Macromolecular Specificity of Collagen Fibrillogenesis

Suprastructures of the extracellular matrix, such as banded collagen fibrils, microfibrils, filaments, or networks, are composites comprising more than one type of macromolecule. The suprastructural diversity reflects tissue-specific requirements and is achieved by formation of macromolecular composites that often share their main molecular components alloyed with minor components. Both, the mechanisms of formation and the final macromolecular organizations depend on the identity of the components and their quantitative contribution. Collagen I is the predominant matrix constituent in many tissues and aggregates with other collagens and/or fibril-associated macromolecules into distinct types of banded fibrils. Here, we studied co-assembly of collagens I and XI, which co-exist in fibrils of several normal and pathologically altered tissues, including fibrous cartilage and bone, or osteoarthritic joints. Immediately upon initiation of fibrillogenesis, the proteins co-assembled into alloy-like stubby aggregates that represented efficient nucleation sites for the formation of composite fibrils. Propagation of fibrillogenesis occurred by exclusive accretion of collagen I to yield composite fibrils of highly variable diameters. Therefore, collagen I/XI fibrils strikingly differed from the homogeneous fibrillar alloy generated by collagens II and XI, although the constituent polypeptides of collagens I and II are highly homologous. Thus, the mode of aggregation of collagens into vastly diverse fibrillar composites is finely tuned by subtle differences in molecular structures through formation of macromolecular alloys.

Suprastructures of the extracellular matrix, such as banded collagen fibrils, microfibrils, filaments, or networks, are composites comprising more than one type of macromolecule. The suprastructural diversity reflects tissue-specific requirements and is achieved by formation of macromolecular composites that often share their main molecular components alloyed with minor components. Both, the mechanisms of formation and the final macromolecular organizations depend on the identity of the components and their quantitative contribution. Collagen I is the predominant matrix constituent in many tissues and aggregates with other collagens and/or fibril-associated macromolecules into distinct types of banded fibrils. Here, we studied co-assembly of collagens I and XI, which co-exist in fibrils of several normal and pathologically altered tissues, including fibrous cartilage and bone, or osteoarthritic joints. Immediately upon initiation of fibrillogenesis, the proteins co-assembled into alloy-like stubby aggregates that represented efficient nucleation sites for the formation of composite fibrils. Propagation of fibrillogenesis occurred by exclusive accretion of collagen I to yield composite fibrils of highly variable diameters. Therefore, collagen I/XI fibrils strikingly differed from the homogeneous fibrillar alloy generated by collagens II and XI, although the constituent polypeptides of collagens I and II are highly homologous. Thus, the mode of aggregation of collagens into vastly diverse fibrillar composites is finely tuned by subtle differences in molecular structures through formation of macromolecular alloys.
Extracellular matrix aggregates, despite their functional and morphological diversity, often contain the same type of macromolecules as their major constituent. In tendons, for example, large and strongly banded fibrils of variable width are aggregated into bundles that resist extreme tensile forces in one dimension. By contrast, thin fibrils with uniform diameter and spacing are organized into orthogonal sheets forming the translucent corneal stromal matrix that withstands high traction in two dimensions. However, the same protein, i.e. collagen I, is the quantitatively predominant component in both suprastructures (1,2). The major cartilage collagen is type II, but this protein, too, is common to fibrils with different structures and functions (3,4). Finally, collagens I and II occur together in fibrocartilage (5), a tissue characterized by a high resistance to both compressive and tensile forces and that contains fibrous bundles in addition to an amorphous extrafibrillar matrix rich in proteoglycans.
Collagens I and II are very similar in their molecular structure. They both contain three polypeptides with a central sequence of 1014 amino acids in which strictly every third residue is glycine. In addition, these glycine residues are frequently preceded by hydroxyproline and/or followed by proline. The (Gly-Xaa-Yaa) n sequences are flanked at both ends by short non-periodic sequences, called telopeptides. Collagen I is a heterotrimer containing two ␣1(I) chains and one ␣2(I) chain, whereas collagen II is a homotrimer of ␣1(II) chains. However, the three primary structures within the (Gly-Xaa-Yaa) n domains of the three polypeptides are more than 90% identical (6), and both collagen I and II form very similar triple helices with a length of 300 nm, turning the molecules into highly elongated rods of considerable stiffness. The molecules are incorporated into fibrils by lateral aggregation and a longitudinal stagger with a periodicity called D yielding the characteristic banding patterns of collagen fibrils observed by electron microscopy.
