Báo cáo khoa học: A novel serine protease highly expressed in the pancreas is expressed in various kinds of cancer cells potx

However, western blot and RT-PCR analyses showed that prosemin is expressed and secreted from various kinds of cancer cells, such as glioma, pancreas, prostate, and ovarian cell lines..

is expressed in various kinds of cancer cells

Shinichi Mitsui1,*, Akira Okui2, Katsuya Kominami2, Eiichi Konishi1, Hidetoshi Uemura2

and Nozomi Yamaguchi1

1 Department of Cell Biology, Research Institute for Geriatrics, Kyoto Prefectural University of Medicine, Kyoto, Japan

2 Research and Development Center, Fuso Pharmaceutical Industries Ltd, Morinomiya, Joto-ku, Osaka, Japan

Serine proteases play important roles in a variety of

physiological processes, such as blood coagulation,

growth factor processing, the digestion of nutrients,

and cell migration, when cells move to the target

organs and reconstitute extracellular matrices,

especi-ally during developmental stages In the central

ner-vous system (CNS), serine proteases are believed to be

involved in various neuronal functions [1] Tissue-type

plasminogen activator is essential for memory forma-tion and seizures [2,3] Knockout mice lacking neuro-psin⁄ KLK8, a member of the kallikrein family, have a reduced number of active synapses [4] On human and mouse genomes, over 500 proteases are encoded [5] and some proteases other than plasminogen activator and neuropsin may contribute to neuronal function

We have recently reported a novel transmembrane

Keywords

cerebellum; chromosome 16; ovarian

cancer; tumor marker

Correspondence

N Yamaguchi, Department of Cell Biology,

Research Institute for Geriatrics, Kyoto

Prefectural University of Medicine,

Kyoto 602-8566, Japan

Fax: +81 75 251 5848

Tel: +81 75 251 5797

E-mail: nozomi@koto.kpu-m.ac.jp

*Present address

Department of Neurobiology and Anatomy,

Kochi Medical School, Okou, Nankoku

783-8505, Japan

Database

The nucleotide sequence reported in this

paper has been entered in the DDBJ ⁄

Gen-Bank ⁄ EMBL databases under accession no.

AB010779

(Received 25 November 2004, revised 4

July 2005, accepted 3 August 2005)

doi:10.1111/j.1742-4658.2005.04901.x

We have isolated a cDNA that encodes a novel serine protease, prosemin, from human brain The cDNA of human prosemin is 1306 bp, encoding

317 amino acids It showed significant homology with the sequence of a chromosome 16 cosmid clone (accession no NT_037887.4) The prosemin gene contains six exons and five introns The amino acid sequence of prose-min shows significant homology to prostasin, c-tryptase, and testisin (43%, 41%, and 38% identity, respectively), the genes of which are also located

on chromosome 16 Northern hybridization showed that prosemin is expressed predominantly in the pancreas and weakly in the prostate and cerebellum However, western blot and RT-PCR analyses showed that prosemin is expressed and secreted from various kinds of cancer cells, such

as glioma, pancreas, prostate, and ovarian cell lines Prosemin is secreted

in the cystic fluid of clinical ovarian cancers Furthermore, immunohisto-chemistry showed prosemin protein localized in the apical parts of ovarian carcinomas Recombinant prosemin was expressed in COS cells and was purified by immunoaffinity chromatography Recombinant prosemin pre-ferentially cleaved benzyloxycarbonyl (Z)-His-Glu-Lys-methylcoumaryl amidide (MCA) and t-butyloxycarbonyl (Boc)-Gln-Ala-Arg-MCA Our results suggest that prosemin is a novel serine protease of the chromo-some 16 cluster that is highly expressed in the pancreas The usefulness

of this serine protease as a candidate tumor marker should be further examined

Abbreviations

Boc, t-butyloxycarbonyl; Bz, benzoyl; CNS, central nervous system; Glu(Obzl), glutamic acid c-benzyl ester; Glp, L -pyroglutamyl; MCA, 4-methylcoumaryl-7-amidide; NHS, N-hydroxysuccinimide; Z, benzyloxycarbonyl.

