Screening a combinatorial peptide library to develop a human

Screening a combinatorial peptide library to developa human glandular kallikrein 2–activated prodrugas targeted therapy for prostate cancer

Samuel Janssen,1 Carsten M. Jakobsen,2

D. Marc Rosen,1 Rebecca M. Ricklis,1

Ulrich Reineke,3 Soeren B. Christensen,2

Hans Lilja,4 and Samuel R. Denmeade1

1Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins,Baltimore, Maryland; 2Danish University of PharmaceuticalSciences, Copenhagen, Denmark; 3Jerini AG, Berlin,Germany; and 4Memorial Sloan Kettering Cancer Center,New York, New York

AbstractObjective: Prostate cancer cells secrete the uniqueprotease human glandular kallikrein 2 (hK2) that repre-sents a target for proteolytic activation of cytotoxicprodrugs. The objective of this study was to identifyhK2-selective peptide substrates that could be coupled toa cytotoxic analogue of thapsigargin, a potent inhibitor ofthe sarcoplasmic/endoplasmic reticulum calcium ATPasepump that induces cell proliferation–independent apopto-sis through dysregulation of intracellular calcium levels.Methods: To identify peptide sequence requirements forhK2, a combination of membrane-bound peptides (SPOTanalysis) and combinatorial chemistry using fluorescence-quenched peptide substrates was used. Peptide substrateswere then coupled to 8-O-(12[L-leucinoylamino]dodeca-noyl)-8-O-debutanoylthapsigargin (L12ADT), a potentanalogue of thapsigargin, to produce a prodrug that wasthen characterized for hK2 hydrolysis, plasma stability,and in vitro cytotoxicity. Results: Both techniquesindicated that a peptide with two arginines NH2-terminalof the scissile bond produced the highest rates ofhydrolysis. A lead peptide substrate with the sequenceGly-Lys-Ala-Phe-Arg-Arg (GKAFRR) was hydrolyzedby hK2 with a Km of 26.5 Mmol/L, kcat of 1.09 s�1, anda kcat/Km ratio of 41,132 s�1 mol/L�1. The GKAFRR-L12ADT prodrug was rapidly hydrolyzed by hK2 and was

stable in plasma, whereas the GKAFRR-L peptide sub-strate was unstable in human plasma. The hK2-activatedthapsigargin prodrug was not activated by cathepsin B,cathepsin D, and urokinase but was an excellent substratefor plasmin. The GKAFRR-L12ADT was selectively cyto-toxic in vitro to cancer cells in the presence ofenzymatically active hK2. Conclusion: The hK2-activatedthapsigargin prodrug represents potential novel targetedtherapy for prostate cancer. [Mol Cancer Ther 2004;3(11):1439–50]

IntroductionNon-organ-confined prostate cancer is uniformly fatal onceit reaches the hormone-refractory state because currenttherapies are unable to completely eliminate the androgen-independent prostate cancer cells present within metastaticsites. In previous studies, our laboratory and others haveshown that metastatic androgen-independent prostatecancer cells have a remarkable low rate of cell proliferation(1, 2). This low proliferative rate could explain the relativeunresponsiveness of these cells to standard antiprolifer-ative chemotherapy. In vitro , however, highly proliferativeandrogen-independent prostate cancer cell lines remainexquisitely sensitive to induction of apoptosis.

In contrast to agents that activate apoptosis in prolifer-ating cells, our laboratory has shown that thapsigargin, apotent inhibitor of the sarcoplasmic/endoplasmic reticu-lum calcium ATPase pumps (3–5), has the dose-responseability to elevate intracellular calcium to sufficient levels toinduce apoptosis in all of the rodent and human androgen-independent prostate cancer cell lines without requiringthe cells to be proliferating (3, 6, 7). The cytotoxicity ofthapsigargin, however, is not prostate cancer cell typespecific (8, 9). Therefore, thapsigargin would be difficultto administer systemically without significant side effects.In addition, thapsigargin is sparingly water soluble due toits high lipophilicity. Therefore, a method is required thatboth better solubilizes thapsigargin and selectively targetsthe cytotoxicity of thapsigargin to metastatic deposits ofandrogen-independent prostate cancer cells systemically(10). To accomplish this, a primary amine-containingthapsigargin analogue can be coupled to a peptide carrierto produce water-soluble inactive prodrug that is selec-tively activated within sites of prostate cancer (8, 10).The peptide carrier in this approach is designed to be aselective substrate for prostate tissue–specific proteasessuch as prostate-specific antigen (PSA) or human glan-dular kallikrein 2 (hK2; ref. 10). PSA and hK2 are onlyproduced in high levels by normal and malignant pros-tate cancer cells (11–14). In addition, metastatic prostate

Received 6/9/04; revised 8/12/04; accepted 8/27/04.

Grant support: NIH Prostate SPORE P50 CA58236 (S.R. Denmeade),Aegon Scholarship in Oncology (S. Janssen), and Danish Cancer Society(C.M. Jakobsen).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely toindicate this fact.

Requests for reprints: Samuel Denmeade, Sidney Kimmel ComprehensiveCancer Center at Johns Hopkins, Bunting Blaustein Cancer ResearchBuilding, 1650 Orleans Street, Baltimore, MD 21231. Phone: 410-502-3941; Fax: 410-614-8397. E-mail: [email protected]

Copyright C 2004 American Association for Cancer Research.

Molecular Cancer Therapeutics 1439

Mol Cancer Ther 2004;3(11). November 2004

on April 4, 2019. © 2004 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

http://mct.aacrjournals.org/

cancer cells continue to secrete enzymatically active PSAand hK2 into the extracellular fluid at high levels (11, 12,15). Once in the extracellular fluid, enzymatically activePSA and hK2 eventually enter the blood where they areinactivated by binding to major serum protease inhibitors[i.e., a1-antichymotrypsin and a2-macroglobulin for PSA(14, 16–19) and a1-antichymotrypsin, a2-antiplasmin, anti-thrombin II, protein C inhibitor, and a2-macrogloblin forhK2 (20, 21)].

Ideally, the thapsigargin analogue prodrug is inactiveuntil the thapsigargin analogue is cleaved in the presenceof enzymatically active PSA or hK2. In previous studies,we synthesized and characterized a series of primaryamine-containing thapsigargin analogues and identified8-O-(12[L-leucinoylamino]dodecanoyl)-8-O-debutanoylth-apsigargin (L12ADT) as a highly potent inhibitor of thesarcoplasmic/endoplasmic reticulum calcium ATPasepump and as equally cytotoxic as thapsigargin (22, 23). Inadditional studies, we identified a six–amino acid peptidesubstrate that is efficiently hydrolyzed by PSA (24). Thepotent thapsigargin analogue and other cytotoxic agentshave been coupled to this peptide to produce prodrugs thatare selectively cytotoxic to PSA-producing prostate cancercells in vitro (8, 10, 24, 25). Significant antitumor effectshave been observed when these PSA-activated prodrugshave been given to animals bearing PSA-secreting prostatecancer xenografts without producing significant hosttoxicity (8, 25).

Enzymatically active PSA is found in high levels in theseminal fluid (0.3–5 mg/mL; refs. 11, 26) and in theextracellular fluid of both normal and malignant prostatecancer cells (i.e., 50–500 Ag/mL; ref. 15). In contrast, levelsof hK2 in the seminal fluid are f1% of those of PSA (26),whereas hK2 levels in the extracellular fluid of prostatecancers have not been reported. Using a chromogenicsubstrate (i.e., Pro-Phe-Arg-pNa), however, Mikolajczyket al. (27) showed that the enzymatic activity of hK2 was20,000-fold higher than that of PSA on a comparablesubstrate containing a Tyr cleavage site. In addition,although PSA and hK2 are both found almost exclusivelyin the prostate, hK2 is more highly expressed by prostatecancer cells than by normal prostate epithelium. UnlikePSA, hK2 expression seems to increase in more poorlydifferentiated cancers, with the strongest staining observedin prostate cancer lymph node metastases (14). Intensity ofstaining for hK2 has been found to increase with increasingGleason grade (14). In contrast, PSA staining tends todecrease with increasing Gleason grade (14). Thus,although hK2 is produced at a lower level than PSA inprostate tissue, the increased production in more poorlydifferentiated cancers coupled with the several orders ofmagnitude higher enzymatic activity suggest that total hK2enzymatic activity in the extracellular fluid may be similaror even greater than that of PSA. Therefore, hK2 representsan attractive alternative candidate for prostate-targetedprodrug activation therapy.

