Collagen fibrillogenesis in vitro: interaction of types I and V collagen regulates fibril diameter

Collagen molecules are synthesized, secreted into the extracellular space and organized into striated fibrils with a tissue-specific organization. A variety of factors have been shown to influence collagen fibrillogenesis and have been postulated to be important in the regulation of fibril architecture. These include the interaction with proteoglycans (Scott, 1984), the sequence and extent of procollagen processing (Fleischmajer et al. 1981, 1983; Miyahara et al. 1984), the collagen type present (Miller, 1976; Birk and Silver, 1984b) and the interaction of different collagen types (Lapiere et al. 1977; Adachi and Hayashi, 1985, 1986).

The embryonic chicken cornea develops through two distinct stages in which several different collagen types are expressed. Collagen types I, II and IX are produced by the corneal epithelium and are present in the primary stroma (Hendrix et al. 1982; Fitch et al. 1988a). Follow-ing the migration of fibroblasts into the stroma, types I, V and VI collagen are synthesized by the fibroblasts and are found throughout the secondary corneal stroma (Hay et al. 1979; Linsenmayer et al. 1984, 1986b).

The collagen types present may be important in the determination of fibril architecture. It has been known for a number of years that in reassembly studies in vitro type I collagen forms thicker fibrils than type HI collagen, while type II collagen fibrils generally have the smallest diameters (Miller, 1976; Lapiere et al. 1977; Birk and Silver, 19846). More recently, different collagen types have been shown to form heterotypic fibrils in a variety of tissues. For example, collagen types I and V are co-assembled in the secondary corneal stroma (Linsenmayer et al. 1985; Fitch et al. 1984, 1988b; Birk et al.1988), types I and II in the primary corneal stroma (Hendrix et al. 1982; Linsenmayer et al. 1985, 1990), types H, IX and XI in cartilage (Vaughan et al. 1988; Mendier et al. 1989) and types I and III in a number of tissues including dermis and tendon (Keene et al. 1987a). We have previously postulated that during fibril-logenesis, the interaction of two or more different collagen types within a fibril may serve to regulate fibril diameter or the interactions of fibrils with the interfibrillar matrix.

Immunoelectron microscopic studies of the corneal stroma with antibodies against type I and V collagen have demonstrated that corneal fibrils are heterotypic, composed of both collagen types I and V (Birk et al. 1986, 1988; Fitch et al. 1988b). The chick cornea is composed predominantly of type I collagen, but also contains a large amount (approximately 20%) type V collagen (Hay et al. 1979; Poschl and von der Mark, 1980; McLaughlin et al.1989). The corneal stroma is considerably enriched in type V collagen relative to other type I-containing connective tissues and, unlike other tissues containing predominantly type I collagen, corneal collagen fibrils have a very narrow range of small-diameter fibrils (∼25nm). This monotony of fibril size is necessary for optical transparency (Cox et al. 1970).

We have suggested that type V collagen may have a regulatory role in the control of fibril architecture through its interactions with type I collagen. Here we have studied the interaction of collagen types I and V using an in vitro self-assembly system. Our data indicate that the interaction of type V collagen with type I collagen is one mechanism capable of modulating fibril diameter and is at least partially responsible for the regulation of collagen fibril diameters in the corneal stroma.

Chick embryos were made lathyritic by sequential injections of 0.1 ml of β-aminopropionitrile fumarate (βAPN) in phosphate-buffered saline (PBS) on days 14 and 15, 1.0 mg, and on day 16, 2.0 mg. The chick embryos were sacrificed on day 17 with a viable yield of 40-60%.

Collagen type I was extracted and purified from lathyritic 17day chick embryo corneas using dilute acetic acid in the presence of protease inhibitors. Type I collagen was purified by differential salt precipitation from acid and neutral pH, chromatography on DEAE-cellulose and low ionic strength precipitation (Birk and Silver, 1984a).

