Unidirectional Binding of Clostridial Collagenase to Triple Helical Substrates*

Histotoxic clostridia produce collagenases responsible for extensive tissue destruction in gas gangrene. The C-terminal collagen-binding domain (CBD) of these enzymes is the minimal segment required to bind to collagen fibril. Collagen binding efficiency of CBD is more pronounced in the presence of Ca2+. We have shown that CBD can be functional to anchor growth factors in local tissue. A 1H-15N HSQC NMR titration study with three different tropocollagen analogues ((POG)10)3, ((GPOG)7PRG)3, and (GPRG(POG)7C-carbamidomethyl)3, mapped a saddle-like binding cleft on CBD. NMR titrations with three nitroxide spin-labeled analogues of collagenous peptide, (PROXYL-G(POG)7PRG)3, (PROXYL-G(POG)7)3, and (GPRG(POG)7C-PROXYL)3 (where PROXYL represents 2,2,5,5-tetramethyl-l-pyrrolidinyloxy), unambiguously demonstrated unidirectional binding of CBD to the tropocollagen analogues. Small angle x-ray scattering data revealed that CBD binds closer to a terminus for each of the five different tropocollagen analogues, which in conjunction with NMR titration studies, implies a binding mode where CBD binds to the C terminus of the triple helix.

Histotoxic clostridia produce collagenases that degrade collagen in connective tissue. Although the enzyme is assumed to be a causative agent for diseases like gas gangrene (1), it is beneficial to remove dead tissue from ulcers or burns and for nonsurgical treatment of Dupuytren's disease (2,3). For collagenases to hydrolyze tissue collagen, the enzymes must 1) anchor themselves onto an insoluble collagen fibril, which is a staggered array of tropocollagen and then 2) isolate a single triple helical molecule from the bundle and finally 3) unwind the triple helix to expose a scissile peptide bond. Clostridium histolyticum produces two classes of collagenases, which contain a catalytic domain belonging to the family M9B, followed by one or two copies of polycystic kidney disease domains and one or two copies of collagen-binding domains (CBD) 2 (4). Each CBD spans ϳ120 amino acid residues and binds specifically to insoluble collagen. CBD also binds to collagenous peptides with triple helical conformation but not to collagenous peptides that lack triple helix or to gelatin (denatured collagen), suggesting that the CBD-collagen interaction is conformation-specific (4,5). Calcium ions enhance the binding at physiological concentration, and x-ray crystal structures of CBD have been solved in the presence and absence of calcium (6).
Since collagen fibrils constitute a major part of the extracellular matrix, bioactive molecules can be anchored with CBD for their prolonged effect. Nishi et al. (7) have demonstrated that growth factors fused to CBD remained at the sites of injection much longer than growth factors alone to induce extended cell proliferation. In order to gain an insight into the anchoring mechanism of CBD, we attempted to co-crystallize CBD and collagenous peptide without success. Also to better address the role of CBD in fibril disruption and transition states from insoluble substrate, solution studies of CBD with the triple helical collagenous peptide became necessary.
NMR titration methods were utilized to identify the collagen binding pocket on CBD. Since it has been shown that most peptidases bind to their substrate in one direction at their catalytic center (8,9), there could be only one direction for the collagen triple helices at the binding site of CBD. On the other hand, CBD might allow bidirectional binding, since it is independent of the catalytic domain. To identify the binding direction, three different NMR titrations were performed with spinlabeled analogues of tropocollagen, where a nitroxide spin label 2,2,5,5-tetramethyl-L-pyrrolidinyloxy (PROXYL) was attached to either the N or C terminus of the collagenous peptide. The nitroxide moiety with an unpaired electron can cause enhancement in paramagnetic relaxation of NMR resonances of CBD via electron-nuclear dipolar coupling, thereby resulting in extreme line broadening of those resonances (10,11). Furthermore, small angle x-ray scattering (SAXS) was utilized to obtain the three-dimensional structure of the CBD-collagenous peptide complex. 15 N-Labeled Protein Production-A C-terminal CBD (Gly 893 -Lys 1008 ) derived from the C. histolyticum class I collagenase (ColG) was expressed as a glutathione S-transferase fusion protein. The glutathione S-transferase tag was cleaved off by thrombin, and CBD was purified as described previously (4). Uniform 15 N isotope labeling was achieved using Tanaka minimal medium containing 40 mM 15 NH 4 Cl. The labeling efficiency was estimated to be 99.6% by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry.
