Biol. Chem., Vol. 393, pp. 331–341, May 2012 • Copyright © by Walter de Gruyter • Berlin • Boston. DOI 10.1515/BC-2011-250

Non-combinatorial library screening reveals subsite cooperativity and identifi es new high-effi ciency substrates for -related peptidase 14

Simon J. de Veer1 , Joakim E. Swedberg 2 , -like serine proteases (Clements et al. , 2001 ; Yousef Edward A. Parker1 and Jonathan M. Harris1, * and Diamandis , 2001 ) transformed perceived associations 1 Institute of Health and Biomedical Innovation , Queensland between this family of enzymes and human (patho)physi- University of Technology, Brisbane, Queensland 4059 , ology (Borgono and Diamandis , 2004 ; Clements et al. , Australia 2004; Hollenberg et al., 2008; Sotiropoulou et al., 2009). 2 Institute for Molecular Bioscience , University of Potential functions both in homeostasis and disease have Queensland, Brisbane, Queensland 4072 , Australia now emerged for many of the newly identifi ed KLKs across a wide range of tissues including the prostate (Takayama * Corresponding author et al. , 2001 ; Michael et al. , 2006 ), skin (Brattsand et al. , 2005 ; e-mail: [email protected] Deraison et al., 2007), and central nervous system (Shimizu et al. , 1998 ; Scarisbrick et al. , 2002 ). The recent isolation of Abstract active KLK14 in extracts of human skin by Stefansson et al. (2006) prompted suggestions that KLK14 formed part of the An array of substrates link the tryptic , protease cascade responsible for maintaining the epidermal kallikrein-related peptidase 14 (KLK14), to physiological barrier (Brattsand et al. , 2005 ; Deraison et al. , 2007 ). This functions including desquamation and activation of signal- view was strengthened by tissue expression profi ling, which ing molecules associated with infl ammation and cancer. revealed that KLK14 was most abundant in the skin, with Recognition of protease cleavage sequences is driven by comparatively lower levels in steroid hormone-sensitive tis- complementarity between exposed substrate motifs and the sues including the breast and prostate (Borgono et al. , 2007b ). physicochemical signature of an enzyme’ s active site cleft. Known substrates for KLK14 have prominent functions in However, conventional substrate screening methods have several of these tissues and include the extracellular cadherin generated confl icting subsite profi les for KLK14. This study link in mature desmosomes, desmoglein 1 (Fortugno et al. , utilizes a recently developed screening technique, the sparse 2011), two members of the protease-activated receptor (PAR) matrix library, to identify fi ve novel high-effi ciency sequences family (PAR2 and PAR4) (Oikonomopoulou et al., 2006), for KLK14. The optimal sequence, YASR, was cleaved with and a plethora of extracellular matrix (Borgono = ± × 6 -1 -1 higher effi ciency ( k cat / KM 3.81 0.4 10 m s ) than favored et al., 2007b). substrates from positional scanning and phage display by The physiological importance of these substrates empha- 2- and 10-fold, respectively. Binding site cooperativity was sizes the need for KLK14 proteolysis to be appropriately prominent among preferred sequences, which enabled optimal controlled. Aberrant KLK14 proteolytic activity is linked interaction at all subsites as indicated by predictive modeling to several skin pathologies by promoting premature prote- of KLK14/substrate complexes. These simulations constitute olysis of desmoglein 1 and prolonged stimulation of proin- the fi rst molecular dynamics analysis of KLK14 and offer a fl ammatory signaling pathways including PAR2 (Stefansson structural rationale for the divergent subsite preferences evi- et al., 2008; Cork et al., 2009). PAR2 activation by KLK14 dent between KLK14 and closely related KLKs, KLK4 and has also attracted recent interest in colon cancer where KLK5. Collectively, these fi ndings highlight the importance KLK14 is more highly expressed than in healthy tissue of binding site cooperativity in protease substrate recognition, and signaling via PAR2 has been shown to stimulate pro- which has implications for discovery of optimal substrates liferation of colon cancer cells in vitro (Gratio et al., 2011). and engineering highly effective protease inhibitors. Additionally, overexpression of KLK14 has been detected in several hormone-dependent cancers, including prostate Keywords: phage display; positional scanning synthetic (Yousef et al., 2003; Rabien et al., 2008), ovarian (Borgono combinatorial library; serine protease; sparse matrix library; et al. , 2003 ), and breast (Borgono et al. , 2003 ; Fritzsche substrate specifi city. et al., 2006), correlating with higher risk of disease progres- sion. Collectively, these observations suggest KLK14 is a Introduction potential point of therapeutic intervention in a variety of pathologies. However, as protease mediated biological func- The discovery of an expanded human kallikrein-related tions are often coordinated by complex interactions with a peptidase (KLK) locus containing 15 homologous (chymo) larger protease network, the relevance of KLK14 proteolytic

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM 332 S.J. de Veer et al.