serine protease, spinesin⁄ TMPRSS5, which is localized

at synapses [6] Motopsin (PRSS12) is a mosaic serine

protease, with a kringle domain and three scavenger

receptor cysteine-rich domains, that is expressed

pref-erentially in motor neurons [7,8] Some novel members

of the kallikrein family have also been reported

Hip-postasin (KLK11) is expressed in hippocampal

pyram-idal neurons and shows kallikrein-like enzyme activity

[9] Another kallikrein-like protease, neurosin (KLK6),

which has trypsin-like activity, is secreted by activated

microglial cells and localized on senile plaques and

Lewy’s bodies [10,11] We have shown that the

alter-native splicing of neuropsin⁄ KLK8 is regulated in a

tissue-specific manner [12] These proteases are thought

to play important roles in neuronal functions via the

digestion of extracellular matrices, the processing of

growth factors, and the activation of cell-surface

recep-tors However, such proteolytic phenomena occur

markedly in tumor cells as well as in the CNS Some

of the proteases described above have been reported to

be tumor markers, although they are expressed

pre-dominantly in the CNS Neuropsin was identified as a

serine protease that is overexpressed in ovarian

carci-nomas, although its mRNA is detected negligibly in

normal ovaries [13,14] Serum neurosin levels are

ele-vated specifically in patients with ovarian cancer [15]

Furthermore, we have previously reported that the

alternative splicing of hippostasin is regulated in a

cancer-specific manner [16] In addition, serum levels

of hippostasin are increased in patients with ovarian

and prostate cancers [17] Other kallikreins are

aber-rantly expressed in many types of cancers and their

expression is often associated with patient prognosis

(reviewed in [18–20])

Here, we describe a novel member of the

chromo-some 16 serine protease family Prosemin cDNA was

isolated from a human brain cDNA pool

Immuno-chemical and RT-PCR analyses revealed that prosemin

is expressed in various kinds of cancer cell lines and in

clinical samples of ovarian carcinomas The

characteri-zation and functions of prosemin are described

Results

cDNA cloning and structure of the prosemin

gene

We isolated a cDNA of 958 bp that encodes a novel

serine protease of 317 amino acids from human brain

cDNA pool The predicted amino acid sequence

includes a serine protease motif, and this protein was

designated ‘prosemin’, as this protein was first detected

in seminal fluid (data not shown) A BLAST search

of the human genome draft sequence showed that this gene is located on chromosome 16p13.3 (NT_037887.4) Although human prosemin cDNA was isolated from the brain, northern hybridization showed that prosemin is expressed strongly in the pancreas, but weakly in the prostate and cerebellum (Fig 4) The nucleotide sequences of the cDNA and the cosmid clone are identical A comparison of the cDNA and ge-nomic sequences clarified the gene structure of human prosemin (Fig 1) The prosemin gene is approximately 6.5 kb, and contains six exons and five introns All exon–intron boundaries are consistent with the GT–

AG rule The exons range in size from 27 to 601 bp The second exon encodes only nine amino acids, which constitute part of the pro-enzyme fragment The initi-ation codon is consistent with Kozak’s consensus sequence [21] Although we did not determine the tran-scription initiation site, the 5¢ UTR appears to be lon-ger than 67 bp because the primer for the first PCR used to isolate the full-length cDNA was designed to hybridize with that position (at 991 in Fig 2) Further, homology search showed that our sequence contained the longest 5¢ UTR when compared with EST sequences Several transcription regulatory elements occur within 1.2 kb upstream from the first exon: three AP-1 sites, a CREB binding site, and a NF-jB binding site There is no typical TATA or CAAT box in the putative promoter region The putative signal sequence

is encoded by the first exon, and the pro-enzyme frag-ment is encoded by exons 1 and 2 The catalytic triad, His, Asp, and Ser, is encoded in exons 3, 4, and 6, respectively Exon 6 contains a polyadenylation signal