Type V collagen also was extracted and purified from lathyritic embryos. The bodies were washed in ice-cold deionized water followed by homogenization in 50 mM Tris-HCl, pH7.5, 4.4M NaCl, 50mM e-amino-n-caproic acid, 10mM n- ethylmaleimide, 1 mM benzamidine-HCl, 5 mM phenylmethylsulfonyl fluoride, 1μgml⁻¹ leupeptin, pH 7.4. The insoluble material was collected by centrifugation at 10 000 revs min⁻¹ for 45 min in a Sorvall G SA rotor, the pellet was washed twice in homogenization buffer and finally with ice-cold water. The pellet was homogenized in 0.5M-acetic acid with lμgml⁻¹ pepstatin A, 1 μgml⁻¹ leupeptin and the pH was adjusted to 2.5 with 6 M HC1. Extraction was with a 20-fold excess of acetic acid for 48 h. The supernatant was collected after centrifugation and type V collagen was purified by differential salt precipitation from acid pH, and dialysis against buffered 2.0M urea followed by a second round of salt precipitations as previously described (Silver and Birk, 1984). The purified type V collagen at approximately 0.5 mg ml ‘was dialyzed exhaustively against 0.5M-acetic acid, lyophilized and stored under liquid nitrogen.

Purified collagens were characterized by SDS-polyacrylamide gel electrophoresis on 5% gels according to the method of Laemmli (1970). After electrophoresis gels were fixed in methanol/acetic acid, stained with Coomassie Brilliant Blue R-250 and destained.

Collagen types I and V were dissolved in 0.01 M HC1 (pH 2.0) at 4°C, dialyzed against the same solvent and centrifuged at 100000 for 1 h. Fibrils were formed from solutions containing type I collagen (0.15 mg ml⁻¹), type V collagen (0.15 mg ml⁻¹) and type I collagen (0.15 mg ml⁻¹) with increasing amounts of type V collagen. These collagen solutions were dialyzed against 0.01 M HC1 (pH 2.0) at 4°C, PBS (pH 7.3) at 4°C and then the temperature was raised to 37°C. Within a single experiment all collagens were from a single extraction and were dialyzed into the same solutions. The products were examined and photographed using a Philips 420 transmission electron microscope after negative staining with 1% phosphotungstic acid, pH 7.0. Diameter measurements were made only on fibrils demonstrating a 67 nm periodicity, except in the case of type V collagen alone where all fibrils were non-striated.

Type V collagen preparations at 25 μgml⁻¹ in acetic acid were mixed with an equal volume of glycerol and sprayed onto freshly cleaved mica. The samples were transferred to an Edwards vacuum evaporator, and rotary shadowed with platinum at 8° followed by carbon from 90° (Birk and Silver, 1983; Silver and Birk, 1984). The replicas were floated onto distilled water, picked up onto mesh grids, examined and photographed using a Philips 420 transmission electron microscope.

Monoclonal antibodies against chick collagen type I (I-BA1 ; Linsenmayer et al. 1986a), type V (V-AB12, V-DH2; Linsen-mayer et al. 1983) and type IV (IV-1A8, 1V-HB12; Fitch et al. 1982) were conjugated to 5 and 10 nm colloidal gold particles (Birk et al. 1988) and used in direct labelling experiments. All antibodies recognize helical epitopes that are collagen typespecific.

For immunoelectron microscopy, fibrils were transferred to Formvar-coated grids and lightly fixed in 4% paraformaldehyde in PBS at 4°C for 15 min. In most cases, the fibrils were pretreated at 4°C for 30-120 min in PBS followed by light fixation in 4% paraformaldehyde as described above. The fixed fibrils were incubated in PBS with sodium borohydride (50mg/100ml) for 60 mm at 4°C to reduce any free aldehydes. The sections were washed several times in PBS and blocked in a 5% solution of non-fat milk powder for 2h at 4°C. The grids were then incubated overnight at 4°C with gold-conjugated antibody diluted in PBS with 1% bovine serum albumin (BSA) and 0.05% Tween 20. The grids were washed five times over lh with PBS/Tween, fixed in 2.5% glutaraldehyde, 0.1M cacodylate, pH 7.4, washed with water and stained with 1% phosphotungstic acid.

Collagen types I and V from lathyritic 17-day chick embryo corneas or whole embryos were analyzed for purity by SDS-polyacrylamide gel electrophoresis (Fig. 1). Both collagen types were extracted using dilute acetic acid in the presence of protease inhibitors. Type I and type V collagen were present predominantly as alpha chains with little multimeric crosslinked material. The type V preparations contained approximately 60-70% of a higher molecular weight form of the αl(V) and α2(V) chains of type V collagen relative to pepsin-extracted or -treated type V collagen. When treated with pepsin these preparations co-migrate with the pepsin-extracted standard (data not shown). No type V procollagens or processing intermediates were present in these preparations. In addition, the type V preparations were completely devoid of detectable type I collagen and the type I preparations also contained no procollagen or contaminating type V collagen.

