 7, C and D). In addition to the structurally abnormal fibrils, the haploinsufficient animals were characterized by a decreased number of collagen fibrils present in the dermis. This decrease in fibril density observed in col5a1ϩ/Ϫ versus wild type animals was found at all postnatal stages examined (postnatal day 10, 45, and 90) (data not shown). At postnatal week 12, col5a1ϩ/Ϫ animals had ϳ46% of the dermal collagen fibrils present in the wild type littermates (Fig. 8A). This difference was significant (p ϭ 0.004) as found by comparing the mean number of dermal fibrils in haploinsufficient (104, 95% CI, 68 to 140) and wild type (194, 95% CI, 158 to 230) animals. The reduced fibril number correlated with a decrease in skin collagen as measured by the amount of hydroxyproline in 8-week-old animals (Fig. 8B). Quantitation of collagen as the amount of hydroxyproline deposited in cultured embryonic fibroblasts from ϩ/Ϫ animals showed a similar reduction in total collagen deposition after 14 days in culture compared with cells from ϩ/ϩ animals; cells from Ϫ/Ϫ embryos had Ͼ95% reduction in hydroxyproline deposition compared with ϩ/ϩ animals (data not shown).
Dermal Fibril Diameter Distributions-The diameter of fibrils from the dermis of mature, 12-week-old wild type and col5a1ϩ/Ϫ mice was analyzed. The wild type dermis contained a roughly symmetric distribution of fibril diameters, whereas the col5a1ϩ/Ϫ dermis was composed of a population of fibrils comparable with the wild type and a second population of fibrils with much larger diameters (Fig. 9). These two populations of fibrils, one with circular contours comparable with the wild type controls and one with large fibrils with irregular contours, were found at ages from embryonic day 10 (Fig. 6B) to postnatal week 20 (data not shown). The finding of irregular fibrils at the earliest stages indicates that this structural phe-notype reflects deficiencies in initial fibril assembly and not merely secondary effects related to fibril growth, stability, or turnover.
Statistical Analyses of Fibril Populations-The main symmetric population of cylindrical fibrils and the frequency of large diameter, structurally abnormal fibrils and outliers in the col5a1 haploinsufficient dermis were compared with the wild type controls. Fibril diameters from five wild type animals (34 negatives, 3-8 diameters/animal) and from five mutant animals (30 negatives, 3-8 diameters/animal) were analyzed. In the wild type group, a total of 221 (3.2%) outliers was observed among 6815 observations, including 129 (1.9%) abnormally large diameters and 7 (0.1%) extremely large ones. In the mutant group, there was a total of 143 (4.3%) outliers among 3340 observations, including 121 (3.6%) abnormally large diameters and 28 (0.8%) extremely large ones. These data demonstrate a significant increase (p ϭ 0.045) of ϳ2-fold in the proportion of abnormally large diameter fibrils, corresponding to the structurally aberrant fibrils seen in the col5a1ϩ/Ϫ dermis. In addition, the main, symmetric fibril populations (free of outliers), corresponding to cylindrical fibrils, were shifted to larger diameters in the col5a1ϩ/Ϫ dermis compared with those observed in the wild type dermis. On average, the mutant medians were 17.3 nm larger (p ϭ 0.003, 95% CI, 7.6 to 27.0) than the wild type medians (Table II). DISCUSSION Murine deficiency of the major type V collagen chain, pro-␣1(V), causes death in early embryogenesis and is associated with the virtual absence of collagen fibrils. Morphological analyses of collagen fibrils in col5a1ϩ/ϩ and col5a1ϩ/Ϫ mice and Ϫ/Ϫ mouse embryos indicate that the collagen fibril number varies directly with col5a1 gene dose, that is, day 10 col5a1Ϫ/Ϫ embryos have virtually no collagen fibrils except for a small FIG. 9. Histogram showing the collagen fibril diameter distribution from the deep dermis of 12-week-old wild type (blue) and col5a1؉/؊ (red) mice. The wild type fibrils show a symmetric distribution with a median of 90.5 nm, a mean of 91.9 nm, and a range of 16 -239 nm (n ϭ 6815). The mutant fibrils have a broader distribution and demonstrate an asymmetric distribution with an increase in the larger diameter fibrils. The col5a1ϩ/Ϫ distribution has a median of 106.9 nm, a mean of 112 nm, and a range of 19 -371 nm (n ϭ 3340). The col5a1ϩ/Ϫ distribution is composed of a broad symmetric distribution and a second population of much larger fibrils.