activity, both in the skin and in cancer progression, remains Results to be completely understood. Recognition of protease cleavage sequences is driven Separate combinatorial analyses do not yield by topological and chemical complementarities between a defi nitive specifi city profi le for KLK14 a substrate and the enzyme active site. These interactions are fundamental to the biological function(s) of a protease As specifi city profi les from PS-SCL have shown variabi- and, in essence, can be recapitulated by screening libraries lity when the same protease is analyzed by independently of shorter model sequences. Accordingly, substrate specifi - produced libraries (Matsumura et al., 2005; Debela et al., city analyses form a common component in strategies 2006b ; Borgono et al. , 2007a ), KLK14 was screened aimed at engineering highly potent and specifi c protease against a second P1-P4 diverse library (including all natu- inhibitors (Goettig et al., 2010; Swedberg et al., 2010) for rally occurring amino acids) with a para -nitroanilide (pNA) further development as therapeutics or research tools such reporter group (Figure 1 ). A clear preference for Arg at P1 as activity-based probes. High-throughput screening can was observed, although cleavage rates for other residues also be used to predict novel substrates and deconvolute were higher than previously reported. At P2, Ser was sequential processing events. These strategies have been highly favored followed by Val, Asn, and Pro, indicating applied to many of the newly identifi ed KLKs (Matsumura small polar or hydrophobic residues were tolerated at this et al., 2005; Debela et al., 2006b; Borgono et al., 2007a), position, whereas larger residues (Gln, Tyr, and Trp) were including KLK14, which has been screened by phage dis- strongly disfavored. Small hydrophobic residues such as play (Felber et al., 2005) and a positional scanning synthetic Val, Leu, and Ala were preferred at P3 as opposed to acidic combinatorial library (PS-SCL) (Borgono et al. , 2007a ). or aromatic residues, whereas KLK14 displayed an equiva- However, the predicted optimal KLK14 substrates from lent preference for Tyr and Trp at P4. The newly generated these studies do not accord and highly divergent sequences specifi city profi les were noticeably different from the previ- VGSLR (Felber et al., 2005) and YAAR (Borgono et al., ous KLK14 PS-SCL screen, which had indicated Ala, Asn, 2007a) were identifi ed. This lack of agreement refl ects and Pro were more favored than Ser at P2; Lys and Ala weaknesses in both techniques. In phage display, the vari- were the most preferred residues at P3, while Tyr was sug- able sequence can modulate biosynthesis effi ciency and gested to be clearly favored over Trp at P4. The lack of con- infectivity, leading to positive and negative selection inde- cordance between the two screens is a likely consequence pendent of phage cleavage by the target protease (Wilson of differing library composition, as distinct methods were and Finlay, 1998). These sources of bias are overcome by used to generate library diversity. positional scanning where library diversity is generated by synthetic chemistry, but the method is unable to identify cooperativity effects, as binding sites are screened inde- Non-combinatorial library screen reveals pendently (Ng et al. , 2009 ; Schneider and Craik , 2009 ; subsite cooperativity Swedberg et al., 2010), and library complexity prevents conclusive validation of its composition. Taken together, To identify individual cleavage sequences, a non- these suggested that the substrate specifi city of KLK14 combinatorial sparse matrix library was produced and was yet to be completely described. screened against KLK14 (Figure 2 ). Across all substrates, Incomplete knowledge of a target protease’ s substrate sequences containing Ala at P3 were hydrolyzed at the high- specifi city is a major obstacle to achieving optimal potency est rates, Tyr and Trp appeared to be similarly preferred at and selectivity for engineered inhibitors. Recent studies on P4, while sequences containing Ser at P2 were generally KLK4 (Swedberg et al. , 2009 ) and (Swedberg and favored, although two substrates cleaved at high rates con- Harris , 2011 ) have overcome this challenge by coupling tained Val at this position. Four sequences were noticeably PS-SCL with a sparse matrix substrate screen to discover preferred by KLK14 and cleaved with substantially higher previously unrecognized high-effi ciency peptide substrates. rates than the remainder of the library: YAVR, WAVR, YASR, This technique builds on existing binding site preferences and WASR. Additionally, several instances of subsite coop- by producing a second peptide library where all sequence erativity were evident that could not have been identifi ed combinations based on known favored residues are indi- following a combinatorial peptide library screen. Within vidually synthesized and analyzed. Here we report the use the substrate sets analyzed, Val at the P3 site always of PS-SCL analysis followed by a sparse matrix substrate resulted in a P4 preference for Trp as opposed to Tyr when screen to elucidate a new tetrapeptide substrate profi le for Ala or Leu occupied P3. Furthermore, substrates with the KLK14, which was subsequently dissected at the structural same P4-P3 pairing were cleaved at equivalent rates in the level using molecular modeling. These fi ndings add further P2 Val and P2 Asn sublibraries, apart from Tyr-Ala and support to the emerging appreciation of subsite coopera- Trp-Ala, which were cleaved with twice the rate when Val tivity in substrate recognition (Ng et al., 2009; Swedberg was present at P2. Indeed, these two sequences (YAVR and et al. , 2009 ; Swedberg and Harris , 2011 ), which has previ- WAVR) showed the highest cleavage rates from the entire ously proven vital for engineered inhibitors achieving opti- library screen, highlighting the need to consider subsite mal effi cacy (Swedberg et al. , 2009, 2011 ; Swedberg and cooperativity when prospecting for highly favored pro- Harris , 2011 ). tease substrates.

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM Identifi cation of novel KLK14 cleavage sequences 333

Figure 1 KLK14 positional scanning analysis using a P1-P4 diverse combinatorial peptide-pNA library. Twenty peptide pools were produced for each nonprimed (P) binding site screen, which consisted of tetrapeptide substrates with one of the naturally occurring amino acids fi xed at the position indicated above each graph. The fi xed residue is shown on the x-axis in single-letter amino acid code with M* being methionine sulfone, while remaining binding sites contain a mixture of residues. Protease activity was detected by release of the pNA reporter [measured by increasing absorbance mOD 405 nm per min] and is expressed as a proportion of the highest rate within each sublibrary. Data are shown as mean ± SEM from three independent experiments performed in triplicate.

Kinetic constants confi rm cooperativity trends Molecular modeling of preferred KLK14 substrates and identify optimal catalytic effi ciencies for sparse matrix substrates Predictive modeling was undertaken to explore the differ- ent protease/substrate interactions formed by selected opti- Determining kinetic constants for KLK14 cleavage of mal substrates (YASR: highest k cat / KM ; WAVR: highest k cat ) favored substrates confi rmed the subsite cooperativity and substrates from PS-SCL and phage display. Molecular trends initially apparent from single concentration rates dynamics (MD) simulations were performed for each pep- (see Figure 2). The preferred residue at P4 was shown to tide-pNA in complex with a homology model of KLK14 (see be dependent on the adjacent residue at P3. Consistent with Methods) followed by calculation of the average structure P3 Ala producing a clear preference for Tyr over Trp at P4, (Figure 3 ). These analyses indicated that only the substrates

YASR was cleaved with 1.33-fold higher k cat and 2.02-fold from the sparse matrix library screen (YASR and WAVR) higher kcat /K M than WASR. The opposite effect was seen formed highly favored interactions at all KLK14 subsites. For with P3 Val, which changed the P4 preference to Trp and YASR (Figure 3A), both P2 Ser and P4 Tyr formed hydro-

WVSR was cleaved with 1.25-fold higher k cat and margin- gen bonds with the protease (with His99 and Arg97, respec- ally higher kcat /K M than YVSR. There was also evidence tively, trypsin numbering), whereas Ala was well suited to for P4-P2 cooperativity. For substrates containing P3 Ala, the small S3 pocket. Tyr at P4 also participated in an aro- replacing Val with Ser at P2 produced an 87% increase in matic π-stacking interaction with Trp215. In contrast, WAVR

k cat /K M when substrates contained P4 Tyr (YAVR vs. YASR), (Figure 3B) relied on favorable nonpolar interactions across while the same substitution with Trp at P4 had essentially the S4-S2 sites with Val binding within the narrow S2 pocket no effect (WAVR vs. WASR). When compared to sequences and Trp accommodated by the large S4 pocket. identifi ed by conventional methods, all fi ve sparse matrix Many of these features were also evident in the substrates substrates were cleaved with higher catalytic effi ciency identifi ed by positional scanning, yet each sequence suffered

( kcat /K M ) than all substrates suggested to be highly preferred from poor complementarity at some point along its length. by combinatorial peptide libraries, YAAR (Borgono et al. , For YAAR (Figure 3C), the S2 pocket was capable of accept- 2007a) and WVSR, or phage display, VGSLR (Felber et al., ing residues larger than Ala, which did not seem to interact 2005 ). Additionally, the substrates hydrolyzed at the highest with the protease (Figure 3F); for example, replacing P2 Ala rate in the sparse matrix screen (YAVR and WAVR) were with Val or Ser produced substantial improvements in both cleaved with higher k cat values than all substrates identifi ed kcat and k cat / KM (Table 1 ). Conversely, while WVSR (Figure by existing specifi city screens. 3D) contained the highly favored Ser at P2, Val was too bulky

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM 334 S.J. de Veer et al.