Steady State Fluorescence Measurements-The tertiary fold of CBD with the addition of collagenous peptide was characterized by its fluorescence spectra measured at 25°C on a Hitachi F-2500 fluorimeter (Tokyo, Japan) with excitation and emission bandwidths at 2.5 and 10 nm, respectively. Measurements were carried out with the excitation wavelength of 280 nm to detect the contribution of both tyrosine and tryptophan residues. The emissions were monitored between 300 and 450 nm. Protein with concentration of 25 g/ml in 10 mM Tris-HCl at pH 7.5 containing 100 mM NaCl and 1 mM CaCl 2 was used. The concentration of ((POG) 10 ) 3 was 63 g/ml.
CD Spectroscopy-The triple helical conformation and the stability of the collagenous peptides were verified using CD spectroscopy. CD spectra were recorded with a J-820 CD spectropolarimeter (JASCO Co., Hachioji, Japan) equipped with a Peltier thermo controller, using a 0.5-mm quartz cuvette and connected to a data station for signal averaging. All peptide samples were dissolved in water (1 mg/ml) and stored at 4°C for 24 h. The spectra are reported in terms of ellipticity units/mol of peptide residues [] mrw . Thermostability of the triple helix was monitored by the [] 225 values of each peptide with increasing temperature at the rate of 0.3°C/min. NMR Spectroscopy-NMR experiments were performed at 25 Ϯ 0.5°C on a Bruker 700-MHz spectrometer equipped with a Cryoprobe TM . The concentration of the protein used was 0.1 mM in 50 mM Tris-HCl (pH 7.5) containing 100 mM NaCl and 20 mM CaCl 2 . A highly concentrated stock solution of the collagenous peptides (4 mM) was prepared and equilibrated at 4°C for 24 h. The concentrated stock solution facilitated the addition of small volumes of peptide(s) on the course of titration; thus, any volume changes were very small and did not affect the intensity measurements of the cross-peaks in the 1 H-15 N HSQC spectrum. Alliquots of collagenous peptide(s) were added to the protein and equilibrated for 5 min before acquiring 1 H-15 N HSQC spectra. The pH of the NMR samples was monitored during the titration, and there was no significant change in the pH at the end of each titration (within Ϯ0.2 units). Changes in the average amide chemical shifts (⌬␦) were calculated using Equation 1, where ␦ 1 H and ␦ 15 N are the amide proton and amide nitrogen chemical shift differences between the free and bound state of the protein (12). NMR chemical shift changes were fit using KaleidaGraph 4.0 (Synergy Software) to a simple bimolecular binding model through a nonlinear regression analysis using Equation 2, where ⌬␦ obs is the observed chemical shift, ⌬␦ max is the maximum chemical shift, [P] 0 is the total concentration of the protein, [L] 0 is the total concentration of collagenous peptide, and K d is the dissociation constant of the complex.