Although the hK2 protein is f80% identical to PSA inprimary structure (13), the two are markedly different in

their enzymatic properties (11). Whereas PSA is the onlymember of the kallikrein family with chymotrypsin-likesubstrate specificity, hK2 displays the trypsin-like specific-ity common to the kallikrein family (13). The goal of thepresent study was to identify specific hK2 peptidesubstrates that could be used to produce prodrugs thatare selectively activated by enzymatically active hK2present in the extracellular fluid of prostate cancer sites.In the present study, we have generated random combina-torial peptide libraries to rapidly screen a large number ofsequences to identify putative hK2 substrates. A leadpeptide substrate that is efficiently hydrolyzed by hK2 wasidentified and used to produce a hK2-activated L12ADTprodrug.

Materials andMethodsMaterialsA mutant form of hK2 was used for these studies in

which the amino acids �5 to �1 of the propeptide sequenceof hK2 (i.e., Leu-Ile-Gln-Ser-Arg) were mutated to Asp-Asp-Asp-Asp-Lys to generate a pro-hK2 protein that can beactivated to functional hK2 by factor Xa (28). Comparedwith wild-type hK2, expression of the propeptide hK2mutant increases the expression levels up to 15- to 40-fold(28). The generation and characteristics of this mutanthK2 have been described previously (28). Fmoc aminoacids were purchased either from Advanced Chemtech(Louisville, KY) or Novabiochem (La Jolla, CA). Reagentswere used without further purification. All other reagentswere from Sigma Chemical Co. (St. Louis, MO) unlessotherwise specified in the text.

Cell LinesThe LNCaP human prostate cancer cell line was

obtained from American Type Culture Collection (Rock-ville, MD). The C4-2B and CWR22R human prostate cancercell lines and TSU human bladder cancer cells wereobtained from Dr. John Isaacs (Johns Hopkins University,Baltimore, MD). These cell lines were maintained by serialpassage in RPMI 1640-10% fetal bovine serum in 5%CO2/95% air at 37jC.

Immobilized Peptide Synthesis and Hydrolysis Deter-mination

Peptides (‘‘Protease Spots’’) were provided by JeriniAG (Berlin, Germany) and were synthesized on contin-uous cellulose assays using the SPOT synthesis tech-nique (29). Each peptide contained a 2-aminobenzoic acid(Abz) moiety at the NH2 terminus. Abz is a fluorescentmolecule with optimal excitation at 325 nm and emissionmaxima at 420 nm. Peptides were synthesized on cellulosemembranes then punched out as small discs into 96-wellmicrotiter plates. Peptides (f8 nmol) are synthesized perspot.

To perform protease assay, spots were rinsed first for 5minutes with methanol to solubilize peptides. Spots werethen rinsed four times for 10 minutes under gentle agitationwith a buffer consisting of 50 mmol/L Tris and 0.1 mol/L

hK2-Activated Prodrug as Therapy for Prostate Cancer1440




NaCl (pH 7.8; buffer A). Fresh buffer A was added to eachwell along with an aliquot of purified protease (i.e., mutanthK2 or trypsin) or 50% human serum in buffer A. The platewas sealed with plastic and reaction was allowed to occurat room temperature without agitation. At describedintervals (i.e., 1, 2, 4, 7, and 24 hours), an aliquot (50 AL)of the reaction mixture was transferred to a new 96-wellmicrotiter plate. Fluorescence was then measured at roomtemperature using a 96-well fluorometric plate reader(Fluoroscan II, ICN Biomedicals, Costa Mesa, CA) withexcitation of 355 nm and emission of 408 nm. Fluorescenceat each point was plotted and reaction rates weredetermined from slope of the best-fit line. Rates areexpressed in relative fluorescence units per hour permilligram of protease.

Combinatorial LibrariesCombinatorial peptides libraries were synthesized as

described previously (30). Peptides were anchored topolyethylene glycol A (PEGA) support resin (Polymerlabs,Amherst, MA, 400 Am, 0.2 mmol/g) without a cleavablelinker.

Amino acid couplings were done according to estab-lished Fmoc/t-butyl protocols using 1-hydroxybenzotria-zole/N ,N V-diisopropylcarbodiimide activation (31) andperforming standard double couplings. Generally, comple-tion of acylation reactions was verified by both ninhydrin(32) and fluorescamine testing (33). The Fmoc protectinggroup was removed with 25% piperidine in dimethylfor-mamide (DMF). N-a-Fmoc-N-h-t-Boc-L-diaminopropionicacid (Novabiochem) was used for the introduction ofdiaminopropionic acid (Dap). Three randomized positionswere introduced using a Labmate Parallel Organic Synthe-sizer (4 � 6 vessels, Advanced Chemtech) according to the‘‘split-and-mix’’ procedure (34). All natural amino acids,except for Cys, were used with the following side chainprotection: Trt (Asn, Gln, and His), tBu (Tyr), OtBu (Asp,Glu, Ser, and Thr), Boc (Lys and Trp), and Pmc (Arg).Amino acid stock solutions (0.5 mol/L with 0.5 mol/L 1-hydroxybenzotriazole) were mixed with N,NV-diisopropyl-carbodiimide for 20 minutes (4 eq. of each). The activatedamino acids were added to the resin and 5% diisopropy-lethylamine (0.15 mL) in DMF was added. After 2 to 3hours, the resin aliquots were washed (3� 1-methyl-2-pyrrolidone, 3� methanol, and 3� DMF) and couplingswere repeated with 2 eq. amino acid for 1 to 2 hours. Aresin sample of each aliquot was subjected to a ninhydrintest and a fluorescamine test, which showed completion ofthe acylation reactions in all cases. Next, the resin aliquotswere pooled [Fmoc-X1-X2-X3-Dap-Phe-K(Abz)-PEGA] anddeprotected with piperidine and the remaining fourconstant residues, Ala, Lys, Gly, and nitrotyrosine (YV,Fluka, Milwaukee, WI) were added as Fmoc amino acids inbatch with 1-hydroxybenzotriazole/N ,NV-diisopropylcar-bodiimide activation as described above. For the finaldeprotection of the side chains, the resin was suspended inreagent K [trifluoroacetic acid (TFA)/thioanisole/water/phenol/1,2-ethanedithiol 82.5:5:5:5:2.5 (v/v), 1� 10minutes, 1� 3 hours]. The resin was washed with 95%

acetic acid (3�), dichloromethane (3�), DMF (3�), 5%diisopropylethylamine in DMF (3�), and DMF (6�). Theresin was stored until screening suspended in DMF at�20jC.

For screening, resin (f1 mL, f65,000 beads) was firstsuspended in methanol in a Petri dish and examined undertransilluminant UV light (302 nm) to detect any false-positive fluorescent beads prior to addition of protease.After removal of 40 to 50 beads, the resin was washedwith water and finally suspended in buffer (10 mL) ina glass Petri dish. After a final screen for false-positives,hK2 was added from a frozen stock solution to make a4 Ag/mL final concentration. Fluorescent beads wereselected and removed with a micropipette; washed withNaCl (1 mol/L), water, DMF, methanol, and water; andstored in methanol at �20jC.

Peptide SequencingPeptide sequencing was completed at the University

of Arizona Laboratory for Protein Sequencing and Analy-ses (Tucson, AZ) using an Applied Biosystems (Foster City,CA) 477A Protein/Peptide Sequencer (Edman chemistry)interfaced with a 120A high-performance liquid chroma-tography (HPLC; C18 phenylthiohydantoin column, re-verse-phase chromatography) analyzer to determinephenylthiohydantoin amino acids.

