Substrate specificity of human kallikrein 2 (hK2) as determined
by phage display technology
Sylvain M. Cloutier
1
, Jair Ribeiro Chagas
2
, Jean-Pierre Mach
3
, Christian M. Gygi
1
, Hans-Jurg Leisinger
1
and David Deperthes
1
1
Urology Research Unit, Department of Urology, Lausanne, Switzerland;
2
Centro Interdisciplinar de Investigacao Bioquimica,
Universidade de Mogi das Cruzes, Brazil;
3
Institute of Biochemistry, University of Lausanne, Switzerland
Human glandular kallikrein 2 (hK2) is a trypsin-like serine
protease expressed predominantly in the prostate epithe-
lium. Recently, hK2 has proven to be a useful marker that
can be used in combination with prostate specific antigen
for screening and diagnosis of prostate cancer. The cleavage
by hK2 of certain substrates in the proteolytic cascade
suggest that the kallikrein may be involved in prostate
cancer development; however, there has been very little
other progress toward its biochemical characterization or
elucidation of its true physiological role. In the present

of 12 new members of the kallikrein family [5–7] could
provide additional prostate cancer markers.
In the seminal plasma, hK2 is mostly recovered com-
plexed with protein C inhibitor [1]. Because hK2 cleaves,
with trypsin-like specificity, certain components of the
semen coagulum (fibronectin and semenogelins), it is
possible that it has a role in the early stages of semen
liquefaction, a biological process which immediately follows
ejaculation [8]. In addition, in vitro studies have shown that
hK2 can activate urokinase-type plasminogen activator [9]
and inactivate plasminogen activator inhibitor-1 [10] leading
to the activation of urokinase system. Moreover, hK2
degrades insulin-like growth factor binding proteins (IGF-
BP) to release IGF, a putative local mitogenic signal for
prostate cancer cells [11].
Despite the in vitro identification of its proteolytic
activities as well as its potential substrates, our understand-
ing of the true physiological role of hK2 remains sketchy.
Much progress has been made toward the characterization
of hK2s serine protease activity using synthetic substrates
derived from reactive serpin loops [12]; however, this type of
approach is limited to known targets and cannot advance
the discovery of new biological substrates.
A system using a monovalent phage library capable of
displaying several million different substrates, which
enabled simultaneous testing of proteolytic specificity, was
developed by Matthews and Wells [13]. Several proteases
including furin [14], PSA [15], membrane type-1 matrix
metalloproteinase [16], and granzyme B [17] have already
been characterized using this approach.

(Roche Biosciences; Amersham Pharmacia), PWO DNA
polymerase and shrimp alkaline phosphatase (Roche Bio-
sciences), T4 DNA ligase (Invitrogen), T4 polynucleotide
kinase (Promega), Ni
2+
-nitrilotriacetic acid agarose, anti-
His antibody, Ni
2+
-nitrilotriacetic acid magnetic agarose
beads and 96-well magnet type A (Qiagen). Mycrosynth
GmbH carried out DNA sequencing and oligonucleotides
synthesis.
Construction of the substrate phage display library
Substrate phage libraries were generated using a modified
pH0508b phagemid [20]. The construction consists of a His
6
tag at either end of a Gly-Gly-Gly-Ser-repeat-rich region
that precedes the carboxyl-terminal domain (codons 249–
406) of the M13 gene III. The random pentamers were
generated by PCR extension of the template oligonucleo-
tides with appropriate restriction sites positioned on both
side of the degenerate codons: 5¢-TGAGCTAGTCTAGAT
AGGTGGCGGTNNSNNSNNSNNSNNSGGGTCGAC
GTCGGTCATAGCAGTCGCTGCA-3¢ (where N is any
nucleotide and S is either G or C) using 5¢ biotinylated
primers corresponding to the flanking regions: 5¢-TGAGC
TAGTCTAGATAGGTG-3¢ and 5¢-TGCAGCGACTGC
TATGA-3¢. PCR templates are digested and purified as
described previously [21], inserted into XbaI/SalIdigested
pH0508b vector, and electroporated into XL1-Blue (F

