The Location of the Ligand-binding Site of Carbohydrate-binding Modules That Have Evolved from a Common Sequence Is Not Conserved*

Polysaccharide-degrading enzymes are generally modular proteins that contain non-catalytic carbohydrate-binding modules (CBMs), which potentiate the activity of the catalytic module. CBMs have been grouped into sequence-based families, and three-dimensional structural data are available for half of these families. Clostridium thermocellum xylanase 11A is a modular enzyme that contains a CBM from family 6 (CBM6), for which no structural data are available. We have determined the crystal structure of this module to a resolution of 2.1 Å. The protein is a β-sandwich that contains two potential ligand-binding clefts designated cleft A and B. The CBM interacts primarily with xylan, and NMR spectroscopy coupled with site-directed mutagenesis identified cleft A, containing Trp-92, Tyr-34, and Asn-120, as the ligand-binding site. The overall fold of CBM6 is similar to proteins in CBM families 4 and 22, although surprisingly the ligand-binding site in CBM4 and CBM22 is equivalent to cleft B in CBM6. These structural data define a superfamily of CBMs, comprising CBM4, CBM6, and CBM22, and demonstrate that, although CBMs have evolved from a relatively small number of ancestors, the structural elements involved in ligand recognition have been assembled at different locations on the ancestral scaffold.

Polysaccharide-degrading enzymes are generally modular proteins that contain non-catalytic carbohydrate-binding modules (CBMs), which potentiate the activity of the catalytic module. CBMs have been grouped into sequence-based families, and three-dimensional structural data are available for half of these families. Clostridium thermocellum xylanase 11A is a modular enzyme that contains a CBM from family 6 (CBM6), for which no structural data are available. We have determined the crystal structure of this module to a resolution of 2.1 Å. The protein is a ␤-sandwich that contains two potential ligand-binding clefts designated cleft A and B. The CBM interacts primarily with xylan, and NMR spectroscopy coupled with site-directed mutagenesis identified cleft A, containing Trp-92, Tyr-34, and Asn-120, as the ligand-binding site. The overall fold of CBM6 is similar to proteins in CBM families 4 and 22, although surprisingly the ligand-binding site in CBM4 and CBM22 is equivalent to cleft B in CBM6. These structural data define a superfamily of CBMs, comprising CBM4, CBM6, and CBM22, and demonstrate that, although CBMs have evolved from a relatively small number of ancestors, the structural elements involved in ligand recognition have been assembled at different locations on the ancestral scaffold.
Microbial plant cell wall hydrolases, in general, have a modular structure in which the catalytic modules are attached, via linker sequences to non-catalytic modules that bind polysaccharides (1). These carbohydrate-binding modules (CBMs) 1 maximize catalytic activity by mediating prolonged and intimate contact between the substrate and the enzyme (2). The ligand specificity of CBMs generally reflects the target substrate of the catalytic module of the enzyme. Thus, cellulose, xylan, mannan, and ␤-1,3-glucan-binding CBMs are located in cellulases (3,4), xylanases (5), mannanases (6), and lichenases (7), respectively, although it should be emphasized that numerous enzymes that do not cleave cellulose contain CBMs that recognize the crystalline form of this polysaccharide (8,9).
CBMs have been classified into more than 28 families based on primary structure similarity (afmb.cnrs-mrs.fr/ϳpedro/ CAZY/cbm.html; see Ref. 10). Ligand specificity can vary both between and within families, demonstrating considerable diversity in polysaccharide recognition in these protein modules. For example, the CBM2 family contains proteins that bind to xylan and crystalline cellulose (2,5), whereas xylan, single cellulose chains, and ␤-1,3-glucan are recognized by different members of CBM4 (4,7,11). The three-dimensional structure (see afmb.cnrs-mrs.fr/ϳpedro/CAZY/cbm.html) of all CBMs characterized to date consist primarily of ␤-strands. Those that bind to crystalline cellulose have a flat ligand-binding surface (3,12), whereas CBMs that interact with single polysaccharide chains contain clefts that accommodate the target carbohydrate (13)(14)(15). Ligand binding in CBMs is dominated by hydrophobic interactions between the sugar rings and aromatic residues on the surface of the protein-binding site (see Ref. 1 for review), and a recent study has shown that the orientation of these aromatic amino acids can play an important role in ligand specificity (16).