The suprastructural diversity of collagen fibrils has been investigated by several approaches, including x-ray diffraction. High resolution fibril diffraction requires regularly ordered, crystalline-like tissue domains of sufficient expansion. These prerequisites are met in only two paradigmatic tissues, i.e. the rat tail tendon and the notochord sheath in a primitive chordate, the lamprey (7). In these cases, the collagen structure has been elucidated in molecular detail (8 -10). Other tissues produced only low resolution diffraction patterns that, however, warranted the conclusion that collagen organizations in skin (11) and cartilage (12) differed from those in rat tail tendons or lamprey notochords. The D-periodicity of skin fibrils was 64 nm (67 nm in rat tail tendon). Likewise, the center-to-center spacings between collagen molecules were larger in cartilage (1.7 nm) than in tendons (1.1 nm). However, further suprastructural details could not be resolved.
The question arises how distinct matrix suprastructures can be formed and/or stabilized. The answers still are incomplete, but it is known that collagen fibrils in situ comprise not only several types of collagens (1,3,13,14) but also additional macromolecules (4,(15)(16)(17)(18). As shown in animals with deficiencies in small leucine-rich proteins, fusion of fibrils into irregularly shaped aggregates is often observed (19,20) suggesting a decisive role of small leucine-rich proteins in fibrillar organizations at the level of tissue architecture. In addition, it has been shown by fibril reconstitution from soluble collagens in vitro that minor collagenous and non-collagenous components can control lateral fibril growth (21)(22)(23)(24). We have demon-strated recently that incorporation of collagen IX is required for the macromolecular alloy stability in prototypic cartilage fibrils (25). The other minor collagen in cartilage fibrils, i.e. collagen XI, forms efficient nuclei of fibrillogenesis and, thereby, stringently controls fibril shape by co-polymerizing with collagen II. However, polypeptides of collagen XI are not unique to cartilage but are present in a wide variety of non-cartilaginous tissues, which in part contain collagen I as the major fibrillar collagen (26 -30). For example, collagen fibrils in the vitreous humor are very similar to cartilage fibrils in that they contain collagen II as the quantitatively major collagen. However, the vitreous humor version of collagen XI contains ␣2(V) chains in place of ␣2(XI) chains (31,32). Presumably for this reason, patients with Stickler syndrome harboring mutations in the ␣2(XI) chains do not exhibit the eye involvement characteristic for Stickler patients with mutations in the ␣1(XI) chains (33). Similarly, the type XI collagen fraction isolated from articular cartilage was reported to contain ␣1(V) chains in amounts increasing with age (34). In addition, these observations suggest that collagens V and XI should not be regarded as separate collagen types. Rather, minor collagens composed of polypeptides derived from the genes encoding ␣ chains of collagens V or XI may exert discrete functions in organ-or tissue-specific fibrillogenesis. Therefore, we addressed the question whether the collagen XI-controlling growth of collagen II-containing cartilage fibrils is effective also in fibrillogenesis of collagen I. Here, we found that mixtures of collagens I and XI aggregated into specific composites, too. Strikingly, however, collagen I/XI fibrils were not only structurally distinct from those of collagens II and XI but also arose by different mechanisms.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were of reagent grade. DEAE-cellulose (DE52) was from Whatman International. Polyclonal anti-collagen I antibodies AB752P were from Chemicon International. Goat anti-rabbit IgG conjugated with horseradish peroxidase was from Sigma. Goat anti-rabbit IgG conjugated with 18-nm gold particles was from Jackson ImmunoResearch.