Small Angle X-ray Scattering Experiments-The SAXS data were collected on solutions of CBD, various collagenous peptides, and the CBD-collagenous peptide complexes in 10 mM Tris-HCl (pH 7.5) containing 100 mM NaCl and 20 mM CaCl 2 at the XOR beamline of sector 12-ID at the Advanced Photon Source (APS) in the Argonne National Laboratory. The main advantage of x-ray scattering is that it can be carried out in solution in nearly physiological conditions (13). All of the samples and buffers were filtered through a 0.22 M filter prior to data collection. To minimize the radiation damage during data collection, the samples were continuously oscillated through a 1.5-mm-wide quartz capillary at a flow rate of 7 l/s. During exposure, the beam was focused to a size of 0.6 mm wide by 0.2 mm high. The energy of the x-ray beam was 12 keV, and the flux was 3 ϫ 10 12 (photons/s). The average exposure time was 1 s. Ten SAXS exposures were collected for each sample with a charge-coupled device x-ray detector at a camera length of 780 mm. The optical configuration used allows a Q range from 0.02 to 0.6 Å Ϫ1 to be measured, where Q ϭ (4 sin()/) is the magnitude of the scattering vector, is the scattering angle, and ϭ 1 Å is the wavelength of the x-ray. All scattering data were acquired at a sample temperature of 10°C. The scattering patterns were circularly averaged with a C program. For further analysis, the program IGOR Pro 5.5 A (WaveMetrics) was used. The scattering profiles of the protein were obtained after subtracting the buffer profiles. The reduced scattering data were plotted as scattering intensity I(Q) versus Q. The radius of gyration, R g , was obtained from the Guinier approximation by linear least squares fitting in the QR g Ͻ 1 region, where the forward scattering intensity I(0) is proportional to the molecular weight of the protein complex (14). An indirect Fourier transformation of I(Q) data using GNOM (15) gave the particle distribution function P(r) in the real space (Fig. 5). The point on the x axis where P(r) reaches zero represents the maximum diameter D max averaged in all orientations. The molecular envelopes were constructed for all of the samples based on the SAXS data after ab initio calculations with the program GASBOR (16). In GASBOR, simulated annealing minimization of randomly distributed dummy atom models represents the protein structure after being tested for the best fit to the I(Q) scattering data. The atomic models are represented as a compact interconnected configuration of beads inside a sphere with diameter D max that fits the experimental data Iexp(s) to minimize discrepancy. Atomic models were fit into ab initio envelopes with the program SUBCOMB (17).
Docking Model-The Protein Data Bank data used to generate the CBD-collagenous peptide complex are from entries 1NQD and 1K6F for CBD and collagenous peptide, respectively. To obtain the complex model, the soft docking algorithm BiGGER (18) was used. CBD-collagenous peptide models generated were filtered using NMR titration data, and the highest scoring model that satisfied NMR, SAXS, and mutation results was chosen.

RESULTS AND DISCUSSION
CBD Retained Its Tertiary Structure upon Binding to Collagenous Peptides-Fluorescence measurements were carried out to monitor the tertiary structural changes on CBD upon binding to the collagenous peptide. The native spectrum of CBD has only one-emission maxima at 317 nm. This comes from solvent-exposed tyrosine residues with the tryptophan being buried inside the core (no emission maxima at 350 nm) (6). The emission spectrum of CBD-collagenous peptide complex was identical to the native CBD spectrum, indicating that the tertiary structure of CBD does not change upon binding to tropocollagen (supplemental Fig. 1).  Tyr 996 , disappeared completely from the 1 H-15 N HSQC spectrum, as seen from the superposition of a region of the two-dimensional NMR spectra of CBD and CBD-collagenous peptide complex (Fig. 1A). Amide resonances of these residues in CBD must be in the close proximity of collagenous peptide upon binding. The intensity decrease of the cross-peaks could be due to chemical exchange between amide resonances of CBD and collagenous peptide on the NMR time scale (Fig. 1B). The amide resonances that disappeared during the course of titration were neighbors to the "hot spot" side chains recognized by mutation studies (6). NMR chemical shifts of few resonances were perturbed when the ratios of ((POG) 10 ) 3 /CBD were increased from 0.1:1 to 0.7:1. Changes in the chemical shifts occur due to the environmental changes on the protein surface as a result of interaction with ligand (11). The amide resonances that shift can be seen in superposition of a region of the two-dimensional 1 H-15 N HSQC NMR spectra of CBD during the titration (Fig. 1C) (Fig. 1D). These residues are located in close proximity to the five amide resonances that broadened and disappeared. Binding curves for residues in CBD (Fig. 2, filled circle) that show significant chemical shift changes upon the addition of increasing amounts of ((POG) 10 ) 3 were obtained using Equation 2. As seen from Fig. 2, binding of ((POG) 10 ) 3 to CBD saturated around a 1:1 molar ratio, indicating 1:1 stoichiometry of the complex. An average value of 57 M was obtained for K d (supplemental Table 1) calculated for the six residues showing a large total chemical shift change. This signifies a moderate binding between CBD and collagenous peptide. The binding constant measured is consistent with the value determined from surface plasmon resonance (57 M) (6). The NMR titration study presented here mapped the tropocollagen binding cleft in solution for the first time (Fig. 3). Residues Leu 924 , Ser 928 , Arg 929 , Thr 957 , Tyr 970 , Gln 972 , Leu 992 , Tyr 994 , Lys 995 , and Tyr 996 of CBD shape the binding cleft. Side chains of these surface residues are within 3 Å of 11 amide groups whose resonances either line-broaden or shift in NMR titration. Mutations of the side chains to alanine were performed for all of the residues except Leu 924 , Ser 928 , Gly 971 , and Lys 995 , and those mutations lowered collagen binding affinity of CBD (6). Gly 925 and Asp 926 may also be involved in binding, but their resonances could not be unambiguously assigned, since they exhibited weakened and broadened signals in all of the spectra collected.