Members of CBM family 6 are found in a variety of enzymes that include ␣-agarases, ␣-1,6-mannanases, xylanases, acetylxylan esterases, cellodextrinases, ␤-1,3-glucanases, and cellulases (afmb.cnrs-mrs.fr/ϳpedro/CAZY/db.html). Currently there is no three-dimensional structural data for any CBM6 module, and there is no information on which amino acids play a key role in ligand binding. To investigate the structural basis for the capacity of family 6 CBMs to bind to polysaccharides, we have used a combination of x-ray crystallography, NMR spectroscopy, and biochemical studies to define the ligand-binding site in CBM6 from Clostridium thermocellum xylanase 11A (Xyn11A, formerly XynU). This modular xylanase consists of an N-terminal catalytic module that belongs to glycoside hydrolase family 11, an internal dockerin module (17), and a C-terminal CBM6 that binds primarily to insoluble xylan but can also interact with insoluble cellulose (18). Here we report the crystal structure of the first representative of CBM family 6, Xyn11A CBM6. This has facilitated the definition of a novel CBM superfamily that encompasses CBM6, CBM4, and CBM22. Surprisingly, the location of the ligand-binding site in CBM6 is clearly different than that of CBM4 and CBM22.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The CBM6 from Xyn11A was produced and purified as described earlier (18). Mutants of the protein were generated by the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. To investigate the ligand-binding site of the protein, the mutants Y34A and N120A were generated. To assist in solving the crystal structure of the protein, four methionine mutants (Y40M, R72M, W92M, and Y112M) were produced, as the native CBM6 does not contain any methionine residues. The seleno-methionine-substituted proteins were expressed as described previously (13) and purified as for the native protein except that 1 mM ␤-mercaptoethanol was included in the lysis and wash buffers, and 10 mM dithiothreitol was added to the dialysis buffer.
To produce 15 N-labeled protein, Escherichia coli BL21 (Novagen) containing pXU (encodes native CBM6) was cultured at 30°C in 6ϫ 400 ml (in 2-liter conical flasks) of growth media comprising 2 mM MgSO 4 , 0.1 mM CaCl 2 , 0.2% glucose, and M9 salts containing 15 N ammonium sulfate (Martek Biosciences Corp.), until the A 600 reached 0.6, at which point expression of the recombinant protein was induced with 1 mM isopropyl-␤-D-thiogalactopyranoside. Cells were harvested after a further 8 h growth and the labeled CBM6 purified as for native protein. The 15 N-13 C-labeled CBM was produced in exactly the same way except that [ 13 C]glucose (Martek Biosciences Corp.) was used instead of unlabeled glucose in the liquid media.
Isothermal Titration Calorimetry (ITC)-ITC measurements were made at 25°C for a range of oligosaccharides and polysaccharides, as described previously (15). The concentration of protein used was such that the c value (K a ϫ molar protein concentration ϫ number of binding sites on protein) was Ͼ3 for all ligands except xylobiose and cellohexaose.
Affinity Gel Electrophoresis (AGE)-The capacity of CBM6 and its derivatives to bind to a range of soluble plant structural polysaccharides was evaluated by AGE. The method was carried out as described by Charnock et al. (13) except that gels were subjected to electrophoresis for 3 h. AGE was also carried out in the presence of a range of oat spelt xylan concentrations (0.1 to 0.005 mg ml Ϫ1 ), and affinity constants for the interaction of native and mutant forms of CBM6 with ligand were determined using equations derived from Takeo (19).
Crystallization-Crystals of CBM6 were grown by vapor phase diffusion using the hanging drop method. Equal amounts of protein (12 mg/ml in 20 mM sodium acetate buffer, pH 5.0) and reservoir solution, containing 30% polyethylene glycol 4000, 100 mM sodium citrate buffer, pH 5.5 and 100 mM (NH 4 ) 2 SO 4 , were mixed. Needle-like crystals grew from microseeds after several weeks at 4°C. The seleno-methioninesubstituted protein crystals for two of the four mutants, namely R72M and Y112M, grew under the same conditions, except that the reservoir additionally contained a 10 mM tris-(2-carboxyethyl)phosphine hydrochloride solution. A 10% glycerol solution was added to the mother liquor in both cases, prior to freezing the crystals. A rayon fiber loop was used to transfer the crystals into a liquid nitrogen cold stream. Preliminary x-ray diffraction analysis revealed that they belong to space group P6 1 22 or P6 5 22 with unit cell dimensions a and b ϭ 60.12 and c ϭ 157.43 Å and with one molecule in the asymmetric unit. The crystals of the R72M seleno-methionine containing CBM6 mutant were larger in size and diffracted to higher resolution and were therefore chosen for the MAD experiment.