Collagen Purification-The cartilage version of collagen XI comprising ␣1(XI), ␣2(XI), and ␣3(XI) chains was purified in native and fibrillogenesis-competent form from cultures of chick embryo sternal chondrocytes in agarose gels as described previously (25). Collagens I and V were obtained from tarso-metatarsal tendons of 17-day-old chick embryos. Tissue from 150 embryos was accumulated, frozen at Ϫ20°C until use, thawed, and extracted overnight with 15 volumes of 0.5 M acetic acid at 4°C. This extraction procedure was repeated, and the extracts were combined. Collagens were precipitated by adding onefourth the extract volume of 4.3 M NaCl. The suspensions were stirred overnight, and, after centrifugation, crude pelleted collagens were dissolved in a minimal amount of buffer A. After extensive dialysis against buffer A, the samples were passed over a DEAE-cellulose column (3.5 ϫ 12 cm) equilibrated in buffer A. Mixtures of collagen I and V were recovered from the breakthrough fraction and were dialyzed exhaustively against buffer B. Final separation of collagens I and V was achieved by re-chromatography on DEAE-cellulose equilibrated in buffer B. Collagen I was recovered from the breakthrough fraction and was pure as judged by SDS-PAGE. Collagen V bound to the column and was eluted from the column by switching the eluent to buffer A. Purified proteins were dialyzed extensively against 0.1 M Tris-HCl, pH 7.4, containing 1 M NaCl, and were precipitated by addition of solid NaCl to a final concentration of 4.5 M. After centrifugation, pellets were redis-solved in storage buffer at appropriate concentrations, dialyzed against storage buffer, and clarified by centrifugation. The purity of the collagens was judged by SDS-PAGE after staining with Coomassie Blue.
Antisera to Native Collagen XI-0.5 mg of purified, native collagen XI was suspended in 1 ml of 0.01 M hydrochloric acid, and was mixed with 1 mg of keyhole limpet hemocyanin (Sigma) in 0.1 ml of H 2 O. After titration to pH 6.8 with 1 M NaOH, conjugation was initiated by addition of 0.05 ml of 1% glutaraldehyde. The reaction was allowed to proceed for 3 h at room temperature, and the sample was exhaustively dialyzed at 4°C against 0.05 M Tris-HCl, pH 7.4, containing 0.15 M NaCl to remove excess glutaraldehyde and to inactivate free aldehyde groups. Antisera were raised by immunizing rabbits with 0.5 mg of conjugate suspended in Freund's complete and incomplete adjuvant following established protocols (35). The reactivity and specificity of the antisera was assessed by ELISA and immunoblotting.
Pepsin Digestion-To obtain molecules without non-helical terminal domains, purified collagen XI was digested with pepsin (1 mg/ml) in 0.5 M acetic acid, pH 2.5, for 24 h at 4°C. The pH was adjusted to pH 7.4 to inactivate pepsin. The digested collagen XI was precipitated with 3 volumes of ethanol and dissolved in SDS-PAGE sample buffer.
Determination of the Collagen Concentration-The collagen concentration was determined using the bicinchoninic acid assay according to the instructions of the manufacturer (Pierce, Rockford, IL). Pepsintreated collagen II in acetic acid was used as standard. The concentrations of the standard solutions were determined by circular dichroism at 221 nm (specific ellipticity [] 221 ϭ 8550 deg ϫ cm 2 ϫ decimole Ϫ1 (36)).
SDS-PAGE and Immunoblots-Protein samples for electrophoresis were prepared by precipitation with cold ethanol. The precipitates were dissolved in SDS-PAGE sample buffer and electrophoresed in 4.5-15% polyacrylamide gradient gels. The proteins were electrotransferred to nitrocellulose filters, and the membranes were treated with blocking buffer for 1 h at room temperature. This was followed by incubation at 4°C overnight with an antiserum to collagen XI diluted 1:1000 in blot buffer, washing, and incubation for 2 h at room temperature with peroxidase-conjugated goat-anti-rabbit IgG (Sigma, Germany) in blot buffer. Immunoreactivity was revealed by chemiluminescence (ECLkit, Amersham Biosciences, UK).
In Vitro Fibrillogenesis-Solutions in storage buffer of pure collagens I and XI, or defined mixtures thereof were degassed under vacuum. 100-l samples were transferred to microcuvettes (Multicell, light path, 1 cm, Beckman, Palo Alto, CA) and were mixed with 100 l of distilled water at 4°C. The cuvettes were sealed and placed immediately into a spectrophotometer (Beckman UV 640, equipped with a Multicell holder, Micro Auto 12), connected to a water bath at 37°C. In some experiments, collagen I solutions in storage buffer were diluted with equal volumes of water and were added to collagen XI fibrils preformed in half-strength storage buffer for 180 min at 37°C. Aggregation was monitored by turbidity development at 313 nm.
