Does the Triple Helical Domain of Type I Collagen Encode Molecular Recognition and Fiber Assembly while Telopeptides Serve as Catalytic Domains?

Over the last several decades, it has been established that proteolytic removal of short, non-helical terminal peptides (telopeptides) from type I collagen significantly alters the kinetics of in vitro fibrillogenesis. However, it has also been observed that the protein is still capable of forming fibers even after complete removal of telopeptides. This study focuses on the characterization of this fibrillogenesis competency of collagen. We have combined traditional kinetic and thermodynamic assays of fibrillogenesis efficacy with direct measurements of interaction between collagen molecules in fibers by osmotic stress and x-ray diffraction. We found that telopeptide cleavage by pepsin or by up to 20 h of Pronase treatment altered fiber assembly kinetics, but the same fraction of the protein still assembled into fibers. Small-angle x-ray diffraction showed that these fibers have normal, native-like D-stagger. Force measurements indicated that collagen-collagen interactions in fibers were not affected by either pepsin or Pronase treatment. In contrast, prolonged (>20 h) Pronase treatment resulted in cleavage of the triple helical domain as indicated by SDS-polyacrylamide gel electrophoresis. The triple-helix cleavage correlated with the observed decrease in the fraction of protein capable of forming fibers and with the measured loss of attraction between helices in fibers. These data suggest that telopeptides play a catalytic role, whereas the information necessary for proper molecular recognition and fiber assembly is encoded in the triple helical domain of collagen.

X-ray diffraction patterns from such reconstituted fibers and from native fibers are similar as well (4 -7). It is commonly believed that type I collagen contains all structural information that is necessary for its self-assembly into fibers, except maybe for some tissue-specific factors (see, for example, Ref. 8). However, the location of the "coding" regions, the nature of this information, and how it is "translated" into intermolecular forces responsible for fibrillogenesis are still poorly understood.
One of the debated issues is the role of telopeptides. Telopeptides form covalent cross-links with triple helical regions on opposing molecules (see, for example, Refs. 9 and 10). It is believed that this occurs after completion of fibrillogenesis and that the cross-links stabilize rather than create appropriate molecular arrangement. It was also suggested that telopeptides are important at earlier stages, i.e. in the process of fibrillogenesis. This hypothesis is based primarily on studies of enzymatically treated collagen. Specifically, partial or complete removal of telopeptides alters the kinetics of collagen fiber formation (11,12). It may also affect fiber morphology, as indicated by electron microscopy (11,(13)(14)(15)(16). Additional evidence for possible telopeptide involvement at early stages of fiber assembly comes from inhibition of intact collagen fibrillogenesis by synthetic peptides that have sequences found in carboxyl-terminal telopeptides (17).
These data clearly establish that telopeptides play some role in fibrillogenesis kinetics. Do they mean, however, that telopeptides contain a code necessary for recognition and appropriate packing of collagen into fibers? This work revisits the question. It builds on recent advances in direct measurement of intermolecular forces by osmotic stress and x-ray diffraction (18 -21). We apply this new technique to measurement of forces between collagen molecules in fibers and combine it with traditional in vitro fibrillogenesis assays. We compare acid-soluble collagen (AcCol) 1 with pepsin-treated collagen (PepCol) and with Pronase-treated collagen (PronCol). We find that marked changes in collagen-collagen interactions and in fibrillogenesis competency occur only after prolonged Pronase treatment. These changes coincide with cleavage of the triple helix rather than with telopeptide removal.

MATERIALS AND METHODS
Collagen Preparation-Type I collagen was extracted from rat tail tendon as described previously (20 -22). Briefly, frozen tails of young rats were purchased from Pel-Freez Biologicals and stored frozen at Ϫ20°C. Tails were thawed in cold (4°C) protease-inhibiting buffer (3.5 * 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 U.S.C. Section 1734 solely to indicate this fact. M NaCl, 10 mM Tris, 20 mM EDTA, 2 mM N-ethylmaleimide, and 1 mM phenylmethylsulfonyl fluoride (pH 7.5)). Tendons were excised from tails, washed in the same buffer for several days at 4°C, and then dissolved in 0.5 M acetic acid (pH 2.8). Solubilized tendons were digested by pepsin (Calbiochem) as described (23) or by Pronase (Calbiochem) as described (13). Pepsin was added directly to the solution of collagen in acetic acid at a ratio of 100 mg of pepsin/1 g of tendons in two doses for 24 h at 4°C each. For Pronase treatment, solubilized tendons were dialyzed against 0.1 M calcium acetate (pH 7.0). Pronase was added to the substrate at a ratio of 1:100 and stirred at 23°C for 68 h. Aliquots of digested collagen were taken at 20, 44, and 68 h after the reaction was started. The reaction was stopped by addition of an equal volume of 0.5 M acetic acid.
