Actinidain-hydrolyzed Type I Collagen Reveals a Crucial Amino Acid Sequence in Fibril Formation*

We investigated the ability of type I collagen telopeptides to bind neighboring collagen molecules, which is thought to be the initial event in fibrillogenesis. Limited hydrolysis by actinidain protease produced monomeric collagen, which consisted almost entirely of α1 and α2 chains. As seen with ultrahigh resolution scanning electron microscopy, actinidain-hydrolyzed collagen exhibited unique self-assembly, as if at an intermediate stage, and formed a novel suprastructure characterized by poor fibrillogenesis. Then, the N- and C-terminal sequences of chicken type I collagen hydrolyzed by actinidain or pepsin were determined by Edman degradation and de novo sequence analysis with matrix-assisted laser desorption ionization-tandem time-of-flight mass spectrometry, respectively. In the C-telopeptide region of the α1 chain, pepsin cleaved between Asp1035 and Phe1036, and actinidain between Gly1032 and Gly1033. Thus, the actinidain-hydrolyzed α1 chain is shorter at the C terminus by three residues, Gly1033, Phe1034, and Asp1035. In the α2 chain, both proteases cleaved between Glu1030 and Val1031. We demonstrated that a synthetic nonapeptide mimicking the α1 C-terminal sequence including GFD weakly inhibited the self-assembly of pepsin-hydrolyzed collagen, whereas it remarkably accelerated that of actinidain-hydrolyzed collagen. We conclude that the specific GFD sequence of the C-telopeptide of the α1 chain plays a crucial role in stipulating collagen suprastructure and in subsequent fibril formation.

Type I collagen is the most abundant protein in connective tissue of all vertebrates. It forms highly ordered fibrils that are generally thought to provide mechanical strength and regulate cell function (1)(2)(3)(4)(5). Collagen is formed from tightly interwoven heterotrimers of two ␣1 chains and one ␣2 chain in a triplehelical coiled-coil structure. The triple helix consisting of Gly-Xaa-Yaa repeats is ϳ1,000 amino acid residues in length and is resistant to proteolysis except for specific collagenases that can cleave the helix (6,7). The high content of 4-hydroxyproline in the Yaa position stabilizes the triple-helical structure (8,9).
Observations of in vitro collagen fibril formation suggest that the collagen molecule has sufficient structural information to assemble spontaneously under suitable conditions. This structural information has been partially revealed by atomic force microscopy, electron microscopy, spectroscopic analysis, and x-ray diffraction (10 -15). From this information, it appears that the telopeptide regions play an important role in the rate of fibril formation and in azimuthal and lateral growth of fibrils. Indeed, fibril formation of pepsin-hydrolyzed collagen (PHCol) 2 that has been partially hydrolyzed at the N-and C-telopeptide domains progresses more slowly than that of acidsoluble collagen (ASCol) (16,17). The recognition and association of collagen N-and C-telopeptides have been partly demonstrated in vitro (18 -25). The results imply that the Nand C-telopeptide regions are required to "seed" the interaction between collagen molecules. It was also concluded that the triple-helical structure is essential for fibril formation (4,26).
Our experimental approach to understand the mechanism of fibril formation was to investigate type I collagen that had been partially hydrolyzed by actinidain protease (EC 3.4.22.14). Actinidain is a cysteine protease found in kiwi fruit (Actinidia deliciosa) and is a member of the papain superfamily with a broad specificity toward a variety of substrates (27)(28)(29). We previously found that actinidain-hydrolyzed collagen (AHCol) of tuna retained its triple-helical structure at pH 4.0 and showed the same thermal stability as that of ASCol, as judged by circular dichroism (CD) spectroscopy. AHCol, however, was found to be essentially monomeric with no cross-linkages, and therefore it provides a unique system for further characterization (30,31).
Knowledge of the exact amino acid sequence is vital for understanding the behavior of collagen, including fibril formation. Although the N-terminal amino acid sequence of collagen can be determined by Edman degradation, there has been no reliable method for determining the C-terminal sequence. Using tandem mass spectrometry (MS/MS) and known amino acid sequence information, we recently determined the amino acid sequences of novel proteins following trypsin digestion (32,33). We employed this method in the current work to iden-tify the C-terminal amino acid sequence of collagen preparations. In this study, by comparing the properties of ASCol, PHCol, and AHCol derived from chicken skin, we identified an amino acid sequence that is required for the fibril formation of type I collagen.