FIG. 8. Fibril number and collagen content in the skin of wild type (؉/؉) and col5a1 (؉/؊). A, col5a1ϩ/Ϫ animals had ϳ54% of the dermal collagen fibrils seen in comparable areas from wild type littermates. At 12 weeks, the haploinsufficient animals had significantly fewer dermal collagen fibrils/unit area than the wild type animals, an average of 104 fibrils/2 m 2 (95% CI, 68 to 140) versus 114 fibrils/2 m 2 (95% CI, 158 to 230), respectively (p ϭ 0.004). B, a reduced fibril number correlated with a reduction in skin collagen measured as the amount of hydroxyproline/mm 2 (p ϭ 0.0002). Error bars represent means Ϯ S.E. number of abnormal structures in the primordial dermal-epidermal junction, whereas col5a1ϩ/Ϫ animals, which survive normally but phenotypically resemble the Ehlers-Danlos syndrome classic type (23), have approximately one-half the number of fibrils that wild type littermates have. Because type V collagen is a minor fibril component, whereas type I comprises over 90% of fibril collagen in most tissues, the data indicate that type V collagen has an essential regulatory role in collagen fibril initiation. This is the first observation that the deficiency of an extracellular matrix component prevents collagen fibril formation.
The failure of collagen fibrillogenesis is not associated with the reduced expression of type I collagen genes. Although the biochemical measurement of collagen from the dermis of col5a1ϩ/Ϫ animals correlated with the observed reduction in fibril number (Fig. 8), the col1a1:col5a2 ratios from RNA harvested from tissues were similar to those in col5a1ϩ/ϩ and col5a1ϩ/Ϫ animals (data not shown). Type I collagen was secreted efficiently from cultured embryonal fibroblasts from col5a1Ϫ/Ϫ cells (Fig. 4) but was not deposited in the cell layer, in contradistinction to wild type cells. A similar pattern of reduced type I collagen utilization in the presence of normal type I collagen biosynthesis and relative type V collagen deficiency was observed in cultured dermal fibroblasts from patients with classic Ehlers-Danlos syndrome who had haploinsufficiency mutations in COL5A1 (4).
Early collagen fibril formation events occur in channels at the cell surface of connective tissue fibroblasts (18). Recently, it has been suggested (24) that this process can begin intracellularly within Golgi-to-plasma membrane carriers during embryogenesis. Fibrils form via an intermediate stage that includes nucleation and unilateral elongation of short primary fibrils. These short fibrils later fuse, resulting in diameter enlargement and bidirectional growth (Fig. 10A, inset) (25)(26)(27). Thus, although the fibrillar irregularities observed in juvenile (postnatal day 10) and adult (postnatal week 12) col5a1ϩ/Ϫ mice may be accounted for in part by abnormal fusion events, abnormal fibril initiation is the only possible explanation for the morphological abnormalities observed in the limited number of fibrils present in col5a1Ϫ/Ϫ embryos at E10, when collagen fibrils are first detected (Fig. 6B, inset). Thus, the data presented in this report support a role for type V collagen in nucleation or initial formation of fibrils. This function is consistent with previous observations by Birk et al. (1) that type V triple helical epitopes are buried within collagen fibrils where they may be among the earliest deposited molecules within a fibril. The retained amino-terminal propeptide, which is thought to project at right angles to the main axis of the fibril (7), may be the means by which the inverse relationship between fibril number and fibril diameter is controlled (see Ref. 28). Under conditions in which type V collagen is not limiting, new fibril formation may be favored over lateral expansion of existing fibrils at a given site (see Ref. 29). The limited number of large, irregular fibrils at the primordial dermal-epidermal junction of E10 Ϫ/Ϫ animals (Fig. 6) and the subpopulation of enlarged fibrils with very irregular borders observed in the dermis of col5a1ϩ/Ϫ animals (Fig. 7) may represent unregulated type I collagen self-assembly when available type V collagen is limiting (Fig. 10, B and C).