AcCol, PepCol, and PronCol were purified by three cycles of salt precipitation and acetic acid resolubilization (23) and stored in 0.5 M acetic acid at 4°C. Samples from each preparation were characterized by SDS-polyacrylamide gel electrophoresis (PAGE) (3% stacking gel and 6% separating gel, stained by Coomassie Blue R-250).
Collagen concentration in solutions was measured by Sircol assay (Accurate Chemical & Scientific Corp.) and/or by optical absorbance in the 215-230 nm region. Both assays were calibrated using a set of standard collagen solutions of different concentrations.
Tyrosine content of collagen was estimated from the optical absorbance at 275.5 nm of collagen denatured in 6 M guanidine hydrochloride following the procedure proposed in Ref. 24. For type I collagen, it was shown previously that this spectroscopic method agrees with direct amino acid analysis within ϳ10% (25,26).
Fibrillogenesis-The kinetics of fiber formation and equilibrium solubility of collagen at 32°C were measured as described (22). Briefly, collagen was dialyzed against 2 mM HCl (pH 2.7). Aliquots from dialyzed collagen solutions (0.5-1.5 mg/ml) were mixed 1:1 on ice with 2ϫ initiation buffer (20 mM sodium phosphate and 0.26 M NaCl with pH adjusted to give pH 7.4 in the mixture). The mixture was degassed for 5 min under vacuum and immediately placed into a Jasco V 560 spectrophotometer, where it was maintained at 32°C. Fibrillogenesis kinetics was monitored by recording the optical density at 450 nm as a function of time. When no further change in the optical density was detected, the precipitate of assembled collagen fibers was spun down from the mixture by centrifugation at 14,000 ϫ g for 5 min. Collagen solubility at 32°C or the percentage of collagen competent to form fibers was evaluated from the protein concentration in the supernatant. The supernatant and precipitate were characterized by SDS-PAGE.
Sample Preparation for Measurement of Intermolecular Forces-Native fibers from rat tail tendons as well as fibers reconstituted from solutions of AcCol, PepCol, and PronCol were used. Native fibers were excised from rat tails, stored in the protease inhibitor buffer as described above, and then directly used for force measurement. Alternatively, fibers were transferred from the protease inhibitor buffer into 0.1 M sodium phosphate and 2 M glycerol (pH 7.5), in which they were washed for several days at 4°C and after that used for force measurement. We refer to the latter fibers as "washed native." Reconstituted collagen fibers were made by slow concentration of AcCol, PepCol, and PronCol in 0.5 M acetic acid. Each solution was dialyzed at 4°C in a Pierce Model 500 microdialyzer system until a solid protein film was formed. Dialysis was performed against 50% polyethylene glycol (PEG; average M r 8000; U. S. Biochemical Corp.) solution in 0.5 M acetic acid. The film was further equilibrated in 40 -50 weight % solution of PEG 8000 in 10 mM Tris and 2 mM EDTA (pH 7.5) for 2-3 days at 4°C, washed in the same buffer to remove PEG, and air-dried at 4°C. It was shown previously that such reconstituted films, prepared either from AcCol or from PepCol, consist of densely packed, native-like collagen fibers (20). The films were cut into small pieces (ϳ1 ϫ 0.5 ϫ 0.5 mm) and prehydrated in 10 mM Tris and 2 mM EDTA (pH 7.5) at 4°C for at least 1 day to ensure more reproducible measurements of interaxial distances. However, this procedure resulted in the loss of integrity and, apparently, partial solubilization of films made from PronCol treated with the enzyme for 44 and 68 h. Thus, we had to skip the prehydration step for the latter samples.