EXPERIMENTAL PROCEDURES
Preparation of ASCol, PHCol, and AHCol from Chicken Skin-ASCol and PHCol were prepared from chicken skin according to the method of Morimoto et al. (30). Actinidain was isolated from kiwi fruit as described (29). Crude ASCol in 20 mM sodium acetate buffer, pH 4.0 (final concentration 2 mg/ml), was incubated with 0.15% actinidain (w/v) at 20°C for 7 days. AHCol was salted out by adding NaCl to a final concentration of 2.5 M. The resulting PHCol and AHCol samples were dialyzed against ultrapure water to remove small molecule contaminants and then lyophilized. The purity of the ASCol, PHCol, and AHCol samples was assessed by 5% SDS-PAGE in the presence of 4 M urea (30). Protein molecular mass markers were obtained from APRO Life Science Institute Inc. (Naruto, Japan). After electrophoresis, proteins were stained with Coomassie Brilliant Blue R-250, and their relative amounts were estimated by ImageJ software (version 1.4.2; National Institutes of Health). The collagen concentration was estimated from the weight of the lyophilized preparation.
CD Spectrum of ASCol, PHCol, and AHCol-The triple-helical structure of each of ASCol, PHCol, and AHCol (prepared as described above) was confirmed by CD spectra as reported (30,31). The far-UV CD spectrum of the collagen preparations was measured by scanning through the range of 200 -250 nm with a CD spectropolarimeter J-820 (JASCO, Tokyo, Japan). Measurement was carried out using a 1.0 mg/ml solution of ASCol, PHCol, or AHCol in 20 mM phosphoric acid (pH 2.2) in a quartz cell with a 0.1-cm optical path length at 20°C.
Morphological Observation of Self-assembly of PHCol and AHCol Using Scanning Electron Microscopy (SEM)-PHCol and AHCol (final concentration 3 mg/ml) were dissolved in 15 mM acetic acid for 20 h at 4°C. Four glass disks (13 mm) were coated with the PHCol or AHCol solution and air-dried at 20°C. Then, the glass disks were placed in 50 mM sodium phosphate buffer, pH 7.4, at 37°C to start self-assembly. After an incubation period of 10, 30, 90, or 120 min, self-assembly was stopped by placing each glass disk in 2% glutaraldehyde at room temperature for 1 h. Sample preparations for SEM observation were carried out according to the methods of Nakayama et al. (34). Collagen self-assembly was observed under an S-900 ultrahigh resolution SEM (Hitachi) with an accelerating voltage of 10 kV.
N-terminal Amino Acid Sequence Analysis of PHCol and AHCol-The ␣1 and ␣2 chains of PHCol and AHCol were separated by 5% SDS-PAGE and electroblotted onto a polyvinylidene fluoride membrane at 180 mV for 1 h. Each protein spot corresponding to the ␣1 or ␣2 chain was used to determine the N-terminal amino acid sequence using a Procise 494 HT Protein Sequencing system (Applied Biosystems, Inc., Carlsbad, CA).
Matrix Metalloprotease-1 (MMP-1) Hydrolysis of PHCol and AHCol-To determine the C-terminal amino acid sequence of PHCol and AHCol using MS/MS analysis, samples were first hydrolyzed with collagenase to obtain shorter polypeptide chains according to the published methods (7,35). MMP-1mediated hydrolysis was performed by the addition of 160 l of 0.5 unit/ml MMP-1 solution (derived from active human fibroblasts; EC 3.4.24.7; Life Laboratory Corporation, Yamagata, Japan) to 200 l of a 1.0 mg/ml PHCol (or AHCol) solution and then incubated at 37°C for 10 days. The resulting fragments of the ␣1 and ␣2 chains were separated by 11% SDS-PAGE and stained with Coomassie Brilliant Blue R-250. The larger 3/4 fragment and the smaller 1/4 fragment generated from PHCol or AHCol by MMP-1 were named TC A and TC B , respectively.
C-terminal Amino Acid Sequence Analysis of PHCol and AHCol with MALDI-MS/MS-Each polypeptide corresponding to the TC B ␣1 and TC B ␣2 chains was digested with trypsin, and the resulting peptides were analyzed using a 4700 MALDI-TOF/TOF MS (Applied Biosystems Inc.) according to our published methods (32,33). MS/MS spectra were measured in the collision-induced dissociation-off mode. Prediction of the amino acid sequence from MS/MS data was performed using DeNovo Explorer TM (version 1.22; Applied Biosystems Inc.) with an MS/MS tolerance of 0.2 Da.