(10
11
) were added to the equilibrated Ni
2+
-nitrilotriacetic
acidresinandallowedtobindwithgentleagitationfor
3hat4°C. The resin was subsequently washed (NaCl/P
i
/
BSA 1 mgÆmL
)1
,5m
M
imidazole, 0.1% Tween 20) to
remove unbound phages and then equilibrated in NaCl/
Pi. The substrate phage was exposed to 27 n
M
(final
concentration) of hK2 for 45 min at 37 °C. A control
selection without protease was also performed. The
cleaved phages released into the supernatant were ampli-
fied using XL1-Blue Escherichia coli andthenusedfor
subsequent rounds of selection. After eight rounds of
panning, about 15 individual clones were picked from the
fifth, sixth and eighth round of selection and plasmid
DNA were isolated and sequenced in the region encoding
for the substrate.
Expression of CFP fluorescent substrate
The construction CFP-X
5

L) with ampicillin (100 lgÆmL) and tetracycline (15 lgÆmL)
antibiotics. Cells were then induced until D
600
¼ 0.5 to
express recombinant fluorescent substrate by addition of
1m
M
of isopropyl thio-b-
D
-galactoside (IPTG) for 16 h at
37 °C. After an additional 16 h of growth, the cells were
harvested by centrifugation and resuspended for 2 h at
room temperature in 6 mL denaturation buffer (6
M
GdN–
HCl in NaCl/P
i
at pH 8.0 containing 10 m
M
2-mercapto-
ethanol) to recover the soluble and insoluble fractions. All
recombinant CFPs were purified in denaturing conditions
to prevent substrate cleavage by endogenous bacterial
proteases. After centrifugation, 100 lLNi
2+
-nitrilotriacetic
acid resin was added to the bacterial cell supernatant and
incubated to bind recombinant proteins. The resin was
subsequently washed with 5
M

conjugated anti-His
6
Ig (Qiagen).
2748 S. M. Cloutier et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Direct determination of the
k
cat
/
K
m
using CFP fluorescent substrates
Refolded CFP-X5-His proteins were fixed to Ni
2+
-nitrilo-
triacetic acid magnetic beads for 2 h at room temperature
and an aliquot was collected for eluting proteins to
determine the specific activity (fluorescence/amount of
protein) and initial substrate concentration [S
0
] for each
CFP. Concentrations were determined by Bradford assay
(Biorad, USA). All of the kinetic assays were carried out at
37 °Cin50m
M
Tris/HCl buffer pH 8.3, containing 0.01%
Tween 20, for 120 min The time course of substrate
hydrolysis was followed by monitoring the fluorescence
released from the beads as the CFPs were cleaved in their
substrate linker. Percentage of hydrolysis was calculated as
the ratio of released CFPs to the initial amount of CFPs

3.2 · 10
6
possible random pentamer sequences were repre-
sented. The sequencing of phages further confirmed the
randomness of the pentamer inserts.
Random selection of hK2 substrates
Although the filamentous phages are considered to be
generally protease resistant, we first verified that hK2
activity had indeed no effect on infectivity. Following eight
rounds of exposure to hK2, 44 individual phage clones were
selected from different rounds; the deduced amino acids
corresponding to the substrate sequences are shown in
Table 1. No phage was selected more than once, indicating
that a large repertoire of susceptible substrates was present
in the pentamer library. DNA sequence analysis reveal that
an arginine appears in 40 clones at the P1 site and only one
peptide is cleaved at a lysine. Among the substrates
hydrolysed at an arginine, 11 different amino acids appeared
at the P¢1 subsite. However, some amino acids were more
frequently recovered at this position with 30% of selected
peptides exhibiting serine and 12% methionine, alanine, or
valine. Interestingly, an evolution of the representation of
scissile bonds emerged during the selection (Fig. 2); the
highest variation was observed for the Arg–Ser scissile bond
with continuously increasing recovery of 14, 32, and 42%,
respectively, for the fifth, sixth, and eighth rounds of
selection. A slight increase was also observed in the Arg–
Met and Arg–Ala motif, while an important decrease was
observed for the Arg–Val motif through the selection, which
completely disappeared after eight rounds. The positions