Structure of human prosemin The predicted amino acid sequence of prosemin shows that this protein belongs to the chymotrypsin (S1) fam-ily and may be synthesized as a prepro form Hydro-pathy plot analysis showed that the 32 amino acids of the N-terminus constitute a putative signal sequence (data not shown) A typical serine protease activation motif is also conserved at Arg49-Val-Val-Gly-Gly, sug-gesting that the 17 amino acids from Ala33 to Arg49 might comprise the pro-fragment of prosemin A pos-sible N-glycosylation site occurs at Asn70 in the cata-lytic domain The essential triad for protease activity was identified at His90, Asp141, and Ser242 An aspar-tate residue six amino acids before the active Ser242 suggests that prosemin has trypsin-like activity In addition to a similarity in gene structure, a homology search of the SWISSPROT protein database showed significant similarity between prosemin and prostasin, c-tryptase, and testisin, with 43%, 41%, and 37.5%

Fig.

identity, respectively (Fig 2) Amino acid alignment of

these proteases shows that 10 of the 12 cysteine

resi-dues in the prosemin zymogen are conserved among

these proteases Cys41 is predicted to form a disulfide

bond with Cys175, as in other S1 proteases, whereas

the a-, b-, and d-tryptases do not contain a cysteine

residue in the enzyme fragment However, pro-semin does not contain a hydrophobic region at the carboxyl terminus This portion is apparently different from the corresponding regions of prostasin, c-tryptase and testisin, which anchor the proteins to the plasma membrane

Fig 2 Amino acid alignment of the chromosome 16 serine protease family White letters indicate amino acids identical to those in human prosemin Asterisks show the amino acids conserved among the seven proteases Dashes denote gaps The essential triad and peripheral sequences are indicated by lines Cys41 in the pro-enzyme fragment may form a disulfide bond with Cys175 in the catalytic domain.

Enzyme activity of human prosemin

To analyze the enzyme activity of prosemin,

recombin-ant prosemin was expressed in COS cells In our system,

the putative catalytic domain was fused with the

prepro-enzyme fragment of human trypsinogen Recombinant

prosemin was secreted into the conditioned medium

and purified by immunoaffinity chromatography SDS⁄

PAGE of purified prosemin produced a single band of

33 000 Da, which reacted with anti-prosemin IgG

(Fig 3A) Purified prosemin still contained the

pro-enzyme fragment of trypsinogen and was activated by

enterokinase (Fig 3B–D) Activated prosemin cleaved a

fluorescent substrate [t-butyloxycarbonyl (Boc)-Gln-Ala-Arg-4-methylcoumaryl-7-amidide (MCA)], whereas preactivated recombinant prosemin and enterokinase showed no activity against the same substrate SDS⁄ PAGE showed that enterokinase treatment of recombinant prosemin caused a decrease in the mole-cular mass from 33 000 to 31 000, indicating successful digestion of the artificial trypsinogen pro-enzyme frag-ment The amount of 31 kDa prosemin increased in pro-portion to the time of incubation with enterokinase Prosemin activity appeared to be related to the amount

of 31 kDa protein, although extended incubation caused some degradation of the recombinant prosemin,

F E

D

Fig 3 Enzymatic characterization of recombinant prosemin (A) SDS ⁄ PAGE of purified recombinant prosemin Anti-prosemin IgG raised against a synthetic peptide recognized this protein (lane 1) Purified prosemin was detected as a single band by silver staining (lane 2) (B) After activation by enterokinase, recombinant prosemin was incubated with Boc-Gln-Ala-Arg-MCA at 37
C for the indicated times White and shadowed bars indicate before and after activation by enterokinase, respectively Black bars indicate the enzymatic activity of entero-kinase on the substrate The closed circles and line indicate prosemin enzyme activity after activation of the protein by enteroentero-kinase; the enzymatic action of enterokinase upon the substrate has been subtracted Mean values and standard deviations (SD; n ¼ 3) are indicated (C) Activation of recombinant prosemin by enterokinase Purified recombinant prosemin was incubated with enterokinase at 37
C for 0 (lane1), 5 (lane 2), 24 (lane 3), 48 (lane 4) h After incubation, the recombinant protein was visualized by silver staining after SDS ⁄ PAGE (D) The enzymatic activity of prosemin activated for various times was measured at 37
C using the substrate Boc-Gln-Ala-Arg-MCA (E) Recom-binant prosemin activated for 24 h was incubated with the indicated substrates for 1 h at 37
C Mean values and SD are shown (n ¼ 3) (F) Activated prosemin was incubated with Boc-Gln-Ala-Arg-MCA at 37
C for 1 h at various pHs Phosphate buffer (0.1 M ) and Tris ⁄ HCl (0.1 M ) buffer were used for pH 5.0–9.0 and pH 9.0–11.0, respectively Mean values and SD are shown (n ¼ 3).

