Antibody-Directed Enzyme Prodrug Therapy with the T268G Mutant of

Jan 18, 1999 - In the current report, prodrugs of the thymidylate synthase inhibitors GW1031 and GW1843 and the dihydrofolate reductase inhibitor ...
0 downloads 0 Views 218KB Size
38

Bioconjugate Chem. 1999, 10, 38−48

Antibody-Directed Enzyme Prodrug Therapy with the T268G Mutant of Human Carboxypeptidase A1: In Vitro and in Vivo Studies with Prodrugs of Methotrexate and the Thymidylate Synthase Inhibitors GW1031 and GW1843 L. A. Wolfe, R. J. Mullin, R. Laethem, T. A. Blumenkopf, M. Cory, J. F. Miller, B. R. Keith, J. Humphreys, and G. K. Smith* Glaxo Wellcome Research and Development Inc., Five Moore Drive, Research Triangle Park, North Carolina 27709. Received May 27, 1998; Revised Manuscript Received November 4, 1998

Antibody-directed enzyme prodrug therapy (ADEPT) is a technique to increase antitumor selectivity in cancer chemotherapy. Our approach to this technology has been to design a mutant of human carboxypeptidase A (hCPA1-T268G) which is capable of hydrolyzing in vivo stable prodrugs of MTX and targeting this enzyme to tumors on an Ep-CAM1-specific antibody, ING1. Through the use of this >99% human enzyme which is capable of catalyzing a completely nonhuman reaction, we hope to increase ADEPT selectivity while decreasing overall immunogenicity of the enzyme-antibody conjugate. In the current report, prodrugs of the thymidylate synthase inhibitors GW1031 and GW1843 and the dihydrofolate reductase inhibitor methotrexate were studied for their wild-type and mutant hCPA enzyme hydrolysis, their in vivo stability, and their use in therapy. Prodrugs with high kcat/Km ratios for mutated versus wild-type hCPA1 were examined in vitro for their stability in human pancreatic juice, and in vivo for their stability in mouse plasma and tissues. In addition, targeting and in vivo enzyme activity studies were performed with an ING1 antibody conjugate of the mutant enzyme (ING1-hCPA1-T268G). Finally, in vivo therapy studies were performed with LS174T tumors to demonstrate proof of principle. Results indicate that prodrugs can be synthesized that are selective and efficient substrates of hCPA1-T268G and not substrates of the endogenous CPA activities; this leads to excellent in vivo stability for these compounds. In vivo conjugate targeting studies showed that the antibody-enzyme conjugate was targeted to the tumor and enzyme was initially active in vivo at the site. Unfortunately therapeutic studies did not demonstrate tumor reduction. Experiments to determine reasons for the lack of antitumor activity showed that the enzyme activity decreased as a result of enzyme instability. The results offer encouragement for additional novel mutant enzyme improvements and additional in vivo studies on this unique approach to ADEPT.

INTRODUCTION

One of the problems associated with cancer chemotherapy is the lack of selectivity of drug for neoplastic cells over normal proliferating cells, and the resulting systemic toxicity (1, 2). In an attempt to circumvent these problems, investigators have examined the potential of using monoclonal antibodies (mAbs)1 that bind to tumorselective antigens localized on the surface of neoplastic cells. This binding specificity can be exploited for tumorspecific delivery of materials conjugated to a targeting antibody. Cytotoxic drugs have been coupled to mAbs (3, 4), and the resulting conjugates are more selective for tumor cells than are drugs alone. Toxicity also is reduced, allowing more drug to be used without affecting highly sensitive normal cells. However, low drug delivery is a problem generally associated with this approach. Efficient drug delivery is compromised by (a) the limitation caused by the number of drug molecules that can be * Correspondence should be addressed to this author at the Department of Molecular Biochemistry, Glaxo Wellcome Research and Development, Venture 300, Research Triangle Park, NC 27709. Phone: (919) 483-1502. Fax: (919) 483-4320. Email: [email protected].