MAD X-ray Diffraction Data Collection and Phasing-A three-wavelength MAD experiment was conducted on beamline ID14-4 at the European Synchrotron Radiation Facility (ESRF) at Grenoble, France, using an ADSC Quantum-4 CCD detector. A single crystal of selenomethionine containing CBM6 was flash-frozen in the laboratory and then preserved in liquid nitrogen. The crystal was transported to the ESRF, where it was remounted on the single axis goniometer. The wavelengths for the MAD experiment were chosen by scanning through the absorption edge of the R72M seleno-methionine containing CBM6 crystal. Data were collected at three wavelengths to maximize the anomalous signal as follows: at minimum fЈ, maximum fЉ, and a reference wavelength at an energy above the absorption edge (Table I). After indexing an initial diffraction image using the program package HKL2000 (20), the program STRATEGY (21) was used to determine the optimal range to collect complete data using a minimal oscillation sweep. A total of 60 images with 1°oscillation was collected at each of the three wavelengths. The data were processed with DENZO as part of the HKL2000 package and scaled with SCALA, part of the CCP4 suite of programs (22).
Because only one selenium was present in the protein, the Patterson maps were deconvoluted by hand, and the site was confirmed by the SOLVE (23) automated Patterson search. The Selenium position was refined, and the phase calculations were performed using the program SHARP (24) in both space groups and in the resolution range between 32 and 3.0 Å. Subsequently, a cycle of phase improvement was applied using the program DM (25). The correct enantiomorph (P6 5 22) was determined by the quality of the resulting electron density maps.  Model Building and Refinement-The electron density map at 3.0 Å, calculated after DM in space group P6 5 22, displayed extensive well defined regions revealing continuous stretches of main chain density, with unambiguous density for carbonyl oxygen atoms and side chains. A model comprising 120 amino acids was built from the initial map with the graphics program TURBO-FRODO (26). This model was refined using the CNS program (27).
Data were processed and reduced as described above. The refinement of the 3.0-Å model was continued using these data at 2.1-Å resolution. Water molecules and alternative conformers were added, and TLS refinement, using the TLS option in REFMAC (28), was performed as deemed appropriate from the behavior of the cross-validation (R free ) subset of reflections (5%). The final electron density map in the region of Tyr-34 and Trp-92 is shown in Fig. 1.
NMR Spectroscopy-Homonuclear NMR experiments were performed on a 2.0 mM unlabeled protein sample at 307 K, unless otherwise indicated, in 90% H 2 O and 10% (v/v) D 2 O containing 20 mM deuterated sodium acetate buffer at pH 5.0. All NMR spectra were recorded on a 500-MHz DRX Bruker spectrometer equipped with a 5-mm triple resonance HCN probe with self-shielded triple axis gradients. DQF-COSY, Clean TOCSY, and NOESY spectra were acquired in a phase-sensitive mode using States-TPPI. Heteronuclear NMR experiments were performed at 307 K on 0.5 mM uniformly 15 N-labeled protein in 20 mM deuterated sodium acetate buffer at pH 5.0 containing 10% (v/v) D 2 O, and on 0.3 mM uniformly 13 C-15 N-labeled protein under the same conditions. Two-dimensional HSQC, two-dimensional HSQC-NOESY, and two-dimensional HSQC-TOCSY as well as three-dimensional NOESY-HSQC, three-dimensional TOCSY-HSQC, and threedimensional HSQC-NOESY-HSQC were acquired using the Fast HSQC scheme described by Mori et al. (29). A three-dimensional HNCA (30), HNcoCA (31), HCCH-TOCSY (32), CBCAcoNH (33), and CBCANH (34) spectra were recorded with the 13 C-15 N-labeled protein. Spectral analysis was performed using the XEASY software, following the standard strategy described by Wü thrich (35) combined with heteronuclear strategy.