including autodegradation These results indicate that

the detected activity derived from the activated

recom-binant prosemin

At pH 8.0, recombinant prosemin the most

preferen-tially cleaved benzyloxycarbonyl

(Z)-His-Glu-Lys-MCA, and it showed high enzymatic activity against

Boc-Gln-Ala-Arg-MCA and

Boc-Glu(Obzl)-Ala-Arg-MCA, which are substrates for trypsin and factor XIa,

respectively, and weakly cleaved

Boc-Val-Pro-Arg-MCA, Pro-Phe-Arg-Boc-Val-Pro-Arg-MCA, l-Pyroglutamyl

(Glp)-Arg-MCA, and Boc-Phe-Ser-Arg-MCA (Fig 3E) The

optimum pH for prosemin activity was investigated

using Boc-Glu-Ala-Arg-MCA as substrate (Fig 3F)

Activated recombinant prosemin showed no enzymatic

activity at pH < 6.0 Prosemin showed maximum

enzymatic activity at pH 8.0–9.0, and showed about

half maximal activity at pH > 10.0

Expression of prosemin mRNA in normal human

tissues

Prosemin mRNA of 1.4 kb was expressed strongly in

the pancreas and weakly in the prostate and

cerebel-lum (Fig 4A) We isolated a cDNA fragment encoding prosemin from a human brain cDNA pool However, prosemin mRNA was detected only in the cerebellum

It was not detected in other regions of the brain, such

as the thalamus, substantia nigra, hippocampus, cor-pus callosum, caudate nucleus, or amygdala (data not shown) However, using RT-PCR, we detected prose-min mRNA in most human tissues except the hippo-campus and liver (Fig 4B) The levels of prosemin expression are very low in these tissues

Prosemin expression in various cancer cell lines Recently, some serine proteases, including testisin and prostasin, have been reported as tumor markers The expression of prosemin in cancer cell lines was ana-lyzed Immunoblotting using anti-prosemin IgG detec-ted a protein of 33 kDa, which is identical to the molecular mass of recombinant prosemin, in condi-tioned medium from some human cancer cell lines (Fig 5A) A 20 kDa protein was detected in some cell lines, such as HPC-Y0, HPC-YS, SKOV-3, OVK-18, DU-145, U373, KHG2, HOG, KP-N-YN, Colo38,

A

B

Fig 4 Expression of prosemin in normal human tissues (A) Northern hybridization was carried out overnight at 65
C Strong signal was detected in the pancreas, and weak signals were observed in the prostate and cerebellum (B) Expression of prosemin mRNA was analyzed

by RT-PCR A specific band of the expected size was detected in most human tissues (arrowhead).

and OUR-10, although the 33 kDa protein was not

detected Either the 33 kDa or the 20 kDa protein was

detected in all 11 pancreatic cancer cell lines and four

colon cancer cell lines An immunoreactive 20 kDa

protein was detected in both of two prostate cell lines,

two of three ovarian cell lines, and one of three lung

cancer cell lines There was no detectable protein in

the conditioned media from some cell lines: SKLU1,

K562, HUE, Ishikawa-EM, and glioma Ishikawa

RT-PCR was used to confirm the expression of prosemin

mRNA in some cell lines A band of the predicted size

(500 bp) of the PCR product was observed for most of

the cell lines tested, including HPC-YO and HPC-YS,

in which only the 20 kDa protein was detected

(Fig 5B) However, no PCR product was detected in

Ishikawa, a glioma line HUE and OVK18 expressed

low levels of prosemin mRNA The transcript levels of

prosemin suggested by RT-PCR analysis were

consis-tent with the results of western blot analysis, although

RT-PCR is relatively nonquantitative A431, DU-145,

and LNCap showed some smaller bands, which may

reflect splicing variants because we identified eight

splicing variants during the PCR cloning process (S Mitsui, A Okui, K Kominami, E Konishi, H Uem-ura and N Yamaguchi, unpublished data) The expres-sion pattern of prosemin in cancer cell lines is summarized in Table 1