Figure 2 Non-combinatorial library screen defi nes the extended substrate specifi city of KLK14 in greater depth and identifi es cooperativity. Sparse matrix substrate screen of individually synthesized peptides based on all combinations of known KLK14 binding site preferences. The P1 residue was kept constant (Arg), while the P4-P2 residues were varied and the sequence for each substrate (P4, P3, P2, P1) is shown on the ± x-axis (M* represents methionine sulfone). Hydrolysis was measured in the change in mOD405 nm per min, and data are expressed as mean SEM from three independent experiments performed in triplicate.

for the S3 pocket resulting in loss of a backbone-backbone backbone conformation (average Cα RMSD between KLK14 hydrogen bond with KLK14 (Gly216). This was consistent and KLK4, 0.77 Å , or KLK5, 0.39 Å ; Figure 4 A) and sequence with the reduction in kcat and k cat / KM (32 % and 25 % , respec- homology at the active site, including the S1-S4 binding sites tively) observed when Val occupied P3 compared with Ala (Figure 4B). However, several regions of sequence divergence (WASR vs. WVSR, Table 1). These residues may be more were evident (Figure 4C), which seemed to justify differences favored with different sequence combinations, which would between the newly identifi ed KLK14 sequences and existing justify their general preference in positional scanning yet subsite preferences for KLK4 and KLK5 (summarized in not as part of the individual substrates cleaved with highest Table 2), particularly at the S2 and S4 subsites. effi ciency. His99 fl anks the S2 pocket of both KLK5 and KLK14, Nonetheless, the substrates from PS-SCL and the sparse and accordingly, preferences at this subsite overlap consider- matrix library were all cleaved with a least 10-fold higher ably (Ser being the most preferred while Asn is also favored). catalytic effi ciency than the optimal substrate identifi ed by However, structural overlay suggested His99 for KLK5 was phage display (VGSLR; Figure 3E). This was in agreement oriented slightly further outward from the protease, increas- with the MD average structure, which revealed that while ing the width of the S2 pocket. This would potentially P2 Leu seemed to be well-tolerated, neither the P3 nor P4 explain the KLK5 preference for Phe or His while KLK14 residues appeared to make favorable contacts with KLK14. favors the smaller side-chain of Val. For KLK4, the same The Ser (P3) side-chain did not form any hydrogen bonds region is occupied by Leu99, hence Val is favored but not with KLK14 and was in fact oriented away from the protease Ser or Asn, as the protease lacks a nearby hydrogen bonding (Figure 3E, F). Additionally, P4 Gly seemed preferred simply residue. The orientation of Leu99 also broadens the entrance to fulfi ll a spacing requirement allowing Val to interact with to the S2 subsite, enabling P2 Gln to access hydrogen bonds the same hydrophobic determinants as aromatic P4 residues, with the side-chains of Asp102 and Ser214 (Debela et al. , particularly Trp (Figure 3E, F). Consistent with this, KLK14 2006a), making Gln the most preferred residue for KLK4 at cleavage of a tetrapeptide sequence lacking Val (GSLR com- this site. pared with VGSLR) yielded lower k and k / K values by cat cat M KLK5 and KLK14 also show a high degree of similar- 3- and 4-fold, respectively (Table 1). ity within the S4 site with Trp215 present in both enzymes Structural comparison of KLK14 with the closely compared with Phe215 for KLK4. However, KLK14 shows related KLK4 and KLK5 an equivalent preference for Tyr and Trp at P4 while KLK5 favors Tyr. This could be due to differences at position 175 One of the hallmarks of the KLK family is a high degree of as Thr175 (KLK14) does not extend as far into the pocket as sequence similarity across the active site cleft (Goettig et al. , Gln175 (KLK5), hence KLK14 is better suited to accommo- 2010; Swedberg et al., 2010), yet unique substrate specifi c- date Trp at P4. Preferences at the KLK4 S4 site are markedly ity profi les have been found for all KLKs screened to date different with smaller, hydrophobic P4 residues (Phe and Ile) (Debela et al., 2006b; Borgono et al., 2007a). KLK14 is par- highly favored as opposed to Tyr and Trp. This is likely to be ticularly homologous to KLK4 and KLK5, both in terms of a combined result of two factors. First, the kallikrein ( ‘ 99 ’ )

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM Identifi cation of novel KLK14 cleavage sequences 335

Figure 3 Molecular modeling of KLK14/substrate complexes. Average structures from MD simulation of KLK14 (homology model, shown as ribbon plot) in complex with the following peptide- pNA substrates (shown as green stick model): (A) Ac-YASR-pNA, (B) Ac-WAVR-pNA, (C) Ac-YAAR-pNA, (D) Ac-WVSR-pNA, and (E) Ac-VGSLR-pNA. KLK14 residues forming interactions with the peptide ligand are shown in gray stick model and hydrogen bonds are shown in purple dashed lines. Residues are labeled in single letter amino acid code with the serine protease catalytic triad indicated by an aster- isk (*). (F) Overlay of the fi ve peptide substrates (YASR: green, WAVR: magenta, YAAR: gray, WVSR: orange, VGSLR: blue) bound to the active site of KLK14 (shown as molecular surface colored according to electrostatic potential: red indicates negative, blue indicates positive). The substrate binding sites S4-S1′ are indicated.

Table 1 Michaelis ( K M ) and rate (k cat ) constants for preferred sparse matrix library (SML) substrates compared with positional scanning (PS-SCL) and phage display (PD) sequences .

μ -1 × 4 -1 -1 Sequence Screen method Theoretical mass Actual mass K M ( m ) k cat (s ) kcat / KM ( 10 M s ) Ac-YASR-pNA SML 657.78 658.87 22.03 ± 1.44 83.85 ± 1.15 380.6 ± 44.6 Ac-YAVR-pNA SML 669.83 670.93 45.01 ± 2.48 93.33 ± 1.38 207.4 ± 26.1 Ac-WASR-pNA SML 680.82 681.63 33.64 ± 1.04 63.24 ± 0.51 188.0 ± 11.2 Ac-WAVR-pNA SML 692.87 693.8151.54 ± 1.55 97.31 ± 0.82 188.8 ± 11.0 Ac-YANR-pNA SML 684.80 685.77 16.78 ± 0.98 53.04 ± 0.58 316.1 ± 33.4 Ac-WVSR-pNA PS-SCL 708.87 709.7430.48 ± 1.63 42.91 ± 0.54 140.8 ± 13.9 Ac-YVSR-pNA PS-SCL 685.83 686.8825.23 ± 1.56 34.57 ± 0.47 137.0 ± 15.3 Ac-YAAR-pNA PS-SCL 641.78 642.64 43.20 ± 1.39 69.37 ± 0.59 160.9 ± 10.0 Ac-VGSLR-pNA PD 692.87 693.68 318.4 ± 22.3 44.03 ± 1 . 5 1 1 3 . 8 3 ± 1.93 Ac-GSLR-pNA PD 593.73 594.58 431.4 ± 19.8 15.80 ± 0 . 3 9 3 . 6 6 ± 0.35

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM 336 S.J. de Veer et al.