NMR Ligand Titration-To identify the residues of CBM6 involved in binding xylohexaose, five 1 H-15 N HSQC spectra were acquired. A first HSQC spectrum was recorded on CBM6 alone at 0.5 mM. Four further 1 H-15 N HSQC spectra were then recorded with 0.2, 0.6, 0.9, and 1.2 mM xylohexaose, respectively.

RESULTS
Ligand Binding Studies-To evaluate the binding specificity of CBM6 in more detail, the protein was subjected to affinity gel electrophoresis. The data (not shown) revealed that the protein binds to oat spelt, birchwood, rye, and wheat xylans, which vary in their degree of substitution, interacts weakly with barley ␤-glucan and soluble forms of cellulose (hydroxyethylcellulose and methylcellulose), but does not associate with sugar beet arabinan, potato, and pectic ␤-1,4-galactan, lime and apple pectin, rhamnogalacturonan, locust bean and carob galactomannan, or carboxymethyl ␤-1,3-glucan.
ITC was used to quantify the affinity of CBM6 for xylan, xylooligosaccharides, and cellohexaose. The full data are presented in Table II. The results show that the protein binds to xylooligosaccharides with a dp of 2 or more, with the affinity increasing up to xylopentaose. The association constant of the CBM for xylopentaose is ϳ50 and 100 times higher than for xylobiose and cellohexaose, respectively. The affinity of the protein for highly substituted arabinoxylans (wheat and rye) or poorly substituted xylans (oat spelt and birchwood xylan) is broadly similar. These data suggest that the protein has five sugar-binding sites and can accommodate highly decorated xylans. The stoichiometry of binding for all oligosaccharides was 1:1, indicating that there is only one ligand-binding site per protein molecule. This is consistent with the other CBMs, such as CBM4 (4), CBM9 (36), and CBM22 (13), which also contain single carbohydrate-binding sites, but is in contrast to the starch-binding modules that can interact with two amylose chains (37). These data suggest that the CBM6 from C. thermocellum Xyn11A is primarily a xylan-binding module.
The ⌬H and T⌬S values were negative for all ligands tested, demonstrating that the binding of CBM6 to soluble carbohydrates is enthalpically driven. The thermodynamics of CBM6ligand interactions are similar to other proteins that bind soluble carbohydrates (4,13,36), whereas favorable entropic forces dominate the binding of CBM2a proteins to crystalline cellulose (38).
Crystal Structure of CBM6 -The crystals of CBM6 belong to the space group P6 5  selected groups of residues showed that the particular loop region including residues 22-33 displayed highly anisotropic librational motion. Refinement statistics are given in Table I.
CBM6 displays a small shallow cleft, lined by two aromatic residues Trp-92 and Tyr-34 (Fig. 2, cleft A). However, when the fold is compared with CBM22 and CBM4, a second putative ligand-binding site becomes apparent on the concave face of the jelly roll (Fig. 2, cleft B). This latter cleft strongly resembles those observed in CBM22 (13) and CBM4 (14), and two aromatic residues are equally located close to the surface, namely Tyr-40 and Tyr-112, the aromatic nature of which are relatively conserved in CBM family 6 (Tyr-40 is replaced by a tryptophan in a number of family members; Fig. 3). Surprisingly, a proline residue of a neighboring loop (residues 73-79) covers up this groove, making the surface aromatic residues inaccessible for xylooligosaccharides.
Two ion sites were identified in the CBM6 crystal structure. The first site displays a typical hepta-coordination geometry to accommodate a calcium ion and links the side chains of Glu-8 and Glu-10 (bidentate coordination) of the N-terminal region to the side chain of Asp-122 of the C-terminal stretch. Two main chain carbonyl groups of Ser-30 and Glu-8 complete the coordination sphere, and three water molecules are in close contact (Fig. 1b). This ion can be considered as a structural calcium ion. It is located at an equivalent position as the structural calcium ion found in CBM22 (13), bridging the N-and C-terminal regions, although neither the residues that interact with the ion nor the number of different ligation types are identical. Structural calcium ions often play a key role in protein stability (40), and recent data have shown that removal of the divalent ion from CBM6 greatly increased the susceptibility of the protein to proteolytic degradation. 2 The coordination of the second ion involves two main chain carbonyl groups (Ile-35 O and Val-119 O) and three main chain amide groups (Ile-35 N, Tyr-34 N, and Val-119 N), as well as one asparagine side chain (Asn-120 OD1), and is completed by two water molecules. Most likely this site is occupied by a sodium ion, although the biological significance of this bound metal ion is unclear.