Immunohistochemical detection of prosemin

in ovarian cancer Western blot analysis and RT-PCR indicated the expression of prosemin in ovarian cancer cell lines Ovarian cancers express some kinds of chromosome 16 serine proteases, including testisin and prostasin We investigated the expression of prosemin in ovarian can-cers using immunohistochemistry Western blot analy-sis showed that prosemin occurred in ovarian cystic fluids (Fig 5C) As shown in Fig 6, immunoreactive prosemin was localized at the apical portion of carci-noma cells, whereas nonimmune serum produced no signal This localization was identical in mucinous and serous adenocarcinomas The strong expression of prosemin in serous carcinomas is noteworthy Three of

A

Fig 5 Secretion of prosemin protein and expression of prosemin mRNA in various kinds of cancer cells (A) Conditioned medium (25 lg) from various cancer cell lines was used for western blot analysis with antibody raised against a synthetic prosemin peptide A 33 kDa immu-noreactive protein (black arrowhead) and a 20 kDa protein (white arrowhead) were detected (B) RT-PCR for prosemin mRNA in cancer cell lines The predicted size of the PCR product was 500 bp and the specific PCR product was detected in various kinds of cell lines It should

be noted that Ishikawa, OVK-18, and HUE showed no band or weak bands, which is consistent with the results of the western blot analysis (C) Western blot analysis of cancerous fluid from clinical ovarian cancers Immunoreactive protein was detected in the cystic fluid as a

30 kDa protein (lanes 2,3), whereas recombinant prosemin produced a 33 kDa band (lane 1).

five serous adenocarcinomas and three of eight

mucin-ous adenocarcinomas were positively stained by

anti-prosemin IgG

Discussion

We have isolated a cDNA encoding a novel serine

pro-tease, prosemin, from a human brain cDNA pool

Homology search to human genomic sequence

indica-ted that the prosemin gene is locaindica-ted on chromosome

16p13.3 In this region, genes for at least eight serine

proteases are clustered: aII-, b-, cI-, cII-, w-tryptases

[22], testisin [23], prostasin [24], and prosemin On the

syntenic region of mouse chromosome 17, there is a tryptase gene cluster consisting of at least four serine proteases [25] Among the chromosome 16 protease family members, the prosemin gene structure is similar

to the structures of prostasin, c-tryptase, and testisin The characteristic feature of these genes is the short second exon, which encodes only six or nine amino acids of the pro-peptide fragment (Fig 1) No TATA box sequence has been identified in the promoter regions of these genes The amino acid sequence of prosemin is up to 43% homologous to those of prosta-sin and c-tryptase (Fig 2) Furthermore, two cysteine residues, Cys41 and Cys161, are conserved only among these four proteases, and not in the other tryptases These results suggest that, of the chromosome 16 serine protease family members, these four proteases at least are derived from a single ancestral gene

The deduced amino acid sequence of prosemin shows the characteristic feature of the S1 family The triad essential for serine protease activity and the peripheral consensus sequence have been identified (Figs 1 and 2)

An aspartic acid residue six amino acids before the act-ive Ser242 suggests that prosemin preferentially cleaves after Arg or Lys residues in a target substrate In fact, recombinant prosemin cleaved Z-His-Glu-Lys-MCA, Boc-Glu(Obzl)-Ala-Arg-MCA, and Boc-Gln-Ala-Arg-MCA Prosemin might be synthesized as a prepro-pro-tein because it contains 32 hydrophobic amino acids at the amino terminal, which constitute a possible signal peptide Prosemin protein was secreted into conditioned medium when the full-length prosemin cDNA was expressed by a baculovirus expression system (data not shown) In fact, western blot analysis showed that prose-min is secreted in the cystic fluids of ovarian cancers (Fig 5C) Prosemin secreted by human cancer cell lines was detected as a 33 kDa protein However, the prose-min zymogen contains 17 aprose-mino acids of the pro-enzyme fragment and 268 amino acids of the mature enzyme, so the molecular mass is calculated to be about