Turbidity Measurements of PHCol and AHCol in the Presence of Five Synthetic Peptides-The "cold start" procedure for fibril formation was used to observe collagen self-assembly (16,36,37). The self-assembly of collagen molecules was detected by monitoring turbidity as observed by an increase in optical density at 313 nm. At 20°C, a 2.0 mg/ml collagen solution in 10 mM sodium acetate buffer, pH 4.0, was diluted 10-fold with 50 mM sodium phosphate buffer, pH 7.4; the final collagen concentration in the neutralized solution was 0.20 mg/ml (0.67 M). The neutralized solution was poured into a cuvette with a 1-cm path length at 20°C, and then heated to 37°C. The turbidity of the solution was measured at 1-min intervals using a UV-2200A spectrophotometer (Shimadzu, Kyoto, Japan). Five synthetic nonapeptides mimicking the sequence around the C terminus of PHCol and AHCol were purchased from Toray Research Center Inc. (Tokyo, Japan). Each individual nonapeptide (PGP-PGPPSG, PGPPSGGFD, PSGGFDFSF, GFDFSFLPQ, and FSFLPQPPQ) was added singly to a neutralized solution of collagen at up to a 400-fold molar excess over collagen, the molecular weight of which was assumed to be 300,000. The turbidity change of the mixture was measured as above. The dose dependence of PSGGFDFSF at 10-, 50-, 100-, or 400-fold molar excess over PHCol or AHCol was examined. The following characterizing parameters were estimated manually from the turbidity progress curve (16,36,37): t lag , the time at the end of the lag phase; t max , the time at the maximum fibril growth rate; and (dA/dt) max , the maximum fibril growth rate.

RESULTS
Preparation of ASCol, PHCol, and AHCol-The components of ASCol, PHCol, and AHCol were analyzed by 5% SDS-PAGE (see supplemental Fig. 1). The molecular mass and relative amount of each component are shown in Table 1. These data indicate that the hydrolysis site for collagen must differ between pepsin and actinidain.
Triple-helical Structure of ASCol, PHCol, and AHCol as Determined by CD Spectroscopy-The CD spectrum for ASCol was typical of a triple-helical structure, with a peak maximum at 221-222 nm. The spectra for AHCol and PHCol were identical to each other and similar to that of ASCol (see supplemental Fig. 2). These observations are consistent with our previous report on tuna collagen (30).
Morphological Observation of PHCol and AHCol Using SEM-The resulting SEM snapshots of self-assembly of PHCol and AHCol are shown in Fig. 1, illustrating the changes in collagen suprastructure at various incubation times. PHCol spontaneously formed typical fibrils, with fibrils increasing their diameter from 10 nm to 500 nm during the 120-min incubation. In addition, the fibrils showed a D-periodic banding pattern. On the other hand, AHCol showed a novel nanoscale meshwork suprastructure with a "spider-web" appearance. No fibril growth was observed during the incubation period.
N-terminal Amino Acid Sequences of PHCol and AHCol Chains as Determined by Edman Degradation-Two N-terminal sequences, H 2 N-YDEKS-and H 2 N-VAVPG-, were found in the ␣1 chain of PHCol, and two N-terminal sequences, H 2 N-VAVPG-and H 2 N-VPGPM-, were found in the ␣1 chain of AHCol. Two N-terminal sequences, H 2 N-ADFGP-and H 2 N-FGPGP-, were also found in the ␣2 chain of PHCol, and three sequences, H 2 N-DFGPG-, H 2 N-GPGPM-, and H 2 N-LMGPR-, were found in the ␣2 chain of AHCol. The N-terminal amino acid sequences therefore showed rather minor differences between PHCol and AHCol.
Comparison of these sequences with the entire amino acid sequence of chicken type I collagen (Gallus gallus) deposited in the UniProtKB/Swiss-Prot data base (P02457, version 80; P02467, version 87) identified the cleavage sites of pepsin and actinidain in the N-telopeptide region ( Fig. 2 and see "Discussion").
