Crystal structure of the catalytic domain of DESC1, a new member of the type II transmembrane serine proteinase family Otto J P Kyrieleis1,*, Robert Huber1,2, Edgar Ong3, Ryan Oehler3, Mike Hunter3, Edwin L Madison3 and Uwe Jacob1,  Max-Planck-Institut fur Biochemie, Martinsried, Germany ă Cardiff University, UK Corvas International, San Diego, CA, USA Keywords squamous cell carcinoma of the head and neck; trypsin-like serine protease; tumor marker; type II transmembrane serine proteinases Correspondence O J P Kyrieleis, Max-Planck-Institut fur ă Biochemie, Abteilung Strukturforschung, Am Klopferspitz 18a, D-82152 Martinsried, Germany Fax: +33 0476207199 Tel: +33 0476207860 E-mail: kyrieleis@embl-grenoble.fr Present address *EMBL Grenoble Outstation, France  SuppreMol GmbH, Martinsried, Germany Database The coordinates and structure factors for DESC1–benzamidine complex have been deposited in the RCSB Protein Data Bank under the accession number 2OQ5 (Received 17 October 2006, revised 31 January 2007, accepted 26 February 2007) doi:10.1111/j.1742-4658.2007.05756.x DESC1 was identified using gene-expression analysis between squamous cell carcinoma of the head and neck and normal tissue It belongs to the type II transmembrane multidomain serine proteinases (TTSPs), an expanding family of serine proteinases, whose members are differentially expressed in several tissues The biological role of these proteins is currently under investigation, although in some cases their participation in specific functions has been reported This is the case for enteropeptidase, hepsin, matriptase and corin Some members, including DESC1, are associated with cell differentiation and have been described as tumor markers TTSPs belong to the type II transmembrane proteins that display, in addition to a C-terminal trypsin-like serine proteinase domain, a differing set of stem domains, a transmembrane segment and a short N-terminal cytoplasmic region Based on sequence analysis, the TTSP family is subdivided into four subfamilies: hepsin ⁄ transmembrane proteinase, serine (TMPRSS); matriptase; corin; and the human airway trypsin (HAT) ⁄ HAT-like ⁄ DESC subfamily Members of the hepsin and matriptase subfamilies are known structurally and here we present the crystal structure of DESC1 as a first member of the HAT ⁄ HAT-like ⁄ DESC subfamily in complex with benzamidine The proteinase domain of DESC1 exhibits a trypsin-like serine proteinase fold with a thrombin-like S1 pocket, a urokinase-type plasminogen activator-type S2 pocket, to accept small residues, and an open hydrophobic S3 ⁄ S4 cavity to accept large hydrophobic residues The deduced substrate specificity for DESC1 differs markedly from that of other structurally known TTSPs Based on surface analysis, we propose a rigid domain association for the N-terminal SEA domain with the back site of the proteinase domain DESC1 is a type II transmembrane serine proteinase (TTSP), an expanding protein family with members differentially expressed in several organs and certain tumors To date, more than 30 mammalian members of this group have been identified and have, according to their sequence similarity, been grouped into four subfamilies: DESC ⁄ human airway trypsin (HAT) ⁄ HATlike type (DESC1–3, HAT, HAT-like 4); matriptase Abbreviations HAI, hepatocyte growth factor activator inhibitor; HAT, human airway trypsin; PAI-1, plasminogen activator inhibitor 1; PCI, protein C inhibitor; SRCR, scavenger receptor cystein-rich; TMPRSS, transmembrane proteinase, serine; TTSP, type II transmembrane serine proteases; uPA, urokinase-type plasminogen activator 2148 FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS O J P Kyrieleis et al type (matriptase 1–3 and polyserase); hepsin ⁄ transmembrane proteinase, serine (TMPRSS) type (hepsin ⁄ MSPL ⁄ enteropeptidase, TMPRSS 2–5); and corin type (corin) The hepsin subfamily includes enteropeptidase [1,2] the best studied member of the TTSPs All TTSPs have an N-terminal cytoplasmic domain of variable length (20–160 amino acids), a short transmembrane region, a modular stem region and a C-terminal, highly conserved trypsin-like serine proteinase TTSPs are oriented in a way that the proteinase domain lies outside the cell, directly exposed to the extracellular matrix Seven structural motifs may be combined in the stem regions (low-density lipoprotein receptor class A, SEA, MAM, Frizzled, CUB, Group A scavenger), and may contribute to activation of the C-terminal proteinase, substrate binding [3] and targeting of the molecule to secondary interaction partners on the cell surface or the extracellular matrix (e.g integrins, sulfated polysaccharides, lipids or proteoglycans) These complex stem regions and the cytoplasmic domain, which may interact with cellular signaling molecules and the cytoskeleton, make it tempting to speculate that TTSPs are key regulators of signaling events on the plasma membrane Their activity is therefore integrated in the networks of much better characterized proteinase systems such as the ADAMs, membrane-type matrix metalloproteinases and the urokinase-type plasminogen activator (uPA) ⁄ uPA-receptor system Gene expression analysis between squamous cell carcinoma of the head and neck and normal tissue led to the identification of a differential expressed squamous cell carcinoma gene (DESC1) The data indicated that expression of DESC1 mRNA was restricted to normal epithelial cells of prostate, skin, testes, head and neck, whereas it was downregulated or absent in the corresponding cells of squamous cell carcinoma of the head and neck [4] It has therefore been proposed as a possible tumor marker Furthermore, Sedghizadeh et al [5] were able to show that DESC1 is upregulated during the induction of terminal keratinocyte differentiation, supporting a role in normal epithelial turnover These results suggest that DESC1 may function in regular epithelial differentiation under normal conditions or in circumventing tumorigenesis under cancer-promoting conditions Recently, the mouse ortholog of DESC1 was identified, and was found to have 72% shared identity with human DESC1 Both proteinases are expressed in similar anatomical locations and are likely to have common functions in the development and maintenance of oral epidermal tissues