31 kDa This discrepancy may be attributable to post-translational modifications, such as N-glycosylation at Asn70 An activation motif typical of serine proteases is also observed at Arg49-Val-Val-Gly-Gly-Glu-Asp The released pro-enzyme fragment is predicted to be linked with the catalytic domain by a disulfide bond between Cys41 and Cys175; this is typical of many members of the S1 family

We confirmed the enzymatic activity of prosemin with a recombinant protein In our system, recombin-ant prosemin was secreted into conditioned medium (Fig 3A), although it was composed of the trypsino-gen prepeptide and the catalytic domain of pro-semin We have developed this chimeric protein system

Table 1 Expression and secretion of prosemin from cancer cell

lines ND, not determined.

Cell line Tissue

Western blot of conditioned media

RT-PCR

33 kDa 20 kDa

to use the attendant advantages, although full-length

cDNA can be expressed in an expression system The

recombinant protein was secreted into the conditioned

medium by the trypsinogen signal peptide, which

facili-tates collection of the protein Furthermore, the

trypsi-nogen prepro-peptide seems to function as a chaperon

for some serine proteases Recently, pro-enzymatic

fragments have been identified as intramolecular

chap-erons for some serine proteases, such as

carboxypepti-dase Y in yeast, subtilisin, a-lytic protease, and

aqualysin in bacteria (reviewed in [26]) When a

histi-dine tag was fused with the prosemin catalytic domain

using the pTrcHis B vector (Invitrogen), no enzymatic

activity was detected because refolding was disrupted

(data not shown) We have reported the enzymatic

activities of chimeric proteins composed of the

trypsi-nogen signal⁄ pro-peptide and the catalytic domains

of the kallikrein-like proteases, neurosin (KLK6) and

hippostasin (KLK11) [9,10] It is also advantageous that the artificial enzyme fragment blocks any pro-teolytic activity acting on the recombinant protein Recombinant prosemin only exhibited its activity when immobilized enterokinase digested away the pro-enzyme fragment (Fig 3B–D) Thus, we can recover the recombinant enzyme with no fear of autodegrada-tion Our system might be suitable for producing recombinant serine proteases

The enzymatic characteristics of prosemin are inter-esting, although its physiological functions are still obscure Prosemin hydrolyzes fluorogenic substrates at

an optimum pH of 8.0–9.0 (Fig 3F), which is similar

to the optimum pH for prostasin activity, pH 9.0 [27] Prostasin preferentially cleaved Pro-Phe-Arg-MCA and Val-Leu-Arg-MCA, but showed slight enzyme activity against Phe-Phe-Arg-MCA Prosemin acted enzymati-cally on Pro-Phe-Arg-MCA and

Boc-Val-Leu-Arg-Fig 6 Immunohistochemistry of prosemin in ovarian cancers Paraffin-embedded sections were stained with antibody directed against recombinant prosemin (A) Mucinous adenocarcinoma (B) and (C) Serous adenocarcinomas, borderline Immunoreactive prosemin was observed in the apical portion (D) Normal rabbit IgG produced no signals in the same sample as shown in (C).

MCA, although it preferentially cleaved Boc-Glu(Obzl)