are amplified in region encoding for the substrate to determine the
sequences cleaved by the enzyme. (7) Gene encoding the random
substrate was subcloned into an expression vector, in order to be
produced as a fusion protein between the CFP protein and a histidine
tag. (8) The CFP-X5-his protein was fixed to Ni
2+
-nitrilotriacetic acid
magnetic beads and (9) treated by the protease hK2. The released CFP
fluorescent protein was measured with a fluorescence reader (10) which
permitted to determine the percentage of hydrolysis, the specificity
constant and the site of cleavage (11).
Ó FEBS 2002 Substrate specificity of kallikrein hK2 (Eur. J. Biochem. 269) 2749
nitrilotriacetic acid beads; the substrate can then be released
by hydrolysis only. By using two CFP–substrates harbour-
ing a substrate that is either cleavable or resistant to hK2,
we showed that the CFP recombinant protein is cleaved
only in the substrate region and not within the CFP
sequence as no fluorescence was detectable with CFP-
resistant. On the other hand, CFP–PCI was efficiently
cleaved with a first-order curve for the product generation
(data not shown) and the specificity constant k
cat
/K
m
was
 20 000
M
)1
Æs
)1

6.11 G V F R S
6.19 G T V R S
5.5 E T K R S
5.2 L G R S L
8.3 R G R S E
6.2 R R S I D
8.11 V L R S P
8.20 L R S R A
8.5 RS GS V
8.9 A R A R S
8.18 RT S D R T A
6.7 K L R T T
8.13 RA R A A M M
5.3 T R A P M
8.17 P G R A P
6.9 V E S R A
6.20 A R A S E
5.19 RV T L Q R V
5.16 R L E R V
5.18 E R V S P
5.12 S S P R V
6.17 RVGP Y
6.4 RM P S A R M
6.14 R G R M A
6.5 T V R M P
8.12 L R M P T
8.14 H R M S S
5.11 RP R P Q E L
6.15 V R P L E
5.7 RL S G R L A

Æs
)1
possess two basic amino acids N-terminal to the scissile
bond except for peptide LRSRA where the second basic
residue was found at P¢2(Table2).
Comparison with natural substrate
When compared to previously reported substrates for hK2,
the peptides selected here had scissile bonds containing the
Arg–Ser motif, which is the same bond cleaved in PCI, a
natural inhibitor of hK2 found in seminal plasma, as well as
semenogelin I, antithrombin III, and kininogen. The Arg–
Thr and Arg–Leu motifs are hydrolysed by hK2 in
semenogelins I and II whereas the Arg–Met motif is cleaved
in the plasminogen activator inhibitor-1 and the Arg–Gln
motif is cleaved in IGF-BP-2. Using each of the 44
pentapeptides substrate sequences,
FASTA
and
BLAST
searches were done to look for new potential human protein
substrates of hK2 (Table 3). Among the 11 identical
matches (data not shown), three putative targets were
identified for hK2: ADAM-TS 8 precursor, cadherin-rela-
ted tumour suppressor homologue precursor, and collagen
(IX) chain precursor matching peptides RGRSE, GVFRS
and PGRAP, respectively.
DISCUSSION
A wide variety of critical processes depend on specific
cleavage of targets by different enzymes so an ability to
discriminate among many potential substrates is crucial to

Phage display does generate libraries that are many times
more diverse,however, than those using other methods such
as combinatorial chemistry [17] or immobilized positional
peptide libraries [27].
In our experiments, randomised pentapeptides were fused
to a truncated form of g3p to produce a library of 1.8 · 10
8
independent recombinant phages where all possible combi-
nations of sequences even the rarest polypeptides, are
represented. The screening of this library with hK2 showed
that no phage was in duplicate which is in contrast to
selections with other types of phage display libraries
(antibody fragments, ligands, or peptide binders) where
selections often identified only the best clones with highest
reactivity [28,29]. Our results are consistent with other
studies using phage display that reported a broad diversity
but good enrichment in the selection of specific enzyme
substrates [13,15,25].
The determination of the specificity constants of the
substrates showed a positive tendency during the selection.
Most of the better substrates were taken the last rounds.
However, this does not preclude that bad substrates could
be conserved throughout the screening process despite
selection pressure. Therefore, selected substrates need to be
further tested in other configuration than that of fused to a
phage. The CFP system developed in the present work
enabled direct determination of specificity constants and the
site of cleavage of the substrate selected by phage display, an
improvement over the previously described semiquantitative
method [13,25] and chemical synthesis of substrates [15,30].