Interaction between Collagen Helices in Hydrated Fibers-Forces between helices in fibers were measured as a function of interaxial distance by the osmotic stress technique and x-ray diffraction (18 -21). Native fibers or pieces of reconstituted films were equilibrated for at least 1 week in gravimetrically prepared PEG 8000 solutions (2-50 weight % PEG in 10 mM Tris and 2 mM EDTA (pH 7.5)) at 5, 20, or 35°C in tightly sealed 1.5-ml microtubes fitted with O-rings. The solutions were refreshed in the middle of the equilibration (after the first 2-3 days).
Each equilibrated sample was sealed in a specially designed cell with a small amount of the solution and placed into an FR590 x-ray diffractometer (Enraf Nonius), where it was maintained at the corresponding temperature during the measurement. This diffractometer was optimized for mid-angle x-ray diffraction (⌬q ϳ 0.05-Å Ϫ1 resolution in reciprocal space). Details of the design of the x-ray equipment were as described (20 -22, 27). The lowest order Bragg spacing (d Br ) for lateral packing of the helices was measured, and interaxial distance (d int ) was calculated from it in the approximation of hexagonal packing (d int ϭ 2d Br /͌3). PEG, because of its large size, does not penetrate inside collagen fibers, whereas water inside fibers freely exchanges with surrounding solution (20,21,28). As a result, PEG osmotically compresses fibers. This action of PEG is counteracted by interaction between collagen helices in fibers that is responsible for fiber swelling. From thermodynamic analysis of this force balance, it was shown that the corresponding force between helices (f) per unit of their length is given by (18) f ϭ ⌸ PEG d int /͌3, provided that lateral packing of collagen can be approximated by a hexagonal lattice and interactions between collagen helices are pairwise additive.
Small-angle X-ray Diffraction-After force measurement, at least three samples of each collagen were reequilibrated in 6% PEG in 10 mM Tris and 2 mM EDTA (pH 7.5) and sealed in the same x-ray cell as for force measurement. The samples were exposed overnight in an Elliot GX-13 x-ray diffractometer equipped with a 100-m focusing cup, a multilayer x-ray lens (45-cm focal length; Osmic), and a single set of slits placed immediately after the lens. Diffraction patterns were captured on CRST-VN image plates (Fuji) and read using a BAS2500 image plate scanner (Fuji). The sample-to-plate distance was ϳ40 cm. The size of the x-ray beam at the focal spot on the plate was ϳ200 ϫ 320 m. The resolution was estimated as ⌬q ϳ 0.003 Å Ϫ1 . The lowest measurable scattering vector q was ϳ0.04 -0.05 Å Ϫ1 . This could not be improved due to limitations associated with the design of the x-ray lens. However, such resolution was sufficient for the purpose of this work.

RESULTS
Enzymatic Treatment of Collagen-We evaluated integrity of collagen molecules by SDS-PAGE and by estimating the number of tyrosine residues/molecule from UV absorption spectra. We assumed that rat tail tendon collagen is similar to mouse type I collagen, whose complete primary sequence is known (29,30). Based on the mouse sequence (␣1(I) chain: CA11_MOUSE, Swiss-Prot accession number P11087; and ␣2(I) chain: CA21_MOUSE, Swiss-Prot accession number Q01149), one expects an intact molecule to contain 14 tyrosines: 12 in telopeptides and 2 in the triple helical region of the ␣2(I) chain.
Judging from SDS-PAGE, AcCol that did not undergo any enzymatic treatment contained ␣1(I) and ␣2(I) chains along with higher molecular weight complexes (Fig. 1, lane 1). The complexes are due to covalent cross-links between two telopeptide chains of the same molecule or between a telopeptide on one molecule and a helical domain on another molecule (9, 10). From UV absorption at 275 nm, the number of tyrosine residues in AcCol was estimated as ϳ10 (Table I). Apparently, some AcCol molecules had damaged telopeptides.
The pepsin treatment of collagen removed higher molecular weight complexes and reduced the molecular weight of ␣1(I) and ␣2(I) chains (Fig. 1, compare lanes 1 and 2). It is believed that this is due to partial cleavage of non-helical telopeptides (13,14,31). The observed decrease in the number of tyrosines/ molecule from ϳ10 in AcCol to ϳ5 in PepCol supports this interpretation, as previously shown (26).