Figure 4 Structural comparison of KLK14 and closely related KLKs (KLK4 and KLK5). (A) Ribbon plot of the KLK14 homology model (yellow) superimposed on the structures of KLK4 (green, PDB ID: 2BDG) and KLK5 (pink, PDB ID: 2PSX). Variable residues around the active site are shown in stick model along with the catalytic triad (*). Residues are labeled with single letter amino acid code in the order KLK4/ KLK5/KLK14 using trypsin numbering. (B) Molecular surface of the KLK14 homology model color coded according to sequence conserva- tion among KLK4, KLK5, and KLK14 (green for identical residues, yellow for conservative sequence variation and red for nonconservative sequence variation). The substrate binding sites S4-S1′ are labeled. Structural orientation between (A) and (B) is identical. (C) Sequence align- ment of KLK4, KLK5, and KLK14 with identical residues and conservative substitutions indicated below with an asterisk (*) or period (.), respectively. Residues are colored according to side-chain properties (red: acidic, blue: basic, green: polar, black: hydrophobic). Additionally, residues within 5 Å of a tetrapeptide substrate bound to the S4-S1 protease binding sites are shaded gray. loop in KLK4 is further removed, which leaves fewer hydro- one or more KLKs in numerous, unrelated physiological pro- gen bonding opportunities for Tyr within the S4 pocket unlike cesses refl ects the requirement for individual members of the KLK14 (Figure 3A, C) or KLK5. Second, the proximity of KLK family to recognize distinct cleavage sequences in spite Phe215 and Leu175 places a size restriction on favored P4 of high levels of sequence and structural conservation. Unique residues; hence, Trp is not easily accommodated. features within the active site cleft are likely to be important for directing separate KLKs to accomplish different tasks, particularly in cases of overlapping tissue expression profi les, Discussion highlighting how knowledge of favored protease cleavage sequences can interface with structural information to aid The repertoire of biological functions under the control of understanding of a target enzyme ’ s biological function(s). KLK proteases is highly diverse and to date includes blood Screening KLK14 against a tailored sparse matrix sub- pressure homeostasis, initiation of infl ammation in several strate library identifi ed several highly preferred cleavage tissues, seminogelin hydrolysis, corneocyte turnover in the sequences that were not predicted by previous positional epidermis, enamel matrix degradation, and maintaining cen- scanning and phage display analyses. Moreover, analyz- tral nervous system plasticity (see Sotiropoulou et al. , 2009 ; ing hydrolysis of individual substrates enabled the substrate Goettig et al., 2010, for recent reviews). The involvement of specifi city of KLK14 to be examined in greater depth, which

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM Identifi cation of novel KLK14 cleavage sequences 337

Table 2 Substrate specifi city analyses for KLK4 and KLK5.

P4 P3 P2 P1 Reference

Sparse matrix library preferences KLK4 F, I, V V, Q, T Q, V, L R Swedberg et al. , 2009 PS-SCL preferences KLK4 I, V, Y V, Q, M Q, V, L R, K Debela et al. , 2006b KLK4 I, V, Y Q, S, V Q, V, L R, K Matsumura et al. , 2005 KLK4 I, V, F S, Q, A Q, L, V R Borgono et al. , 2007a KLK5 G, Y, V M, Y, K S, N, T R Debela et al. , 2006b KLK5 Y, G, V R, K, Y F, N, S R Borgono et al. , 2007a

-1 μ -1 -1 Substrate k cat (s )K M ( m )k cat / KM (m s ) Kinetic constants for individual substrates KLK4 FVQR-pNA 15.7 679.9 2.31 × 104 Swedberg et al. , 2009 KLK4 IVQR-pNA 2.10 183.1 1.15 × 104 Swedberg et al. , 2009 K L K 4 B z - F V R - p N A 1 . 8 9 4 2 . 5 4 . 1 5 × 104 Swedberg et al. , 2009 KLK5 Boc-VPR-AMC 3.28 200 1.64 × 104 Michael et al. , 2006 KLK5 Boc-FSR-AMC 2.83 190 1.49 × 10 4 Michael et al. , 2006 Preferences from sparse matrix library and PS-SCL screens are shown in single-letter amino acid code and in descending order from left (most preferred) to right. in turn revealed prominent effects of binding site coopera- substrates can also be harnessed to design engineered inhibi- tivity. KLK14 preferences at several subsites were found to tors with optimal potency and selectivity. Substituting the most be dependent on the neighboring sequence, which allowed preferred substrate for KLK4 into the contact surface of sun- residues that seemed only modestly preferred to become fl ower produced a variant highly selective for highly favored under certain conditions. This was particu- KLK4 over KLK5 and KLK14 (Swedberg et al. , 2009, 2011 ). larly evident in substrates containing Val at the P2 site, which A similar approach enabled the production of high-affi nity pep- were only highly preferred when Trp-Ala or Tyr-Ala occu- tide aldehyde inhibitors for plasmin, which displayed selectiv- pied P4-P3. These two substrates (WAVR and YAVR) were ity over (Swedberg and Harris , 2011 ). cleaved with higher k cat values than all other substrates exam- The affi nity and selectivity of these engineered inhibitors ined. Conversely, since the biological activity of each KLK14 further illustrate the merit in considering ligands as a single subsite was not autonomous, the substrates based on the most entity rather than the sum of separate binding interactions as preferred residue at each binding site from separate PS-SCL proposed by combinatorial chemistry. The optimal substrate screens were not cleaved with optimal effi ciency. This was for plasmin was identifi ed by screening a non-combinatorial supported by predictive modeling of KLK14/substrate com- library (Swedberg and Harris , 2011 ) and outperformed the plexes, which indicated that in these sequence combinations, divergent sequences found by separate PS-SCL analyses interaction with at least one protease binding site was not (Backes et al., 2000; Harris et al., 2000). Similarly, three inde- optimal. In contrast, the substrates identifi ed by the sparse pendent PS-SCL screens for KLK4 (Matsumura et al., 2005; matrix substrate screen were cleaved with higher effi ciency Debela et al. , 2006b ; Borgono et al. , 2007a ) predicted different and formed highly favored interactions at all binding sites. substrates that were all cleaved at lower rates than a sequence Extending molecular modeling analyses to related KLKs discovered by a sparse matrix library screen (Swedberg enabled known specifi city differences among KLK4, KLK5, et al., 2009). In this study, YASR was found to be the optimal and KLK14 to be explored. Subtle active site differences were substrate for KLK14 (Table 1), not YAAR (Borgono et al., identifi ed for each KLK, which were consistent with highly 2007a) or WVSR (Figure 1) from separate PS-SCL screens. dissimilar optimal cleavage sequences for KLK14 and KLK4: Combinatorial methods such as PS-SCL enable rapid, high- YASR (Table 1) and FVQR (Swedberg et al. , 2009 ), respec- throughput analysis and have been widely used to establish tively. The signifi cance of each KLK active site possessing a distinct specifi city profi les among members of the kallikrein- unique physicochemical fi ngerprint extends to recognition related peptidase (Debela et al. , 2006b ), caspase (Thornberry of endogenous substrates and inhibitors. For example, the et al. , 1997 ), (Choe et al. , 2006 ), and plasmepsin P4-P1 residues of the skin-expressed, single-domain Kazal (Beyer et al., 2005) protease families. However, positional inhibitor SPINK6 (YCTR) are strikingly similar to the opti- scanning has equally been unable to conclusively resolve the mal substrate for KLK14 (YASR) and complementary to the substrate specifi city of numerous protease targets (see above) PS-SCL profi le for KLK5 but not KLK4. Consistent with this, and the method ’ s shortcomings are not widely appreciated. SPINK6 is a more effective inhibitor of KLK5 and KLK14 by Paradoxically, the combinatorial nature of the library is the 20- and 60-fold compared with KLK4 (Meyer -Hoffert et al., most notable drawback of PS-SCL: while it confers practical 2010 ; Kantyka et al. , 2011 ). Knowledge of highly preferred advantages in terms of library synthesis and screening, its use