Automated Synthesis of Fluorescence-QuenchedPeptides

For validation of the Edman results, peptides wereresynthesized using a Rainin PS3 peptide synthesizerwith O-(benzotriazol-1-yl)-N ,N ,N V,NV-tetramethyluroniumhexafluorophosphate/4-methylmorpholine activation. Pep-tides were synthesized on PEGA resin for on-bead analysisor on Fmoc-Lys(4-methyltrityl)-Wang resin for solutionassays. The Fmoc-Lys(4-methyltrityl)-Wang resin wasfirst deprotected with 2% TFA in DCM (3� 2 minutes).The q-amine of Lys was then acylated with Boc-Abz.Deprotection/cleavage was done in TFA/triisopropylsi-lane/water [95:2.5:2.5 (v/v)] for 2 to 3 hours. Peptideswere ether precipitated, dried, and purified by C18-HPLCusing a linear gradient of acetonitrile (0.1% TFA), lyoph-ilized, and dissolved in DMSO. Peptide identities wereconfirmed by analysis on a PerSeptive Voyager DE-STRmatrix-assisted laser desorption ionization-time of flight(MALDI-TOF) using dihydroxybenzoic acid as a matrix.Fluorescence measurements were done on a Fluoroscan II96-well plate reader (ICN Biomedicals, excitation 355 nm,emission 460 nm). Kinetic variables were calculated asdescribed earlier (24).

Peptide-Prodrug SynthesisThe peptide sequence Gly-Lys-Ala-Phe-Arg-Arg

(GKAFRR)-L was synthesized on a Rainin PS3 automatedpeptide synthesizer on Fmoc-Leu-Wang resin (100-Amolscale). The same protecting groups were used as duringthe combinatorial synthesis, except for the Lys, whichwas orthogonally protected with the 1-(4,4-dimethyl-2,6-dioxocyclohex-1-ylidene)-3-methylbutyl (ivDde) group(Novabiochem). After deprotection of the NH2-terminalGly, the amine was acetylated with acetic anhydride and





4-methylmorpholine. Deprotection of the acid-labile pro-tecting groups and purification were done as outlinedabove. Boc-12ADT was synthesized as described previ-ously (23). TFA treatment followed by semiprep HPLCand lyophilization afforded the amine-containing 12ADT.The protected peptide (Ac-GK(ivDde)AFRRL) was coupledto 12ADT after 1-hydroxybenzotriazole/N,NV-diisopropyl-carbodiimide activation. After completion of the reaction,the ivDde group was removed by adding hydrazine tothe reaction mixture (2% final concentration, 30 minutes).Semipreparative HPLC yielded Ac-GKAFRR-L12ADT typ-ically in 60% to 70% yield. Product was confirmed byMALDI-TOF analysis.

Plasma StabilityAssaysMouse plasma was obtained from cardiac puncture of

anesthetized mice prior to euthanization by CO2 overdoseaccording to protocols approved by the Johns HopkinsAnimal Care and Use Committee. Human plasma wasobtained from discarded, pooled, and unlabeled clinicalsamples. Plasma was diluted to 50% in Tris buffer[50 mmol/L Tris-HCl, 100 mmol/L NaCl (pH 7.8)]. Tothis, test peptides/prodrugs were added to 250 or 500Amol/L final concentration. After the incubation period,1 volume of 1% TFA in acetonitrile was added toprecipitate the protein fraction and the tube was centri-fuged at 14,000 rpm for 2 minutes. The supernatant wasanalyzed by C18-HPLC and the collected peaks wereanalyzed by MALDI-TOF as described above.

CytotoxicityAssaysClonal survival of TSU (5 � 104 cells) following 48-hour

exposure to varying concentrations of hK2-activatedprodrugs or vehicle control with or without exogenouslyadded hK2 was done as described previously (9). Percent-age inhibition of clonal survival was calculated from theratio of number of colonies observed in treated group tonumber of colonies in control group. Cytotoxic responseto 7-day exposure to varying concentrations of hK2-activated prodrug in hK2-producing and nonproducingcell lines was determined using the Promega Cell Titer96 Nonradioactive Cell Proliferation Assays (Promega,Madison, WI) according to the manufacturer’s instructionsas described previously (8).

ResultsIdentification of a hK2-Specific Peptide SubstrateThe critical requirement for a hK2-activated prodrug is

identification of a peptide substrate that is efficiently andselectively hydrolyzed by the enzyme. Therefore, the firsttask was to define the peptide substrate requirements ofhK2. Previously, Lovgren et al. (20) identified several hK2cleavage sites within the human seminal proteins semeno-gelin I (SgI) and semenogelin II (SgII). Based on this hK2cleavage map for SgI and SgII, small peptides weresynthesized containing the amino acid sequence on theNH2-terminal side of the scissile bond (i.e., P6-P1, whereP1 is the amino acid at the cleavage site). These peptides

were synthesized using a technique called the SPOTtechnique (for a recent review, see ref. 29) in which theCOOH terminus of the peptide is tethered to a cellulosemembrane and the amine of the NH2 terminus of thepeptide is labeled with the fluorophore Abz. hK2 digestionreleases the Abz-containing portion of the peptide and thedegree of hydrolysis can thus be monitored by measuringthe increase in fluorescence of the supernatant over time.

Using this SPOT technique, the peptides representinghK2 cleavage sites within SgI and SgII were synthesizedand screened for hK2 hydrolysis. Within the group ofsemenogelin cleavage sites defined by Lovgren et al. (20),the amino acid Leu was present in the PV1 position in 5 ofthe 10 sites. Therefore, for this assay, the PV1 position(COOH-terminal of scissile bond) was fixed as Leu. Glywas introduced at the NH2 terminus of all peptidesto provide a linker between the Abz and the P6 positionof the semenogelin cleavage site peptide sequence. Therelative hydrolysis rates of these SgI and SgII native hK2cleavage site peptides were then obtained by incubatingenzymatically active hK2 with the cellulose membranepeptides and measuring fluorescent activity released intothe supernatant following hK2 hydrolysis of the peptide(Table 1). The semenogelin sequences showed a rangeof hydrolysis rates and not all semenogelin sequencesshowed significant activity. As noted previously, many ofthe hK2 cleavage sites within semenogelin protein containtandem basic amino acid residues at position P1 and P2,with the P1 position being Arg in all but one sequence andthe P2 being either Arg, His, or Lys in 5 of 11 sites (20).In our analysis, the most active native SgI/SgII sequence(GGKAHRL) contains besides the Arg at P1 two more basicresidues (Lys and His), suggesting a preference for positivecharge by hK2 (see Table 1).

The occurrence of Arg in the P1 position is a commonphenomenon in the natural hK2 substrates (11). Therefore,in an attempt to produce better hK2 substrates, the nativesequences (Table 1, bold) were modified by substitutingthe amino acid in the P2 position for Arg, therebyproducing a R-R cleavage site. In each case, substitutionof R-R for the native P2-P1 sequence resulted in peptidesthat were for the most part better (i.e., 1.2- to 74-fold) hK2substrates. Next, based on hK2 hydrolysis rates, the twobest substrates (i.e., GHEQKRRL and GGGKARRL) weresubjected to further sequence modifications that includedsubstitution of other amino acids in the PV1 position,substitution with D-amino acids, and substitution of Alain each position (i.e., Ala screen; data not shown). Fromthese experiments, it could be concluded that at thePV1 position Leu is preferred over Asp and Ser (data notshown). Substitution of L-Arg over D-Arg in P1 and/orP2 markedly decreased activity (i.e., dR-R = 4.4-fold, R-dR= 43-fold, and dR-dR = 130-fold decreased activity),whereas substitution of D-Leu in the PV1 position onlydecreased activity f2-fold (data not shown).

To determine if there was a strict requirement for Arg inthe P1 position of the SgI/SgII–based peptide sequences,additional peptides were synthesized with His substituted





for Arg in the P1 position (Table 1). In all cases, these P1His-containing substrates were markedly poorer substratesfor hK2 hydrolysis (Table 1). In additional studies, solublefluorescence-quenched peptide substrates based on theSgI/SgII sequence GSKGHFRL were produced in whichthe P1 Arg was substituted with the positively chargedamino acid Lys (i.e., GSKGHFKL and GSKGPFKL). TheArg-free sequence GSKGHFHL identified as native sub-strate from the SgI/SgII hK2 cleavage map was alsosynthesized for testing in solution. In these studies, noneof these three Arg-free peptides were appreciably digestedby hK2 even after prolonged incubation. These resultsfurther support results from earlier studies using smallpeptide substrates and phage display and show thathK2 has a strong preference for Arg in the P1 position ofpeptide substrates. Overall, from these studies, it appearedthat hK2 prefers polar, positively charged substrates witha monobasic or dibasic RR motif at the cleavage site.