After 20 h of Pronase treatment (20-h PronCol), the major ␣1 and ␣2 bands on SDS-PAGE were similar to those in PepCol (Fig. 1, compare lanes 2 and 3). From UV absorption, we found ϳ2 tyrosines/20-h PronCol molecule. Most likely, these are the tyrosines located within the triple helical domain of the ␣2 chain. Apparently, all telopeptide tyrosines are removed by 20 h of Pronase treatment, including those at the interface between the triple helix and carboxyl-terminal telopeptides. This indicates complete telopeptide digestion, in agreement with previous reports (12,13,32).
In contrast to pepsin, Pronase treatment led to two addi-tional bands that appeared on the gel close to the ␣1 and ␣2 bands. Evidently, Pronase not only digests telopeptides, but also cleaves collagen at some specific site that lies within the triple helical domain. These bands must be products of the cleavage of the ␣1 and ␣2 chains, and we designated the bands as ␣1Ј and ␣2Ј, respectively. This is consistent with earlier observations that Pronase slowly cleaves collagen even after complete digestion of telopeptides (13,32). Longer Pronase digestion (44 h (Fig. 1, lane 5) and 68 h (lane 7)) led to intensification of the ␣1Ј and ␣2Ј bands and to the appearance of two more bands labeled as ␣1Љ and ␣2Љ. The latter bands most likely result from cleavage of the ␣1 and ␣2 chains at yet another site within the triple helical domain. The tyrosine content of collagen was further reduced to Ͻ2/molecule (Table I), apparently because of tyrosine removal together with a chunk of collagen triple helix at the carboxyl-terminal end.
The kinetics of AcCol, PepCol, and PronCol self-assembly are shown in Fig. 2a (at 0.3-0.8 mg/ml collagen and 10 mM phos-phate in the fibrillogenesis buffer). We used higher collagen and lower phosphate concentrations than in traditional protocols (3,11,12) to accelerate fiber nucleation. (Note that phosphate is a potent fibrillogenesis inhibitor (3,22).) Even at these conditions, we did not achieve equilibrium for 44-and 68-h PronCol samples.
We found that Pronase treatment substantially increased the fibrillogenesis lag time (nucleation time) compared with AcCol and PepCol, consistent with previously reported data (11,12). However, we also found that the lag time changed drastically only after prolonged (Ͼ20 h) Pronase treatment, i.e. only when digestion of triple helical domains became pronounced.
Fibrillogenesis Competency of Enzymatically Treated Collagen-After each kinetic experiment (24 h), we separated assembled fibers and the soluble fraction by centrifugation and measured collagen concentration in the soluble fraction. A significant amount of collagen may be present in this fraction either because collagen becomes assembly-incompetent (as a result of enzymatic cleavage) or because fibrillogenesis is incomplete even after 24 h (as a result of very slow kinetics). We found that virtually all AcCol and PepCol assembled into fibers, with only trace amounts remaining in the soluble fraction. Thus, the equilibrium solubility of assembly-competent collagen is negligibly small.
After 20 h of Pronase treatment, the kinetics was still sufficiently fast (Fig. 2a) so that fiber assembly was complete at the time of fraction separation and concentration measurement (24 h). Thus, in equilibrium, ϳ80% of 20-h PronCol formed fibers, whereas the remaining protein stayed in solution. As indicated by SDS-PAGE, 20-h PronCol contained some molecules with cleaved triple helical regions (Fig. 1, lane 3). The soluble fraction consisted virtually only of damaged molecules that had the cleaved ␣1Ј and ␣2Ј chains instead of the normal ␣1 and ␣2 chains (Fig. 1, lane 4).
The fraction of assembly-competent collagen decreased with the length of Pronase treatment as shown in Fig. 2b. This coincided with enhancement of the ␣1Ј and ␣2Ј bands and the appearance of the ␣1Љ and ␣2Љ bands on SDS-PAGE of PronCol (Fig. 1, lanes 5 and 7).
Long Pronase treatment (44 and 68 h) resulted in extremely slow fibrillogenesis kinetics so that the process was not complete after 24 h. Thus, although the soluble protein fraction contained primarily ␣1Ј, ␣1Љ, ␣2Ј, and ␣2Љ chains, it also had some intact ␣1 and ␣2 chains (Fig. 1, lanes 6 and 8).