and the male reproduction tract [6] Human DESC1 has a short 20-amino acid cytoplasmic region followed by 14 residues of a putative trans- Crystal structure of the catalytic domain of DESC1 Scheme Domain organization of human DESC1 membrane region The extracellular part of DESC1 consists of a 120-amino acid SEA domain followed by the C-terminal trypsin-like serine proteinase domain, as shown in Scheme DESC2 and DESC3 were subsequently identified by database searches [2] In contrast to DESC1, many TTSPs are overexpressed by tumor cells (e.g matriptase, hepsin) The frequent association between cancer and TTSP expression suggests that development of specific inhibitors of individual TTSPs may provide insight into the molecular mechanisms of carcinogenesis as well as the normal biological roles of this interesting, emerging class of cell-surface proteases Structural information on the protein domains of the TTSP subfamilies of the hepsin ⁄ TMPRSS (hepsin) [7], enteropeptidase [8] and matriptase (matriptase) [9] types exists, but no crystal structure data on the remaining subgroups of the HAT ⁄ HAT-like ⁄ DESC and corin subfamilies exists We therefore cloned, expressed and purified the serine proteinase domain of DESC1 and solved the crystal structure of the complex of this protease with benzamidine Results and Discussion Overall structure The catalytic domain of DESC1 resembles an oblate ˚ ellipsoid with diameters of 38 and 48 A Similar to other trypsin-like proteinases, two adjacent b-barrel domains each formed by six antiparallel b-strands are connected by three trans-domain segments The catalytic triad is located along the junction between the two barrels, whereas the active site cleft runs perpendicular to this junction (Fig 1) Loops The crystal structures of enteropeptidase, hepsin, matriptase and DESC1 can be structurally superimposed ˚ with r.m.s.d values < 0.8 A The highest topological similarity to DESC1 is seen with hepsin (r.m.s.d ¼ ˚ 0.70 A) with 229 Ca atoms of topologically equivalent residues, of which 96 are topologically identical The next best fit is found with matriptase and enteropeptidase, FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS 2149 Crystal structure of the catalytic domain of DESC1 O J P Kyrieleis et al Fig Stereo ribbon representation of DESC1 in complex with benzamidine (white) The residues of the catalytic triad are shown in ball and stick form (Ser195, His57 and Asp102) The termini are labeled and hydrogen bonds are shown as yellow dashed lines The figure was generated using MOLSCRIPT [30] and RASTER3D [31] ˚ both with r.m.s.d values of 0.75 A Matriptase shares 111 topologically identical residues with DESC1, whereas enteropeptidase has 87 topologically identical residues The topological equivalence of the four TTSPs forms the basis for the sequence alignment shown in Fig The numbering in the alignment refers to the chymotrypsin numbering Despite the high topological similarity found among these proteinases, significant differences exist within the loop structures that confer specificity to the enzymes for the interactions with the differing substrates and binding partners These loop regions surround the active site and are named according to the residue in the midpoint of the respective loop, as shown in Fig To the east of the active site the 37- and 60-loops border the S2¢ pocket of the proteinase The observed differences in the 37-loop result from interactions between the differing side chains in this region, which directly influence the architecture of the prime site (see below) The 60-loops of the TTSPs vary markedly in length, as well as in the conformation of the Ca trace Particularly in matriptase, this loop is distorted due to a four-residue insertion, which leads to thrombin-like shielding of the prime site [9] DESC1 carries a oneresidue deletion (Fig 3) compared with the other TTSPs, and as a consequence the prime site of DESC1 is the most narrowed by the 60-loop Important differences between the TTSPs are found in the 99-loop, which protrudes from the north rim into the active site creating a roof-like structure on top of the active site cleft Residue 99 directly limits the space for the P2 and P4 residues of the substrate peptide and contributes significantly to specificity generation This loop is six residues longer in hepsin and four residues preceding Asn99 were found to be disordered in the crystal structure The length of this loop is identical in the other TTSPs, but its conformation varies significantly due to the pronounced sequence heterogeneity found at this position (see below) The southern boundary of the active site cleft of DESC1 is formed by the 145 autolysis loop The backbone of this loop differs markedly from the other serine proteinases, making the active site cleft much narrower in DESC1 because of residues Tyr149 and Ser150, which point directly towards the active site cleft Adjacent to this autolysis loop and behind the 37-loop resides the 70-loop This binds the calcium ion in the calcium-dependent pancreatic serine proteinases via the carboxylate groups of Glu70 and Glu80 The first half (71–75) (Fig 3) of this loop is deleted in DESC1 Val70 and Lys80 replace the calcium-binding residues in DESC1 Whereas in the other TTSPs, residue 80 is hydrophobic and interacts Fig DESC1 (light blue) superimposed with the catalytic domains of human matriptase (yellow) (9), human enteropeptidase (red) (8) and human hepsin (dark blue) (7) The active site residues of DESC1, Asp189 and the bound benzamidine are shown as ball-and-stick models The termini and the important loops discussed in the text are labeled The figure was generated using MOLSCRIPT [30] and PYMOL [32] 2150 FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS O J P Kyrieleis et al Crystal structure of the catalytic domain of DESC1 Fig Structure-based sequence alignment of the human DESC1 catalytic domain with human DESC2 [2], human DESC3 [2], HAT, HAT-like [6], human matriptase (MTSP1) [9], human enteropeptidase (ENTK) [8] and human hepsin (TMPRSS1) [7] The indicated numbers correspond to the chymotrypsin numbering scheme Red arrowheads indicate the residues of the catalytic triad Blue, cyan and green arrowheads indicate residues, which confer specificity to the subsites