Type V collagen fractions were examined further using rotary shadowing. The electron micrographs show the type V molecules to be long semiflexible rods. These molecules have a slightly longer helical length than type I collagen. A large percentage (60-80% depending on the preparation) of these molecules have a terminal globular region (Fig. 2A-C). The percentage of molecules with terminal extensions was consistent with the amount of the high molecular weight form of type V collagen demonstrated by electrophoresis. There appears to be a flexible region between the helix and the terminal domain, in that this region is often observed at extreme angles relative to the helical domain. This is presumably a pepsin-sensitive domain, since pepsin-treated type V collagen does not display this terminal region (Fig. 2D).

The in vitro formation of homotypic type I and V fibrils was studied using purified types I and V collagen (Figs 3 and 4). In electron micrographs of negatively stained preparations, type I collagen formed typical 67 nm striated fibrils while type V collagen formed considerably thinner fibrils with no obvious periodicity. The fibrils formed from type I collagen had a broad distribution of relatively large diameters (157±46nm; Fig. 4A). In contrast, fibrils formed from purified type V collagen were composed of loosely associated filaments (Fig. 3B). The fibrils had small diameters with a narrow distribution (25±8nm, Fig. 4E). Using highly purified type V collagen we observed only fibrillar structures with no apparent periodicity. However, when small amounts (2-5%) of type I collagen were added, a large number of well-formed striated fibrils were observed as well as numerous poorly formed fibrils with clear periodicity (data not shown).

To examine the possible interactions of these two different collagens during fibrillogenesis, heterotypic fibrils were formed by mixing type I collagen with increasing amounts of type V collagen (Figs 3 and 4). All mixtures studied formed typical 67 nm cross-striated fibrils (Fig. 3C). However, at higher type V collagen concentrations there were small numbers of non-striated fibrils in the background. While type I collagen alone formed fibrils with a broad range of diameters, in the presence of increasing amounts of type V collagen the mean diameters progressively decreased (10%, 144±39nm; 20%, 101±30nm; 30%, 85±21nm, Fig. 4A-D). The decrease at each step was significant when analyzed using the Cochran t-test (P<0.005). The variance in diameters within each population also was decreased as the percentage of type V collagen in the mixtures increased. This decrease in variance was significant to at least the level of P<0.01 at each step when analyzed using the Fisher variance ratio. These experiments were repeated with four different preparations of type V collagen using concentrations from 5 to 50% (data not shown) and the data were completely consistent with those presented. These experiments were also done with type I collagen isolated from whole embryos, with the same results. However, when these experiments were repeated with pepsin-extracted type V collagen, the modulation of fibril diameter was not as pronounced and 30% pepsin-extracted type V collagen was required to produce a measurable decrease in mean fibril diameter (data not shown). These results indicate that collagen types I and V interact in vitro, this interaction is at least partially responsible for the control of collagen fibril diameter and the terminal pepsin-sensitive extension of the type V collagen molecule is required for the full effect.

To determine whether types I and V collagen are coassembled within the fibrils assembled in vitro, immunoelectron microscopy using collagen type-specific monoclonal antibodies was performed on fibrils formed from type I collagen or mixtures of types I and V collagen. Type I collagen fibrils reacted positively with the antibodies against type I collagen (Fig. 5A), but demonstrated no reactivity with antibodies against type V (Fig. 5B) or type IV collagen (data not shown). Fibrils formed from a mixture of type I and type V collagen reacted positively with the antibodies against type I collagen, but only weakly with the antibodies against type V collagen (data not shown).

We have previously demonstrated that immunohistochemical localization of type V collagen within corneal fibrils in situ required unmasking by fibril disruption; for example, by cold temperature in lathyritic tissues. We performed similar unmasking in the fibrils formed in vitro. When the fibrils formed from mixtures of types I and V collagen were pretreated in cold PBS followed by fixation and labelling, the fibrils reacted strongly with antibodies against type V collagen as well as type I collagen. While the cold-treated fibrils showed similar reactivity’ for type I collagen (small arrows, Fig. 6A) when compared with the non-treated fibrils (data not shown), the reactivity for type V collagen (large arrows, Fig. 6A) was significantly enhanced in the cold-treated specimens. Such cold-treated fibrils remained unreactive with irrelevant antibodies (Fig. 6B).