MMP-1 Hydrolysis of PHCol and AHCol-The SDS-PAGE result of AHCol after MMP-1 hydrolysis was similar to that of PHCol (see supplemental Fig. 3). Two fragments, TC A ␣1 (96 kDa) and TC B ␣1 (29 kDa), were obtained from each ␣1 chain. The ␣2 chain was also hydrolyzed into two fragments, TC A ␣2 (86 kDa) and TC B ␣2 (26 kDa). Only the TC B fragments were subjected to further analysis, however, because they were thought to have been derived from the C terminus of PHCol and AHCol, by analogy with previous reports (7,35).
C-terminal Amino Acid Sequences of PHCol and AHCol as Determined by MS/MS-The C-terminal amino acid sequence was determined by de novo sequence analysis using MALDI-TOF/TOF. Fig. 3, A and B, shows the MALDI-TOF MS spectra of the trypsin digests of TC B ␣1 chains of PHCol and AHCol, respectively. To identify the specific peak originating from the C-terminal peptide, the two spectra were compared, and peaks at m/z 2654.32 for PHCol and m/z 2335.11 for AHCol were identified as unique to each collagen preparation. The amino acid sequences of these peaks were further analyzed by MALDI-TOF/TOF and identified, by y-and b-series ions, to be GRTGEVGPVGPOGPOGPOGPOGPPSGGFD for the peak at m/z 2654.32 and GRTGEVGPVGPOGPOGPOGPOGPPSG for the peak at m/z 2335.11 (see supplemental Fig. 4), where O refers to hydroxyproline. In both cases, four hydroxyproline residues were identified at the same positions as those in the data base sequence.
The tryptic digest of the TC B ␣2 chain showed the same peak profile in MALDI-TOF MS for both PHCol and AHCol (Fig. 4).
We identified the C-terminal fragment as the peak at m/z 2598.16, because this was the only peak that did not contain a C-terminal K or R, as determined by a preliminary MS/MS analysis. This peak was applied to MS/MS de novo sequencing analysis, and the amino acid sequence was determined to be GSHGSQGPAGPOGPOGPOGPOGODGGGYE (see supplemental Fig. 5). Thus, the C-terminal amino acid sequences of both TC B ␣2 chains were similar using the same sequencing method, and five hydroxyproline residues were identified in the sequence. The putative pepsin and actinidain cleavage sites of the C-telopeptide region are shown in Fig. 5. Therefore, a clear

. SEM snapshots of PHCol and AHCol suprastructures after various incubation times. Solutions of PHCol and AHCol prepared in 15 mM
acetic acid (unfolding conditions) were placed onto glass disks and air-dried. Self-assembly of the collagen was initiated by incubating the disks in 50 mM sodium phosphate buffer, pH 7.4, at 37°C, and then assembly was followed for various periods. Upper and lower micrographs are of PHCol and AHCol, respectively. Scale bar, 300 nm.  difference between PHCol and AHCol was found in the GFD sequence at the C terminus of the ␣1 chain. Turbidity Measurements for the Self-assembly of PHCol and AHCol in the Presence of Mimic Nonapeptides-To investigate the role of the GFD sequence in self-assembly further, the turbidity of separate PHCol and AHCol solutions in the presence of a nonapeptide was kinetically monitored. Five synthetic nonapeptides (described under "Experimental Procedures") were tested. Without the addition of the nonapeptides, PHCol, which formed typical fibrils as seen by SEM (Fig.  1), showed a high rate of increase in turbidity, with t lag of 32 min, t max of 57 min, and (dA/dt) max of 12.3 ϫ 10 Ϫ3 OD⅐min Ϫ1 . However, AHCol showed a longer lag time and a lower rate with a t lag of 100 min, t max of 155 min, and (dA/dt) max of 5.0 ϫ 10 Ϫ3 OD⅐min Ϫ1 (Fig. 6A).
The addition of four of the five nonapeptides did not affect the selfassembly of either PHCol or AHCol. Interestingly, however, one nonapeptide, PSGGFDFSF (400-fold molar concentration), weakly inhibited PHCol self-assembly but markedly accelerated AHCol self-assembly (Fig. 6A, c and cЈ). The turbidity parameters of PHCol in the presence of this nonapeptide changed to: t lag , 35 min; t max , 67 min; and (dA/dt) max , 9.3 ϫ 10 Ϫ3 OD⅐min Ϫ1 . On the other hand, those parameters for AHCol in the presence of this nonapeptide were: t lag , 50 min; t max , 85 min; and (dA/dt) max , 8.7 ϫ 10 Ϫ3 OD⅐min Ϫ1 . Moreover, the effect of the relative molar excess of PSGGFDFSF (10 -400-fold) on the self-assembly of PHCol and AHCol was examined (Fig. 6, B and C). Substantial differences were observed in the values of the t lag and t max for AHCol (see supplemental Table 1). This nonapeptide appeared to affect the initial step in AHCol self-assembly.