Collagen Packing in Reconstituted Films-To assess changes in structure and interaction between collagen molecules in fibers, we prepared reconstituted protein films as described under "Materials and Methods." The structure of these films can be determined by comparing their small-angle x-ray diffraction patterns with a similar pattern from native rat tail tendons, as shown in Fig. 3.
In native tendon fibers, a characteristic set of multiple reflections can be clearly seen (Fig. 3, curve a). These reflections are higher orders of diffraction arising from d ϭ 670-Å axial periodicity in collagen fibers. The positions of maxima in the q-space are given by q(n) ϭ 2n/d, where n is the order of the diffraction and q is the scattering vector. Note that the 5th, 9th, 12th, 20th, and 21st orders are significantly stronger than their neighboring peaks. This pattern is a signature of the native D-stagger of collagen molecules (4, 6, 7, 33).
The small-angle diffraction pattern from reconstituted AcCol films (Fig. 3, curve b) is very similar to the pattern from native fibers in terms of both positions and relative intensities of the diffraction peaks, consistent with what was reported previously (4, 7). However, the ordering of collagen molecules in reconsti-  a The values were obtained using a 1500 M Ϫ1 cm Ϫ1 extinction coefficient for Tyr at 275.5 nm, as described (24). For type I collagen, this spectroscopic method agrees with direct amino acid analysis within ϳ10% (25,26). In our case, the most likely source of error was base-line subtraction. For AcCol, PepCol, and 20-h PronCol, the uncertainty in base-line subtraction did not exceed 20%. For 44-and 68-h PronCol, the base-line subtraction was unreliable, and only upper bounds for the tyrosine content could be estimated. tuted films was not as good as in native tendons so that the 5th order (that was close to the low-angle limit of our camera) was overwhelmed by central scattering originating from imperfections of the x-ray lens, and it could not be reliably resolved. Still, well resolved 9th and 12th diffraction orders and the data reported previously (4, 7) allow us to conclude that films reconstituted from AcCol consist of fibrils that have the same basic structure as fibrils in native tail tendons.
The diffraction patterns from PepCol and 20-h PronCol films (virtually indistinguishable from each other) contain diffraction peaks at the same positions as in native tendons and AcCol films (Fig. 3, curves c and d). Moreover, as in native tendons, the 9th, 12th, 20th, and 21st diffraction peaks are stronger than their neighboring peaks. However, unlike native tendons, the 6th order in PepCol and 20-h PronCol becomes very strong. From this, we can conclude that although packing of collagen molecules remains the same as in native fibrils, the electron density profile along the fibril axis changes. Most likely this is a result of telopeptide removal from PepCol and 20-h PronCol molecules and consequent widening of gap regions in fibrils formed by them.
We could not detect any low-angle diffraction peaks from 44and 68-h PronCol (Fig. 3, curves e and f). The small bump at the position of the 6th order in 44-h PronCol may, in principle, be a residual 6th order peak. However, because of its small amplitude and closeness to the low-angle limit of our setup, a reliable assignment could not be made. In principle, intensity of diffraction peaks depends on the degree of molecular ordering inside films and on the size of ordered domains. Their absence suggests that 44-and 68-h PronCol films are much less ordered than PepCol and 20-h PronCol films and is an indication that the native-like D-stagger may be lost after prolonged Pronase treatment. Fig. 4 for native and reconstituted fibers. Note that native tendon fibers contain collagen and other tissue-specific components. We found that an overnight wash in 0.1 M phosphate and 2 M glycerol (pH 7.5) resulted in the loss of tissue-specific crystal-like lateral packing of collagen (observed in tendons from some (but not all) species and not observed in skin and other tissues (5, 6)). The collagen-collagen forces in washed native fibers were identical to forces in reconstituted AcCol, PepCol, and 20-h PronCol fibers (Fig. 4a).

Effect of Protease Treatment on Helix-Helix Interaction in Fibers-The net force between collagen helices is plotted versus average interaxial separation (d int ) in
We found previously that the net force between collagen helices in fibers is a sum of repulsive and attractive interactions (20,21). The balance between the repulsion and attraction determines the equilibrium separation between molecules. The repulsion dominates at shorter and the attraction dominates at larger interaxial separations. The attraction is responsible for molecular recognition between collagen helices and for their spontaneous assembly from solution into native-like fibers. The repulsion is shown in Fig. 4b by the dashed straight line. The attraction leads to a downward deviation of the net force curve from this straight line. The attraction increases with increasing temperature (Fig. 4b), resulting in temperature-favored fiber assembly.