Electron Microscopy and Immunoelectron Microscopy-10-l aliquots of reconstitution mixtures were spotted onto sheets of Parafilm. Copper grids coated with Formvar/carbon were floated on the drops for 5 min to allow adsorption of aggregates and were washed with distilled water. For transmission electron microscopy, fibrils on grids were negatively stained with 2% uranyl acetate for 10 min. For immunoelectronmicroscopy, the grids were treated for 30 min with 2% (w/v) dried skim milk in PBS and, subsequently, for 2 h with the same buffer containing antiserum to collagen XI or polyclonal antibodies to collagen I (dilution 1:100 and 1:400, respectively). After extensive washing with PBS, the grids were put on drops of 0.2% dried skim milk in PBS, containing colloidal gold particles (18 nm) coated with goat antibodies to rabbit immunoglobulins at dilutions recommended by the manufacturer (Jackson ImmunoResearch). Finally, the grids were washed with distilled water and negatively stained with 2% uranyl acetate for 10 min. Negative controls were done without first antibody treatment. Electron micrographs were taken at 60 kV with an EM 410 electron microscope (Philips) or at 80 kV with a CM 10 electron microscope (Philips).

RESULTS
Purification of Collagens-Native, fibrillogenesis-competent collagen XI was purified from cultures of chick embryo sternal chondrocytes embedded into agarose and cultured in the presence of the lysyl oxidase inhibitor ␤-aminopropionitrile. As described previously (25), the pure protein produced electrophoretic patterns in SDS-PAGE that corresponded to several variants of ␣1(XI)-polypeptides, as well as the ␣2(XI) and ␣3(XI) chains (Fig. 1, lane 1). This complex polypeptide composition results from both alternative splicing (37,38) as well as incomplete proteolytic processing of the amino-terminal propeptide domain (39,40). Collagen I was purified by ion exchange chromatography on DEAE-cellulose from a mixture of collagens extracted from leg tendons of 17-day-old chicken embryos. The purified collagen I yielded the expected banding pattern in SDS-PAGE and consisted mostly of monomeric ␣1(I) and ␣2(I) chains, in addition to small amounts of cross-linked ␤-components (Fig. 1, lane 5). Thus, the preparations were considered as pure and, in particular, free of collagen V.
Antisera to Native Collagen XI-Because collagens V and XI are highly similar proteins, they tend to elicit cross-reactive immune responses. In addition, because the ␣3(XI) chain is genetically identical to the ␣1(II) chain, most antisera to collagen XI also recognize collagen II. Previously, we have immunized rabbits with pepsin-treated collagen XI coupled to keyhole limpet hemocyanin, and, even without affinity purification, the resulting antisera did not react with collagen II. However, a potential cross-reaction with collagen V was not investigated (3). Therefore, rabbits were immunized with a hemocyanin-conjugate of native collagen XI retaining its nontriple helical domains. As demonstrated by ELISA, the crude antiserum from one of several animals specifically recognized chicken collagen XI, but not the types I, II, or V (Fig. 2). The predominant reactivity in immunoblotting was associated with the ␣1(XI) chains (Fig. 1, lane 2). A similar reactivity of collagen XI chains was observed after treatment with pepsin (Fig. 1,  compare lanes 3 and 4).
In Vitro Fibrillogenesis-All aggregation studies described here were conducted with native collagens that had not previously been subjected to limited proteolysis by pepsin. Collagens I or XI were exhaustively dialyzed against 100 mM Tris HCl, pH 7.4, containing 400 mM NaCl (storage buffer). To adjust for identical final concentrations of collagen I, mixtures of the two collagens were appropriately diluted with storage buffer. The concentrations of the stock solutions were determined by a colorimetric assay calibrated with pepsin-treated collagen II at concentrations evaluated by circular dichroism at 221 nm ([⌰] 221 nm ϭ 8550 deg ϫ cm 2 ϫ dmol Ϫ1 (36)). Fibrillogenesis was initiated by dilution with an equal volume of distilled water followed by immediate warming of the reaction mixtures to 37°C. Equilibration of these aggregation conditions was achieved within less than 3 min. The kinetics of fibrillogenesis was monitored by turbidity development at 313 nm. At given intervals, the newly formed aggregates were examined in parallel samples by transmission electron and/or immunoelectron microscopy.