Characterization of Tropocollagen
Unidirectional Binding of Collagenous Peptide on CBD-To identify the orientation of a tropocollagen on CBD, a collagenous peptide was modified to accommodate a nitroxyl group in the N terminus, (PROXYL-G(POG) 7 PRG) 3 (supplemental Fig.  2). The PROXYL group being paramagnetic enhances the relaxation rate of the NMR resonances of the protein under study via electron-nuclei dipolar coupling (10). An arginine residue was placed in the C-terminal triplet to increase solubility while stabilizing the triple helix to a similar extent as a hydroxyproline residue (25). The spin-labeled collagenous peptide is stable and retained the triple helical conformation under our experimental conditions as evident from CD spectra (sup- plemental Figs. 3 and 4). NMR titrations with this spin-labeled analogue were carried out at the molar ratios described earlier. Five residues broadened, and six residues underwent chemical shift perturbation just as observed upon ((POG) 10 ) 3 titration. In addition, the amide resonance corresponding to Val 973 appreciably line-broadened and vanished from the 1 H-15 N HSQC spectrum at a 0.33:1 (PROXYL-G(POG) 7 PRG) 3 /CBD ratio (Fig.  4, A and B). Resonances of no other residues were perturbed due to the PROXYL group even at higher (PROXYL-G(POG) 7 PRG) 3 /CBD ratios. In a three-dimensional space, Val 973 is located at an edge of the ␤-sandwich fold at the opposite end of Ca 2ϩ -binding site (Fig. 4F). This in turn implied that the N terminus of the tropocollagen is pointed away from the Ca 2ϩ -binding site. To further confirm the direction of the tropocollagen, a shorter peptide, (PROXYL-G(POG) 7 ) 3 (supplemental Fig. 2), was synthesized and titrated with 15 Nlabeled CBD. Again, five residues extensively broadened, and six residues underwent chemical shift perturbation, as seen in the previous two titrations. In addition to Val 973 , amide resonances corresponding to Gly 975 and Ser 979 were also extensively line-broadened and vanished from 1 H-15 N HSQC spectrum during the titration (Figs. 4, D and E). The titration results not only confirm that the N terminus of tropocollagen is pointed away from the Ca 2ϩ binding site but also that the distance between the N terminus and CBD became shorter with (PROXYL-G(POG) 7 ) 3 .