Note that the helix-helix forces in washed native fibers, in reconstituted PepCol fibers, and in reconstituted 20-h PronCol fibers were virtually identical at all temperatures (Fig. 4b).
Even complete telopeptide removal had no visible effect on helix-helix interaction.
Longer Pronase treatment (44 and 68 h) strongly suppressed the attractive component of the net force so that the downward deviation from pure helix-helix repulsion was much weaker FIG. 2. a, fibrillogenesis kinetics measured by light absorbance. Curve 1, 0.35 mg/ml AcCol; curve 2, 0.25 mg/ml PepCol; curve 3, 0.25 mg/ml 20-h PronCol; curve 4, 0.5 mg/ml 44-h PronCol; curve 5, 0.8 mg/ml 68-h PronCol. The inset shows initial stages of the process up to 60 min. The optical density is normalized to give the ratio of collagen in the precipitate to total collagen after the end of each kinetic experiment. (The absolute value of the optical density is poorly reproducible and may be misleading since it depends on the size of forming aggregates.) b, effect of the length of Pronase treatment on the fraction of collagen that forms fibers at 32°C. Each point corresponds to the ratio of collagen in the precipitate to total collagen measured after the end of the corresponding kinetic experiment shown in a. The error bars represent estimated errors of collagen concentration measurements (see "Materials and Methods").

FIG. 3. Small-angle x-ray diffraction patterns from native collagen fibers (curve a), reconstituted AcCol fibers (curve b), Pep-Col (curve c), 20-h PronCol (curve d), 44-h PronCol (curve e), and 68-h PronCol (curve f).
For clarity, the patterns are shown after subtraction of background scattering, fitted separately for each sample by a stretched exponential curve. No data filtering or noise reduction procedures were used. The number of the diffraction order corresponding to each relatively strong peak is shown at the top of the graph. (Fig. 5a). To verify that this is the suppression of the attractive component rather than an enhancement of the repulsive component of the net force, we used glycerol, which is a specific attraction inhibitor (22). We found that 2 M glycerol reduced the attraction between PepCol helices, but it had almost no effect on interaction between 44-h PronCol helices (Fig. 5b). In other words, the observed collagen-collagen attraction is sensitive to cleavage of the triple helical domain, but not to telopeptide removal.

Effect of Telopeptides on Kinetics of Fiber Nucleation and
Growth-Several mechanisms of collagen fibrillogenesis that address the role of different parts of the collagen molecule and the effect of the environment (pH, ionic strength, specific ions, and various additives) were proposed in the literature (see, for example, Ref. 8). It is now commonly accepted that fibrillogenesis involves at least two distinct steps: nucleation and fiber growth. It was suggested that fibril nuclei may exist as metastable intermediates that have a long life-time at low temperature. By analyzing fibrillogenesis kinetics during heating/ cooling cycles, it was shown that collagen solution has a long "thermal memory." Repeated heating results in fibrillogenesis with a considerably shorter lag time. Proteolytic cleavage of telopeptides abolishes the thermal memory (34). It also leads to an increase in the fibrillogenesis nucleation time (11,16). The larger the fraction of the amino-terminal telopeptides that is removed, the longer the nucleation delay (12) (see also Fig. 2a). Hence, it was postulated that amino-terminal telopeptides accelerate fiber nucleation.
Evidence was also reported that removal of carboxyl-terminal telopeptides from collagen slows down fibril growth (12). Consistently, synthetic peptides containing amino acids found in carboxyl-terminal telopeptides were shown to inhibit fibril assembly when added to a solution of intact collagen, apparently by competing with carboxyl-terminal telopeptides for binding sites on collagen (17). Thus, it was suggested that carboxyl-terminal telopeptides are essential for the fibril growth step.
Effect of Telopeptides on Fiber-forming Competency of Collagen, on Fiber Structure, and on Fiber Stability-In this work, we found that even complete removal of telopeptides by 20 h of Pronase treatment had no significant effect on the fiber-forming competency of collagen. Although fibrillogenesis kinetics was altered (Fig. 2a), in agreement with previous reports, virtually all molecules with intact triple helical domains assembled into fibers as if the telopeptides were present (Figs. 1  and 2b). Only those molecules whose triple helical domains were cleaved by 44 or 68 h of Pronase treatment lost their ability to assemble into fibers.