and Gln-Ala-Arg-MCA as well as

Z-His-Glu-Lys-MCA unlike prostasin (Fig 3E) Such enzymatic

characteristics may reflect the amino acid similarities

between prosemin and prostasin The optimum pH for

prosemin is also similar to that for pancreatic digestive

serine proteases, including CTRL-1 [28], trypsin, and

pancreatic elastase [29] Northern blot analysis showed

that, of the normal tissues, prosemin is abundantly

expressed in the pancreas (Fig 4) All the pancreatic

cancer cell lines tested secreted prosemin into the

con-ditioned medium (Fig 5) These results suggest that

the prosemin produced in the pancreas might function

as a digestive enzyme, although confirmation that

prosemin is secreted into the pancreatic fluid is

required In other tissues, prosemin may not be

involved in the reconstitution of the extracellular

mat-rix, because recombinant prosemin does not digest

fibronectin, laminin, type-V collagen, or gelatin,

despite its enzyme activity (data not shown) The levels

of prosemin expressed in normal tissues are too low

for the reconstitution of the extracellular matrix Nerve

growth factor and brain-derived neurotrophic factor

have recently been shown to undergo processing by

serine proteases in extracellular regions [30] Prosemin

might be involved in the processing of some proteins

In contrast to the specific expression of prosemin in

normal tissues, various kinds of tumor cell lines secrete

prosemin (Fig 5) The immunoreactive 20 kDa protein

might be a degradation product of 33 kDa prosemin

formed during the preparation process, because

RT-PCR detected prosemin mRNA in cells that

expressed only the 20 kDa protein (Fig 5B and

Table 1) It should be noted that prosemin protein was

detected in many ovarian, lung, and colon cancer cell

lines, whereas the corresponding normal tissues

expressed low levels of prosemin mRNA There are

some reports that tumorigenesis causes the aberrant

expression of serine proteases Neurosin⁄ KLK6,

expressed predominantly in the brain, is up-regulated

in ovarian carcinomas [10,15] Neuropsin⁄ KLK8

mRNA is not detected in the normal ovary, but is

overexpressed in ovarian carcinomas [13]

Further-more, hippostasin⁄ KLK11 is up-regulated in ovarian

and prostate cancers [17] Recent experimental data

suggest that human kallikreins promote or inhibit

tumor growth, angiogenesis, invasion, metastasis by

proteolytic processing of growth factor binding

pro-teins, activation of other proteases and growth factors,

degradation of extracellular matrix proteins (reviewed

in [18–20]) Prosemin may also have other roles

inclu-ding the processing of growth factors and other

pro-teases, as described above, especially in the CNS

We have focused on ovarian cancer because other members of the chromosome 16 serine protease family, testisin and prostasin, are also expressed in ovarian car-cinomas [31,32] Immunohistochemical analysis demon-strated the expression of prosemin in clinical ovarian carcinomas, regardless of cancer type (Fig 6) The local-ization of prosemin in the apical portion suggests that prosemin is secreted into the lumen Western blot analy-sis that detected prosemin in cystic fluids confirms this possibility (Fig 5C) It should be noted that prosemin occurs significantly in borderline serous adenocarcino-mas, because there is still no available marker for the early diagnosis of ovarian cancer (Fig 6)

Prosemin might be a good candidate tumor marker, although it remains to be confirmed that prosemin protein secreted from carcinomas flows into the blood

We are developing an ELISA to measure prosemin, and seeking physiological substrates to better under-stand the physiological functions of the protein and its potential as a tumor marker

Experimental procedures

cDNA cloning of human prosemin cDNA cloning of human prosemin was performed by PCR techniques, as described previously [9] PolyA+RNA from human brain (Clontech Laboratories Inc., Palo Alto, CA) was reverse-transcribed using an oligo-dT primer attached

to a NotI adaptor sequence, GGCCACGCGTCGACTAG TAC(T)17, using Superscript II reverse transcriptase (Invi-trogen, Carlsbad, CA) Degenerate RT-PCR was performed using a primer designed to hybridize with the serine prote-ase motif (Supplementary Fig S1) The sequences of PCR products were analyzed using an automatic sequencer (DSQ-1000, Shimadzu, Kyoto, Japan) after cloning into pGEM-T Easy vector (Promega, Madison, WI) Primers for 5¢ and 3¢ RACE were designed based on the sequence

of the product of degenerate PCR For 3¢ RACE, nested PCR was carried out with primer 2 and the adaptor primer, using the PCR product amplified by primer 1 and the adap-tor primer as template 5¢ RACE was performed with primers 3, 4, AP1, and AP2 using a Marathon RACE Amplification Kit (Clontech) according to the manufac-turer’s instructions Full-length cDNA was isolated by nes-ted RT-PCR between primers 5 (GCCATGGTGGTTTC TGGAGC) and 6 (CTGAATTCCTAGGAGCGCGCGGC GGCC) using the PCR product generated by primers 7 (TACACACCCTGACCCGCATC) and 6 as template

Sequence analysis The sequence of the cDNA was analyzed using genetyx software (Software Development Co Ltd, Tokyo, Japan)