Ligand-induced Structural Changes of the CD44 Hyaluronan-binding Domain Revealed by NMR*

CD44, a major cell surface receptor for hyaluronan (HA), contains a functional domain responsible for HA binding at its N terminus (residues 21-178). Accumulating evidence indicates that proteolytic cleavage of CD44 in its extracellular region (residues 21-268) leads to enhanced tumor cell migration and invasion. Hence, understanding the mechanisms underlying the CD44 proteolytic cleavage is important for understanding the mechanism of CD44-mediated tumor progression. Here we present the NMR structure of the HA-binding domain of CD44 in its HA-bound state. The structure is composed of the Link module (residues 32-124) and an extended lobe (residues 21-31 and 125-152). Interestingly, a comparison of its unbound and HA-bound structures revealed that rearrangement of the β-strands in the extended lobe (residues 143-148) and disorder of the structure in the following C-terminal region (residues 153-169) occurred upon HA binding, which is consistent with the results of trypsin proteolysis studies of the CD44 HA-binding domain. The order-to-disorder transition of the C-terminal region by HA binding may be involved in the CD44-mediated cell migration.

CD44, a major cell surface receptor for hyaluronan (HA), contains a functional domain responsible for HA binding at its N terminus (residues 21-178). Accumulating evidence indicates that proteolytic cleavage of CD44 in its extracellular region (residues 21-268) leads to enhanced tumor cell migration and invasion. Hence, understanding the mechanisms underlying the CD44 proteolytic cleavage is important for understanding the mechanism of CD44-mediated tumor progression. Here we present the NMR structure of the HA-binding domain of CD44 in its HA-bound state. The structure is composed of the Link module (residues 32-124) and an extended lobe (residues 21-31 and 125-152). Interestingly, a comparison of its unbound and HA-bound structures revealed that rearrangement of the ␤-strands in the extended lobe (residues 143-148) and disorder of the structure in the following C-terminal region (residues 153-169) occurred upon HA binding, which is consistent with the results of trypsin proteolysis studies of the CD44 HA-binding domain. The order-to-disorder transition of the C-terminal region by HA binding may be involved in the CD44-mediated cell migration.
CD44 is a type I transmembrane glycoprotein with diverse functions and is expressed on the surface of many cell types (1,2). CD44 recognizes hyaluronic acid (HA) 2 (3) and participates in various biological processes, such as lymphocyte rolling, tumor cell migration, and invasion (2).
HA, a major component of the extracellular matrix, is a very high molecular mass glycosaminoglycan, composed of a repeating disaccharide, D-glucuronic acid (GlcA) (␤1 3 3) N-acetyl-D-glucosamine (GlcNAc) (␤1 3 4) (4). Although HA exists as a high molecular mass polymer, HA fragments of various molecular sizes can be generated in vivo by a variety of mechanisms, and they exhibit different biological activities (5,6).
The cell surface CD44 is proteolytically cleaved at the extracellular region (residues 21-268) (7). Cleavage of CD44 has been suggested to play an important role in tumor cell migration along extracellular matrix components (8,9). Cleavage of the extracellular region is promoted by various pathways, such as those activated by extracellular Ca 2ϩ influx, protein kinase C, Rho family small GTPases, and Rac and Ras oncoproteins (8,9). Recently, it was reported that small HA oligosaccharides, generated by tumor cells constitutively expressing hyaluronidases, could induce the cleavage of the extracellular region and promote tumor migration in a CD44-dependent fashion (6,10).
CD44 has an N-terminal functional domain that interacts with HA. This ligand recognition domain contains the link protein homology region or the Link module (residues 32-124). Link modules are found in extracellular matrix molecules, such as link protein, aggrecan, versican, neurocan, and brevican, and the protein product of tumor necrosis factor-stimulated gene-6 (11). In addition, CD44 requires N-and C-terminal extensions (residues 21-31 and 125-178, respectively) for proper folding and HA binding activity (12,13). Recently, the topology of the secondary structures and the three-dimensional structure of CD44 HABD in the unbound state were revealed, and the Nand C-terminal extensions were found to form an additional structural lobe that intimately contacts with the Link module (14,15). Moreover, we found from cross-saturation experiments (16) for the CD44 HABD-HA complex that the contact residues of CD44 HABD for HA are distributed in both the consensus fold for the Link module and the extended lobe (14). Interestingly, we found that the residues with large chemical shift changes induced by HA binding were mostly localized in the C-terminal extension and the first ␣-helix, and they generally differed from the contact residues revealed by the crosssaturation experiment. These results led us to hypothesize that the first ␣-helix and the C-terminal extension undergo significant conformational changes upon HA binding. * This work was supported by a grant from the Japan New Energy and Industrial Technology Development Organization and Ministry of Economy, Trade and Industry. 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. In the present study, the solution structure of CD44 HABD in the HA-bound form has been determined by NMR. Comparisons of the structures and the backbone flexibility of CD44 HABD between its unbound and HA-bound states revealed that the C-terminal region of CD44 HABD became disordered upon HA binding. Concomitant with this structural analysis, trypsin digestion experiments of CD44 HABD in the presence and absence of HA showed that the proteolytic susceptibility in the C-terminal region increased upon HA binding. This ligand-induced order-to-disorder transition of the C terminus of CD44 HABD provides a plausible explanation for the mechanism underlying the HA-induced CD44-mediated cell migration.