DISCUSSION
Collagen fibril formation is a multistep process occurring in a sequential manner between adjacent molecules. The intrinsic propensity of collagen to undergo fibrillogenesis must be connected to the amino acid sequence of the N-and C-terminal domains because truncation of either or both of the N-and C-terminal telopeptide regions leads to impaired collagen fibrillogenesis (19 -22, 24 -26). In this study, we compared the structural features of two types of protease-hydrolyzed collagen molecules, PHCol and AHCol, obtained from limited hydrolysis by pepsin and actinidain, respectively, from chicken skinderived ASCol. These collagen preparations were chosen because PHCol and AHCol derived similarly from tuna skin show interesting structural differences (30,31). Full sequence information of tuna skin collagen is not available. Therefore, we   chose chicken skin as the source of collagen for this study because the chicken collagen sequence is known.
Components of ASCol, PHCol, and AHCol as Determined by SDS-PAGE-From SDS-PAGE analysis (see supplemental Fig. 1), PHCol showed at least five bands, corresponding to ␣1, ␣2, ␤11, ␤12, and ␥ polypeptide chains, whereas AHCol exhibited mainly two bands, corresponding to chains ␣1 and ␣2. The relative content of ␣1 and ␣2 chains increased, and that of ␤11 and ␤12 chains decreased following the limited hydrolysis by pepsin or actinidain (Table 1). AHCol can be considered as an essentially monomeric collagen in which 58% of molecules are composed of two ␣1 chains and 32% of ␣2 chains, with few cross-linkages between ␣ chains. These results are consistent with our previous data on tuna AHCol (30,31).
Morphological Change during Self-assembly as Observed by SEM-As shown in Fig. 1, after a 10-min incubation, collagen molecules appeared to adhere to adjacent molecules at specific binding positions. No preferential lateral orientation was observed, indicating that the early stage of self-assembly was controlled by the binding between triple helices. After 30 min, the PHCol molecules assembled with each other and clearly formed filaments, and finally typical D-periodic banding pattern appeared in the fibril. In contrast, we observed a novel meshwork suprastructure for AHCol upon self-assembly after incubation times of 10 -120 min, and the feature of self-assembly remained constant over the 120-min incubation. The SEM observations suggest that the meshwork suprastructure may be an intermediate in the early stage of filament reconstitution.
Cabral et al. (24) and Kuznetsova and Leikin (26) suggested that molecular recognition and fibril formation are controlled by the triple-helical domain. We demonstrated that the triplehelical structure of the AHCol molecule is identical to that of PHCol (see supplemental Fig. 2). AHCol, therefore, should have the same initial propensity for self-assembly as ASCol or PHCol. An understanding of the subsequent maturation phase of self-assembly requires an analysis of azimuthal and lateral growth. We observed that AHCol can grow in azimuthal orientations and that there was no specific packing arrangement (Fig.  1). Using molecular modeling, Malone et al. (25,38) recently concluded that the N-and C-telopeptide structures and the ordered helical structures were strongly related to the molecular packing of collagen molecules. Such a difference in AHCol structure suggests that the N and C termini must differ significantly from those of PHCol. These results prompted us to determine the amino acid sequences of these termini.
N-terminal Sequence Determination and Hydrolysis Sites for Pepsin and Actinidain-Pepsin cleaves the ␣1 chain at two sites, Gly 5 -Tyr 6 and Gly 12 -Val 13 . Thus, the PHCol ␣1 chain has at least two components, one of which contains Lys 9 . Pepsin cleavage of the ␣2 chain at Ala 7 -Ala 8 and Asp 9 -Phe 10 removes Lys 6 from both components. Therefore, the possibility to crosslink the ␣1 chain via Lys 9 is partly retained in PHCol. Actinidain, on the other hand, cleaves the ␣1 chain at two sites (Gly 12 -Val 13 , Ala 14 -Val 15 ) and the ␣2 chain at three sites (Ala 8 -Asp 9 , Phe 10 -Gly 11 , Gly 16 -Leu 17 ). AHCol, therefore, has no remaining cross-linking residues because the ␣1 chain Lys 9 and ␣2 chain Lys 6 are in the leaving N-telopeptide fragments. These results are consistent with the observed lower content of ␤11 and ␤12 chain in AHCol ( Table 1). The H 2 N-(YDEK-SAG)(VA)V-sequence of the ␣1 chain and the H 2 N-(A)(D)(F)-(GPGPMG)L-sequence of the ␣2 chain correspond to the differences in the N-terminal sequences of PHCol and AHCol (Fig. 2).