The role of type V collagen in fibril nucleation is also consistent with the hypothesis that COL5A1 also is a more ancient gene than COL1A1 (15, 30). Tillet et al. (31) have compared a narrow fibril collagen from the sea-pen Veretillum cynomorium  10. Schematic diagram of collagen fibril formation in col5a1؉/؉ (A), col5a1؉/؊ (B), and col5a1؊/؊ (C) animals. A, under normal conditions, type V collagen fibrils are initiated at the surface of connective tissue cells to form fibril intermediates, which enlarge by lateral and end-to-end fusion events (inset). B, when type V collagen molecules (blue) are limiting because of haploinsufficiency, a reduced number of normal fibrils are nucleated through a regulated fibril assembly mechanism. Excess type I collagen molecules (black) may form morphologically abnormal aggregates through unregulated self-assembly so that both normal and abnormal populations of collagen fibrils are present (inset). C, in the state of complete type col5a1 deficiency, only the self-assembly pathway may be present, leading to the presence of a very limited number of fibrils, all of which are morphologically abnormal (inset). and noted structural similarities to vertebrate type V collagen, including the voluminous N-propeptide domain and the distribution of polar residues. Such collagens may represent ancestral forms of the minor vertebrate fibril collagens, and fibril initiation may be a residual function of type V collagen.
The death of col5a1Ϫ/Ϫ embryos at early organogenesis was associated with pooling of blood in the yolk sac, indicating the presence of cardiovascular abnormalities. It is difficult to determine whether this pooling is caused by cardiac defects or the loss of the gene within the vessels. Initial morphological studies show that the ventricles in col5a1-deficient embryos at E10 demonstrate a reduction in normal trabeculation and collagen content in contradistinction to wild type littermates. There was no apparent growth retardation at the time of death, and col5a1Ϫ/Ϫ embryos exhibited normal patterns of vessel formation, based on whole mount immunohistochemical staining with platelet endothelial cell adhesion molecule (CD31, BD Biosciences) (data not shown), although col5a1 was highly expressed in umbilical vessels and was present in the vessels in the yolk sac and the embryo (Fig. 4). It seems likely that in the absence of a collagenous component in the vessel wall, blood vessels do not have sufficient wall integrity to maintain a functioning yolk sac circulation, which is critical to survival of the embryo.
Collagen fibril formation is a complex process that requires participation of collagenous and non-collagenous elements. As expected, collagen fibrils fail to form in the absence of the major component (type I collagen) (32). However, complete murine deficiency of type III collagen, which, like type V, is a quantitatively minor component of collagen fibrils, does not prevent collagen fibrillogenesis (33). col3a1Ϫ/Ϫ animals perish from vascular rupture late in gestation, although some animals survive birth. Surprisingly, type I collagen-deficient mov-13 mice survive to postfertilization day 13 (34), compared with the col5a1Ϫ/Ϫ animals described in this report, which died ϳ3 days earlier. The differences in survival could be related to strain differences or possibly because of the residual type I collagen expression observed in some tissues in mov-13 mice (35). Another explanation for the earlier demise of col5a1Ϫ/Ϫ embryos compared with mov-13 mice may be that in the former embryonic demise is caused by the disruption of basement membrane formation rather than collagen fibril formation. Type V collagen has been localized to basement membranes (36), and embryonal disruption of basement membranes in col4a1 and col4a2 doubly deficient animals is also associated with fetal demise at approximately 10 days (37).
Defects in collagen fibril size, morphology, and content have also been observed in mice deficient in small leucine-rich repeat proteoglycans with collagen binding properties, including decorin, biglycan, fibromodulin, and lumican (38). Skin fragility phenotypes in decorin-and lumican-deficient mice have been associated with abnormal collagen fibril morphology, possibly as a consequence of abnormal lateral fusion/growth of collagen fibrils (39 -41). Murine deficiency of tenascin X is also associated with abnormalities in fibril morphology and moderately reduced collagen content of skin (10). A comparison of the data presented in this report with the substantial body of knowledge of extracellular matrix deficiency states in mice indicates that no matrix component (save type I collagen itself) appears to exert the same level of control of fibril initiation as type V collagen. We propose that the control of collagen fibril formation is complex and that, in addition to transcriptional control of constituent molecules, post-translational control is exerted at the level of fibril nucleation; this process appears likely to be affected by the availability of type V collagen molecules.