Structure Calculation-NOE interproton distance restraints were derived from the three-dimensional 15 N-and 13 C-edited NOE experiments with mixing times of 60 and 120 ms, respectively. Cross-peak intensities were used to evaluate the target distances. Dihedral and angles were obtained as described (19,20). Structures were generated with CYANA, using the CANDID method (21) for NOE spectroscopy cross-peak assignment and calibration. Two hundred structures were calculated at each iteration. The 20 lowest energy structures have been submitted to the Protein Data Bank (code 2I83). { 1 H}-15 N NOE experiments were done using the pulse sequences adapted from standard schemes (22).
Trypsin Digestions of CD44 HABD-Trypsin digestions were performed in 50 mM Tris-HCl, 10 mM CaCl 2 at pH 7.4, at 20°C in the presence and absence of HA. The size of the added HA was 1000 -2000-mer. The concentration of the CD44 HABD was 34 M. Digestion was started by the addition of an aqueous solution of trypsin (final concentration, 850 nM). Aliquots were removed periodically, mixed with 2ϫ denaturing loading dye, boiled for 5 min, and analyzed by Tris-Tricine SDS-PAGE. Gels were stained with Coomassie Brilliant Blue. The digests were also analyzed by reverse phase chromatography and mass spectrometry. The digests after 20 min were applied to a PEGASIL-300 C8P column (Senshu Pak), equilibrated in 0.1% trifluoroacetic acid in H 2 O. The proteins were eluted by a gradient of 0 -65% acetonitrile in 0.09% trifluoroacetic acid in H 2 O. The peaks were fractionated and mixed with a saturated solution of 3,5-dimethoxy-4-dihydroxycinnamic acid in 50% acetonitrile in H 2 O, 0.1% trifluoroacetic acid and subjected to MALDI-TOF mass spectroscopy.

Structure Determination of CD44 HABD in the HA-bound
Form-CD44 HABD comprises the Link module (residues 32-124) and the N-and C-terminal extensions (residues 21-31and 125-178, respectively) (Fig. 1A). Our previous studies, based on cross-saturation and chemical shift perturbation experiments, showed that the structural elements consisting of the N-and C-terminal extensions of CD44 HABD contain part of the contact surface for HA and undergo significant conformational changes upon HA binding (14). To elucidate the structural changes of CD44 HABD, the solution structure of CD44 HABD in the HA-bound state was determined by heteronuclear multidimensional NMR measurements. A summary of the structural statistics for the CD44 HABD coordinates (residues 21-152) is described in Table 1, and a superimposition of the backbone atoms for the ensemble of 10 simulated annealing structures is shown in Fig. 1B. The 10 structures were superposed in a folded region, and residues 21-152 were defined, whereas residues 153-178 were poorly defined. The defined structure comprises a consensus fold for the Link module and an additional structural lobe formed by the N and C termini.
We performed two-dimensional 13 C-filtered NOE spectroscopy experiments to observe the intermolecular NOE between CD44 HABD and HA (23). A few intermolecular NOEs were observed in the aromatic (CD44 HABD)/methyl (HA) and methyl (CD44 HABD)/ring proton (HA) regions. However, the assignments of these signals were ambiguous, because of signal overlap and broadening. We checked that the intermolecular NOEs were not included in the assigned NOE peaks derived from the three-dimensional 13 C-edited or 15 N-edited NOE experiments, on the basis of the chemical shifts of the observed signals in the filtered NOE spectroscopy experiment.