attached to the antibody, (b) antigen density, (c) the limited diffusibility of conjugates compared to drugs alone due to differences in their sizes, and (d) poor internalization of conjugate. Because of these problems, investigators continue to look for alternate modes of therapy. ADEPT has been proposed as a means to solve these problems (5-10). This technology involves selectively delivering an activating enzyme to the tumor site by conjugating it to a targeting antibody. This activating enzyme is then used to convert systemically administered 1 Abbreviations: GW1031, N-(2-fluoro-4(((1,2-dihydro-3-methyl-1-oxobenzo(f )quinazolin-9-yl)methyl)amino)benzoyl)-L-glutamic acid, 1031U89; GW1843, (S)-2-(5-(((1,2-dihydro-3-methyl-1oxobenzo(f )quinazolin-9-yl)methyl)amino)-1-oxo-2-isoindolinyl)glutaric acid, 1843U89; mtx, methotrexate; mAbs, monoclonal antibodies; ADEPT, antibody-directed enzyme prodrug therapy; CPA, carboxypeptidase type A; CPG2, carboxypeptidase type G2; CPA1, carboxypeptidase type A1; HPA, hippuryl-L-phenylalanine; HPL, hippuryl-D,L-phenyllactate; HPLC, high-performance liquid chromatography; CPG, carboxypeptidase type G; HOAc, acetic acid; FBS, fetal bovine serum; sc, subcutaneous; PBS, Dulbecco’s phosphate-buffered saline; ip, intraperitoneal; iv, intravenous; SIP, small intestine/ pancreas; HCl, hydrochloric acid; MTD, maximum tolerated dose; DHFR, dihydrofolate reductase; TS, thymidylate synthase.

10.1021/bc980057z CCC: $18.00 © 1999 American Chemical Society Published on Web 01/18/1999

ADEPT with T268G-hCPA and MTX Prodrugs

prodrug to drug at the tumor site. The catalytic potential of a targeted enzyme and the high potential prodrug concentration may overcome the low drug delivery problem. The toxic drug would be concentrated around the tumor, and its comparatively low molecular weight should allow for better tumor penetration than antibodydrug conjugates. In addition, this regional tumor-specific activation should lower systemic toxicity. An important decision that needs to be made with this type of therapy is what types of prodrugs and enzymes are to be employed. Although the need to find new active drugs still exists, it would also be of benefit to look at prodrugs of drugs currently used in the clinic since their toxicity and efficacy profiles are already understood. Enzymes that have been considered for ADEPT include both endogenous and nonendogenous types (11). One endogenous enzyme being investigated is alkaline phosphatase for the hydrolysis of phosphorylated prodrugs of phenol mustard (12), etoposide (13-15), and mitomycin C (14). Another enzyme is carboxypeptidase A for the hydrolysis of methotrexate prodrugs (16). Some nonendogeneous enzymes being used include cytosine deaminase for the conversion of 5-fluorocytosine to 5-fluorouracil (17) and the bacterial enzyme CPG2 for the cleavage of the glutamic acid moiety from prodrugs of alkylating agents (18). The ADEPT approach is not without its problems. Use of endogenous enzymes may result in the hydrolysis of prodrugs at sites distal to the tumor site, thus increasing toxicity. By contrast, nonendogenous enzymes may generate an immune response which would interfere with multiple rounds of therapy. To address these problems, we recently reported an alternate approach to ADEPT. This strategy in its most generalized form uses an endogenous enzyme mutated at the active site to perform an otherwise nonendogenous reaction to convert an in vivo stable prodrug to active drug only at the targeted tumor. In our prototype example of the concept, the “novel endogenous enzyme” was based upon human CPA1. Binding pocket mutations were introduced into hCPA1 to alter the substrate specificity of the enzyme to allow for hydrolysis of novel prodrugs that were not substrates for wild-type hCPA (19). This “novel endogenous enzyme” should be less immunogenic than nonendogenous enzymes, and prodrug hydrolysis at sites distal to tumors should be better controlled than in systems based upon endogenous activities. Simultaneously we have attempted to match mutant enzyme constructs with prodrugs of methotrexate (19) and two promising active drugs, GW1031 (20) and GW1843 (20-25). In the current report, the most promising of the previously reported MTX prodrugs and novel prodrugs of two thymidylate inhibitors were studied in vitro and in vivo to determine the therapeutic potential of this approach. MATERIALS AND METHODS

Chemicals. GW1031, GW1843, and all prodrugs were synthesized at Glaxo Wellcome Inc., Research Triangle Park, NC. ING1-TMT was obtained from Sterling Winthrop Immunoconjugates (Philadelphia, PA). 90Y was purchased from Dupont NEN (Milford, MA). Methotrexate, HPA, trypsin, HPL, and L-glutamine were purchased from Sigma Chemical Co. (St. Louis, MO). RPMI-1640 tissue culture medium was purchased from Gibco BRL (Grand Island, NY), and FBS was purchased from JRH Biosciences (Lenexa, KY). Human serum was prepared in house from blood donors.