Pure collagen I produced sigmoidal turbidity curves at all concentrations tested, essentially as described previously (24,41). Unlike collagen II, which precipitates from solution (25), collagen I formed homogeneous viscous suspensions allowing an unambiguous determination of final plateau levels of turbidity. Aggregate structures recognizable by electron microscopy were not formed during the initial lag phase of turbidity development (see also Fig. 6A). The length of this lag period decreased, whereas the final plateau level increased with the concentration of collagen I. After completion of fibrillogenesis, the protein was incorporated into large fibrils with a strong D-periodic banding pattern of 67 nm visualized by electron microscopy after negative staining (Fig. 3, A and B). At 90 g/ml or at 200 g/ml collagen I, the fibrils exhibited a broad diameter distribution ranging from 50 to 300 nm (solid bars in Fig. 4, A and B, respectively). An example of a very large fibril characteristically formed by pure collagen I is shown in Fig. 3B. Therefore, we concluded that in vitro fibrillogenesis of collagen I was not subject to a stringent lateral growth control. Instead, fibrils were formed by stochastical nucleation followed by propagation accretion of single collagen molecules, consistently with the model proposed by Silver et al. (42).
Confirming our earlier results (25), collagen XI alone produced turbidity curves without lag periods and with plateau levels linearly depending on concentration (see also Fig. 5). Unlike collagen I, the protein formed long flexible filaments immediately after reduction of the salinity. In the electron microscope, the filaments displayed a uniform width of 26.34 Ϯ 1.9 nm, but no banding pattern (Fig. 6B). They were subsequently compacted into fibrils of a similarly homogeneous width (24.6 Ϯ 1.6 nm, Fig. 3C) and a weak banding pattern (inset in Fig. 3C).
Next, mixtures of collagens I and XI were subjected to in vitro fibrillogenesis. In a given set of experiments, the concentration of either collagen I or XI was kept constant, whereas the concentration of the partner was varied. Thus, the concentrations of collagen I in the reconstitution mixtures ranged between 90 and 200 g/ml and those of collagen XI between 0.16 and 224 g/ml, and molar fractions ƒ I/XI ϭ [collagen I]/[collagen XI] were adjusted to values between 0.4 to 1250 at numerous total collagen concentrations.
The kinetics of fibrillogenesis were monitored by turbidimetry. In the experiments presented in Fig. 5A, the mixtures consistently contained 200 g/ml of collagen I in addition to variable amounts of collagen XI. Kinetically unresolved onsets of turbidity occurred with small amplitudes depending on collagen XI content. This was followed by lag phases whose duration increased in a non-linear fashion with ƒ I/XI (Fig. 5B). Thereafter, large increases in turbidity occurred that eventually approached plateau levels. The final turbidity values were lowered after addition of collagen XI even though total collagen concentrations were increased (Fig. 5A). These results suggested that the basic mechanism of fibril formation in mixtures of collagens I and XI was similar to that followed by pure collagen I. A slow step of nucleation is followed by a more rapid accretion of further collagen monomers. In pure collagen I, however, nucleation is inefficient, which results in extended lag phases, whereas, in mixtures of collagens I and XI, nucleation is greatly facilitated, which entails a strong reduction of the lag periods even when the collagen XI content is less than 1%. In this respect, collagen fibrillogenesis resembles formation of cytoskeletal filaments in which helper proteins can strongly enhance aggregation. For example, the Arp2/3 complex or the protein Scar favor actin filament formation or filament branching (43,44). However, collagen fibrillogenesis differs from the aggregation of actin or intermediate filaments (45) or even microtubules (for review, see Ref. 46), in that collagen fibrils not only grow longitudinally but, to a widely variable degree, also in their lateral dimensions. This necessitates the existence of further control elements in collagen fibrillogenesis that are not intrinsic properties of the major aggregating subunits, already.