We previously identified the contact residues of CD44 HABD for HA by cross-saturation experiments (14). The residues with reduction ratios larger than 0.20 comprise Thr 76 , Cys 77 , Arg 78 , Tyr 79 , Gly 80 , Ile 96 , Cys 97 , Ala 98 , Ala 99 , Asn 100 , Asn 101 , Leu 107 , Asp 115 , and Gly 152 , and those with ratios from 0.15 to 0.20 include Leu 24 , Asn 25 , Tyr 105 , Ile 106 , Asn 149 , and Asp 151 (Fig. 1A). Surface representations of the structure with the HA-binding sites are shown in Fig. 1D. These contact residues are distributed on one side of the molecule and on both the Link module (residues 32-124) and the N-and C-terminal extensions (residues 21-31 and 125-178, respectively). The length of the contact surface for HA was about 30 Å, which corresponds to a hexameric HA, consisting of three repeating disaccharides (GlcA-GlcNAc-GlcA-GlcNAc-GlcA-GlcNAc), the minimum size of HA that efficiently binds to CD44 (24).
Structural Differences of CD44 HABD in Its Unbound and HA-bound States-We compared the HA-bound structure described here with the previously reported unbound crystal structure (15) and found that significant rearrangements of the ␤-strands occur in the lobular extension of CD44 HABD upon HA binding. Ribbon representations of the unbound and HAbound states of CD44 HABD are shown in Fig. 2 (A and B,  respectively), without the unfolded regions (residues 170 -178 in the unbound state and residues 153-178 in the HA-bound state). A comparison of the secondary structures between its unbound and HA-bound states revealed that the formation of the ␤-strand in the region corresponding to ␤9 of the unbound state was not observed in the HA-bound state (Fig. 2, A and B). The backbone interstrand NOEs that correspond to those observed between ␤8 and ␤9 in the unbound state were not observed in the HA-bound state (Fig. 2, C and D). Because residues 158 -162 go under the loop between ␤7 and ␤8 in the unbound state, we supposed that the undefined C-terminal region (residues 153-178) goes under a loop between ␤7 and ␤8. However, this supposition could not be confirmed, because of the limitation of observable NOEs.
In addition, ␤8 became rearranged relative to ␤0 upon HA binding by two residues. As shown in Fig. 2C, the backbone interstrand NOE connectivities revealed that residues 22-25 of ␤0 partner with residues 145-148 of ␤8 in parallel, in the unbound state. However, residues 22-25 of ␤0 partner with residues 143-146 of ␤8 in parallel, in the HA-bound state (Fig. 2D).
Heteronuclear { 1 H}-15 N NOE experiments were performed in the unbound and HA-bound states to compare the backbone flexibility, and significant decreases of the heteronuclear NOEs at residues 162-169 were observed upon HA binding (Fig. 2, E  and F). To facilitate our discussion, we refer to this region as the anchoring region. The anchoring region runs beside the ␣1 helix and is ordered in the unbound state. The decreases of the heteronuclear NOEs upon HA binding in the anchoring region suggest that the region became unstructured upon HA binding. Indeed, no NOEs were observed between the ␣1 helix and the anchoring region in the HA-bound state, and the anchoring region was unstructured in the HA-bound state (Fig. 2, A  and B).
Trypsin Digestions of CD44 HABD in Its Unbound and HAbound States-To confirm that the anchoring region of CD44 HABD became unstructured upon HA binding, a combination of proteolysis and mass spectrometry was used. We performed time course proteolytic digestions of CD44 HABD in both the absence and presence of HA (Fig. 3A). At 60 min, two major cleavage products were observed in the absence of HA on SDS-PAGE. To identify the cleavage products, the sample at 20 min was analyzed by MALDI-TOF mass spectroscopy. On the basis of a comparison between the experimental and theoretical masses, these two fragments were found to be C-terminally truncated fragments (amino acids 21-154 and 21-162) (data not shown).
On the other hand, in the presence of HA, only one cleavage product was observed at 60 min. Analysis of the fragment revealed that it spanned from amino acids 21 to 162. Despite the decreased digestion at Arg 154 , predigested CD44 HABD (residues 21-178) in the HA-bound state disappeared rapidly, as compared with the unbound state. These results are summarized in Fig. 3B. Therefore, upon HA binding, the cleavage at the peptide bond between Arg 154 and Tyr 155 was suppressed and that between Arg 162 and Thr 163 was enhanced. The peptide bond between Arg 154 and Tyr 155 is in close proximity to the HA contact residues (i.e. Asn 149 , Asp 151 , and Gly 152 ), suggesting that the decrease in the proteolytic cleavage was due to hindrance by the bound HA. In contrast, the peptide bond between Arg 162 and Thr 163 , with increased the proteolytic susceptibility upon HA binding, is located in the anchoring region. These results confirmed the HA-induced order-to-disorder transition of the anchoring region.