Finally, to determine whether these SgI/SgII–basedpeptide substrates were selective for hK2 hydrolysis, eachsequence was incubated with equimolar amount of trypsin.A comparison of hydrolysis rates for each individualpeptide for hK2 versus trypsin showed that all of thesepeptides were better substrates for trypsin (Table 1). hK2hydrolysis rates or this series of peptides ranged from 1%to 78% of trypsin hydrolysis rates.

Based on these studies, the GKAFR peptide wasselected for further analysis based on high relative hK2versus trypsin hydrolysis rates. A fluorescent substratewas synthesized by coupling the fluorophore 7-amino-4-methyl coumarin (AMC) to the COOH terminus of thepeptides to produce the substrate Mu-GKAFR-AMC(where Mu is morpholinocarbonyl). Rates of hydrolysisby hK2 and trypsin were determined and compared withthe kallikrein substrate PFR-AMC. In addition, stability tononspecific hydrolysis in human plasma was also assayed(Table 2). These results show that the GKAFR-AMCsubstrate is a better substrate for hK2 than PFR-AMC;however, neither substrate was selective for hK2 hydrolysisnor were these substrates stable to hydrolysis in humanplasma (Table 2).

Screening a Combinatorial Peptide Library to IdentifyhK2 Substrates

Trypsin and trypsin-like proteases have a definedpreference for the amino acid Arg or Lys at the site ofhydrolysis. There are a large number of trypsin-likeproteases present in human blood including human kalli-krein 1, plasmin, thrombin, and other members of theclotting factor cascade. The proteolytic activity of theseproteases in the blood is tightly regulated and these pro-teases are present in the bloodstream predominantly asinactive zymogens. hK2 is also a trypsin-like protease with

Table 1. Relative hydrolysis rates of hK2 substrates by hK2 as determined by release of NH2-terminal conjugated Abz

P7 P6 P5 P4 P3 P2 P1 PV1 Type hK2 rate(fluorescence units/h/Ag)

Trypsin rate(fluorescence units/h/Ag)

RatiohK2/trypsin

Abz G G G K A H R L Native SgI/SgII 7.52 17.00 0.44Abz G G G K A R R L Dibasic RR 8.77 11.20 0.78Abz G G G K A H H L His substituted 0.13 18.60 0.01Abz G H E Q K G R L Native SgI/SgII 1.19 39.60 0.03Abz G H E Q K R R L Dibasic RR 10.94 22.20 0.49Abz G H E Q K G H L His substituted 0.01 7.20 0.00Abz G K D V S G R L Native SgI/SgII 0.45 12.40 0.04Abz G K D V S R R L Dibasic RR 4.25 11.90 0.36Abz G K D V S G H L His substituted 0.12 11.30 0.01Abz G P A H Q D R L Native SgI/SgII 0.14 13.20 0.01Abz G P A H Q R R L Dibasic RR 7.36 18.20 0.40Abz G P A H Q D H L His substituted 0.09 0.60 0.15Abz G S K G H F H L Native SgI/SgII 0.31 11.00 0.03Abz G S K G H R R L Dibasic RR 4.86 16.30 0.30Abz G S K G H F R L Arg substituted 5.55 11.40 0.49Abz G S N T E K R L Native SgI/SgII 0.47 22.60 0.02Abz G S N T E R R L Dibasic RR 1.09 18.10 0.06Abz G S N T E K H L His substituted 0.15 13.30 0.01Abz G S Q N Q V R L Native SgI/SgII 0.13 18.62 0.01Abz G S Q N Q R R L Dibasic RR 3.85 7.80 0.49Abz G S Q N Q V H L His substituted 0.06 0.50 0.12Abz G S Y E E R R L Native SgI/SgII 0.60 24.20 0.02Abz G S Y E E R H L His substituted 0.08 14.60 0.01Abz G S Y P S S R L Native SgI/SgII 2.67 10.90 0.24Abz G S Y P S R R L Dibasic RR 3.78 17.30 0.22Abz G S Y P S S H L His substituted 0.05 0.50 0.10Abz G P L I L S R L PSA-propeptide 0.60 6.00 0.10





a preference or perhaps a requirement for Arg at theP1 hydrolysis site (21). As our experience with the peptidesequences defined above indicated, peptide substratescontaining Arg seemed relatively unstable to nonspecifichydrolysis in the blood. Therefore, in an attempt to identifymore selective hK2 peptide substrates and to determine ifhK2 indeed has a strict requirement for Arg at site ofcleavage, a fluorescent-quenched combinatorial peptidelibrary was synthesized containing random amino acid inpositions P3-P1 to allow for screening of all possible aminoacid combinations (i.e., 193 or 6,859 sequences).

Previously, St Hilaire et al. (30) showed that proteasesubstrate requirements can be routinely mapped by on-bead (i.e., resin) digestion of short peptides. By followingthe ‘‘split-and-mix’’ approach (34), a peptide library isgenerated on polymeric solid-phase synthesis resin‘‘beads,’’ so that each bead contains at the end a uniquebut random peptide sequence. These peptides are bracket-ed by a fluorophore at the COOH terminus (Abz coupledto the q-amino group of Lys) and a pairing quencher moietyat the NH2 terminus (3-nitrotyrosine; Fig. 1). On hydrolysisof any backbone amide bond, the quencher-containingNH2-terminal part of the peptide is liberated and diffusesinto the solution, resulting in bright fluorescence dueto unquenching of the remaining COOH-terminal part, stilllinked to the bead (30). The polymeric support has to swellsufficiently in water to allow diffusion of the protease intothe bead. Our previous data using the SPOT-basedpeptides suggested that the positively charged tripeptideGKA would be a close-to-optimal P6-P4 amino acidsequence and that the COOH-terminal positions wereof more significance for defining selectivity of hK2 activity.Therefore, this library was biased in that between the NH2-terminal Lys-Abz fluorophore and the COOH-terminalnitrotyrosine quencher a constant tripeptide (GKA) wasinserted in positions P6-P4 followed by random aminoacids in positions P3-P1 (i.e., GKAXXX; Fig. 1). PEGA (amix of polyacrylamide and polyethylene glycol) resin waschosen as the solid-phase support based on preliminarystudies, demonstrating superiority of this resin overalternative resin supports (e.g., TentaGel) for this applica-tion (data not shown).

Additionally, Thorpe and Walle (35) published data froma small combinatorial library to find linkers for the optimaldisplay of peptide ligands to various protein targets. Theyreported that the insertion of a dipeptide consisting of acationic residue together with a hydrophobic residue

presents a general method for optimizing peptide displayon solid-phase beads. To test this observation in our ownsystem, we incorporated a linker described by Thorpe andWalle (35) in which the cationic residue was Dap and thehydrophobic residue was phenylalanine (F). We synthe-sized the test peptide YVGKAFRL-Dap-F-KV-PEGA andobserved that the time required to generate clearlydetectable fluorescence on the beads was reduced to 4 to5 hours compared with 10 to 12 hours for the same peptidesequence lacking the Dap-F linker. Based on these results,the Dap-F dipeptide linker was included in all subsequentlibraries.

Therefore, the final library used for screening with hK2contained the general sequence YVGKAXXX-Dap-F-KVPEGA, where X is any of 19 amino acids (Cys wasexcluded from library) and contained 193 peptide sequen-ces on f50,000 beads (i.e., f7 beads for each uniquepeptide sequence; Fig. 1). After carefully removing anyfalse-positive fluorescent beads from the library (f40–50beads), purified, enzymatically active hK2 was added at afinal concentration of 4 Ag/mL. After 1 hour, the firstpositive bead was removed. Over the subsequent 3 hours, 9more beads were selected. In total, 14 beads were selectedover a period of 24 hours. Positive beads were sequencedby Edman degradation.