We have demonstrated that type V collagen co-polym-erizes with type I collagen in an in vitro self-assembly assay. This interaction produces fibrils with smaller diameters; increasing the amounts of type V collagen in the reaction mixture effects a progressive decrease in the mean fibril diameter. These studies along with other studies involving the co-polymerization of collagen types I and III in vitro (Lapiere et al. 1977) implicate heterotypic assembly as a general regulatory mechanism in the control of fibril structure. In vivo immunochemical localization studies have shown that collagen types I and V (Birk et al. 1988; Fitch et al. 19886; Linsenmayer et al. 1990), types I and II (Hendrix et al. 1982; Linsenmayer et ail. 1985, 1990), types II and IX (Vaughanet al. 1988), types II and XI (Mendier et al. 1989) and types I and III (Keene et al. can be present as heteroty pic fibrils. Covalent peptides containing intermolecular crosslinked sequences from collagen types I and III (Henkel and Glanville, 1982) as well as II and IX (van der Rest and Mayne, 1988) have been isolated, which further supports the existence of such fibrils.

Like most dense stromal connective tissues, the cornea contains predominantly type I collagen (Hay et al. 1979). However, unlike other tissues containing predominantly type I collagen, the corneal fibrils have uniformly small diameters (∼25nm). This unique morphology may be related to the fact that in several species, it has been shown that the corneal stroma contains significantly more type V collagen than do other tissues whose stromal matrix collagen is predominantly type I (Davison et al. 1979; Hong et al. 1979; Welsh et al. 1980; Cintron et al. 1981; Fessler et al. 1982; Tseng et al. 1982). Immunohistochemical studies have demonstrated that type V collagen is an interstitial collagen distributed uniformly throughout the corneal stroma with type I collagen (von der Mark and Ocalan, 1982; Linsenmayer et al. 1983, 1984, 1985; Fitch et al. 1984; Birkei al. 1986). Collagen types I and V physically interact as heterotypic fibrils, and the type V collagen is arranged in the quarterstaggered array typical of fibrillar collagens (Birk et al. 1988; Fitch et al. 19886). These biochemical and immunohistochemical observations suggest that the interaction of type V collagen with type I collagen may regulate corneal fibril diameter. The results from in vitro assembly assays support this conclusion.

There are a number of ways in which a quantitatively minor collagen type could function to regulate fibril diameter. It is noteworthy in this regard that our results show that the terminal globular domain is necessary for the full effect of type V collagen in modulating fibril diameter. Type V collagen may inhibit the lateral growth of the corneal fibrils because of its longer helix (Bachinger et al. 1982; Silver and Birk, 1984) and/or because of the persistence of a terminal globular domain (Fessler et al. 1982; Broek et al. 1985) even after assembly into a fibril. These features may permit the co-assembly of type V collagen with type I collagen only until a certain ‘critical’ type V concentration is reached, after which the continued addition of type I collagen is inhibited by the longer ‘non-fitting’ type V molecule and/or its aminoterminal domain. Another possibility is that type V collagen forms thin filaments that serve as nucleation sites for type I collagen assembly. The presence of more nucleation sites for a given quantity of type I collagen might result in smaller fibril diameters than in a tissue where there were fewer nucleation sites for the same amount of type I collagen.

Our results demonstrate that, comparable to the corneal fibrils in situ, the fibrils co-assembled from types I and V collagen in vitro require unmasking of the type V epitopes by partial disruption of fibril structure for their localization. These data indicate that the substructure of the heterotypic fibrils formed in vitro is similar to that seen in vivo.

The in vitro studies described here and elsewhere (Broek et al. 1985) also indicate that type V collagen can form filaments with no apparent periodicity in the absence of type I collagen. These data indicate that a type V collagen core filament serving as a nucleation site for type I collagen is possible. However, our immunoelectron microscope studies indicate that this is probably not the case, at least within the chick corneal stroma (Fitch et al. 19886). After limited hydrolysis with type I collagenase, which has no activity for type V collagen, distinct subassemblies containing both collagen types I and V are present. Collagen types I and V co-assemble to form heterotypic fibrils, but additional data regarding the structural organization of the different collagen molecules within a heterotypic fibril as well as on the initial intra- and extracellular assembly steps is needed to distinguish among the possible mechanisms.