Bioconjugate Chem., Vol. 10, No. 1, 1999 39

Kinetic Assays. Determination of the activity of hCPA1 and hCPA1-T268G on prodrugs was done by kinetic assays. Kinetic assays either were performed spectrophotometrically as described by Smith et al. (19) or were HPLC-based. HPLC assays were carried out in 0.5 mL reactions at 25 °C in assay buffer (25 mM TrisHCl, 100 mM NaCl, pH 7.4). Reactions with varying concentrations of prodrug in buffer were preincubated at 25 °C for 3 min before reactions were initiated by addition of a known amount of human CPA enzyme. Aliquots (75 µL) of the reaction were withdrawn at predetermined times and stopped by the addition of 75 µL of 1 M perchloric acid. An aliquot of the acidified reaction (20 µL) was injected onto a Waters Nova-Pak C18 column, and the prodrug and drug were resolved with the following step gradient (solvent A ) 0.1% aqueous trifluoroacetic acid, solvent B ) 0.1% trifluoroacetic acid in acetonitrile): 18% B for 3 min, 18-50% B over 0.1 min, and 50% B for 3.9 min. Kinetic rates were determined from the linear portion of product vs time plots. HPLC Analysis. HPLC analysis of in vivo samples consisted of using a 715 WISP autoinjector (Waters, Milford, MA), a constaMetric 4100 solvent delivery system (LDC Analytical, Riviera Beach, FL), a variablewavelength spectroMonitor D detector (LDC) set at the absorbance maximum of each individual prodrug, and a 3.9 × 300 mm C18, 4 µm Nova-Pak or a C18, 10 µm µBondapak HPLC column equipped with a C18 guard column. Data acquisition and analysis were performed with a DS-80 microcomputer (Digital Specialities, Chapel Hill, NC). Prodrugs of GW1031 and GW1843 and the MTX prodrug 2-carboxyPhe were injected onto the Novapak column equilibrated at 5-20% acetonitrile containing 2% HOAc, pH 3.0, and run at a flow rate of 1 mL/min. A linear gradient was immediately run over 25-35 min to a final concentration of 50% acetonitrile, 2% HOAc, pH 3.0. Absorbance was monitored at 259-266 nm. All other prodrugs were injected onto the µBondapak column equilibrated with 16% acetonitrile, 5 mM Pic A, 10 mM NH4H2PO4. A linear gradient of 16-50% acetonitrile was run over 30 min at 1 mL/min, and compounds were detected at 307 nm. Stability in Human Pancreatic Fluid. Prodrugs were evaluated for their stability in activated human pancreatic fluid as described by Smith et al. (19). Briefly, prodrug stability was determined in diluted activated pancreatic fluid. After selected incubation times, prodrug hydrolysis was determined by HPLC. Animals and Cell Lines. CD-1 nu/nu female mice (4-6 weeks old, 20-24 g) were purchased from Charles River Laboratories (Wilmington, MA). CB17 and Swiss nude female mice were purchased from Taconic Farms, Inc. (Germantown, NY). Mice were housed in microisolator cages (Lab Products Inc., Rockville, MD) in groups of up to four mice per cage. The human colon adenocarcinoma cell lines (WiDr and LS174T) were purchased from the American Type Culture Collection (Rockville, MD). Wein-133 cells, a B-cell lymphoma, were kindly provided by Dr. John Tite (The Wellcome Research Laboratories, Beckenham, U.K.). WiDr cells were grown in RPMI 1640, low-folate media supplemented with 10% charcoal-dialyzed FBS and with 10 nM calcium leucovorin (Glaxo Wellcome Inc., RTP, NC). Wein-133 cells were grown in RPMI 1640 containing 10% FBS with an additional 2.05 mM L-glutamine added. Tumor Implantation. WiDr cells were implanted sc as a 0.15 mL solution of 1.5 × 106 cells in PBS. LS174T and Wein-133 cells were implanted as a 0.20 mL solution of 2 × 106 cells in 50% low-endotoxin matrigel (Col-