A possible mechanism of fibrillogenesis of collagen I/XI mixtures could be as follows: Fibril formation is initiated by preformation of cores of collagen XI aggregates, which subsequently grow by facilitated apposition of collagen I. To test this hypothesis, collagen XI was allowed to aggregate for 120 min. Collagen I was then added at a 7-fold molar excess, and fibrillogenesis was monitored by turbidimetry. Under these conditions, the lag period was longer than in experiments starting with unaggregated mixtures of the two collagens (Fig. 5C, compare "120 min" with "0 min"). Thus, preformed collagen XI fibrils were poorer nuclei of fibrillogenesis than those formed in fresh mixtures. This result is inconsistent with the notion that collagen XI fibrils act as nucleation sites but, in fact, is best explained by the low propensity of collagen XI fibrils to accrete collagen I. Still, the lag phase was shorter than in aggregation of pure collagen I (Fig. 5C). This may well be due to residual monomeric (or oligomeric) collagen XI, which, upon addition of a large excess of collagen I, triggers heterotypic fibrillogenesis. To follow up on this idea, we allowed collagen XI assembly to proceed for 5 min only. At this stage, fully aggregated collagen XI fibrils presumably still are scarce or inexistent, whereas extensive formation of collagen XI nuclei already had occurred. A 7-fold excess of collagen I was then added to the aggregation mixtures. Unexpectedly, this resulted in a further lengthening of the lag period well beyond that observed for pure collagen I (Fig. 5C, compare "collagen I alone" with "5 min"). It is possible, therefore, that collagen XI nuclei formed under these conditions retain their capability to incorporate collagen I. However, such collagen I/XI composites apparently are incompetent to grow further by accretion of collagen I. For reliable interpretation, these phenomena obviously require further investigation, which will be presented elsewhere. However, they clearly dismiss the concept of collagen XI aggregates nucleating heterotypic collagen I/XI fibrils. Instead, effective nuclei of fibrillogenesis already must comprise both collagens.
To further elucidate the distinctions between nuclei formed by pure collagens I and XI, or mixtures thereof, we examined by electron microscopy aggregates formed during early phases of fibrillogenesis. After 5 min, pure collagen XI produced long and flexible filaments (Fig. 6B), and there was only scant, if any, evidence of aggregation of pure collagen I (Fig. 6A). By contrast, stubby aggregates were specifically formed by mixtures of collagens I and XI (Fig. 6, C and D) even when the content of collagen XI was very low. Interestingly, their exact dimensions depended on ƒ I/XI , in that their width increased from 23.24 Ϯ 1.75 nm at ƒ I/XI ϭ 0.4 to 45.24 Ϯ 4.95 nm at ƒ I/XI ϭ 7 (90 g/ml of collagen I). Their length was reduced at high ƒ I/XI values but could not accurately be determined, because the ends were poorly defined in the electron micrographs (Fig. 6, C and D). Nevertheless, these observations corroborated that specific aggregates containing both collagen types were formed already during early phases of fibrillogenesis. The stubby aggregates had no counterpart in the assembly of collagens II, IX, and XI, which formed flexible and loosely organized filaments during early phases of fibrillogenesis (47). In conclusion, the final fibrillar organization is specified by alloyed nuclei during the most initial phases of fibrillogenesis.
The morphology of the fibrillar aggregates emerging after completion of turbidity were examined by electron microscopy. In all mixtures of collagens I and XI, two distinct fibril populations arose. Large fibrils with a strong D-periodic banding pattern co-existed with thin weakly banded fibrils (Fig. 3D) that were absent in the reconstitution products of collagen I alone (Fig. 3, A and B). Diameters were measured in transmission electron micrographs of negatively stained fibrils, and the number of fibrils falling into distinct diameter classes was counted. For the following determination of mass-weighted fibril frequencies, constant density was assumed for all fibrils, and, because fibrils were much longer than the observable grid sizes, their lengths were also considered as constant. Thus, neither parameter was further taken into account. The numbers of the fibrils per diameter classes were multiplied with the square of the mean diameter within the classes, and the products were divided by the sum of all products in the experiment. Representative histograms of mass-weighted diameter distributions of fibrils formed by mixtures with constant collagen I concentrations are shown in Fig. 4 (A (90 g/ml) and B (200 g/ml)). In comparison with pure collagen I (solid bars), the diameter distribution of fibrils formed by the mixtures were bimodal. A first single peak represented the small fibrils with a weak banding pattern, and a second peak with a broad diameter distribution corresponded to large, strongly banded fibrils. Although absolute values varied somewhat with the total collagen concentrations, the mean diameter of the fibrils and the distribution width inversely depended on ƒ I/XI . The main peak occurring at diameters between 80 and 100 nm at a collagen XI concentration of 13 g/ml and a molar proportion of ƒ I/XI ϭ 7 (Fig. 4A, gray bars) was gradually changed to values of 60 -80 nm when the collagen XI concentration was 53 g/ml and f I/XI ϭ 1.7 (Fig. 4A, open bars). When the concentration of collagen I was raised to 200 g/ml, addition of collagen XI also led to a reduction of the mean fibril diameters (Fig. 4B). We concluded that heterotypic collagen I/XI fibrils were formed and their lateral growth was progressively reduced by the amount of collagen XI. However, unlike mixtures of collagens II The concentration of collagen I was 200 g/ml. For comparison, the turbidity development of collagen XI alone (50 g/ml) is also shown. B, times of half-maximal turbidity development as a function of the logarithm of ƒ I/XI . The concentration of collagen I was 200 g/ml. C, time course of sequential fibrillogenesis of collagens XI and I. Preliminary formation of collagen XI aggregates was allowed to proceed for 0, 5, or 120 min before collagen I was added to the mixtures. Only turbidity development after addition of collagen I is shown. and XI, fibril shape changes occurred continuously rather than at discrete values of ƒ I/XI , and an exclusive formation of a uniform population of prototypic 20-nm fibrils (25) was not observed at any ƒ I/XI value.