Despite this observation, hK2 seems to be dependent on a
more extended site of binding than R–S bond for an efficient
catalysis as some Arg–Ser peptides possess lower specificity
constants. Nonetheless, the observation that the best three
peptides are cleaved as efficiently as the sequence of PCI–
peptide obtained by the classic iterative methods indicates
the impressive ability of substrate phage technology to
elucidate optimal subsite occupancy for proteases from
large banks of randomly selected candidates.
Interestingly, the Arg–Ser scissile bond found in numer-
ous natural substrates like PCI, semenogelins I and II,
fibronectin and kininogen as well as other preferential
cleavage sites like Arg–Thr or Arg–Met in seminal coagu-
lum proteins and in plasminogen activator inhibitor-1,
respectively; is also preferentially selected by hK2 using
phage display substrates thus confirming the success of
phage display substrate selection.
Finally, a SwissProt database search with selected
sequences identified three potential human protein sub-
strates for hK2. Regions identified in different substrates are
extracellular and thus accessible to proteases. These poten-
tial substrates are not yet well characterized, but are
suspected to be involved in cancer progression. For
example, the desintegrin-like and metalloprotease domain
with thrombospondin type I modules 8 (ADAM-TS8)
Table 2. Comparaison of specificity constant (k
cat
/K
m
) values and the percentage hydrolysis of CFP-X5-his based on selected substrates with hK2.

5.7 SGRflLA 5.78 E-05 30.5 3042
8.11 VLRflSP 5.24 E-05 33.1 2756
8.19 TRDSR 4.84 E-05 30.9 2548
5.10 IMSRflQ 4.77 E-05 27.1 2512
6.5 TVRflMP 4.35 E-05 24.6 2289
6.20 ARflASE 4.10 E-05 23.5 2158
6.6 KTRflSN 3.64 E-05 26.2 1917
6.1 MTRflSN 3.37 E-05 22.9 1772
6.3 LTTSKfl 3.24 E-05 19.6 1705
5.4 TGSRflD 2.69 E-05 16.1 1417
6.15 VRflPLE 2.37 E-05 13.1 1248
5.14 RflNDKL 2.27 E-05 19.1 1196
5.18 ERflVSP 1.89 E-05 11.2 99
MTMQS ND ND ND
QTSLS ND ND ND
Rst AIKFF ND ND ND
Table 3. Identification of potential physiological substrate of hK2 using
the SwissProt data base.
HK2 selected
peptides Sequences Potential protein substrate (residues)
8.3 RGRflSE ADAM-TS 8 precursor (646–50)
6.19 GVFRflS Cadherin-related tumour suppressor
homologue precursor (2473–77)
8.17 PGRflAP Collagen alpha (IX) chain precursor
(753–57)
2752 S. M. Cloutier et al. (Eur. J. Biochem. 269) Ó FEBS 2002
could act as a tumour suppressor through its antiangiogenic
activity [32,33]. Cadherin-related tumour suppressor homo-
logue precursor [34] and collagen alpha (IX) chain precur-
sor, a minor cartilage nonfibrillar collagen associated with