To further illustrate that the C terminus of the tropocollagen is directed toward the Ca 2ϩ -binding site in CBD, another spinlabeled collagenous peptide, (GPRG(POG) 7 C-PROXYL) 3 was synthesized, where the PROXYL spin label was attached to a cysteine residue placed at the C terminus (supplemental Fig. 2). An arginine residue at the third position was made to increase the solubility of the peptide. NMR titrations with this peptide were repeated in a similar manner as with ((POG) 10 ) 3 . Line broadening and chemical shift perturbation of residues involved in collagen binding were consistent, as observed with ((POG) 10 ) 3 titration. In addition, amide resonances of Ser 906 , Arg 929 , Ser 997 , and Gly 998 excessively line-broadened and disappeared from the 1 H-15 N HSQC spectrum of CBD at greater than a 0.3:1 ratio of (GPRG(POG) 7 C-PROXYL) 3 /CBD (Fig. 4, G  and H). In the three-dimensional structure, these residues are located close to the Ca 2ϩ -binding site of CBD (Fig. 4F). Resonances of no other residues in the collagen binding cleft were disturbed even at the higher ratios of (GPRG(POG) 7 C-PROXYL) 3 /CBD. To rule out any ambiguities arising from the arginine or cysteine residue in the collagenous peptides ((PROXYL-G(POG) 7 PRG) 3 and (GPRG(POG) 7 C-PROXYL) 3 ) interacting with CBD, two more peptides, (G(POG) 7 PRG) 3 and (GPRG(POG) 7 C-carbamidomethyl) 3 (supplemental Fig. 2), that lack the PROXYL groups were synthesized, and NMR titrations were repeated. The titration results with these peptides were nearly identical with those of ((POG) 10 ) 3. On the other hand, Val 973 that disappeared due to the PROXYL group remained intact even at a 1:1 (G(POG) 7 PRG) 3 /CBD ratio (Fig.   4C). Ser 906 , Arg 929 , Ser 997 , and Gly 998 that disappeared due to the PROXYL group also remained unchanged at a 1:1 (GPRG(POG) 7 -C-carbamidomethyl) 3 /CBD ratio (Fig. 4I). Results of these NMR titrations collectively demonstrate that all six tropocollagen analogues influenced the amide resonances of CBD in nearly identical fashion. Line broadening of additional residues was due to the presence of PROXYL. When the PROXYL group was reduced by ascorbic acid in the (PROXYL-G(POG) 7 -PRG) 3 -CBD complex, Val 973 peak emerged (supplemental Fig. 5A), further confirming that the amide resonance of Val 973 was line-broadened by the PROXYL moiety. When the [PROXYL-G(POG) 7 ] 3 -CBD complex was reduced with ascorbic acid, amide resonances corresponding to Val 973 , Gly 975 , and Ser 979 reappeared (supplemental Fig. 5B). Similarly, amide resonances for Ser 906 , Arg 929 , Ser 997 , and Gly 998 emerged in the (GPRG(POG) 7 C-PROXYL) 3 -CBD complex when PROXYL was reduced with ascorbic acid (supplemental Fig. 5C).
Binding curves for the six residues (Fig. 2) in CBD that show significant chemical shift changes with increasing amounts of different collagenous peptides were obtained using Equation 2. Slopes of the binding curves for all six residues in CBD upon titration(s) with six different collagen peptides were similar and consistent with moderate binding affinity (average K d ϳ52-57 M) (supplemental Table 1).
The PROXYL group is known to induce line broadening of nuclear resonances in a distance-dependent manner (26). Since more residues are line-broadened with PROXYL at the C terminus than at the N terminus (Fig. 4F), CBD plausibly binds closer to the C terminus of the tropocollagen analogue. Spin quantization of electron spin resonance spectra in the (GPRG(POG) 7 C-PROXYL) 3 is 2.2 times larger than that of (PROXYL-G(POG) 7 PRG) 3 and (PROXYL-G(POG) 7 ) 3 , which could be due to a more dynamic nature of the PROXYL moiety attached to cysteine at the C terminus (supplemental Fig. 6). More residues line-broadened upon NMR titration with (GPRG(POG) 7 C-PROXYL) 3 might also be accredited to the dynamics of the PROXYL moiety. The experiment, however, clearly demonstrated unidirectional binding of collagenous peptide on CBD.
Earlier binding studies using extensive mutagenesis of surface residues postulated a Tyr-rich interface for CBD-collagen binding (6). Three binding modes were proposed, in which two modes oriented collagenous peptide perpendicular to a Tyrrich ␤-sheet region in CBD and the third mode placed it diagonally. Two tropocollagen binding modes that predicted it to bind perpendicularly to the ␤-sheets of CBD are consistent with the NMR titration studies done with various spin-labeled collagenous peptides. Residues that were extensively line-broadened due to collagen binding and proximity of the spin label are  in accordance with the perpendicular binding mode. The binding mode that placed tropocollagen interacting diagonally with CBD can be eliminated, since no residues that could account for the diagonal binding mode are line-broadened with the PROXYL moiety at either terminus. The tropocollagen binding surface in CBD is narrowed to a 10-Å-wide and 25-Å-long cleft (Fig. 3). The width of the binding cleft in CBD matches with the diameter of the triple helix (27) (supplemental Fig. 7), and its length could accommodate ((POG) 3 ) 3 .