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM 338 S.J. de Veer et al.

entails a compromise in that only the identity of the fi xed resi- 2005 ), the prevalence of subsite cooperativity effects may be due is known within each substrate pool. Consequently, the considerably greater than is currently recognized. neighboring sequence, which can strongly infl uence subsite preferences, is largely unknown, preventing identifi cation of cooperativity effects (Schneider and Craik , 2009 ; Swedberg Materials and methods et al. , 2010 ). This limitation is potentially compounded by variable composition within positional scanning libraries. Reagents The common contention that PS-SCL contains all theoretical sequence combinations in equimolar concentrations (Harris et All synthesis reagents were obtained from Auspep (Melbourne, al., 2000; Matsumura et al., 2005; Borgono et al., 2007a) due Australia) and all solvents from Merck (Melbourne, Australia) un- to the use of isokinetic amino acid mixtures (Ostresh et al., less stated otherwise. Thermolysin was obtained from R&D Systems (Minneapolis, MN). 1994 ) is in confl ict with a comprehensive study by Boutin et al. (1997) , which did not fi nd chemical compensation effective during competitive dipeptide couplings. Additionally, more Production of recombinant KLK14 than 25% of theoretical sequences could not be detected in the KLK14 was produced in stably transfected Sf9 insect cells fi nal peptide mixture by tandem LC/MS when the experiment containing the pro-KLK14 open-reading frame inserted into the was expanded to tripeptide and tetrapeptide libraries. pIB/V5-His vector Invitrogen (Melbourne, Australia) (Swedberg et al. , Even with these shortcomings, there can be no doubt 2009). After dialysis, pro-KLK14 was purifi ed from conditioned super- that PS-SCL provides useful information as seen by the natant by Ni-NTA superfl ow agarose (Qiagen, Melbourne, Australia) wealth of existing studies where unique subsite preferences according to the manufacturer’ s instructions. Expression and purity were discerned even between closely homologous pro- were confi rmed by Western blot analysis (against the poly-His epitope) teases (Thornberry et al. , 1997 ; Harris et al. , 2000 ; Choe et and Coomassie-stained SDS-PAGE, respectively. Pro-KLK14 was al., 2006; Debela et al., 2006b). Indeed, both PS-SCL sub- activated with thermolysin at 37° C also resulting in the removal of the strates for KLK14 were cleaved with 10-fold higher effi - C-terminal tag. Following activation, thermolysin was inhibited with ciency than the best sequence from phage display. Rather, 25 mm EDTA. The concentration of mature KLK14 was determined by active site titration using 4-methylumbelliferyl-p -guanidinobenzo- this and related studies (Takeuchi et al. , 2000 ; Swedberg et ate hydrochloride (Sigma-Aldrich, Sydney, Australia). al., 2009; Swedberg and Harris, 2011) highlight that PS-SCL data should be interpreted within the limitations of the method ’ s capabilities. Combinatorial peptide library screens Synthesis of peptide substrate libraries have proven very effective for identifying individual sub- Peptide pNA substrates were synthesized on para -phenylenediamine site preferences for numerous proteases. However, the abil- (Sigma-Aldrich) derivatized 2-chlorotrityl resin (0.13 mmol/g) using ity of PS-SCL to determine optimal tetrapeptide substrates fl uorenylmethyl carbamate (Fmoc)-protected amino acids (Abbenante can be compromised by binding site cooperativity effects et al., 2000). Specifi c conditions for coupling, deprotection, cleavage and whether preferred residues are compatible in combi- from the solid support, oxidation of the pNA reporter, and removal of nation. These challenges can be overcome by secondary side-chain-protecting groups have been previously described elsewhere screening techniques (Schneider and Craik , 2009 ; Swedberg (Abbenante et al. , 2000 ; Swedberg et al. , 2009 ; Swedberg and Harris , et al., 2010), such as a non-combinatorial sparse matrix library 2011 ). Individually synthesized substrates were assembled by sequen- (Swedberg et al. , 2009 ; Swedberg and Harris , 2011 ), or high- tial coupling reactions using four equivalents Fmoc-protected amino throughput proteomic approaches, which enable simultane- acids. For the P1-P4 diverse combinatorial library, fi xed positions were coupled as above while degenerate positions were added using a two- ous analysis of prime and nonprime specifi cities (Schilling step limited loading approach to compensate for the widely varying and and Overall , 2008 ). context-dependent reaction rates of individual amino acids (Ragnarsson Additionally, the fi ndings from this study add further et al., 1971, 1974; Mutter, 1979; Boutin et al., 1997). These positions support to the emerging signifi cance of subsite cooperativ- were fi rst coupled using an equimolar mixture of all amino acids total- ity in protease substrate specifi city (see Ng et al. , 2009 , for ing the molar loading capacity of the resin followed by a second round a review). Conditional preferences defi ned by cooperativity of coupling with four equivalents mixed amino acids to ensure oc- effects may contribute to the fi ne-tuning that enables indi- cupancy at all available sites. Substrates for Michaelis-Menten kinetic vidual proteolytic enzymes to discern unique cleavage sites analysis were purifi ed by reverse-phase high-performance liquid chro- from a predominating background of nontarget sequences. matography (rp-HPLC) using a Jupiter 4μ Proteo 90A C-18 column Appreciably, proteolysis of large substrates is consid- (Phenomenex, Sydney, Australia) across a gradient of 10 % – 100 % iso- erably more complex, and there are additional requirements propanol containing 0.1 % TFA. This was followed by MALDI-TOF/ MS analysis for purity and mass validation using a Biorad ProteinChip for favorable structural context and complementary interac- System: Personal Edition (Biorad, Sydney, Australia). tions beyond the active site. However, cooperativity is likely to be particularly relevant to optimal potency and selectiv- PS-SCL and sparse matrix library assays ity of engineered peptide-based inhibitors where the contact interface is considerably smaller. Given the result from a Kinetic assays were carried out in 96-well transparent low-binding large-scale specifi city analysis of 13 serine proteases and 11 plates (Corning, Lowell, MA) using 0.1 m Tris-HCl pH 8.0, 0.1 m cysteine proteases, which revealed sequence-dependent bind- NaCl, 25 mm EDTA, 0.005% Triton X-100 assay buffer (fi nal vol- ing site preferences in all enzymes screened (Gosalia et al. , ume 250 μ l). Activity was detected by the change in absorbance at 405