S2, S3 ⁄ S4 and S1¢ ⁄ S2¢, respectively The secondary structural elements correspond to the crystal structure of DESC1 The figure was generated using CLUSTALX [33,34] and ESPRIPT [35] with its counterpart in position 70, the DESC1 Lys80 points in the opposite direction to interact with the carboxylate group of Glu24 Active site At first glance, the structures of the four TTSPs appear very similar (Fig 2) Closer inspection, however, reveals that the most similar regions of these proteinases mediate interaction of the two b-barrels, formation of the catalytic machinery and structures required for binding of the main chain of the substrate peptide and proper positioning of the scissile bond with respect to the catalytic serine Specificity is generated by both the physico- chemical properties of the substrate-binding subsites (e.g S4–S2¢) and the differing loops that surround the active site, which are optimized for recognition of the variable part of the substrates (side chains) Examination of the individual subsites S3–S2¢ strongly suggests that at least the structurally solved members of the four subtypes of the TTSPs will recognize largely nonoverlapping substrates Consequently, these TTSPs have differing potential to activate or inactivate the proteolytic systems of matrix metalloproteinases and uPA together with their receptors and inhibitors that have been shown to be involved in cancer-associated tissue remodeling and angiogenesis In addition, it should be possible to exploit the structural features underlying the specificity FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS 2151 Crystal structure of the catalytic domain of DESC1 differences among individual TTSPs to develop potent, selective small-molecule inhibitors that may represent an interesting new class of anticancer compounds The following analysis of the active site pockets and the key residues is based on the structurally solved members of the TTSP subfamilies Within the individual subfamilies these residues are not conserved, leading to even more pronounced diversification Detailed sequence-based information on all TTSPs can be obtained from the recent comprehensive reviews [1,2,10] S1 The following segments border the S1-specificity pocket of DESC1: Asp189–Gln192 (the basement of the pocket), Ser214–Gly219 (the entrance frame), Lys224– Tyr228 (the back of the pocket) and the disulfide bridge Cys191–Cys220 (the front of the pocket) (Fig 4A) The backbones of these segments form a deep hydrophobic pocket with the negatively charged Asp189 at its bottom Asp189 at the bottom of the pocket determines the specificity of the S1 pocket for basic residues Arg and Lys at position P1 of the substrate Consequently, in the DESC1 complex structure the bound benzamidine points with its amidino group towards the carboxylate group of Asp189 forming the canonical two-O ⁄ two-N salt bridge One additional hydrogen bond is found between the amidinonitrogen and the Asp219 carbonyl oxygen The peptide planes of the bonds between Trp215–Gly216 and Cys191–Gln192 sandwich the phenyl ring of benzamidine The DESC1 S1 pocket resembles the thrombin S1 pocket type because of the presence of an Ala rather than a Ser at position 190 The S1 specificity pockets of the TTSPs belong to the Ala190-type (DESC1, hepsin) and serine190-type (matriptase and enteropeptidase) and only one sequence displays a threonine at this position (TMPRSS4) DESC1 and other Ala190-type serine proteases prefer Arg in the P1 position versus Lys, because of the enlarged S1 pocket and the lack of a hydrogenbonding partner for P1 Lys substrates due to the Ser190Ala substitution, which compares well with the preliminary substrate-specificity analysis presented in Hobson et al [6] The190-exchange has only limited influence upon substrate discrimination, as shown by site-directed mutagenesis [11], but can be exploited for the design of small molecular mass inhibitors [7] S2 The S2 pocket is found next to the S1 pocket of DESC1 It is formed and limited by the imidazol rings 2152 O J P Kyrieleis et al of His57 and His99, which are orientated edge-to-face The S2 pocket is similar to that of uPA, which also carries a histidine at position 99, and is shaped to accept small residues like glycine or maximally alanine [12] Position 99 is the critical residue that separates the S2 from the S3 ⁄ S4 site and the chemical nature of this residue in combination with its flexibility determines the cross talk of the P2 and P3 ⁄ P4 residues bound to both pockets All TTSP proteinases differ in this residue, which is His, Phe, Lys and Asn in DESC1, matriptase, enteropeptidase and hepsin, respectively Thus, the S2 pocket of matriptase is almost closed (Fig 4B) and there will be a strong preference for glycine in the corresponding substrate residue DESC1 may accommodate alanine residues as stated, is wide open and shaped as a rather shallow depression with no exact borders In hepsin (Fig 4C) the 99-position is occupied by asparagine, which is markedly pulled out of the active site, so that the S2 site merges directly into the S3 ⁄ S4 site Compared with the other TTSPs hepsin displays the largest S2 site giving space for bulky polar residues that can interact with the carbonyl oxygen as well as with the amino group of Asn99 In comparison with hepsin in enteropeptidase (Fig 4D) Lys99 clearly separates the S2 and S3 ⁄ S4 subsites Whereas DESC1, matriptase and hepsin are shown in Fig 4A–C complexed with benzamidine (DESC1 and matriptase) and with a derivative of benzamidine (hepsin), enteropeptidase is shown in complex with the trypsinogen-activation peptide Val-(Asp)4-Lys-chloromethylketone The synthetic benzamidine-based inhibitors of DESC1, matriptase and hepsin display nicely the interaction of the S1 site, but not interact with the S2 site of these proteinases By contrast, the aspartates in positions P2 and P3 of the chloromethylketone occupy the S2 and S3 ⁄ S4 cavity in enteropeptidase Moreover, the side chain of Lys99 separates both cavities, generating specificity for these acidic residues in position P2 as well as in position P3 ⁄ P4 (Fig 4D) S3 ⁄ S4 Trp215 in DESC1, which is conserved in matriptase and enteropeptidase, but replaced by Phe in hepsin, and loop residues 173–175, build the bottom of the S3 ⁄ S4 site Central to the pocket is a flat hydrophobic area comprising residues Trp215, Tyr174 and Ala175 This can accommodate large hydrophobic residues, but some polar interactions are also possible, and these can be exploited by the design of specific inhibitors To the west, the pocket is limited by Lys224 side chain The flexibility of this side chain is reduced FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS O J P Kyrieleis et al Crystal structure of the catalytic domain of DESC1 Fig Close up of the active site of (A) human DESC1 in complex with benzamidine, (B) human matriptase in complex with benzamidine, (C) human hepsin in complex with the inhibitor 2-(2-hydroxy-phenyl)1Hbenzoimidaxole-5-carboxamidin and (D) human enteropeptidase in complex with the trypsinogen activation peptide Val-(Asp)4Lys-chloro-methylketon in stereo representation All inhibitors are represented in ball-and-stick models in black Residues discussed in the text are labeled, and hydrogen bonds are drawn as dashed black lines The figure was generated using GRASP [36] and PYMOL [32] because of an ionic interaction with the carboxylate group of Asp217 The three backbone carbonyl oxygens of residues 173–174a represent possible hydrogen- bond acceptors and point towards the S4 pocket Strong variability in length and conformation between the different TTSPs is seen in the 174-loop, which FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS 2153 Crystal structure of the catalytic domain of DESC1 limits the S3 ⁄ S4 pocket In comparison with DESC1, the matriptase and hepsin S3 ⁄ S4 pockets are significantly smaller because of the Ala175Gln substitution Structural distinctions among these three TTSPs in the 174-loop, combined with the presence of differing residue 99s (His ⁄ DESC1, Phe ⁄ matriptase, Asn ⁄ hepsin, Lys ⁄ enteropeptidase) that line the S3 ⁄ S4 pocket to the east, suggest clearly distinct P3 preferences for substrate and inhibitor recognition DESC1 prefers large hydrophobic residues with the capability to interact specifically with His99 to the east at P3 Similar to DESC1, and because of the presence of Phe99, matriptase binds preferably large hydrophobic residues at P3, but with the difference that these residues are able to interact specifically with Gln175 to the west By contrast to DESC1 and matriptase, the S3 pocket of hepsin is best suited for polar interactions to the west (Gln175) and east (Asn99) In enteropeptidase, this pocket is very narrow because of the tyrosine at position 174a and Lys99, but, depending on the residue bound to the S2 pocket, the lysine may reorient to create a broader S3 ⁄ S4 pocket The aspartate side chain of the bound chloromethylketon (Fig 4D) stacks between the aromatic side chain of Tyr175 and Lys99 The amino group of Lys99 therefore generates the specificity for acidic residues at P3 position in enteropeptidase, but the hydroxyl group of Tyr175 may also be a possible interaction partner for P3 residues S1¢ ⁄ S2¢ The S1¢ ⁄ S2¢ site is located east of the active site Ser195 It is limited by the 60- (north), 37- (east) and 145-loops (south) The bottom of the hydrophobic S1¢ ⁄ S3¢ pocket is built by the conserved disulfide bridge Cys42–Cys58 Tyr60g and Arg41 close the east site of this pocket The pocket is shielded in the north by the 60-loop residues Thr60 and Thr60a The S1¢ ⁄ S2¢ pocket of DESC1 is narrow in comparison with other TTSPs because of the one-residue deletion in the 60-loop and the Arg41 side chain, which points directly into the active site and which is stabilized in this conformation by hydrogen bonding to the Tyr60g hydroxyl group As seen in the structure-based sequence alignment (Fig 3), the residues at position 41 in the other TTSPs are significantly smaller and more hydrophobic than the Arg41 side chain in DESC1, i.e Ile (matriptase), Val (enteropeptidase) and Leu (hepsin) The S1¢ ⁄ S3¢ pockets of matriptase, hepsin and enteropeptidase are therefore more open because of the missing hydrogen bonding to the 37-loop The S2¢ pocket is formed mainly by the 145-loop In the observed conformation of Tyr149 in DESC1, the 2154 O J P Kyrieleis et al entrance to the active site is significantly restricted from the south, but this residue is completely solvent exposed and may rotate out of the way during interaction with bigger substrates The exposed hydroxyphenyl group of Tyr149 might even represent a secondary binding site for substrates or inhibitors Substrate specificity of DESC2, -3, HAT and HAT-like Comparison of the primary sequences of DESC2, -3, HAT and HAT-like with DESC1 reveals that the residues, which confer specificity to subsites S3 ⁄ 4, S2 and S1¢ ⁄ 2¢, differ markedly in the members of this subfamily, as shown in Fig By contrast, the S1 subsite is characterized by the conserved residues Asp189 and Ala190 of the Ala190-type of serine proteases which prefer Arg to Lys at position P1 Also conserved are residues Trp215, Lys224 and Trp174 forming the flat hydrophobic area at the bottom of the S3 ⁄ subsite Differences are found in the 174-loop residues, which represent the interacting partners for P4 residues In combination with the different residues for DESC2, -3, HAT and HAT-like in the 99-position it is therefore likely that the five known members of this subfamily have different preferences for residues bound to subsites S3 ⁄ S4 and S2 With regard to the S1¢ ⁄ 2¢ subsite, the residues of the 60-loop mainly determine the substrate specificity The alignment in Fig clearly shows that these residues differ not only in their chemical nature, but also in the flexibility of the different members of the HAT ⁄ HAT-like ⁄ DESC subfamily In conclusion, it is possible to summarize that not only the members of the four TTSP subfamilies display different substrate specificity, but also members within the subfamily recognize largely nonoverlapping substrates Surface Hepsin was crystallized as a complete extracellular domain including, in addition to the proteinase domain, an N-terminal scavenger receptor cystein-rich (SRCR) domain, which was rigidly bound to the back of the proteinase domain in the crevice between the C-terminal helix, the 204-loop and the 126-loop [7] The C-terminus of hepsin is elongated by 11 residues in comparison with the other structurally known