(1997) Human glandular kallikrein 2 (hK2) expression in prostatic
intraepithelial neoplasia and adenocarcinoma: a novel prostate
cancer marker. Urology 49, 857–862.
4. Darson, M.F., Pacelli, A., Roche, P., Rittenhouse, H.G., Wolfert,
R.L., Saeid, M.S., Young, C.Y., Klee, G.G., Tindall, D.J. &
Bostwick, D.G. (1999) Human glandular kallikrein 2 expression in
prostate adenocarcinoma and lymph node metastases. Urology 53,
939–944.
5. Yousef, G.M., Obiezu, C.V., Luo, L.Y., Black, M.H. &
Diamandis, E.P. (1999) Prostase/KLK-L1 is a new member of the
human kallikrein gene family, is expressed in the prostate and
breast tissues, and is hormonally regulated. Cancer Res. 59, 4252–
4256.
6. Diamandis, E.P., Yousef Clements, G.M., Ashworth, J., Yoshida,
L.K., Egelrud, S., Nelson, T., Shiosaka, P.S., Little, S., Lilja, S.,
Stenman, H., Rittenhouse, U.H., & Wain, H.G. (2000) New
nomenclature for the human tissue kallikrein gene family. Clin.
Chem. 46, 1855–1858.
7. Yousef, G.M. & Diamandis, E.P. (2001) The new human tissue
kallikrein gene family: structure, function, and association to
disease. Endocrine Rev. 22, 184–204.
8. Deperthes, D., Frenette, G., Brillard-Bourdet, M., Bourgeois,
L., Gauthier, F., Tremblay, R.R. & Dube, J.Y. (1996) Potential
involvement of kallikrein hK2 in the hydrolysis of the
human seminal vesicle proteins after ejaculation. J. Androl. 17,
659–665.
9. Frenette, G., Tremblay, R.R., Lazure, C. & Dube, J.Y. (1997)
Prostatic kallikrein hK2, but not prostate-specific antigen (hK3),
activates single-chain urokinase-type plasminogen activator. Int. J.
Cancer 71, 897–899.

ficity by using combinatorial fluorogenic substrate libraries. Proc.
Natl Acad. Sci. USA 97, 7754–7759.
18. Frenette, G., Deperthes, D., Tremblay, R.R., Lazure, C. & Dube,
J.Y. (1997) Purification of enzymatically active kallikrein hK2
from human seminal plasma. Biochim. Biophys. Acta 1334, 109–
115.
19. Knight, C.G. (1995) Fluorimetric assays of proteolytic enzymes.
Methods Enzymol. 248, 18–34.
20. Lowman, H.B., Bass, S.H., Simpson, N. & Wells, J.A. (1991)
Selecting high-affinity binding proteins by monovalent phage
display. Biochemistry 12, 10832–10838.
21. Smith, G.P. & Scott, J.K. (1993) Libraries of peptides and proteins
displayed on filamentous phage. Methods Enzymol. 217, 228–257.
22. Laemmli UK (1970) Cleavage of structural proteins during
the assembly of the head of bacteriophage T4. Nature 227, 680–
685.
23. Elmoujahed, A., Gutman, N., Brillard, M. & Gauthier, F. (1990)
Substrate specificity of two kallikrein family gene products
isolated from the rat submaxillary gland. FEBS Lett. 265,
137–140.
24. Mikolajczyk, S.D., Millar, L.S., Kumar, A. & Saedi, M.S. (1998)
Human glandular kallikrein, hK2, shows arginine-restricted spe-
cificity and forms complexes with plasma protease inhibitors.
Prostate 34, 44–50.
25. Smith, M.M., Shi, L. & Navre, M. (1995) Rapid identification of
highly active and selective substrates for stromelysin and
matrilysin using bacteriophage peptide display libraries. J. Biol.
Chem. 270, 6440–6449.
26. Lowman, H.B., Bass, S.H., Simpson, N. & Wells, J.A. (1991)
Selecting high-affinity binding proteins by monovalent phage

Chin, W.G.S.M., Zhao, Q., Beverley, P.C. & Owen, M.J. (1995)
Molecular cloning and tissue expression of FAT, the human
homologue of the Drosophila fat gene that is located on chro-
mosome 4q34-q35 and encodes a putative adhesion molecule.
Genomics 30, 207–223.
35.Muragaki,Y.,Kimura,T.,Ninomiya,Y.&Olsen,B.R.
(1990) The complete primary structure of two distinct forms of
human alpha 1 (IX) collagen chains. Eur J. Biochem. 192, 703–708.
2754 S. M. Cloutier et al. (Eur. J. Biochem. 269) Ó FEBS 2002