DISCUSSION
Our previous cross-saturation experiments showed that the contact residues of CD44 HABD for HA were located on both the Link module and the flanking regions. In addition, the residues with significant chemical shift perturbations upon HA binding were different from the contact residues, and they were distributed in the first ␣-helix and the C-terminal extension. On the basis of these results, we concluded that significant conformational changes occur in the C-terminal extension upon HA binding. In this work, we elucidated the solution structure of CD44 HABD in the HA-bound state. The resultant structure comprises a consensus fold for the Link module and the extended lobe, including N-and C-terminal extensions. The residues in the C-terminal extension exhibit high root mean square deviation values, as compared with those at the N-terminal extension and the Link module, because of the intrinsic flexibility of the C-terminal extension.
Tumor necrosis factor-stimulated gene-6, which possesses the amino acid sequence of the Link module but not that of the   (26). The comparison of the structures of CD44 HABD in its unbound and HA-bound forms showed that the overall folding of the solution structure of the HA-bound form was the same as that of the unbound form. However, significant structural differences were present in the C-terminal extension. First, the rearrangement of the parallel ␤-sheet formation with ␤8 and ␤0 occurred. It should be noted that three isoleucine residues, spaced every two residues, are located on the ␤8 strand: Ile 143 , Ile 145 , and Ile 147 . This periodicity of amino acid residues may facilitate the rearrangement of ␤8 and ␤0. LYVE-1 was recently identified as a homologue of CD44. LYVE-1 is a major receptor for HA on the lymph vessel wall and may be involved in cell migration within the lymphatic system (12,25). Like CD44, LYVE-1 has a single Link module and a similar flanking region corresponding to the N-and C-terminal extensions of CD44 HABD (25). Interestingly, a sequence alignment of CD44 HABD and LYVE-1 revealed that LYVE-1 HABD has a periodic sequence in a region corresponding to ␤8 of CD44 HABD (Thr 153 , Thr 155 , and Thr 157 ) (Fig. 4A). A ␤-sheet rearrangement upon ligand binding may occur in the C-terminal extension of LYVE-1 HABD as well as in that of CD44 HABD.
A second structural change we observed was that the anchoring region of CD44 became disordered upon HA binding. We previously demonstrated that the contact residues of CD44 HABD for HA undergo small chemical shift changes upon HA binding and that the residues with large chemical shift changes reside in the C-terminal extension and the ␣1 helix. The results presented in this paper directly explain the observed large chemical shift changes. Unfortunately, we were not able to obtain structural information about the bound HA in this study, because of significant overlaps of the 1 H NMR signals from HA, as commonly seen in NMR spectra of carbohydrates. Stable isotope labeling for HA is required for further analysis of the bound conformation of HA to CD44 (27).
Although the physiological significance of the order-to-disorder transition of the anchoring region upon HA binding remains to be elucidated, the structural change might contribute to CD44-dependent cell migration. For example, the disordered state of CD44 HABD might be favored for multimerization of CD44 on cells, leading to the activation of intracellular signaling pathways related to cell migration. Furthermore, the order-to-disorder transition of CD44 HABD may be responsible for in vivo proteolytic cleavage in the anchoring region, because the transition enhances the proteolytic susceptibility of the anchoring region as shown in the trypsin proteolysis experiments (Fig. 4B). CD44 is reportedly proteolytically cleaved by metalloproteases, such as MT1-MMP (28), ADAM-10, and ADAM-17 (7). Although the anchoring region (residues 162-169) does not contain previously reported cleavage sites (Gly 192 -Tyr 193 , Gly 233 -Ser 234 , and Ser 249 -Gln 250 ) in the stem region (29), the transition may affect the proteolytic susceptibility of the cleavage sites by the metalloproteases, because we confirmed that the transition also occurs in CD44 HABD with the stem region by NMR (data not shown). We are now investigating whether the structural change of CD44 HABD contributes to proteolytic cleavage, based on cell biology studies.

CONCLUSIONS
We determined the solution structure of CD44 HABD in the HA-bound form. A comparison of the bound state of CD44 HABD with the unbound state of the protein revealed that a rearrangement of ␤-strands occurred in the extended lobe of  CD44 HABD. Moreover, the C-terminal region of CD44 HABD became disordered and was released from the structural domain. This study will shed light on the HA-induced promotion mechanisms for CD44-dependent cell migration.