Seven of 14 peptides contained one or more Arg residues.The peptides lacking any Arg did not show a specificamino acid preference. One sequence (FRR) was similar todouble Arg motif sequences identified previously asoptimal in the SPOT assay.

To confirm that the selected sequences represented truehK2 substrates and not false-positives, the majority of thefluorescence-quenched peptides were resynthesized,

Figure 1. Chemical structure of fluorescence-quenched combinatorial‘‘one bead, one peptide’’ library (YVGKAXXX-Dap-F-KV-PEGA, where YV isnitrotyrosine and KV is Abz-substituted Lys).

Table 2. Relative hydrolysis rates of hK2 substrates by hK2 and trypsin as determined by release of NH2-terminal AMC

Substrate hk2 (5 Ag/mL) Trypsin (5 Ag/mL) 50% Plasma

pmol AMC/min pmol AMC/min/nmol protease

pmol AMC/min pmol AMC/min/nmol protease

(pmol AMC/min)

Mu-GKAFR-AMC 28.5 1,584 1,356 85,300 28.3PFR-AMC 15.6 867 436 27,400 16.6





cleaved from the resin, and tested for hydrolysis by hK2 insolution. After resynthesis, none of the soluble non-Arg-containing peptides were hydrolyzed by hK2 (data notshown), confirming that the Arg-free sequences were nothK2 substrates but false-positives. In contrast, each of theresynthesized Arg-containing peptide substrates was read-ily hydrolyzed by hK2. The combinatorial screen identifiedseven Arg-containing peptides. Four of these were resyn-thesized (X1-X2-X3 = RAF, KPR, FRR, and MRQ, respec-tively). All four Arg-containing sequences that wereresynthesized reproduced fluorescence when these pep-tides were digested on-bead with hK2. For a morequantitative analysis, the fluorescence-quenched peptideswere cleaved off the resin and purified by HPLC. The rateof hydrolysis was quantified by measuring increase influorescence (Fig. 2). The best substrate has proven tobe the sequence with Arg at P1 and P2 (i.e., YVGKAFRR-Dap-F-KV). In <5 minutes, >50% of the peptide weredigested (500 Amol/L peptide, 4 Ag/mL hK2). For theother peptides, digestion of the same amount of peptidetook 19 to 29 minutes. Maximum digestion was only 70% to75%, a value that was reached with YVGKAFRR-Dap-F-KV in<15 minutes. In subsequent studies, hydrolysis rates usingthe YVGKAFRR-Dap-F-KV peptide were analyzed by Line-weaver-Burke reciprocal plots. The Michaelis-Mentenconstant (Km) was determined at 26.5 Amol/L, the kcat

was 1.09 s�1, and the kcat/Km ratio was 41,132 s�1 mol/L�1.These results compare favorable with those reportedpreviously for the Pro-Phe-Arg-AMC substrate used toassay hK2 activity (Km, 40 Amol/L; kcat, 0.92 s�1; kcat/Km,22,916 s�1 mol/L�1) and were superior to the GKAFR-AMC substrate we generated based on results of SPOTanalysis (Km, 146 Amol/L; kcat, 0.13 s�1; kcat/Km, 895 s�1

mol/L�1).

Stability of hK2 Peptides in PlasmaArg-containing peptides are potential substrates for the

wide variety of other trypsin-like proteases that are presentin the blood and may have residual activity in the blood.The plasma stability of a hK2 peptide substrate maytherefore be limited and this would have significant conse-quences related to the development of a hK2-activatedprodrug. Therefore, two fluorescence-quenched hK2 pep-tide substrates were incubated in 50% mouse or humanplasma (diluted in PBS buffer) to determine stability usinga plate reader. Similar to results with Arg-containing pep-tides from the SPOT analysis, none of these Arg-containing,fluorescence-quenched hK2 peptide substrates identifiedfrom the combinatorial screening were stable in human ormouse plasma (data not shown). Fluorescence-quenchedYVGKAFRR-Dap-GKV and YVGKAFRR-LGKV (500 Amol/Leach) were hydrolyzed when incubated in 50% mouse orhuman plasma (Fig. 3). Mouse plasma degraded the pep-tides faster than human plasma (Fig. 3). The Leu-containingpeptide was less stable than the Dap-containing peptide inboth plasma types. HPLC analysis of the peptides after3 hours of incubation confirmed that the fluorescencegenerated during this plate reader assay corresponded withproteolysis, and in each case, several degradation productswere generated; almost no parent peptide remained after3 hours in mouse plasma (f5%), whereas in humanplasma f25% of intact peptide remained after 3 hours.Overnight incubation in human plasma resulted in com-plete degradation of both peptides. To determine whetheracetylation of the NH2 terminus of the hK2 peptide wouldenhance stability, the acetylated fluorescence-quenchedpeptide, Ac-YVGKAFRR-LGKV, was synthesized and plasmastability was compared with that of the nonacetylated pep-tide (YVGKAFRR-LGKV). Hydrolysis of these two peptides

Figure 2. Hydrolysis of variouslead substrates (250 Amol/L) by hK2(8 mg/mL) in PBS buffer. Note thatthe double Arg substrate (YVGKAFRR-Dap-FKV) by far exceeded all othersubstrates. A substrate without Arg(YVGSKGHFKL-Dap-F-KV) did not showany hydrolysis. To determine 100%digestion, trypsin was added (50mg/mL) and samples were incubatedto 37jC for 30 minutes and the fluo-rescence was determined.





was completely identical as judged from the generationof fluorescence in the plate reader assay and by HPLCanalysis. These results indicate that nonspecific hydrolysisof the hK2 peptide substrate is not due to degradation byplasma aminopeptidases.

Synthesis and Characterization of a hK2-ActivatedThapsigargin Prodrug

The preceding results would suggest that the develop-ment of a hK2-activated peptide-based prodrug mightnot be feasible due to poor stability of Arg-containingpeptides in plasma. However, the possibility remained thatthe introduction of a bulky hydrophobic moiety likean analogue of thapsigargin might alter the relative rateof hK2 hydrolysis and/or stability in plasma. Therefore,we proceeded to synthesize a putative hK2-activatedprodrug by coupling the GKAFRR peptide sequence to aprimary amine-containing thapsigargin analogue (Fig. 4).

In previous studies, we had identified L12ADT as a potentamino acid–containing thapsigargin analogue that was ascytotoxic as thapsigargin with a LD50 value of f30 nmol/Lagainst human prostate cancer cells in vitro (23). Previ-ously, the L12ADT analogue has been coupled to a PSA-specific peptide to produce a prodrug that is selectivelycytotoxic to PSA-producing prostate cancer cells in vitroand in vivo (8).

Based on these results, the prodrug Ac-GKAFRR-L12ADT was synthesized (Fig. 4). This putative hK2prodrug was incubated with enzymatically active hK2(4 Ag/mL) to determine extent of hydrolysis over time.HPLC analysis of aliquots of the incubation mixtureindicated that the hK2 prodrug is rapidly cleaved by hK2(Fig. 5). MALDI-TOF analysis of the digestion productsindicated that cleavage occurred after each Arg residuegenerating both Arg-Leu-12ADT (RL12ADT) and L12ADT.In 25 minutes, 50% was hydrolyzed; after 1 hour, >80%of the starting prodrug were hydrolyzed (Fig. 5). The ratioof the products RL12ADT/L12ADT was 1:1.8 as deter-mined by HPLC integration. In previous studies, the IC50

of L12ADT against human TSU bladder cancer cell line wasf30 nmol/L. R-L12ADT was f5-fold less potent againstthis cell line in growth inhibition assays in vitro (data notshown).

To determine the plasma stability for the hK2 prodrug,we incubated the prodrug in 50% fresh heparinized(1%) plasma for 24 hours at room temperature, precipitatedthe proteins, and analyzed the supernatant by HPLC.Unexpectedly, HPLC analysis after 24-hour incubation inplasma yielded only a single peak corresponding to thehK2 prodrug. MALDI-TOF mass spectrometry of the iso-lated single peak confirmed the plasma stability of theprodrug.

Figure 4. Chemical structure of hK2 prodrug, AcGKAFRR-L12ADT.HK2 cleavage sites are indicated. The ratio of R-L12ADT/L12ADTgenerated by hK2 digestion was 1:1.8.

