Tài liệu Báo cáo khoa học: Crystal structure of the catalytic domain of DESC1, a new member of the type II transmembrane serine proteinase family pptx

The proteinase domain of DESC1 exhibits a trypsin-like serine pro-teinase fold with a thrombin-like S1 pocket, a urokinase-type plasminogen activator-type S2 pocket, to accept small resi

member of the type II transmembrane serine proteinase family

Otto J P Kyrieleis1,*, Robert Huber1,2, Edgar Ong3, Ryan Oehler3, Mike Hunter3,

Edwin L Madison3and Uwe Jacob1,

1 Max-Planck-Institut fu¨r Biochemie, Martinsried, Germany

2 Cardiff University, UK

3 Corvas International, San Diego, CA, USA

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 sub-families: DESC⁄ human airway trypsin (HAT) ⁄ HAT-like type (DESC1–3, HAT, HAT-HAT-like 4); matriptase

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 fu¨r

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 cur-rently 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 associ-ated with cell differentiation and have been described as tumor markers TTSPs belong to the type II transmembrane proteins that display, in addi-tion to a C-terminal trypsin-like serine proteinase domain, a differing set of stem domains, a transmembrane segment and a short N-terminal cytoplas-mic 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 benza-midine The proteinase domain of DESC1 exhibits a trypsin-like serine pro-teinase fold with a thrombin-like S1 pocket, a urokinase-type plasminogen activator-type S2 pocket, to accept small residues, and an open hydro-phobic S3⁄ S4 cavity to accept large hydrophobic residues The deduced substrate specificity for DESC1 differs markedly from that of other struc-turally 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

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.

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

polysaccha-rides, lipids or proteoglycans) These complex stem

regions and the cytoplasmic domain, which may

inter-act with cellular signaling molecules and the

cytoskele-ton, 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 1 (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

downregu-lated 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

Further-more, Sedghizadeh et al [5] were able to show that

DESC1 is upregulated during the induction of terminal

keratinocyte differentiation, supporting a role in

nor-mal epithelial turnover These results suggest that

DESC1 may function in regular epithelial

differenti-ation 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

cytoplas-mic region followed by 14 residues of a putative

trans-as shown in Scheme 1

DESC2 and DESC3 were subsequently identified by database searches [2] In contrast to DESC1, many TTSPs are overexpressed by tumor cells (e.g matrip-tase, 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 carcino-genesis 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 cata-lytic triad is located along the junction between the two barrels, whereas the active site cleft runs perpen-dicular to this junction (Fig 1)

Loops The crystal structures of enteropeptidase, hepsin, matri-ptase 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.d¼ 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,

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 3 The numbering in the alignment

refers to the chymotrypsin numbering Despite the

high topological similarity found among these

protein-ases, significant differences exist within the loop

struc-tures 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

mid-point of the respective loop, as shown in Fig 2 To

the east of the active site the 37- and 60-loops border

the S2¢ pocket of the proteinase The observed

differ-ences 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

Parti-cularly 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

one-residue 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 differ-ences 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 contri-butes significantly to specificity generation This loop

is six residues longer in hepsin and four residues pre-ceding Asn99 were found to be disordered in the crys-tal 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 mark-edly from the other serine proteinases, making the act-ive 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 pancre-atic 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 1 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 RASTER 3 D [31].

Fig 2 DESC1 (light blue) superimposed with the catalytic domains of human matrip-tase (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].

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

medi-ate 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

cata-lytic 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) Examina-tion 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 nonover-lapping substrates Consequently, these TTSPs have dif-fering 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

Fig 3 Structure-based sequence alignment of the human DESC1 catalytic domain with human DESC2 [2], human DESC3 [2], HAT, HAT-like 4 [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 arrow-heads 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].

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

indi-vidual 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

bot-tom 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

benza-midine points with its amidino group towards the

carb-oxylate 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

pro-teases prefer Arg in the P1 position versus Lys, because

of the enlarged S1 pocket and the lack of a

hydrogen-bonding 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

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 deter-mines 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 pref-erence for glycine in the corresponding substrate resi-due 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 giv-ing 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 entero-peptidase (Fig 4D) Lys99 clearly separates the S2 and S3⁄ S4 subsites Whereas DESC1, matriptase and hep-sin are shown in Fig 4A–C complexed with benza-midine (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 benz-amidine-based inhibitors of DESC1, matriptase and hepsin display nicely the interaction of the S1 site, but

do 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 cav-ity in enteropeptidase Moreover, the side chain of Lys99 separates both cavities, generating specificity for these acidic residues in position P2 as well as in posi-tion 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

because of an ionic interaction with the carboxylate

group of Asp217 The three backbone carbonyl

oxy-gens 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

Fig 4 Close up of the active site of (A)

human DESC1 in complex with

benzami-dine, (B) human matriptase in complex with

benzamidine, (C) human hepsin in complex

with the inhibitor

2-(2-hydroxy-phenyl)1H-benzoimidaxole-5-carboxamidin and (D)

human enteropeptidase in complex with the

trypsinogen activation peptide Val-(Asp) 4

-Lys-chloro-methylketon in stereo

repre-sentation All inhibitors are represented in

ball-and-stick models in black Residues

discussed in the text are labeled, and

hydro-gen bonds are drawn as dashed black lines.