TTSPs, which leads to elongation of helix H2 and an additional loop structure that interacts with the core of the proteinase Also in DESC1, a noncharged surface broken only by the guanidyl group of Arg120 in the center of this surface is found at this position with FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS O J P Kyrieleis et al Crystal structure of the catalytic domain of DESC1 Fig Solid-surface representations of human hepsin (A), human DESC1 (B), human matriptase (C) and human enteropeptidase (D) The enzymes are rotated around a vertical axis for 180° in comparison with the standard orientation in Fig Hepsin (A) is shown bound to the SRCR domain, which is drawn as golden Ca-trace Hydrophobic residues are in blue, and polar residues are in red The corresponding residues Ile244 (DESC1), Val244 (matriptase) and Thr244 (hepsin) are shown in ball-andstick models The figure was prepared using GRASP [36] and PYMOL [32] several hydrophobic residues exposed to the solvent (Tyr114, Tyr208, Ile206) (Fig 5B) This surface would be well suited to an interaction with the N-terminal SEA-domain of DESC1, as seen in hepsin with the SRCR domain But in comparison with hepsin, DESC1 carries an Ile244 on the shorter C-terminus The conformation of this residue is changed in DESC1 and matriptase (Val244) in a way that the hydrophobic side chain of this residue can fill a hydrophobic hole that is occupied by Leu51 of the SRCR domain in hepsin (Fig 5A) This conformation does not seem to be an artifact of the missing N-terminal domain because in DESC1, as well as in matriptase, the conformation of this residue is stabilized by a salt bridge of the C-terminal carboxylate group with the guanidyl group of Arg235 In hepsin the less hydrophobic Thr244 replaces the Ile244 side chain of DESC1 As part of the additional loop structure in hepsin, Thr244 is shifted to the north of the hydrophobic interaction surface so that the Leu51 side chain of the SRCR domain can bind into the hydrophobic depression This interaction is not possible in DESC1 and matriptase because of the above-mentioned position of the C-terminal Ile244 (DESC1) and Val244 (matriptase) In DESC1, the exposed Arg120 side chain in the center of the interaction surface may serve as an interaction partner for negatively charged residues of the SEA domain, in addition to interactions of the surrounding hydrophobic residues Coloring of the surfaces according to hydrophobic and polar residues clearly shows that the hydrophobic interaction surface positioned at the backside of the proteinase domain is a conserved feature of all structural known TTSPs Moreover, Fig 5A shows that the C-terminus of the SCRC domain runs in a hydrophobic canyon connecting the left lower part of the hydrophobic interaction surface with the front site of the molecule This canyon, as well as the binding mode of the C-terminus, is also conserved across all members, as seen in Fig 5B–D Remarkable on the surfaces of matriptase and enteropeptidase is a second interaction surface positioned above the first to the right In matriptase a small hydrophobic channel connects both interaction surfaces and could probably harbor a linker peptide between two N-terminal domains In enteropeptidase, a bridge of polar residues separates both interaction surfaces In both DESC1 and hepsin the surface region of the second interaction surface is formed by a mixture of hydrophobic and polar residues, which not create a continuous polar or hydrophobic surface The second interaction surface is therefore missing in DESC1 and hepsin This fits well with the domain organization in the extracellular stem region of known TTSP structures Whereas in the stem regions of FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS 2155 Crystal structure of the catalytic domain of DESC1 DESC1 ⁄ hepsin, beside the proteinase domain, only a single SEA or SRCR domain is found, the stem regions of matriptase and enteropeptidase are extended to six (matriptase) and seven (enteropeptidase) additional domains [2] These additional domains may represent the interaction partners of the second interaction surfaces observed in matriptase (Fig 5C) and enteropeptidase (Fig 5D) Conclusions O J P Kyrieleis et al The preferred serine at P2 and the non-natural d-residue present at position P3 in the screened inhibitors is also observed for peptidomimetic inhibitors of uPA These related inhibitors have been crystallized in complex with urokinase [12] and a related binding mode of the found inhibitor to DESC1 may be expected In the uPA complex structures, the P2 serine binds to the small S2 pocket which, as in DESC1, is separated by His99 from the S4 pocket; the P3 side chain of the uPA inhibitors interacts due to its d-configuration with the S3 ⁄ S4 pocket Substrate specificity The substrate specificity of DESC1’s proteinase domain, as deduced from the analysis of this crystal structure with large hydrophobic residues in P4 ⁄ P3, for small residues in P2, Arg or Lys in P1 and hydrophobic residues in P1¢ and P3¢ is in agreement with the work of Hobson et al [6] The authors found the highest enzymatic activity of DESC1 with chromogenic substrates containing Ala in positions P4 and P3 and Pro in position P2, followed by substrates containing Phe and Gly in positions P3 and P2 Acidic residues in position P3 are still processed, but with much lower enzymatic activity [6] Taken together, the predicted substrate sequence differs markedly from other known TTSP structures This unique fine structure of the binding pockets could consequently be exploited in a mixture-based peptidic inhibitor library screen, arrayed in a positional scanning format (Corvas International, personnel communication) This screening suggested that DESC1 prefers hydroxyproline, proline, and serine at P2; phenyl glycine, d-phenyl glycine and d-benzylserine at P3, but can also accommodate well d-lysine and d-serine at this position; and 5-phenylthiophene-2,5-disulfonyl 3,5-dichlorobenzene sulfonamide, 3-nitrobenzene sulfonamide and 4-biphenylsulfonamide at P4 Based on these data, 47 DESC1 inhibitors were synthesized The most potent of these inhibitors (3,4-diCl,2-O(4,5-dihydroxyPent))PhEt-CO-M(O2)-S-(2-amdn)thiophene-5MeAm (Fig 6), had a Ki value of 6.4 nm for DESC1 Fig Structural