Figure 3. Hydrolysis of hK2 peptidesubstrates in plasma. Arg-containinglead hK2 substrates (500 Amol/L) wereincubated in 50% mouse or humanplasma. Generation of fluorescence indi-cated that the fluorescence-quenchedpeptides are unstable in plasma. Hydro-lysis of the substrates was confirmed byHPLC. Comparison of mouse and humanplasma for the same substrate suggestshigher proteolytic activity in mouseplasma compared with human plasma.





hK2 Levels in ProstaticTissues and Human ProstateCancer Cell Lines

The levels of hK2 production in benign and malignantprostate tissues have not been as well characterized asPSA levels. In a previous study, Lovgren et al. showedthat the average hK2 levels in the seminal plasma was 6.4Ag/mL (26). In this study, the level of PSA in the seminalplasma was 820 Ag/mL (26). We have isolated prostaticfluid taken directly from radical prostatectomy specimens(i.e., without contamination from seminal vesicle fluid) andmeasured PSA and hK2 levels. In this prostatic fluid, PSAlevels were 696 F 305 Ag/mL (n = 4), whereas hK2 levelswere 165-fold lower at 4.2 F 0.5 Ag/mL (n = 5). Haese et al.(36) evaluated preoperative PSA and hK2 levels in menwith stage T2 or T3 prostate cancers undergoing radicalprostatectomy. Men were selected with PSA values <10 ng/mL. In this study, the PSA level for men with T2 versus T3was not significantly different (f6 ng/mL). In contrast,hK2 levels in the serum of these patients were 50-fold (stageT3) to 75-fold (stage T2) lower (i.e., 0.08–0.12 ng/mL,respectively; ref. 36).

In contrast, hK2 levels in conditioned medium of humanprostate cancer cell lines are several orders of magnitudelower (i.e., 0.1–60 ng/mL after 4 days of conditioning). Forexample, in one experiment, LNCaP human prostate cancercells (6 � 106) produce hK2 levels in the conditionedmedium of f16.7 ng/mL, whereas PSA levels were 166ng/mL after 4 days of conditioning (data not shown). Thus,in vitro assessment of hK2-selective cytotoxicity of a hK2-activated prodrug is problematic because the availablecell models produce much lower levels (i.e., f100-fold) ofhK2 than estimated levels in extracellular fluid of humanprostate cancers.

Cytotoxicity of hK2-Activated Prodrug In vitroTo determine the in vitro efficacy and selectivity of the

hK2-activated prodrug, cytotoxicity against a series of hK2-

producing cell lines (i.e., CWR22R, LNCaP, and C4-2B) wascompared with cytotoxicity against TSU a non-hK2-producing human bladder cancer cell line (Fig. 6). Ina previous study, the levels of hK2 production by thesehuman prostate cancer cell lines were determined (37). Thelowest levels were found in the CWR22R line, whichproduced 1.1 F 0.2 ng hK2/106 cells/d (37). LNCaP cellsproduced 2.2 F 0.6 ng hK2/106 cells/d and C4-2Bcells produced 14.9 F 3.0 ng hK2/106 cell/d, the highestlevel of all the cell lines tested (37). In this experiment, thehK2 prodrug had a similar inhibitory effect on cell growthafter 7-day exposure at concentrations z1.25 Amol/L inall cell lines tested (Fig. 6). A modest difference in effectwas observed for all hK2-producing cell lines at lowerconcentrations of prodrug. C4-2B cells, the line thatproduces highest levels of hK2, seemed to be the mostsensitive to the prodrug (Fig. 6). The estimated IC50 for TSUin this study was f1.25 Amol/L, whereas the IC50 for thehighest hK2-producing line (C4-2B) was f0.3 Amol/L.

The modest f4-fold difference in activity between ahK2-producing and a nonproducing cell line in thisexperiment may be secondary to production of low levelsof hK2 by these human prostate cancer cell lines. Therefore,to determine the relative efficacy and specificity of a hK2-activated prodrug requires conditions that more closelymimic levels of enzymatically active hK2 found inextracellular fluid of human prostate cancers. In previousstudies, we showed that PSA levels in extracellular fluid ofhuman prostate cancers was 69 F 12 Ag/mL (15). However,hK2 levels in the extracellular fluid of human prostatecancers have not been reported. To determine approximatelevels of hK2 in relation to PSA, we measured hK2 and PSAlevels in the human PC-82 prostate cancer xenograft model.PC-82 does not grow as a cell line in vitro and is maintained

Figure 5. HK2-mediated hydrolysis of Ac-GKAFRR-L12ADT prodrug.Hydrolysis of the prodrug (125 Amol/L) was analyzed by HPLC andquantified by HLPC integration of appropriate peak areas. Percentagehydrolysis was calculated from the ratio of area of starting materialsubstrate at indicated time points/peak area of substrate at time 0 min.

Figure 6. Cytotoxicity of Ac-GKAFRR-L12ADT against hK2-producingCWR22R, LNCaP, and C4-2B human prostate cancer cell lines and hK2nonproducing TSU human bladder cancer cell line. Cell lines wereincubated in the presence of indicated concentrations of prodrug for 7days at which point cell proliferation assay (Promega) was done todetermine viable cell number. Columns, percentage growth inhibitioncompared with vehicle-treated controls from each cell line; bars, SE.Results of eight replicate wells from duplicate experiments.





through serial passage in nude mice. In previous studies,we determined that PSA levels in PC-82 xenografts wereonly f3-fold lower than those found in human prostatecancers (15). In contrast, in LNCaP xenografts, the levelsof PSA were f45-fold lower than those observed in humanprostate cancers (15). Therefore, the PC-82 represents amore relevant model for estimating levels of hK2 in humanprostate cancers. PC-82 xenografts were harvested andhomogenized; then, total PSA and hK2 levels in tumorlysates were determined by ELISA assay (Hybritech, SanDiego, CA). The concentration of PSA in these PC-82lysates was 5.15 F 0.35 Ag/g of tissue, whereas the levels ofhK2 were f12-fold lower at 0.41 F 0.06 Ag/g of tissue.

On this basis, we estimated that a level of hK2 of f1Ag/mL in the conditioned medium would roughly approx-imate levels of enzymatically active hK2 found in extracel-lular fluid of human prostate cancer xenografts. Therefore,in the next series of experiments, non-hK2-producinghuman TSU cells were treated in serum-containing medi-um in the absence or presence of purified, enzymaticallyactive hK2 (1 Ag/mL). Clonal survival assays were thendone after 5-day exposure to fhK2-activated prodrug(Fig. 7). In these experiments, there was f10-fold enhance-ment of efficacy (i.e., IC50 = 0.5 Amol/L in the presence ofhK2 versus f5 Amol/L in the absence of hK2) of the hK2-activated drug in the presence of enzymatically active hK2(1 Ag/mL) in the serum-containing tissue culture medium.

Hydrolysis of hK2 Peptide and Prodrug by OtherPotentialTumor-Associated Proteases

Due to their short length, the hK2 peptide substratescould potentially be substrates for other trypsin-likeproteases as well. Although a rather specific protease

substrate can be defined with seven amino acids, there isa lack of higher-order structural information by whichnatural protein substrates normally impose high specificity.Cancer progression is often correlated with increasedprotease activity (38). These activities could be potentiallybeneficial because they could broaden the scope ofapplications for protease-activated prodrugs. To test thehypothesis that other tumor-associated proteases couldactivate the Ac-GKAFRR-L12ADT prodrug, we selectedseveral known proteases (Table 3) implicated in cancerprogression to determine if our lead substrate, GKAFRR-L,could be efficiently hydrolyzed by any of these proteases.For this analysis, we selected trypsin-like Ser proteasesplasmin and urokinase, cathepsin B, and cathepsin D. Weanalyzed hydrolysis of both the fluorescence-quenchedpeptide substrate and the thapsigargin prodrug. Noappreciable hydrolysis of either substrate was observedfollowing incubation with cathepsin D or cathepsin B.Urokinase showed low activity on the fluorescence-quenched peptide substrate but not on the prodrug (Table3). Plasmin had a >10-fold slower rate of hydrolysis of thepeptide substrate than hK2. However, with the prodrug,plasmin had f6-fold higher hydrolysis rate than hK2.Analysis of the cleavage products showed that, withplasmin, proteolysis occurs between the two arginines,generating the less potent cytotoxin Arg-Leu-12ADT.Plasmin therefore could be a valid target for selectiveactivation of the GKAFRR-L12ADT prodrug in other typesof cancer where plasmin activation may play an importantrole. Further studies are under way in our laboratory tocharacterize in vitro activation of the GKAFRR-L12ADTprodrug using non–prostate cancer cell lines that displayincreased plasmin activity.