40 Bioconjugate Chem., Vol. 10, No. 1, 1999

laborative Biomedical Products, Bedford, MA) in PBS. Wein-133 cells were implanted into 400 rad irradiated nude mice (Gammacell 40 irradiator, Nordion International Inc., Ontario, Canada). Tumors were allowed to grow to approximately 200-400 mg before animals were used for study. In Vivo Drug Stability and Biodistribution. Mice bearing either WiDr, Wein-133, or LS174T tumors were dosed ip bolus or iv at 50 mg/kg by injecting 0.01 mL/g mouse with a 5 mg/mL solution of prodrug in PBS, pH 6-8. Animals were sacrificed after 30 min and 2 h, and the plasma, liver, kidney, small intestine and pancreas, large intestine, spleen, cecum, feces, and tumor tissues were collected. Plasma was frozen at -70 °C, and tissues were snap-frozen in liquid nitrogen and stored at -70 °C. Duplicate animals were sacrificed for each time point. Plasma was prepared for extraction by dilution (4-fold) with ice-cold 0.1 N HCl. Tissues were prepared for extraction by homogenization with a Polytron (Brinkmann Instruments, Westbury, NY) equipped with a PTA 7 generator in 5 volumes of ice-cold 0.1 N HCl. Drugs were subsequently extracted by adding 5 volumes of -20 °C acetonitrile to 2 volumes of the HCl homogenate. After a 5 min incubation on ice, samples were centrifuged at 12400g for 15 min. Supernatants were collected and diluted 3.57-fold with PBS to reduce the final acetonitrile concentration to 20%. Diluted supernatants were then analyzed by HPLC as described above. To account for drug recovery, control homogenates were spiked with drug, prodrug, or PBS prior to the acetonitrile extraction step. After correction for recovery of these standards, the levels of drug and prodrug and the stability of the prodrug were calculated. Toxicity of MTX and Prodrugs in Mice. MTX prodrugs demonstrating good kinetic and in vivo stability profiles were compared to MTX for toxicity. Prodrugs were dosed iv into mice for 5 days, and a prodrug was considered toxic if weight loss exceeded 15%. CD-1 nu/ nu female mice were dosed with MTX (2, 4, or 6 mg/kg), MTX-Phe (5, 15, or 45 mg/kg), MTX-2-cyclopentylPhe, MTX-3-cyclopentylPhe, or MTX-3-cyclopentylPhe (10, 20, or 40 mg/kg). Swiss nude mice were dosed with MTX (2.5, 4, or 5.5 mg/kg), MTX-cyclopentylTyr, or MTX-3-cyclobutylPhe (30, 40, or 50 mg/kg). Radiolabeling of ING1 Antibody. 90Y-Antibody conjugate, [90Y]-TMT-ING1, was prepared from TMT-ING1 as follows. Using metal-free containers and pipets thoughout, 3.8 µL of 90Y (0.075 mCi, diluted with Ultrex water) was combined with 50 µL of ING1-TMT at 5 mg/mL (250 µg). The reaction was mixed and allowed to stand 60 min at 25 °C; 5 µL of 10 mM diethylenetriaminepentaacetic acid (DTPA) was added, mixed, and allowed to stand another 30 min. The reaction was purified through two 1 mL Biospin 30 columns (Biorad) equilibrated with Dulbecco’s phosphate-buffered saline. Quality control was determined by thin-layer chromatography of a small aliquot on silica gel 60 (0.2 mm thickness; catalog no. 5735; E. Merck; mobile phase 5 g of ammonium acetate, 50 mL of H2O, and 50 mL of methanol). The percent of the Biospin-purified material as chelate was typically over 95%. The purified material was diluted into 4.9 mL of 5 mg/mL ascorbate. Animals were dosed with 200 µL of ascorbate-diluted material. Conjugation of Mutant hCPA to ING1. Antibodyenzyme conjugate ING1-hCPA1-T268G was prepared as described previously (19). Briefly, the process was carried out in three steps. (A) Enzyme Modification. Four milligrams of mutant carboxypeptidase A was combined with 0.15 mg of Sulfo-

Wolfe et al.