formula of (3,4-diCl,2-O(4,5-dihydroxyPent))PhEtCO-M(O2)-S-(2-amdn)thiophene-5-MeAm Residues P1 to P4 are indicated in bold 2156 Inhibition of DESC1 Regulation of proteolytic activity by Kunitz-type inhibitors is commonly observed in trypsin-like serine proteinases, including the TTSP matriptase [13] Although it remains unclear whether physiologically relevant regulation of DESC1 involves interaction with Kunitztype inhibitors, it is clear that DESC1 exhibits a high affinity for BPTI (unpublished data), a prototypical Kunitz domain Matriptase is efficiently inhibited by hepatocyte growth factor activator inhibitor (HAI)-1, a transmembrane protein, which consists of 478 residues and contains two Kunitz-type domains [14] Only the first Kunitz-type HAI-1 has inhibitory properties on matriptase [9], and, as expected, the reactive center loop of this Kunitz domain, which is Gly12(I)Arg13(I)-Cys14(I)-Arg15(I)-Gly16(I)-Ser17(I)-Phe18(I) [using the BPTI nomenclature, with Arg15(I) | Gly 16(I) as the scissile bond], matches optimal subsite occupancy for matriptase relatively well, contributing to the tight binding of the enzyme [9] The efficient inhibition of matriptase by HAI-1 appears to represent a key regulatory constraint on matriptase activity in vivo However, the distinct specificities of matriptase and DESC1 suggest that it is unlikely that DESC1 is a physiologically relevant target for HAI-1; neither the first nor the second Kunitz-type domain match the reported substrate specificity of DESC1 [14] Other Kunitz-type inhibitors present in human plasma include HAI-2 [15], amyloid b protein precursor [16] and tissue factor pathway inhibitor [17,18], but the existence and ⁄ or identity of (the) physiologically relevant inhibitor(s) of DESC1 remain uncertain Another commonly observed type of inhibition for serine proteinases is the inhibition by serine proteinase inhibitors (serpins) The serpins form a family of homologous, large (glyco-) proteins comprising about 400 amino acid residues Serpin inhibitors interact with their cognate serine proteinases via an exposed binding loop, which acts as a potential substrate [19,20] Expression of human DESC1 and its mouse ortholog in oral epidermal and FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS O J P Kyrieleis et al male reproductive tissues suggest a similar role of both proteinases Mouse DESC1 was found to form efficiently inhibitory complexes with the serpins plasminogen activator inhibitor (PAI-1) and protein C inhibitor (PCI), which both are present in DESC1expressing tissue [6] The stable inhibitory complexes of mouse DESC1 with PAI-1 and PCI indicate that serpins might be critical regulators of the proteolytic activity of DESC1 The reactive site loop sequences of PAI-1 and PCI fit to the active site geometry of human DESC1, without matching the optimal docking sequence of DESC1 The reactive loop sequences are Val-Ser-Ala-ArgflMet-Ala-Pro and Phe-Thr-Phe-ArgflSer-Ala-Arg for PAI-1 and PCI, respectively [6] By contrast, the reactive site loops of a1-antichymotrypsin and heparin cofactor II contain leucine instead of arginine as P1 residues, which explains why the formation of a stable inhibitory complexes of these serpins is not possible with DESC1 [6] However, predictions of serpin–proteinase interactions are notoriously difficult because of the flexible nature of their reactive site segment and ⁄ or possible exosite binding [21] Domain structure Surface analysis of DESC1 suggests a possible rigid domain association between the N-terminal SEA domain and the back site of the proteinase domain This interaction would fix the SEA domain in a location on the opposite side of the proteinase domain from the active site cleft It seems very unlikely, therefore, that the SEA domain would directly influence the binding of either substrates or inhibitors into the active site cleft of the DESC1 Instead, because SEA domains are proposed to bind O-glucosidic-linked proteoglycans present in the carbohydrate-rich environments [2,22] of the extracellular matrix, it seems more likely that the SEA domain functions by orienting the active site cleft of DESC1 toward plasma and ⁄ or extracellular spaces and away from the cell surface and ⁄ or the extracellular matrix The SEA domain may also contribute to the adhesion properties of DESC1-expressing cells and might localize ‘shed’ DESC1 in appropriate microenvironments Corresponding surface analysis of other structurally investigated TTSPs suggests that rigid association with at least one N-terminal domain appears to be a common structural feature of TTSPs Moreover, it suggests that orientation of the active site towards soluble factors and away from the cell surface may be generally important for the function of members of this intriguing and emerging subfamily of serine proteases Crystal structure of the catalytic domain of DESC1 Experimental procedures Cloning, expression and purification Cloning, expression and purification of the DESC1 catalytic domain was performed at Corvas International (San Diego, CA) (to be published) In short, human umbilical vein endothelial cells (HUVEC P145) were purchased from Clonetics (CC-2519) All subsequent cell manipulations were carried out according to the manufacturer’s instructions Cells were allowed to grow to $ 90% confluence RNA was isolated and enriched for poly(A+) RNAs on oligo(dT) beads (Oligotex, Qiagen, Chatsworth, CA, USA) The HUVEC poly (A+) RNAs were converted to single-stranded cDNA and subjected to PCR using primers that correspond to two highly conserved regions in all trypsin-like serine proteinases that resulted in the expected PCR products ranging from 400 to 500 bp Purified DNA fragments were cloned and sequenced To obtain the cDNA that encodes the entire proteinase domain of DESC1, rapid amplification of cDNA ends reactions were performed on a human prostate Marathon-Ready cDNA (Clontech, Mountain View, CA, USA) Two fragments were isolated and confirmed by Southern analysis using the internal cDNA fragment as the probe and by DNA sequence analysis The cDNA encoding DESC1 was cloned into a derivative of the Pichia pastoris vector pPIC9K (Invitrogen, Carlsbad, CA, USA) Pichia clones transformed with DNA encoding DESC1 were screened for production of the