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM Identifi cation of novel KLK14 cleavage sequences 339 nm with linear rates measured for 7 min (10 s reading interval) using References a Biorad Benchmark Plus multi-well spectrophotometer. All data are shown as the mean rate ± SEM from three independent assays carried out Abbenante, G., Leung, D., Bond, T., and Fairlie, D.P. (2000). An in triplicate. Substrate pools for positional scanning were solubilized effi cient Fmoc strategy for the rapid synthesis of peptide para- in 50 % isopropanol at 3.75 mm (by pool average molecular weight). nitroanilidies. Lett. Pept. Sci. 7 , 347 – 351. Assays were performed using 7.5 nm KLK14 and approximately Backes, B.J., Harris, J.L., Leonetti, F., Craik, C.S., and Ellman, J.A. 150 μ m pooled substrate per well. For the sparse matrix substrate (2000). Synthesis of positional-scanning libraries of fl uorogenic screen, individually synthesized peptides were solubilized in 40 % iso- peptide substrates to defi ne the extended substrate specifi city of propanol and adjusted to apparent equal molarity by total hydrolysis of plasmin and . Nat. Biotechnol. 18 , 187 – 193. the pNA moiety (Swedberg et al., 2009). Library analysis was carried Beyer, B.B., Johnson, J.V., Chung, A.Y., Li, T., Madabushi, A., μ out using 5 nm KLK14 and approximately 250 m substrate per well. Agbandje-McKenna, M., McKenna, R., Dame, J.B., and Dunn, B.M. (2005). Active-site specifi city of digestive aspartic pepti-

Determination of Michaelis (K M ) and rate (k cat ) dases from the four species of Plasmodium that infect humans constants using chromogenic combinatorial peptide libraries. Biochemistry 44 , 1768 – 1779. Highly favored substrates from the sparse matrix library (YAVR, Borgono, C.A. and Diamandis, E.P. (2004). The emerging roles of WAVR, YASR, WASR, YANR), independent PS-SCL screens human tissue in cancer. Nat. Rev. Cancer 4 , 876 – 890. (YAAR: Borgono et al., 2007a; WVSR: this study) and phage dis- Borgono, C.A., Grass, L., Soosaipillai, A., Yousef, G.M., Petraki, play (VGSLR: Felber et al. , 2005 ) were synthesized and purifi ed by C.D., Howarth, D.H., Fracchioli, S., Katsaros, D., and Diamandis, rp-HPLC. Assays were carried out at room temperature (298 K) in E.P. (2003). Human kallikrein 14: a new potential biomarker for 250 μl buffer (as above) using a constant concentration of 700 pm ovarian and breast cancer. Cancer Res. 63 , 9032 – 9041. KLK14 and serial dilutions of purifi ed substrates (600 – 18.75 μ m ). Borgono, C.A., Gavigan, J.A., Alves, J., Bowles, B., Harris, J.L., Initial rates were measured by the change in absorbance (405 nm) Sotiropoulou, G., and Diamandis, E.P. (2007a). Defi ning the for 7 min over three independent assays carried out in triplicate. extended substrate specifi city of kallikrein 1-related peptidases.

Kinetic constants (K M , k cat ) were determined by nonlinear regression Biol. Chem. 388 , 1215 – 1225. in GraphPad Prism 5 (GraphPad Software) and are reported ± SEM, Borgono, C.A., Michael, I.P., Shaw, J.L., Luo, L.Y., Ghosh, M.C., ± whereas k cat / KM is given 95 % confi dence interval from the calcu- Soosaipillai, A., Grass, L., Katsaros, D., and Diamandis, E.P. lated propagation of error. (2007b). Expression and functional characterization of the cancer-related serine protease, human tissue kallikrein 14. MD simulations of KLK14 substrate complexes J. Biol. Chem. 282 , 2405 – 2422. Boutin, J.A., Gesson, I., Henlin, J.M., Bertin, S., Lambert, P.H., As a KLK14 structure is currently not available, a homology model Volland, J.P., and Fauchere, J.L. (1997). Limitations of the cou- was produced in Swiss model (Guex et al. , 2009 ) based on the KLK pling of amino acid mixtures for the preparation of equimolar structure with the highest sequence similarity to KLK14 (KLK5: peptide libraries. Mol. Divers. 3 , 43 – 60. PDB ID 2PSX, 50 % homology). KLK14 substrates were positioned Brattsand, M., Stefansson, K., Lundh, C., Haasum, Y., and Egelrud, T. at the active site by overlay with the KLK5/leupeptin complex (2005). A proteolytic cascade of kallikreins in the stratum cor- (C α RMSD 0.39 Å ) using SPBDV v4.1 (Guex et al. , 2009 ). neum. J. Invest. Dermatol. 124 , 198 – 203. All subsequent molecular modeling manipulations were carried Choe, Y., Leonetti, F., Greenbaum, D.C., Lecaille, F., Bogyo, M., using YASARA Dynamics v10.7.8 (Krieger et al., 2002) and the Bromme, D., Ellman, J.A., and Craik, C.S. (2006). Substrate pro- AMBER03 force fi eld (Duan et al. , 2003 ). KLK14/substrate com- fi ling of cysteine proteases using a combinatorial peptide library plexes were solvated with TIP3P water (pH 8.0) and neutralized by identifi es functionally unique specifi cities. J. Biol. Chem. 281 , Na + /Cl - counter ions to a fi nal concentration of 100 mm (approxi- 12824 – 12832. mating the conditions of kinetic assays), generating a system of ap- Clements, J., Hooper, J., Dong, Y., and Harvey, T. (2001). The proximately 35,000 atoms, which included 10,000 water molecules. expanded human kallikrein (KLK) family: genomic organi- For long-range electrostatics, the particle mesh Ewald algorithm was sation, tissue-specifi c expression and potential functions. Biol. used with nonbonded interactions truncated at 10.5 Å . Solvated com- Chem. 382 , 5 – 14. plexes were subjected to conjugate gradient minimization before 500 Clements, J.A., Willemsen, N.M., Myers, S.A., and Dong, Y. (2004). ps MDs while applying fi xed Cα atoms. This was followed by 1 ns The tissue kallikrein family of serine proteases: functional roles production simulations without constraints from which the average in human disease and potential as clinical biomarkers. Crit. Rev. simulation structure was determined. Molecular surfaces were cre- Clin. Lab. Sci. 41 , 265 – 312. ated using CCP4MG (Potterton et al., 2004) and color-coded accord- Cork, M.J., Danby, S.G., Vasilopoulos, Y., Hadgraft, J., Lane, M.E., ing to electrostatic potential (calculated by the Poisson-Boltzmann Moustafa, M., Guy, R.H., Macgowan, A.L., Tazi-Ahnini, R., solver within CCP4MG using a 1.4 Å probe radius) or sequence and Ward, S.J. (2009). Epidermal barrier dysfunction in atopic conservation. dermatitis. J. Invest. Dermatol. 129 , 1892 – 1908. Debela, M., Magdolen, V., Grimminger, V., Sommerhoff, C., Messerschmidt, A., Huber, R., Friedrich, R., Bode, W., and Acknowledgements Goettig, P. (2006a). Crystal structures of human tissue kallikrein 4: activity modulation by a specifi c zinc binding site. J. Mol. This work was supported by grants from the Prostate Cancer Biol. 362 , 1094 – 1107. Foundation of Australia (grant PR09) and the Institute of Health Debela, M., Magdolen, V., Schechter, N., Valachova, M., Lottspeich, and Biomedical Innovation: MCR scheme (awarded to Jonathan F., Craik, C.S., Choe, Y., Bode, W., and Goettig, P. (2006b). M. Harris). Simon J. de Veer receives funding from the Queensland Specifi city profi ling of seven human tissue kallikreins reveals indi- Government Smart Futures Fund. vidual subsite preferences. J. Biol. Chem. 281 , 25678 – 25688.