The figure was generated using GRASP [36]

and PYMOL [32].

limits the S3⁄ S4 pocket In comparison with DESC1,

the matriptase and hepsin S3⁄ S4 pockets are

signifi-cantly 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

sub-strate 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,

matrip-tase 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

con-trast to DESC1 and matriptase, the S3 pocket of

hep-sin 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

posi-tion 174a and Lys99, but, depending on the residue

bound to the S2 pocket, the lysine may reorient to

cre-ate 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

spe-cificity for acidic residues at P3 position in

enteropepti-dase, 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

(hep-sin) 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

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 interac-tion with bigger substrates The exposed hydroxy-phenyl group of Tyr149 might even represent a secondary binding site for substrates or inhibitors

Substrate specificity of DESC2, -3, HAT and HAT-like 4

Comparison of the primary sequences of DESC2, -3, HAT and HAT-like 4 with DESC1 reveals that the residues, which confer specificity to subsites S3⁄ 4, S2 and S1¢ ⁄ 2¢, differ markedly in the members of this sub-family, as shown in Fig 3 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⁄ 4 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 4 in the 99-position it is therefore likely that the five known members of this subfamily have different preferences for residues bound to sub-sites S3⁄ S4 and S2 With regard to the S1¢ ⁄ 2¢ subsite, the residues of the 60-loop mainly determine the sub-strate specificity The alignment in Fig 3 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

do the members of the four TTSP subfamilies dis-play 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 sur-face broken only by the guanidyl group of Arg120 in the center of this surface is found at this position with

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

con-formation 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

matrip-tase 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 accord-ing 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 entero-peptidase is a second interaction surface positioned above the first to the right In matriptase a small hydrophobic channel connects both interaction surfa-ces 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 mix-ture of hydrophobic and polar residues, which do 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

Fig 5 Solid-surface representations of

human hepsin (A), human DESC1 (B),

human matriptase (C) and human

entero-peptidase (D) The enzymes are rotated

around a vertical axis for 180 in comparison

with the standard orientation in Fig 1

Hep-sin (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

resi-dues Ile244 (DESC1), Val244 (matriptase)

and Thr244 (hepsin) are shown in

ball-and-stick models The figure was prepared using

GRASP [36] and PYMOL [32].

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)

addi-tional domains [2] These addiaddi-tional domains may

represent the interaction partners of the second

inter-action surfaces observed in matriptase (Fig 5C) and

enteropeptidase (Fig 5D)

Conclusions

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

res-idues in P2, Arg or Lys in P1 and hydrophobic resres-idues

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

con-taining 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

communica-tion) 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

sul-fonamide and 4-biphenylsulsul-fonamide at P4 Based on

these data, 47 DESC1 inhibitors were synthesized

The most potent of these inhibitors

(3,4-diCl,2-O(4,5-di-hydroxyPent))PhEt-CO-M(O2

)-S-(2-amdn)thiophene-5-MeAm (Fig 6), had a Kivalue of 6.4 nm for DESC1

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 com-plex 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

Inhibition of DESC1 Regulation of proteolytic activity by Kunitz-type inhib-itors is commonly observed in trypsin-like serine pro-teinases, including the TTSP matriptase [13] Although

it remains unclear whether physiologically relevant regulation of DESC1 involves interaction with Kunitz-type 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 inhib-itor(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 pro-teinases 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

Fig 6 Structural formula of

(3,4-diCl,2-O(4,5-dihydroxyPent))PhEt-CO-M(O2)-S-(2-amdn)thiophene-5-MeAm Residues P1 to P4 are

indicated in bold.

sequence of DESC1 The reactive loop sequences are

Val-Ser-Ala-ArgflMet-Ala-Pro and

Phe-Thr-Phe-Argfl-Ser-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

forma-tion 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

seg-ment 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

loca-tion on the opposite side of the proteinase domain

from the active site cleft It seems very unlikely,

there-fore, 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

microenvi-ronments 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

mem-bers of this intriguing and emerging subfamily of

serine proteases

beads (Oligotex, Qiagen, Chatsworth, CA, USA) The HUVEC poly (A+) RNAs were converted to single-stran-ded cDNA and subjected to PCR using primers that corres-pond to two highly conserved regions in all trypsin-like serine proteinases that resulted in the expected PCR prod-ucts 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, Moun-tain 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

American Diagnostica, Stanford, CT, USA)

Details of the expression and purification of multimilli-gram amounts of human DESC1 will be published sepa-rately 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-Seph-arose column (Amersham Biosciences, Inc., Piscataway,

NJ, USA) and on a Source 15Q column (Amersham Bio-sciences) Fractions containing protein were pooled, and benzamidine was added to a final concentration of 10 mm

extinction coefficient of 2.012 mgÆA280 1)

DESC1⁄ benzamidine crystals

Cloning, expression and purification yielded milligram quantities of highly purified, mature DESC1 catalytic domain Fractions of the enzyme were inhibited with