protein by assaying conditioned media for enzymatic activity against Spectrozyme t-PA (CH3SO2-D-HHT-Gly-Arg-pNA*HCl; American Diagnostica, Stanford, CT, USA) Details of the expression and purification of multimilligram amounts of human DESC1 will be published separately Briefly, the protein was expressed in the SMD 1168 strain of P pastoris using a variant of the pPIC9K vector Cells were grown in 5-L fermentation vessels, supernatant was clarified and collected, and DESC1 was purified by using affinity chromatography on a benzamidine column followed by anion exchange chromatography on a Q-Sepharose column (Amersham Biosciences, Inc., Piscataway, NJ, USA) and on a Source 15Q column (Amersham Biosciences) Fractions containing protein were pooled, and benzamidine was added to a final concentration of 10 mm The protein purity was examined by SDS ⁄ PAGE, and the protein concentration was determined at A280 (using an À extinction coefficient of 2.012 mgỈA280 ) DESC1 ⁄ benzamidine crystals Cloning, expression and purification yielded milligram quantities of highly purified, mature DESC1 catalytic domain Fractions of the enzyme were inhibited with benzamidine, concentrated to mgỈmL)1 and subjected to FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS 2157 Crystal structure of the catalytic domain of DESC1 O J P Kyrieleis et al screenings of crystallization conditions Initial crystal needles appeared after days After optimization of the conditions, rod-like crystals of benzamidine-inhibited DESC1 were grown from 0.1 m Tris pH ¼ 8.5, 8% (m ⁄ w) PEG 8000 at 18 °C using the sitting drop vapor diffusion technique They belong to orthorhombic space group P21212 ˚ ˚ ˚ with the cell constants a ¼ 47.9 A, b ¼ 70.2 A, c ¼ 80.2 A, a ¼ b ¼ c ¼ 90 ° and diffract X-rays to a limiting resolu˚ tion of 1.6 A with one molecule in the asymmetric unit Structure determination and crystallographic refinement ˚ A complete native data set to 1.6 A resolution was collected at room temperature from a single crystal of the DESC1– benzamidine complex mounted on a rotating anode generator (Rigaku, Tokyo, Japan) equipped with an image plate detector (Mar Research, Hamburg, Germany) These data were integrated with the mosflm package [23] and scaled with scala from the ccp4 [24] program suite (Table 1) To determine the position of DESC1 molecules within the asymmetric unit rotation and translation searches were car˚ ried out with amore using data from 20 to 3.5 A, and an enteropeptidase search model The best solution had a correlation factor of 0.36 and an R-factor of 0.46; the corresponding values of the next best solution were 0.22 and 0.50 Crystallographic refinement was carried out over several cycles consisting of model building performed with O [25– 27] and conjugate gradient minimization and simulated annealing with the cns [28] program suite, using the target parameters of Engh and Huber [29] The refinement procedure leads to a model with excellent parameters (Table 1) In the final model building ⁄ refinement cycles water molecules were inserted and individual restrained atomic B-values were refined We omitted 4.3% of all reflections from the refinement to calculate the Rfree The final R and Rfree values of the model are 0.21 and 0.22 for the complete data ˚ set up to 1.6 A The electron density of the whole main chain of the catalytic domain (B-chain) of DESC1 is well defined Only a few side chains on the surface of the molecule are partially undefined in the electron density The occupancy of all undefined atoms was set to zero Quality of the model The backbone of the complex is completely well defined in terms of electron density In the benzamidine-inhibited DESC1 structure, only the four surface-exposed side chains of Tyr114, Glu129, Lys137 and Glu186 are not defined by electron density The DESC1–benzamidine complex shows excellent stereochemistry, with 98.5% of all residues in the most favored and favored regions of the Ramachandran ˚ plot and r.m.s.d values for bond and angle of 0.005 A and 1.37 ° as shown in Table References Table Data collection and refinement statistics of DESC1 Numbers in parentheses are for the outermost shell of the data Data collection Space group ˚ Unit cell dimesions (A) ˚ Wavelength (A) ˚ Resolution of data (A) Completeness (%) Rmerge Multiplicity Refinement ˚ Resolution range (A) Non-hydrogen atoms Water molecules Rcryst (%)a Rfree (%)b Reflections (working ⁄ test) ˚ Average B-factors (A) [2] RMS deviations from ideal stereochemistry ˚ Bond lengths (A) Bond angles (°) P21212 a ¼ 47.9 b ¼ 70.2 c ¼ 80.2 a ¼ b ¼ c ¼ 90° 1.54 20–1.6 98.1 (89.8) 0.079 (0.186) 4.6 (2.1) 20–1.6 1930 130 21.13 (24.3) 22.14 (26.4) 29853 ⁄ 1556 28.39 0.005415 1.36707 Crystallographic R-factor ¼ Shkl||Fobs| ) k|Fcalc|| ⁄ Shkl|Fobs| R-factor ¼ ShklTest||Fobs| ) k|Fcalc|| ⁄ ShklTest|Fobs| a 2158 b Free Szabo R, Wu QY, Dickson RB, Netzel-Arnett S, Antalis TM & Bugge TH (2003) Type II transmembrane serine proteases Thromb Haemost 90, 185–193 Netzel-Arnett S, Hooper JD, Szabo R, Madison EL, Quigley JP, Bugge TH & Antalis AM (2003) Membrane anchored serine proteases: a rapidly expanding group of cell surface proteolytic enzymes with potential roles in cancer Cancer Metastasis Rev 22, 237–258 Lu DS, Yuan X, Zheng XL & Sadler JE (1997) Bovine proenteropeptidase is activated by trypsin, and the specificity of enteropeptidase depends on the heavy chain J Biol Chem 272, 31293–31300 Lang JC & Schuller DE (2001) Differential expression of a novel serine protease homologue in squamous cell carcinoma of the head and neck Br J Cancer 84, 237–243 Sedghizadeh PP, Mallery SR, Thompson SJ, Kresty L, Beck FM, Parkinson EK, Biancamano J & Lang JC (2006) Expression of the serine protease Desc1 correlates directly with normal keratinocyte differentiation and inversely with head and neck squamous cell carcinoma progression Head Neck J 28, 432–440 Hobson JP, Netzel-Arnett S, Szabo R, Rehault SM, Church FC, Strickland DK, Lawrence DA, Antalis TM & Bugge TH (2004) Mouse DESC1 is located within a FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS O J P Kyrieleis et al 10 11 12 13 14 15 16 17 cluster of seven DESC1-like genes and encodes a type II transmembrane serine protease that forms serpin inhibitory complexes J Biol Chem 279, 46981–46994 