C-terminal Sequence Determination of PHCol and AHCol-Our results were in good agreement with the registered sequence data of the ␣1 chain of chicken type I collagen and revealed that pepsin and actinidain cleaved the ␣1 polypeptide chain at Asp 1035 -Phe 1036 and Gly 1032 -Gly 1033 , respectively (Fig. 5). The difference in the molecular mass of the two peaks at m/z 2654.32 and m/z 2335.11 in Fig. 3 is m/z 319.21, which corresponds to the mass of the intervening tripeptide HN-GFD-CO (m/z 319.33).
The results for the ␣2 chain C-telopeptide showed that both pepsin and actinidain cleaved the polypeptide at Glu 1030 -Val 1031 . In addition, we identified five hydroxyproline residues in a GPP repeat sequence, at positions 1013, 1016, 1019, 1022, and 1024. There are no annotation data related to hydroxyproline in the sequence of the chicken type I collagen ␣2 chain, according to the UniProt entry (http://www.uniprot.org/uniprot/ P02467). We demonstrated here that both PHCol and AHCol have four GPO repeat sequences in each ␣1 and ␣2 chain (see supplemental Figs. 4 and 5). These are likely to be essential for the stability of chicken collagen.
Turbidity Measurement for the Self-assembly of PHCol and AHCol in the Presence of Mimicking Nonapeptides-The result for PHCol (Fig. 6A) is in qualitative agreement with published data (16,36,37). The t lag of AHCol, however, was much longer than that of PHCol, and the fibril growth rate decreased to ϳ41% of that of PHCol. Despite the fact that the amino acid sequence of AHCol is generally similar to that of PHCol, the kinetics of self-assembly differ markedly between these collagens. It has been suggested that the C-terminal sequence of collagen plays a substantive role in fibril formation (20,22,24,25). We have demonstrated that the AHCol ␣1 chain has a unique C terminus (Fig. 5). It is therefore worthwhile to investigate AHCol self-assembly further. Interestingly, we found that only one nonapeptide, PSGGFDFSF, among five that mimic the C-telopeptide region of the ␣1 chain, influenced the turbidity kinetics. The N-terminal six residues of PSGGFDFSF overlap the C-terminal sequence of PHCol and thus may compete with PHCol for the binding site on adjacent collagen molecules (Fig.  6B). On the other hand, for AHCol, in which the sequence GFD is missing and only three residues at the C terminus overlap the N-terminal sequence of the nonapeptide, the nonapeptide seems to behave as if a piece of mending tape by filling the vacant space made by truncation (Fig. 6C). The seemingly opposite effects of this nonapeptide on self-assembly may provide an important clue to understanding the initial step of fibrillogenesis, although the molecular mechanism remains to be elucidated.
This study has directly shown that the primary structure of the N and C termini of chicken type I collagen influences the mechanism of fibrillogenesis and the resulting suprastructure and that the GFD sequence of the ␣1 chain is required for lateral growth of fibrils. Interestingly, the GFD sequence in the C-telopeptide region of the ␣1(I) chain is universally conserved in human (P02452), dog (Q9XSJ7), chicken (P02457), African clawed frog (Q802B5), Japanese common newt (Q9YIB4), and zebrafish (B1WB89) collagen. Similarly, a GYD sequence is conserved in cow (P02453), rat (P02454), and mouse (P11087) collagen. In conclusion, several lines of evidence support our hypothesis that the presence of the ␣1 chain telopeptide region containing a GFD or GYD sequence is especially important for conferring orientation specificity. PHCol and AHCol from chicken type I collagen, for which the N-and C-terminal amino acid sequences have been determined, are excellent choices for the study of self-assembly and suprastructure formation.