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM 340 S.J. de Veer et al.

Deraison, C., Bonnart, C., Lopez, F., Besson, C., Robinson, R., (2010). Isolation of SPINK6 in human skin: selective inhibitor of Jayakumar, A., Wagberg, F., Brattsand, M., Hachem, J.P., kallikrein-related peptidases. J. Biol. Chem. 285 , 32174 – 32181. Leonardsson, G., et al. (2007). LEKTI fragments specifi cally inhibit Michael, I.P., Pampalakis, G., Mikolajczyk, S.D., Malm, J., KLK5, KLK7, and KLK14 and control desquamation through a Sotiropoulou, G., and Diamandis, E.P. (2006). Human tissue kal- pH-dependent interaction. Mol. Biol. Cell 18 , 3607 – 3619. likrein 5 is a member of a proteolytic cascade pathway involved Duan, Y., Wu, C., Chowdhury, S., Lee, M.C., Xiong, G., Zhang, in seminal clot liquefaction and potentially in prostate cancer W., Yang, R., Cieplak, P., Luo, R., Lee, T., et al. (2003). A progression. J. Biol. Chem. 281 , 12743 – 12750. point-charge force fi eld for molecular mechanics simulations Mutter, M. (1979). Studies on the coupling rates in liquid-phase pep- of proteins based on condensed-phase quantum mechanical tide synthesis using competition experiments. Int. J. Pept. Protein calculations. J. Comput. Chem. 24 , 1999 – 2012. Res. 13 , 274 – 277. Felber, L.M., Borgono, C.A., Cloutier, S.M., Kundig, C., Kishi, T., Ng, N.M., Pike, R.N., and Boyd, S.E. (2009). Subsite cooperativity Ribeiro Chagas, J., Jichlinski, P., Gygi, C.M., Leisinger, H.J., in protease specifi city. Biol. Chem. 390 , 401 – 407. Diamandis, E.P., et al. (2005). Enzymatic profi ling of human Oikonomopoulou, K., Hansen, K.K., Saifeddine, M., Tea, I., kallikrein 14 using phage-display substrate technology. Biol. Blaber, M., Blaber, S.I., Scarisbrick, I., Andrade-Gordon, P., Chem. 386 , 291 – 298. Cottrell, G.S., Bunnett, N.W., et al. (2006). Proteinase-activated Fortugno, P., Bresciani, A., Paolini, C., Pazzagli, C., El Hachem, M., receptors, targets for kallikrein signaling. J. Biol. Chem. 281 , D’ Alessio, M., and Zambruno, G. (2011). Proteolytic activation 32095 – 32112. cascade of the -defective protein, LEKTI, Ostresh, J.M., Winkle, J.H., Hamashin, V.T., and Houghten, R.A. in the epidermis: implications for skin homeostasis. J. Invest. (1994). Peptide libraries: determination of relative reaction rates Dermatol. 131 , 2223 – 2232. of protected amino acids in competitive couplings. Biopolymers Fritzsche, F., Gansukh, T., Borgono, C.A., Burkhardt, M., Pahl, S., 34 , 1681 – 1689. Mayordomo, E., Winzer, K.J., Weichert, W., Denkert, C., Jung, Potterton, L., McNicholas, S., Krissinel, E., Gruber, J., Cowtan, K., et al. (2006). Expression of human kallikrein 14 (KLK14) K., Emsley, P., Murshudov, G.N., Cohen, S., Perrakis, A., and in breast cancer is associated with higher tumour grades and Noble, M. (2004). Developments in the CCP4 molecular- positive nodal status. Br. J. Cancer 94 , 540 – 547. graphics project. Acta Crystallogr. D Biol. Crystallogr. 60 , Goettig, P., Magdolen, V., and Brandstetter, H. (2010). Natural and 2288 – 2294. synthetic inhibitors of kallikrein-related peptidases (KLKs). Rabien, A., Fritzsche, F., Jung, M., Diamandis, E.P., Loening, S.A., Biochimie 92 , 1546 – 1567. Dietel, M., Jung, K., Stephan, C., and Kristiansen, G. (2008). Gosalia, D.N., Salisbury, C.M., Ellman, J.A., and Diamond, S.L. High expression of KLK14 in prostatic adenocarcinoma is asso- (2005). High throughput substrate specifi city profi ling of serine ciated with elevated risk of prostate-specifi c antigen relapse. and cysteine proteases using solution-phase fl uorogenic peptide Tumour Biol. 29 , 1 – 8. microarrays. Mol. Cell. Proteomics 4 , 626 – 636. Ragnarsson, U., Karlsson, S., and Sandberg, B. (1971). Studies on Gratio, V., Loriot, C., Duke Virca, G., Oikonomopoulou, K., Walker, the coupling step in solid phase peptide synthesis. Some prelimi- F., Diamandis, E.P., Hollenberg, M.D., and Darmoul, D. (2011). nary results from competition experiments. Acta Chem. Scand. Kallikrein-related peptidase 14 acts on proteinase-activated 25 , 1487 – 1489. receptor 2 to induce signaling pathway in colon cancer cells. Am. Ragnarsson, U., Karlsson, S.M., and Sandberg, B.E. (1974). Studies J. Pathol. 179, 2625–2636. on the coupling step in solid phase peptide synthesis. Further Guex, N., Peitsch, M.C., and Schwede, T. (2009). Automated com- competition experiments and attempts to assess formation of ion parative protein structure modeling with SWISS-MODEL and pairs. J. Org. Chem. 39 , 3837 – 3841. Swiss-PdbViewer: a historical perspective. Electrophoresis Scarisbrick, I.A., Blaber, S.I., Lucchinetti, C.F., Genain, C.P., Blaber, 30 (Suppl. 1), S162 – S173. M., and Rodriguez, M. (2002). Activity of a newly identifi ed Harris, J.L., Backes, B.J., Leonetti, F., Mahrus, S., Ellman, J.A., serine protease in CNS demyelination. Brain 125 , 1283 – and Craik, C.S. (2000). Rapid and general profi ling of protease 1296. specifi city by using combinatorial fl uorogenic substrate libraries. Schilling, O. and Overall, C.M. (2008). Proteome-derived, database- Proc. Natl. Acad. Sci. USA 97 , 7754 – 7759. searchable peptide libraries for identifying protease cleavage Hollenberg, M.D., Oikonomopoulou, K., Hansen, K.K., Saifeddine, sites. Nat. Biotechnol. 26 , 685 – 694. M., Ramachandran, R., and Diamandis, E.P. (2008). Kallikreins Schneider, E.L. and Craik, C.S. (2009). Positional scanning synthetic and proteinase-mediated signaling: proteinase-activated recep- combinatorial libraries for substrate profi ling. Methods Mol. tors (PARs) and the pathophysiology of infl ammatory diseases Biol. 539 , 59 – 78. and cancer. Biol. Chem. 389 , 643 – 651. Shimizu, C., Yoshida, S., Shibata, M., Kato, K., Momota, Y., Matsumoto, Kantyka, T., Fischer, J., Wu, Z., Declercq, W., Reiss, K., Schroder, K., Shiosaka, T., Midorikawa, R., Kamachi, T., Kawabe, A., et al. J.M., and Meyer-Hoffert, U. (2011). Inhibition of kallikrein- (1998). Characterization of recombinant and brain neuropsin, related peptidases by the serine protease inhibitor of Kazal-type a plasticity-related serine protease. J. Biol. Chem. 273 , 11189 – 6. Peptides 32 , 1187 – 1192. 11196. Krieger, E., Koraimann, G., and Vriend, G. (2002). Increasing the Sotiropoulou, G., Pampalakis, G., and Diamandis, E.P. (2009). precision of comparative models with YASARA NOVA – a Functional roles of human kallikrein-related peptidases. J. Biol. self-parameterizing force fi eld. Proteins 47 , 393 – 402. Chem. 284 , 32989 – 32994. Matsumura, M., Bhatt, A.S., Andress, D., Clegg, N., Takayama, T.K., Stefansson, K., Brattsand, M., Ny, A., Glas, B., and Egelrud, T. Craik, C.S., and Nelson, P.S. (2005). Substrates of the prostate- (2006). Kallikrein-related peptidase 14 may be a major contribu- specifi c serine protease prostase/KLK4 defi ned by positional- tor to trypsin-like proteolytic activity in human stratum corneum. scanning peptide libraries. Prostate 62 , 1 – 13. Biol. Chem. 387 , 761 – 768. Meyer-Hoffert, U., Wu, Z., Kantyka, T., Fischer, J., Latendorf, T., Stefansson, K., Brattsand, M., Roosterman, D., Kempkes, C., Hansmann, B., Bartels, J., He, Y., Glaser, R., and Schroder, J.M. Bocheva, G., Steinhoff, M., and Egelrud, T. (2008). Activation