Spin System and Sequential Assignment-More than 200 cross-peaks could be seen in the HSQC spectrum for only 132 amino acids complicating the spectral assignment. Starting from fingerprint cross-peaks of the COSY spectra ( 15 N-1 H-HSQC and 13 3. Sequence alignment and phylogenetic tree of the modules CBM family 6. a, sequence alignment of selected sequences. The numbering corresponds to the sequence of CBM6 from C. thermocellum Xyn11A (formerly XynU). The secondary structure elements (exclusively ␤-strands) are marked below the sequences in blue when part of the structural jelly roll fold and in green for two ␤-strands that are not part of the ␤-sandwich. The three residues, Tyr-34, Trp-92, and Asn-120, which have been shown to be involved in ligand binding, are marked by red arrowheads above the sequences. The pink circles above the sequences mark the position of two aromatic residues situated in cleft B. The flexible loop region, adjacent to the ligand-binding site cleft A, is indicated above the sequences. The alignment was performed with ClustalW, and the color coding of homology (yellow boxes, cut off of 5; red boxes, cut off of 9; the maximum value for strictly conserved residues is 10) was calculated with an ALSCRIPT option. The figure was produced with ALSCRIPT (41). b, phylogenetic tree for all known members of family CBM6, based on the alignment produced with ClustalW. The red numbers above the branches indicate the glycoside hydrolase (GH) family of the catalytic domain to which the corresponding CBM6 is attached, as reported by afmb.cnrs-mrs.fr/ϳpedro/CAZY/db.html (see Ref. 10). From top to bottom: xylanase U (Xyn11A) from C. thermocellum (Swiss-Prot accession number O52780); xylanase V from C. thermocellum (Swiss-Prot accession number O52779); xylanase Z from C. thermocellum (Swiss-Prot accession number P10478); m-1, m-2, and m-3 of xylanase A from C. stercorarium strain NCIB11754 (GenBank TM accession number AF417638); ORF SC2H12.11c from Streptomyces coelicolor (GenBank TM accession number AL359215); cellodextrinase D from Microbispora bispora (Swiss-Prot accession number Q59506); ORF BH1908 from Bacillus halodurans (GenBank TM accession multiple amide proton chemical shifts. This is exemplified by the arrows in Fig. 4a. The complete assignment of intra-residue HN-H ␣ and HN-N and C ␣ -N cross-peaks could not be achieved because of the multiple conformations that induced many ambiguities. We were, however, able to assign some regions of the sequence (residues 6 -12, 21-26, 28 -40, 67-70, 91-96, and 112-117). Fig. 4a provides an example of the sequential assignment encompassing residues 28 -40. Fig. 4b illustrates the loop formed by these residues, adjacent to cleft A.
CBM6 Titration Experiments-The 1 H-15 N HSQC spectra of CBM6 titrated with increasing amounts of xylohexaose show that the protein and its ligand are in fast exchange between the free and bound states. This allowed us to identify the backbone CBM6 amide groups that are affected by oligosaccharide binding (Fig. 4c).
The chemical shift variations of the N amide and the HN proton of the residues that are close to the xylohexaose-binding site are listed in Table III. Among the 23 cross-peaks that are shifted upon ligand binding, only 9 were sequentially identified using the conventional method (see "Experimental Procedures"). The remaining 14 cross-peaks fortunately corresponded to known spin systems, although they were not sequentially assigned. This problem was solved by first determining the type of amino acid corresponding to a given spin system by virtue of its 13 C␣ and/or 13 C␤ chemical shift value and second by looking for sequential connectivity on HNCA/NHcoCA and CBCANH/CBCAcoNH groups of spectra between the given amino acid and the preceding and following residues. Following this strategy, all 23 amino acids, whose chemical shifts were influenced by ligand binding and are thus likely to be at, or close, to the ligand-binding site, could unambiguously be assigned (Table III). These residues include Trp-92, Tyr-34, and Asn-120, located in cleft A, indicating that they are part of the ligand-binding site. None of the residues situated in cleft B are affected by the titration experiment.