To substantiate the notion that collagens I and XI formed heterotypic fibrils, immunoelectron microscopy was performed using collagen type-specific antibodies. Our antiserum to collagen XI predominantly labeled the thin but not the thick and banded fibrils (Fig. 7A), and polyclonal antibodies to collagen I produced the opposite labeling pattern (Fig. 7B). However, these differential labeling patterns may have resulted from immunochemical masking, particularly of collagen XI epitopes in the thick, strongly banded fibrils. To exclude this possibility, we examined in greater detail the labeling of the tapering or, occasionally, damaged ends of the fibrils, because labeling with antibodies to collagen XI was restricted to analogous regions of authentic cartilage fibrils (3). Indeed, the tapering ends of large fibrils displayed an immunoreaction not only with antibodies to collagen I (Fig. 7D) but also with antibodies to collagen XI (Fig.  7C). We concluded that mixtures of collagens I and XI formed two types of fibrils, i.e. a homogeneous population of thin, weakly banded fibrils and a heterogeneous mixture of strongly banded fibrils of variable diameter. The thin fibrils were collagen XI fibrils devoid of collagen I, but large diameter fibrils were heterotypic. However, large fibrils were composites with heterotypic cores and a coat of collagen I. In the central portions of the fibrils, the collagen XI epitopes became gradually masked by the jacket of collagen I. DISCUSSION Collagen I is the most abundant structural macromolecule in vertebrates. It is the quantitatively major fibril constituent in virtually all interstitial extracellular matrices and is the prototype member of the collagen protein family. It is surprising, therefore, that the mechanisms of collagen I assembly into functional aggregates still are poorly understood. The molecular organizations of collagen I-containing fibrils vary in a tissue-specific manner, but the causes of this diversity still are mostly enigmatic. Here, we have provided further evidence that fibrillar organizations are decisively determined by quantitatively minor components that form discrete macromolecular composites with more abundant collagens. In the paradigm studied, collagens I and XI can aggregate into composite fibrils following a mechanism of nucleation and propagation, essentially as proposed by Oosawa and Asakura (48) for other polymerizing protein systems. When mixed with collagen I, a part of collagen XI forms homotypic aggregates resembling prototypic cartilage fibrils with a uniform diameter of about 20 nm and a weak banding pattern. In addition, collagens I and XI can rapidly co-assemble into stubby nuclei of fibril formation. Such stubs are generated by neither protein alone but uniquely by their mixtures. Their precise shapes depend on the molar fractions of the two collagen types, but they are homogeneous under any given condition of aggregation. Thus, collagen I/XI nuclei are macromolecular alloys that determine the organization of collagen I/XI fibrils growing by apposition of collagen I but no further collagen XI. Therefore, collagen I/XI fibrils are suprastructural composites with alloyed core regions and are distinct from the fibrillar alloys formed by collagens II and XI at ƒ II/XI Ͻ 8 (25). This is surprising, because collagens I and II are highly homologous proteins (6). At all molar proportions The primary antibodies were: A, rabbit antiserum to collagen XI (see Figs. 1 and 2); B, affinity-purified polyclonal rabbit antibodies to collagen I. Note: collagen XI labeling predominantly occurred in thin but not wide fibrils with strong banding patterns. Tapering ends of the large fibrils were labeled with antibodies to collagen XI (C) and collagen I (D, only the thin end of a banded fibril is shown at high magnification). Bars, 200 nm. and total collagen concentrations examined, collagens I and XI did not give rise to the uniform, weakly banded fibrils of a width of 20 nm resembling cartilage fibrils containing collagens II and XI. Collagen I/XI fibrils gradually became thinner as their content of collagen XI increased. In addition, the mixtures formed cartilage-like 20-nm fibrils that, however, only contained