Asymmetric CBD-Collagen Complex-To further confirm the solution structure of the CBD-collagenous peptide complex, SAXS studies were conducted. Using SAXS and an ab initio shape determination procedure, the three-dimensional structure of the macromolecule can be constructed. Linear Guinier plots from SAXS data for CBD, (G(POG) 7 PRG) 3 , and CBD-(G(POG) 7 PRG) 3 complex (Fig.  5A) yielded the radius of gyration (R g ) values ( Table 1). The pair-distance distribution function P(r) and the maximum diameter D max averaged in all orientations were computed ( Fig. 5B and Table 1). The R g values computed from P(r) are in excellent agreement with those obtained from Guinier approximation ( Table 1). The three-dimensional molecular envelopes constructed for CBD and the triple helical collagen peptide from SAXS data after ab initio calculation and simulated annealing minimization (for details, see "Experimental Procedures") are in excellent agreement with the overall dimensions and shape of the crystal structure (Fig. 6,   A and B). The overall shape of the tropocollagen analogue in solution is "8" or "∞", which could be due to the more dynamic nature of collagen peptide at the ends as evident from its high B-factors in crystal structure (27). The threedimensional SAXS envelope calculated for CBD-(G(POG) 7 PRG) 3 showed asymmetric binding (Fig.  6C). Three-dimensional structures calculated by SAXS for four other CBD-collagenous peptide complexes also confirmed the asymmetric binding mode (supplemental Fig. 8). Combining NMR titration and SAXS data, CBD plausibly binds closer to the C terminus of the collagenous peptide. It is interesting to note that the scattering density of the collagenous peptide changes to a rod shape upon binding to CBD. Changes in dynamics of the collagenous peptide due to binding could influence the appearance of the SAXS envelopes (Fig. 6, B and C). If the collagenous peptide is placed in accordance with its orientation determined by NMR titration studies, CBD binds to C-terminal (POG) 3 repeats (i.e. the sixth to eighth POG repeats in (POG) 8 ) (Fig. 7). The dynamics of collagen does not appear to dictate CBD binding. However the non-triple helical nature of the tropocollagen at the carboxyl end (27,28) might lure CBD toward the C terminus. The distance from the N-terminal spin label of (PROXYL-G(POG) 7 PRG) 3 to Val 973 in the CBD would be ϳ37 Å, and the distances from the N terminus of (PROXYL-G(POG) 7 ) 3 to Val 973 , Gly 975 , and Ser 979 are about 29 Å. Whereas Ser 906 , Arg 929 , Ser 997 , and Gly 998 are about 15 Å from the C-terminal spin label of (GPRG(POG) 7 C-PROXYL) 3 . Paramagnetic relaxation is usually observed in distances ranging between 15 and 24 Å (29); however, in our case, when the PROXYL is attached in the main chain at the N terminus, three PROXYL moieties are clustered together (supplemental Fig. 2), and long range relaxation effects might be pronounced. The SAXS envelope calculated for the (PROXYL-G(POG) 7 PRG) 3 -CBD complex shows an additional strong density at the N terminus, indicating that the PROXYL moieties at the N terminus are clustered (supplemental Fig. 8B). There is no such strong scattering density when the PROXYL moiety is attached to the side chain (supplemental Fig. 8C).
CBD binds unidirectionally to the C-terminal region of collagenous peptides. However, whether CBD binds only at the C-terminal region of tropocollagen remains uncertain. Though CBD targets the least ordered region of the collagenous peptides, CBD could also target partly unwound regions even in the middle of a tropocollagen. The catalytic domain is situated at the N terminus of CBD. Bacterial collagenolysis may involve optimal orientation of the catalytic domain with respect to tropocollagen. Furthermore, the collagenolysis could initiate from partly unwound regions in tropocollagen. Since the complexes of CBD and various col-  lagenous peptides are structurally homogeneous, co-crystals are probably attainable.