Site-directed Mutagenesis-Amino acids located on the surface of the two clefts of CBM6 were mutated, and the ligand binding capacity of the resultant proteins was assessed. Mutations to Tyr-40, Arg-72, or Tyr-112, located in cleft B, did not result in a significant decrease in the affinity of the protein for xylan (Table IV). In contrast, the mutants W92M, Y34A, and N120A exhibited greatly reduced affinity for oat spelt xylan. The location of Trp-92, Asn-120, and Tyr-34 in cleft A indicates that this region composes the ligand-binding site of CBM6. DISCUSSION This report shows that CBM6 binds preferentially to xylan and xylooligosaccharides of at least two xylose moieties, exhibiting maximum affinity for the pentasaccharide. CBM6 is thus primarily a xylan-binding module, and by analogy it is likely that the other CBM6s, which are components of xylanases, also recognize xylan. These modules, however, are also found in glycoside hydrolases that act on a diversity of plant polysaccharides other than xylan. Thus, it is likely that, similar to the CBM families 2 and 4 (3,4,7,11,16), different members of CBM6 will recognize distinct polysaccharides. This view is supported by a recent study that showed that the N-terminal CBM6 in xylanase 11 A from Clostridium stercorarium strain NCIB11754 binds specifically to xylan, whereas the other two CBM6 modules in this protein bind cello-and xylooligosaccharides with similar affinity (42). The crystal structure of CBM6 combined with NMR titration and site-directed mutagenesis experiments show that the shallow groove formed by Tyr-34 and Trp-92 is part of the ligand-binding site. Surprisingly, these residues are not invariant within CBM family 6 (Fig. 3). However, this lack of conservation is plausible when we as- sume that family 6 CBMs linked to glycoside hydrolases other than xylanases display specificity for the saccharides cleaved by their respective catalytic modules. It is interesting to note that the cleft between Tyr-34 and Trp-92 can only accommodate a single sugar unit as the aromatic residues are directly opposite each other, and therefore this does not account for the observed maximum affinity for oligosaccharides of at least 5 sugar units. The multiple conformations observed by NMR for the adjacent loop, encompassing residues 22-33, suggest that there might be further residues exposed to the surface in solution. Indeed, some amino acids, i.e. Ile-32, Ile-121, Asp-122, and Tyr-123, which are buried by the flexible loop, undergo a change in their chemical shift upon addition of xylohexaose. This observation indicates that the loop conformation trapped in the crystal structure is not the one present in the ligandprotein complex in solution, although this view remains to be firmly established.
An unexpected and interesting feature of this study is the observation that proteins that share a common ancestor (as evident from high structural similarity) have ligand-binding sites in different locations on the protein scaffold. Thus, data presented in this report show that cleft A is the ligand-binding site for CBM6, whereas cleft B binds to the target oligosaccharides in CBM4 and CBM22 (43,44), which share a common evolutionary origin with CBM6. There is, however, precedence for such a phenomenon. CBM10 from Pseudomonas xylanase 10A, which binds to crystalline cellulose has an OB-fold, but the ligand-binding site in the CBM is in a different location to oligonucleotide-binding proteins that have the same fold (12).
It is rather more surprising that proteins that recognize the same ligand, and have evolved from a common progenitor sequence, bind to xylan in different locations. It should be noted that CBM6 contains a surface loop that partially occludes cleft B and thus prevents access to the ligand. It is possible that for other members of the CBM6, the loop that extends from Leu-73 to Thr-79 is in a different conformation, making cleft B accessible to xylan and xylooligosaccharides. Support for this view comes from the sequence alignment, displayed in Fig. 3, which shows that, apart from one exception, the CBM6 modules that do not have a tyrosine in the equivalent position to Tyr-34 (cleft A) have a tryptophan in the position equivalent to Tyr-40 (situated in cleft B) and actually lack the proline residue that covers up cleft B in Xyn11A CBM6. Another intriguing possibility is that the location of the ligand-binding site in CBM6 modules varies depending on ligand specificity, a view strengthened by the lack of conservation, in family 6 members, of residues that are known to play a key role in ligand binding in Xyn11A CBM6. This suggests that the CBM6 scaffold displays a remarkable structural and functional plasticity.