Small-angle x-ray diffraction showed that inside reconstituted collagen films, AcCol, PepCol, and 20-h PronCol molecules organized into fibers whose D-staggered packing was identical to that of native fibers (Fig. 3). The films were reconstituted under acidic conditions (see "Methods and Methods") that suppress collagen-collagen recognition. To ensure their stability, the films had to be equilibrated at neutral pH under osmotic stress for at least several days. Apparently, collagen slowly reorganizes into native-like fibers during this equilibra- 5. a, effect of the length of Pronase treatment on force per unit of molecular length between collagen helices at 20°C; b, effect of 2 molal glycerol (2m gly; the specific inhibitor of collagen-collagen attraction) on measured interactions between enzymatically treated collagen molecules. As described in the legend to Fig. 4, the solid lines are drawn only to guide the eye.
FIG. 4. Force per unit of molecular length between collagen helices measured in native and reconstituted fibers at different temperatures. Note that the "dry" collagen diameter is ϳ12 Å so that the surface-to-surface separation can be obtained from the interaxial spacing by subtracting 12 Å. The straight dashed line in b shows the repulsive component of collagen-collagen interaction, measured as described (21). The solid lines are drawn to guide the eye; they are not a fit to any theory. tion process. Somewhat lower ordering of AcCol in these films (compared with PepCol and 20-h PronCol) is probably related to re-formation of covalent cross-links that hinder this molecular rearrangement. Even without telopeptides, the molecules had no trouble establishing their native-like fiber packing despite being preassembled into a "wrong" structure and crowded together in a film. When their triple helical domains were cut, collagen molecules (40-and 68-h PronCol) lost this ability to rearrange.
Finally, forces measured between collagen molecules in washed native fibers and in reconstituted films from AcCol, PepCol, and 20-h PronCol were all virtually identical at all temperatures (Fig. 4b). The same temperature-favored attraction responsible for fiber self-assembly could be clearly seen in the presence and absence of telopeptides. The attraction was lost only after the cleavage of triple helical domains (44-or 68-h PronCol) (Fig. 5).
Catalytic Versus Recognition and Structural Role of Telopeptides-Thus, measurements of collagen solubility after kinetic experiments, small-angle x-ray diffraction, and force measurement all give entirely consistent data suggesting that telopeptides are not essential for the ability of collagen to form nativelike, D-staggered fibers. The information that is essential for noncovalent interactions holding fibers together appears to be encoded in the triple helical domain of the protein. In contrast, covalent interactions, which are responsible for fiber stability with respect to severe stress, involve telopeptides (see, for example, Refs. 2, 9, and 10).
This interpretation may seem to be at odds with previous reports that AcCol, PepCol, and 20-h PronCol fibers have different morphologies when studied by electron microscopy (11,(13)(14)(15)(16). However, staining by heavy metal ions used in electron microscopy strongly affects fiber structure, as shown by x-ray diffraction from native rat tail tendons (35). Drying, high vacuum, and high energy electron beam impose additional stresses. It is possible that only AcCol fibers, stabilized by intermolecular covalent cross-links, are able to retain their basic structure under such conditions. Note that previously reported (11,12,16,34) and our own ( Fig. 2a) kinetic data suggest that telopeptides accelerate collagen fiber nucleation. However, the exact nature of the nuclei and the mechanism of telopeptide effect on their formation remain unknown. It was shown that telopeptides bind to specific sites on collagen molecules (12,17), but the exact role of this binding in fibrillogenesis is unclear. We can only speculate that binding of floppy, non-helical peptides may occur faster than proper alignment of long, rigid molecules. The binding may be transient, and/or it may be energetically insignificant compared with the measured interaction (Figs. 4 and 5) be-tween triple helical domains. Still, it may accelerate fibrillogenesis by placing the molecules in correct register with each other (36).
In other words, some questions remain unresolved. However, regardless of what their exact role is, telopeptides appear to perform a catalytic function rather than a recognition or an energetic function. In our opinion, this hypothesis gives the most consistent explanation to the existing experimental data.