collagen XI. This readily explains the occurrence of discrete fibril populations in fibrocartilage tissues, such as the annulus fibrosus of intervertebral discs or the menisci in knee joints. In these tissues, collagens I, II, and XI co-polymerize to yield large banded fibrils affording resistance to very high tensile loads. Another population of fibrils similar to those of immature joint cartilage is generated in fibrocartilage by the same proteins, if they occur at different molar proportions. The concept of tissue-specific composites also may apply to bone. In this tissue, minor collagen molecules containing ␣1(XI) chains (26) are likely to contribute to the nucleation of large diameter fibrils. In this case, the overall structure of the fibrillar alloy is controlled by co-aggregation of collagens I, V/XI, and very small amounts of collagen III (49). The structural properties of these fibrils are adapted to the requirements of osseous tissues, including binding of bone-specific protein mediators of mineralization and deposition of calcium hydroxyapatite mineral into the gap regions.
The concept of inappropriate composite formation by distorted mixtures of extracellular matrix macromolecules also may help to explain several puzzling details of tissue alterations seen in transgenic animals or patients with hereditary connective tissue diseases. Patients with a normal and a nullallele COL1A1 usually survive into adulthood (50) but have mild forms of osteogenesis imperfecta (OI, type I). It is generally thought that haploinsufficiency leads to inadequate quantities of fibrils due to reduced collagen I synthesis. However, fibrils in connective tissues of these patients had abnormal shapes, too. Taking into account earlier reports on fibrillogenesis by mixtures of collagens I and V (2) and the results presented here, we propose that formation of inappropriate collagen composites and, hence, abnormal fibrillar organization underlie the functional deficiencies of the OI tissues. Interestingly, the absence of ␣2(I) chains in a patient with homozygous null-allele COL1A2 leads to a more severe form of OI, i.e. type III (51,52). Nevertheless, the condition is compatible with survival well beyond birth, because the patient's tissues contained a variant collagen I with three ␣1(I) chains, which has a molecular stability similar to that of the wild type protein.
However, fibrillogenesis of this collagen I is abnormal. Again, it is plausible that the major consequence of the absence of ␣2(I) chains is a structural fibril anomaly caused by formation of aberrant composites containing ␣1(I) trimers.
Another pertinent case is the collagen II-deficient mouse with inactivated Col2a1 alleles (53,54). Cartilages of these animals not only lack collagen II but also normal collagen XI, because that protein contains ␣3(XI) chains, i.e. another Col2a1 gene product. Homozygous fetuses do not survive until birth, because their overabundant, but functionally poor cartilage tissues are almost completely devoid of fibrils. This is paradoxical, because, in a frustrated attempt at functional compensation, considerable quantities of seemingly normal collagen I are produced in place of collagen II in the mutant cartilages. As shown here, mixtures of collagens I and XI can efficiently form fibrils, albeit with an ultrastructure deviating from that of normal cartilage fibrils. Thus, we reason that the almost complete absence of fibrils in collagen II-deficient mice results from a deficiency of collagen XI rather than of collagen II.
Matrix suprastructures other than collagen fibrils also con-tain more than one molecular constituent. Thus, other extracellular matrix aggregates, including basement membranes and microfibrils containing fibrillin (55) or collagen VI (56,57), are likely to be macromolecular alloys as well. It is even tempting to speculate that the concept applies to biological aggregates in general and, thus, may help to explain tissue-specific functions. It will be challenging to investigate from this point of view biological suprastructures such as chromatin, transcriptional machineries, the various forms of the cytoskeleton, ribosomes, proteasomes, multienzyme complexes, or lipid rafts.