DiscussionThe peptide substrate requirements of hK2 have beencharacterized to identify peptides that are readily cleavedby this trypsin-like Ser protease that could be used toproduce prodrugs that are targeted to hK2-producing

Figure 7. Cytotoxicity of Ac-GKAFRR-L12ADT in the presence of hK2-containing medium. Human TSU bladder cancer cells were exposed for5 days to indicated concentrations of prodrug in either standardserum-containing medium or same medium plus enzymatically active hK2(1 Ag/mL). Clonal survival was then compared with survival of vehicle-treated controls.

Table 3. Relative hydrolysis rates normalized to hydrolysis rateof cathepsin B

Protease Relative hydrolysis rate

Abz-GKAFRR-LYV hK2 Prodrug

Cathepsin B 1 1Cathepsin D 1 1Plasmin 36 750hK2 353 125Urokinase 17 1

NOTE: Enzyme concentrations were 1 Ag/mL. Concentration Abz-GKAFRR-LYV was 500 Amol/L and concentration hK2 prodrug (Ac-GKAFRR-L12ADT) was 100 Amol/L. A relative hydrolysis rate of 1corresponds to f0.1% digestion in 1 hour.





prostate cancers. Previous to our studies, several groupshave evaluated the substrate requirements for hK2.Mikolajczyk et al. (21) used a few peptide substrates andconcluded that hK2 had P1 Arg-restricted specificity. Morerecently, Cloutier et al. (39) recently used phage displaytechnology to study the substrate specificity of hK2 usingmany peptide sequences and also reported that hK2 has astrict preference for Arg in the P1 position. Substitution ofthe P1 Arg with Lys or His normally abolished all hK2hydrolysis. Despite many attempts, including a combina-torial peptide library screen, we did not identify an Arg-free hK2 substrate, confirming the findings of theseprevious investigators. We further showed that hK2 prefersdibasic Arg-Arg at the P1-P2 positions, an observation thatwas initially reported by Lovgren et al. Additionally, weobserved that substitution of D-Arg in the P1 and/or P2position significantly decreases hK2 hydrolysis.

Previously, several studies have shown that inactivationof peptide hormones such as vasopressin as well as majorclotting factors occurs by cleavage after Arg by proteolyticactivity present in plasma (40, 41). Because these hK2-selective peptides will be incorporated into prodrugs thatwill need to be given systemically via the blood, our mainconcern was that Arg-containing peptide substrates fora trypsin-like protease such as hK2 would have limitedstability in plasma. This would severely limit the feasibilityof a hK2-targeted prodrug. Our initial results werediscouraging in that all of the Arg-containing hK2fluorescence-quenched peptide substrates were rapidlyhydrolyzed in non-hK2-containing human and mouseplasma. However, the L12ADT-containing hK2 prodrugwas completely stable in human and mouse plasmapossibly due to binding of this prodrug to serum proteins,making it inaccessible to active plasma proteases. Current-ly, studies are under way in our laboratory to elucidatethe mechanisms for this paradoxical plasma stability of theprodrug versus the unconjugated peptide substrates.

Several laboratories are currently developing protease-targeted prodrugs using similar types of combinatorialor phage-based screens to identify peptide substrates. Ourresults show that the peptide carrier may behave differ-ently when unconjugated to drug than the subsequentprodrug in terms of hydrolysis rates and stability inplasma. This difference most likely is due to certaincharacteristics (e.g., hydrophobicity, molecular weight,and degree of binding to serum proteins) of the drug beingtargeted. Therefore, the prescreening of lead peptidesunconjugated to drug for plasma stability may not beuseful for predicting plasma stability of the subsequentprodrugs.

In this study, there was only f4-fold difference incytotoxicity between the highest hK2-producing cell line(C4-2B) and a non-hK2-producing cell line (TSU). Thislimited differential cytotoxicity most likely is due to the lowlevel of hK2 production by the human prostate cancer celllines evaluated in this study (37). In this regard, when non-hK2-producing TSU cells were incubated in the presence ofhigher levels of hK2 that are more representative of levels

that may be present in extracellular fluid of human prostatecancers, the differential cytotoxicity increased to 10-fold.Additionally, Kumar et al. (42) showed that, in serum-containing conditioned medium from LNCaP cells, mostof the hK2 produced was present as enzymatically inactivepro-hK2, whereas, if these cells were grown in serum-freemedium supplemented with androgen, all of the hK2 waspresent as mature hK2. Finally, enzymatically activehK2 rapidly forms complexes with serum protease inhib-itors such as a2-macroglobulin, a1-antichymotrypsin, a2-antiplasmin, protein C inhibitor (20, 21). Thus, in vitro andin vivo assessment of hK2-selective cytotoxicity of the hK2-activated prodrug will be difficult because the availablecell models produce much lower levels (i.e., f100-fold) ofenzymatically active hK2 than estimated levels in extracel-lular fluid of human prostate cancers, which in serum-containing medium may be enzymatically inactive eitherthrough lack of proper processing (i.e., pro-hK2) or throughbinding to serum protease inhibitors. Therefore, we arecurrently measuring levels of hK2 in the extracellular fluidof human prostate cancers to generate human prostatecancer cell lines that produce similar levels of enzymati-cally active hK2 that can then be used to better assess thecytotoxicity of the hK2-activated prodrug both in vitro andin vivo .

References

1. Pinski J, Parikh A, Bova GS, Isaacs JT. Therapeutic implications ofenhanced G(0)/G(1) checkpoint control induced by coculture of prostatecancer cells with osteoblasts. Cancer Res 2001;61:6372–6.

2. Berges RR, Vukanovic J, Epstein JI, et al. Implication of cell kineticchanges during the progression of human prostatic cancer. Clin CancerRes 1995;1:473–80.

3. Rasmussen U, Brogger Christensen S, Sandberg F. Thapsigargine andthapsigargicine, two new histamine liberators from Thapsia garganica L.Acta Pharm Suec 1978;15:133–40.

4. Thastrup O, Cullen PJ, Drobak BK, Hanley MR, Dawson AP.Thapsigargin, a tumor promoter, discharges intracellular Ca2+ stores byspecific inhibition of the endoplasmic reticulum Ca2(+)-ATPase. Proc NatlAcad Sci U S A 1990;87:2466–70.

5. Davidson GA, Varhol RJ. Kinetics of thapsigargin-Ca(2+)-ATPase(sarcoplasmic reticulum) interaction reveals a two-step binding mechanismand picomolar inhibition. J Biol Chem 1995;270:11731–4.

6. Furuya Y, Lundmo P, Short AD, Gill DL, Isaacs JT. The role of calcium,pH, and cell proliferation in the programmed (apoptotic) death of androgen-independent prostatic cancer cells induced by thapsigargin. Cancer Res1994;54:6167–75.

7. Lin XS, Denmeade SR, Cisek L, Isaacs JT. Mechanism and role ofgrowth arrest in programmed (apoptotic) death of prostatic cancer cellsinduced by thapsigargin. Prostate 1997;33:201–7.

8. Denmeade SR, Jakobsen CM, Janssen S, et al. Prostate-specificantigen-activated thapsigargin prodrug as targeted therapy for prostatecancer. J Natl Cancer Inst 2003;95:990–1000.

9. Jackisch C, Hahm HA, Tombal B, et al. Delayed micromolar elevation inintracellular calcium precedes induction of apoptosis in thapsigargin-treated breast cancer cells. Clin Cancer Res 2000;6:2844–50.