Others have reported that type V collagen can form D periodic fibrils in vitro (Adachi and Hayashi, 1985, 1987), a phenomenon that we cannot confirm. This discrepancy could be explained by the use of pepsin-extracted type V collagen, which has no terminal globular region and/or by contamination with small amounts of type I collagen. We have demonstrated that only small amounts (2%) of type I collagen are sufficient to form periodic fibrils. In addition, the absence of a non-helical domain may promote the assembly of D periodic fibrils (Adachi and Hayashi, 1985, 1986). In situ, type V collagen has been localized to thin fibrils with no apparent periodicity in the subepithelial region of the amnion (Modesti et al. 1984). While, it is unclear whether these fibrils also contain type I collagen, the data presented here suggest that they probably do not.

This study indicates that the interaction between molecules of different collagen types during the coassembly into heterotypic fibrils is an important regulatory mechanism in collagen fibrillogenesis. We have demonstrated that the interaction of type V collagen with type I collagen decreases the diameters of fibrils formed in vitro. However, such fibrils are still substantially wider than the 25 nm fibrils characteristic of the corneal stroma in vivo. Therefore other mechanisms may also operate to regulate fibril structure. For example, in vitro assembly is not as controlled as the fibril-forming process is in situ. In vitro there are no cell-defined domains characteristic of connective tissues actively assembling collagenous matrices (Birk and Trelstad, 1984, 1986; Birk et al. 1989). Cellular control of the mixing of different macromolecules and post-depositional processing probably play an important role in the control of fibril formation. However, at present little is known about the intracellular and extracellular mechanism(s) controlling these interactions.

Non-collagenous matrix components, including proteoglycans and glycoproteins, exert a major influence over fibril assembly and tissue organization. The relative content of specific proteoglycans decreases as fibril diameter increases in some tissues and these changing patterns may be partially responsible for the resulting fibril diameters (Scott, 1984). The secondary corneal stroma contains keratan sulfate, heparan sulfate and chondroitin sulfate proteoglycans, some of which interact with specific regions on the collagen fibril (Scott and Haigh, 1985, 1988). The effects of proteoglycans on collagen fibril formation have been studied in vitro using proteoglycans from cornea or sclera (Birk and Lande, 1981) and were shown to affect the kinetics of fibril formation differently.

Another possible control point in the regulation of fibril structure may reside in the processing of the procollagen propeptides during fibrillogenesis. This has been implicated in the regulation of collagen fibril formation both in vivo and in vitro. In dermatosparatic cattle the amino-terminal propeptide is not removed and the collagen fibrils are abnormally thin (Lenaers et al. 1971). The differential processing of the amino and carboxy propeptides has been studied in developing skin using immunoelectron microscopy (Fleischmajer et al. 1981, 1983). In these studies the amino-terminal propeptides were associated only with the thinner and presumably more recently formed fibrils while the carboxy propeptides were not associated with any fibrils in a regular manner, suggesting that the processing of these regions may be important in the control of fibril assembly. Studies on propeptide processing in vitro have indicated that the temporal order and extent of procollagen processing may be important determinants of fibril structure, specifically, the initial determination of fibril diameter (Miyahara et al. 1982, 1984). When the car-boxy-terminal propeptide was removed prior to the removal of the amino-terminal peptide, thin fibrils formed. However, if the order of propeptide processing was reversed, thick fibrils formed.

In summary, heterotypic fibril assembly is one important regulatory mechanism in fibril formation. However, this mechanism is likely to be integrated with other regulatory mechanisms in fibrillogenesis. For example, the temporal and spatial mixing of macromolecules within different intra- and extracellular compartments; processing of type V and other procollagens in heterotypic assemblies; and whether homotypic versus heterotypic fibrils interact with other matrix components differently during assembly or as mature fibrils. The further elucidation of these and other questions will be required for an understanding of the assembly of tissue specific extracellular matrices.

This work was supported by NIH grants EY 05129 and EY 05191.