Somoza JR, Ho JD, Luong C, Ghate M, Sprengeler PA, Mortara K, Shrader WD, Sperandio D, Chan H, McGrath ME et al (2003) The structure of the extracellular region of human hepsin reveals a serine protease domain and a novel scavenger receptor cysteine-rich (SRCR) domain Structure 11, 1123–1131 Lu DS, Futterer K, Korolev S, Zheng XL, Tan K, Waksman G & Sadler JE (1999) Crystal structure of enteropeptidase light chain complexed with an analog of the trypsinogen activation peptide J Mol Biol 292, 361–373 Friedrich R, Fuentes-Prior P, Ong E, Coombs G, Hunter M, Oehler R, Pierson D, Gonzalez R, Huber R, Bode W et al (2002) Catalytic domain structures of MT-SP1 ⁄ matriptase, a matrix-degrading transmembrane serine proteinase J Biol Chem 277, 2160–2168 Hooper JD, Clements JA, Quigley JP & Antalis TM (2001) Type II transmembrane serine proteases – insights into an emerging class of cell surface proteolytic enzymes J Biol Chem 276, 857–860 Sichler K, Hopfner KP, Kopetzki E, Huber R, Bode W & Brandstetter H (2002) The influence of residue 190 in the S1 site of trypsin-like serine proteases on substrate selectivity is universally conserved FEBS Lett 530, 220–224 Zeslawska E, Schweinitz A, Karcher A, Sondermann P, Sperl S, Sturzebecher J & Jacob U (2000) Crystals of the urokinase type plasminogen activator variant beta c-uPA in complex with small molecule inhibitors open the way towards structure-based drug design J Mol Biol 301, 465–475 Lin CY, Anders J, Johnson M & Dickson RB (1999) Purification and characterization of a complex containing matriptase and a Kunitz-type serine protease inhibitor from human milk J Biol Chem 274, 18237–18242 Shimomura T, Denda K, Kitamura A, Kawaguchi T, Kito M, Kondo J, Kagaya S, Qin L, Takata H, Miyazawa K et al (1997) Hepatocyte growth factor activator inhibitor, a novel Kunitz-type serine protease inhibitor J Biol Chem 272, 6370–6376 Kawaguchi T, Qin L, Shimomura T, Kondo J, Matsumoto K, Denda K & Kitamura N (1997) Purification and cloning of hepatocyte growth factor activator inhibitor type 2, a Kunitz-type serine protease inhibitor J Biol Chem 272, 27558–27564 Seguchi K, Kataoka H, Uchino H, Nabeshima K & Koono M (1999) Secretion of protease nexin-II ⁄ amyloid beta protein precursor by human colorectal carcinoma cells and its modulation by cytokines ⁄ growth factors and proteinase inhibitors Biol Chem 380, 473–483 Kataoka H, Uchino H, Asada Y, Hatakeyama K, Nabeshima K, Sumiyoshi A & Koono M (1997) Analy- Crystal structure of the catalytic domain of DESC1 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 sis of tissue factor and tissue factor pathway inhibitor expression in human colorectal carcinoma cell lines and metastatic sublines to the liver Int J Cancer 72, 878–884 Kataoka H, Itoh H & Koono M (2002) Emerging multifunctional aspects of cellular serine proteinase inhibitors in tumor progression and tissue regeneration Pathol Int 52, 89–102 Gettins PGW (2002) Serpin structure, mechanism, and function Chem Rev 102, 4751–4803 Bode W & Huber R (1992) Natural protein proteinase –inhibitors and their interaction with proteinases Eur J Biochem 204, 433–451 Huber R & Carrell RW (1989) Implications of the 3-dimensional structure of alpha-1-antitrypsin for structure and function of serpins Biochemistry 28, 8951– 8966 Bork P & Patthy L (1995) The sea module – a new extracellular domain associated with O-glycosylation Protein Sci 4, 1421–1425 Leslie AGW (1992) Recent changes to the MOSFLM package for processing film and image plate data Joint CCP4 + ESF-EAMCB Newsletter on Protein Crystallography 26, 11–20 Bailey S (1994) The CCP4 suite – programs for protein crystallography Acta Crystallogr D Biol Crystallogr 50, 760–763 Jones TA, Zou JY, Cowan SW & Kjeldgaard M (1991) Improved methods for building protein models in electron-density maps and the location of errors in these models Acta Crystallogr A 47, 110–119 Jones TA (1978) A graphic model building and refinement system for macromolecular structure determination J Appl Crystallogr 11, 268–272 Jones TA & Thirup S (1986) Using known substructures in protein model-building and crystallography EMBO J 5, 819–822 Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS et al (1998) Crystallography and NMR system: a new software suite for macromolecular structure determination Acta Crystallogr D Biol Crystallogr 54, 905–921 Engh RA & Huber R (1991) Accurate bond and angle parameters for X-Ray protein-structure refinement Acta Crystallogr A 47, 392–400 Kraulis PJ (1991) Molscript – a program to produce both detailed and schematic plots of protein structures J Appl Crystallogr 24, 946–950 Merritt EA & Murphy MEP (1994) Raster3d Version 2.0 – a program for photorealistic molecular graphics Acta Crystallogr D Biol Crystallogr 50, 869– 873 DeLano WL (2002) The PyMOL Molecular Graphics System DeLano Scientific, Palo Alto, CA FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS 2159 Crystal structure of the catalytic domain of DESC1 33 Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F & Higgins DG (1997) The CLUSTAL_X Windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 25, 4876–4882 34 Jeanmougin F, Thompson JD, Gouy M, Higgins DG & Gibson TJ (1998) Multiple sequence alignment with Clustal X Trends Biochem Sci 23, 403–405 2160 O J P Kyrieleis et al 35 Gouet P, Courcelle E, Stuart DI & Metoz F (1999) ESPript: analysis of multiple sequence alignments in PostScript Bioinformatics 15, 305–308 36 Nicholls A, Bharadwaj R & Honig B (1993) Grasp – graphical representation and analysis of surface-properties Biophys J 64, A166–A166 FEBS Journal 274 (2007) 2148–2160 ª 2007 Max-Planck-Insitute of Biochemistry Journal compilation ª 2007 FEBS ... structure of the catalytic domain of DESC1 Scheme Domain organization of human DESC1 membrane region The extracellular part of DESC1 consists of a 120-amino acid SEA domain followed by the C-terminal... growth factors and proteinase inhibitors Biol Chem 380, 473–483 Kataoka H, Uchino H, Asada Y, Hatakeyama K, Nabeshima K, Sumiyoshi A & Koono M (1997) Analy- Crystal structure of the catalytic domain. .. that the SEA domain functions by orienting the active site cleft of DESC1 toward plasma and ⁄ or extracellular spaces and away from the cell surface and ⁄ or the extracellular matrix The SEA domain