10. Denmeade SR, Isaacs JT. Enzymatic activation of prodrugs byprostate-specific antigen: targeted therapy for metastatic prostate cancer.Cancer J Sci Am 1998;4 Suppl 1:S15–21.

11. Lilja H, Piironen T, Rittenhouse H, Mikolajczyk S, Slawin K. Prostate-specific Antigen. In: Vogelzang NJ, Shipley WU, Scardino PT, Coffey DS,Miles BJ, editors. Comprehensive textbook of genitourinary oncology.Philadelphia: Lippincott Williams and Wilkins; 2000. p. 638–50.





12. Deperthes D, Chapdelaine P, Tremblay RR, et al. Isolation of prostatickallikrein hK2, also known as hGK-1, in human seminal plasma. BiochimBiophys Acta 1995;1245:311–6.

13. Diamandis EP, Yousef GM, Clements J, et al. New nomenclature forthe human tissue kallikrein gene family. Clin Chem 2000;46:1855–8.

14. Darson MF, Pacelli A, Roche P, et al. Human glandular kallikrein 2(hK2) expression in prostatic intraepithelial neoplasia and adenocarcinoma:a novel prostate cancer marker. Urology 1997;49:857–62.

15. Denmeade SR, Sokoll LJ, Chan DW, Khan SR, Isaacs JT. Concentra-tion of enzymatically active prostate-specific antigen (PSA) in theextracellular fluid of primary human prostate cancers and human prostatecancer xenograft models. Prostate 2001;48:1–6.

16. Christensson A, Laurell CB, Lilja H. Enzymatic activity of prostate-specific antigen and its reactions with extracellular serine proteinaseinhibitors. Eur J Biochem 1990;194:755–63.

17. Akiyama K, Nakamura T, Iwanaga S, Hara M. The chymotrypsin-likeactivity of human prostate-specific antigen, g-seminoprotein. FEBS Lett1987;225:168–72.

18. Lilja H, Christensson A, Dahlen U, et al. Prostate-specific antigenin serum occurs predominantly in complex with a1-antichymotrypsin. ClinChem 1991;37:1618–25.

19. Otto A, Bar J, Birkenmeier G. Prostate-specific antigen formscomplexes with human a2-macroglobulin and binds to the a2-macroglob-ulin receptor/LDL receptor-related protein. J Urol 1998;159:297–303.

20. Lovgren J, Airas K, Lilja H. Enzymatic action of human glandularkallikrein 2 (hK2). Substrate specificity and regulation by Zn2+ andextracellular protease inhibitors. Eur J Biochem 1999;262:781–9.

21. Mikolajczyk SD, Millar LS, Kumar A, Saedi MS. Human glandularkallikrein, hK2, shows arginine-restricted specificity and forms complexeswith plasma protease inhibitors. Prostate 1998;34:44–50.

22. Christensen SB, Andersen A, Kromann H, et al. Thapsigarginanalogues for targeting programmed death of androgen-independentprostate cancer cells. Bioorg Med Chem 1999;7:1273–80.

23. Jakobsen CM, Denmeade SR, Isaacs JT, et al. Design, synthesis, andpharmacological evaluation of thapsigargin analogues for targetingapoptosis to prostatic cancer cells. J Med Chem 2001;44:4696–703.

24. Denmeade SR, Lou W, Lovgren J, et al. Specific and efficient peptidesubstrates for assaying the proteolytic activity of prostate-specificantigen. Cancer Res 1997;57:4924–30.

25. Khan SR, Denmeade SR. In vivo activity of a PSA-activateddoxorubicin prodrug against PSA-producing human prostate cancerxenografts. Prostate 2000;45:80–3.

26. Lovgren J, Valtonen-Andre C, Marsal K, Lilja H, Lundwall A.Measurement of prostate-specific antigen and human glandular kallikrein2 in different body fluids. J Androl 1999;20:348–55.

27. Mikolajczyk SD, Millar LS, Marker KM, et al. Ala217 is important forthe catalytic function and autoactivation of prostate-specific humankallikrein 2. Eur J Biochem 1997;246:440–6.

28. Lovgren J, Tian S, Lundwall A, Karp M, Lilja H. Production andactivation of recombinant hK2 with propeptide mutations resulting in highexpression levels. Eur J Biochem 1999;266:1050–5.

29. Kramer A, Reineke U, Dong L, et al. Spot synthesis: observations andoptimizations. J Pept Res 1999;54:319–27.

30. St Hilaire PM, Willert M, Juliano MA, Juliano L, Meldal M.Fluorescence-quenched solid phase combinatorial libraries in the charac-terization of cysteine protease substrate specificity. J Comb Chem 1999;1:509–23.

31. Chan WC, White PD, editors. Fmoc solid phase peptide synthesis,a practical approach. New York: Oxford University; 2000.

32. Kaiser E, Colescott RL, Bossinger CD, Cook PI. Color test for detectionof free terminal amino groups in the solid-phase synthesis of peptides.Anal Biochem 1970;34:595–8.

33. Felix AM, Jimenez MH. Rapid fluorometric detection for completenessin solid phase coupling reactions. Anal Biochem 1973;52:377–81.

34. Houghten RA. General method for the rapid solid-phase synthesisof large numbers of peptides: specificity of antigen-antibody interaction atthe level of individual amino acids. Proc Natl Acad Sci U S A 1985;82:5131–5.

35. Thorpe DS, Walle S. Combinatorial chemistry defines generalproperties of linkers for the optimal display of peptide ligands for bindingsoluble protein targets to TentaGel microscopic beads. Biochem BiophysRes Commun 2000;269:591–5.

36. Haese A, Graefen M, Steuber T, et al. Human glandular kallikrein2 levels in serum for discrimination of pathologically organ-confined fromlocally-advanced prostate cancer in total PSA-levels below 10 ng/mL.Prostate 2001;49:101–9.

37. Denmeade SR, Sokoll LJ, Dalrymple S, et al. Dissociation betweenandrogen responsiveness for malignant growth vs. expression of prostatespecific differentiation markers PSA, hK2, and PSMA in human prostatecancer models. Prostate 2003;54:249–57.

38. Koblinski JE, Ahram M, Sloane BF. Unraveling the role of proteasesin cancer. Clin Chim Acta 2000;291:113–35.

39. Cloutier SM, Chagas JR, Mach JP, Gygi CM, Leisinger HJ, DeperthesD. Substrate specificity of human kallikrein 2 (hK2) as determined byphage display technology. Eur J Biochem 2002;269:2747–54.

40. Schwaab R, Ludwig M, Kochhan L, et al. Detection and character-ization of two missense mutations at a cleavage site in the factor VIII lightchain. Thromb Res 1991;61:225–34.

41. Grzonka Z, Kasprzykowski F, Lubkowska L, Darlak K, Hahn TA,Spatola AF. In vitro degradation of some arginine-vasopressin analogs byhomogenates of rat kidney, liver and serum. J Pept Res 1991;4:270–4.

42. Kumar A, Mikolajczyk SD, Hill TM, Millar LS, Saedi MS. Differentproportions of various prostate-specific antigen (PSA) and human kallikrein2 (hK2) forms are present in noninduced and androgen-induced LNCaPcells. Prostate 2000;44:248–54.





2004;3:1439-1450. Mol Cancer Ther Samuel Janssen, Carsten M. Jakobsen, D. Marc Rosen, et al. therapy for prostate cancer

activated prodrug as targeted−human glandular kallikrein 2 Screening a combinatorial peptide library to develop a

Updated version

http://mct.aacrjournals.org/content/3/11/1439

Access the most recent version of this article at:

Cited articles

http://mct.aacrjournals.org/content/3/11/1439.full#ref-list-1

This article cites 39 articles, 10 of which you can access for free at:

Citing articles

http://mct.aacrjournals.org/content/3/11/1439.full#related-urls

This article has been cited by 1 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://mct.aacrjournals.org/content/3/11/1439To request permission to re-use all or part of this article, use this link



http://mct.aacrjournals.org/content/3/11/1439.full#ref-list-1

http://mct.aacrjournals.org/content/3/11/1439.full#related-urls

http://mct.aacrjournals.org/cgi/alerts

mailto:[email protected]



Screening a combinatorial peptide library to develop a human

Documents