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM Identifi cation of novel KLK14 cleavage sequences 341

of proteinase-activated receptor-2 by human kallikrein-related membrane-type serine protease 1 and identifi cation of protease- peptidases. J. Invest. Dermatol. 128 , 18 – 25. activated receptor-2 and single-chain -type plasmino- Swedberg, J.E., Nigon, L.V., Reid, J.C., de Veer, S.J., Walpole, gen activator as substrates. J. Biol. Chem. 275 , 26333 – 26342. C.M., Stephens, C.R., Walsh, T.P., Takayama, T.K., Hooper, J.D., Thornberry, N.A., Rano, T.A., Peterson, E.P., Rasper, D.M., Clements, J.A., et al. (2009). Substrate-guided design of a potent Timkey, T., Garcia-Calvo, M., Houtzager, V.M., Nordstrom, and selective kallikrein-related peptidase inhibitor for kallikrein P.A., Roy, S., Vaillancourt, J.P., et al. (1997). A combinato- 4. Chem. Biol. 16 , 633 – 643. rial approach defi nes specifi cities of members of the cas- Swedberg, J.E., de Veer, S.J., and Harris, J.M. (2010). Natural and pase family and B. Functional relationships engineered kallikrein inhibitors: an emerging pharmacopoeia. established for key mediators of apoptosis. J. Biol. Chem. 272 , Biol. Chem. 391 , 357 – 374. 17907 – 17911. Swedberg, J.E. and Harris, J.M. (2011). Plasmin substrate binding Wilson, D.R. and Finlay, B.B. (1998). Phage display: applica- site cooperativity guides the design of potent peptide aldehyde tions, innovations, and issues in phage and host biology. Can. J. inhibitors. Biochemistry 50 , 8454 – 8462. Microbiol. 44 , 313 – 329. Swedberg, J.E., de Veer, S.J., Sit, K.C., Reboul, C.F., Buckle, A.M., Yousef, G.M. and Diamandis, E.P. (2001). The new human tissue and Harris, J.M. (2011). Mastering the canonical loop of serine kallikrein gene family: structure, function, and association to protease inhibitors: enhancing potency by optimising the internal disease. Endocr. Rev. 22 , 184 – 204. hydrogen bond network. PLoS One 6 , e19302. Yousef, G.M., Stephan, C., Scorilas, A., Ellatif, M.A., Jung, K., Takayama, T.K., McMullen, B.A., Nelson, P.S., Matsumura, M., Kristiansen, G., Jung, M., Polymeris, M.E., and Diamandis, E.P. and Fujikawa, K. (2001). Characterization of hK4 (prostase), a (2003). Differential expression of the human kallikrein gene 14 prostate-specifi c serine protease: activation of the precursor of (KLK14) in normal and cancerous prostatic tissues. Prostate 56 , prostate specifi c antigen (pro-PSA) and single-chain urokinase- 287 – 292. type plasminogen activator and degradation of prostatic acid phosphatase. Biochemistry 40 , 15341 – 15348. Takeuchi, T., Harris, J.L., Huang, W., Yan, K.W., Coughlin, S.R., and Craik, C.S. (2000). Cellular localization of Received November 7, 2011; accepted December 5, 2011

Brought to you by | University of Queensland - UQ Library Authenticated Download Date | 9/14/15 2:57 AM