SMCC (Pierce Chemical Co.) in 400 µL of PBS and was stirred for 45 min at 25 °C. The modified enzyme was then purified through a 1 × 13 cm G-25 medium column equilibrated with PBS. (B) Antibody Modification. The full-length IgG1 antibody ING1 was modified with the amine-reactive reagent 2-iminothiolane; 1.8 mg of antibody was combined with 0.22 mg of 2-iminothiolane in 1.3 mL of 50:50 PBS/0.1 M triethanolamine hydrochloride, 2 mM EDTA, pH 8.0, under anaerobic conditions. This was allowed to react with stirring for 1 h and 45 min, and the modified antibody was purified through a 1 × 13 cm G-25 medium column equilibrated with 0.02 M sodium acetate, 0.1 M NaCl, pH 5.8, bubbled with and maintained under a He atmosphere. (C) Coupling of Modified Antibody with Modified Enzyme. The modified antibody was collected directly from the G-25 column into the solution of the modified carboxypeptidase A. The resulting mixture was then adjusted to pH 7.4 with NaOH, made anaerobic, and allowed to react with stirring at 4 °C for 18 h. After treatment with 0.3 mM mercaptoethanolamine at room temperature for 1 h, the conjugate was then concentrated to 1 mL on an Amicon PM10 membrane and then purified on Superose 12 HR 10/30. In Vivo Targeting Studies. Targeting studies were done with radiolabeled antibody (ING1) alone and with antibody-enzyme conjugate (ING1-hCPA1-T268G). CB17 SCID female mice bearing LS174T tumors sc were dosed iv with 10 µg of [90Y]-TMT-ING1. At 24, 48, 72, or 144 h after antibody administration, 100 mg tissue fragments were dissolved in 1 mL of Solvable tissue solubilizer solution (NEN Research Products, Boston, MA) at 55 °C for 48 h, followed by counting in 20 mL of Formula 989 scintillation cocktail (NEN). The second targeting experiment was designed to assay targeted CPA activity in tissues. Swiss nude female mice bearing LS174T tumors sc were dosed iv with 300 µg of ING1-hCPA1-T268G conjugate. At 6, 24, 48, or 72 h following dosing, plasma and tissues were collected and snap-frozen in liquid N2 and stored at -80 °C. Plasma and tissues were diluted or homogenized in assay buffer and incubated (without centrifugation) in 20 µM MTX-3-cyclopentylTyr for varying times. Reactions were stopped with the addition of perchloric acid, and supernatants were analyzed by HPLC. In Vivo Therapy Studies. Swiss nu/nu female mice bearing LS174T sc implants were dosed with 300 µg of ING1-hCPA1-T268G. After 24, 48, or 72 h, mice were dosed iv for 5 days with 30 mg/kg MTX-3-cyclopentylTyr. Mice were monitored for tumor regression. Conjugate Activity (Post-Tumor Injection). Swiss nude female mice bearing LS174T tumors were dosed with 300 µg of ING1-hCPA1-T268G conjugate at three sites directly into the tumor. Mice were then dosed iv with 100 mg/kg MTX-3-cyclopentylTyr 30 min prior to sacrificing at 0.5, 6, 24, 48, or 96 h post-conjugate administration. Plasma and tumor were extracted and frozen in liquid N2 and stored -80 °C. Samples were processed for prodrug and drug levels as described above. In Vitro HCPA1 And Hcpa1-T268G Stability Studies. Wild type (hCPA1) and mutant (hCPA1-T268G) enzymes were studied for their stability at 0.1 mg/mL in PBS or human serum. Blood was collected from a healthy, normal volunteer. Serum was generated at room temperature, collected by centrifugation, and stored as aliquots at -70 °C. Enzyme was diluted from 1 mg/mL stocks (in 50 mM Tris, pH 7.4) directly into PBS, or

ADEPT with T268G-hCPA and MTX Prodrugs

Bioconjugate Chem., Vol. 10, No. 1, 1999 41

Figure 1. Prodrug structures.

human serum. Samples were then sterile-filtered into sterile screw cap centrifuge tubes for denaturation reactions. Sterility was maintained throughout the experiment. At specified times, samples were removed and assayed for CPA activity with hippurylphenylalanine as described previously (19).

RESULTS

Prodrugs under investigation were classified into four catagories (Figure 1 and Table 1). Class I prodrugs have natural, nonpolar amino acids and are substrates for wild-type CPA1. Class II prodrugs have natural, acidic amino acids. Class III prodrugs have nonnatural, bulky,

42 Bioconjugate Chem., Vol. 10, No. 1, 1999

Wolfe et al.

Table 1. Kinetics of Prodrug Hydrolysis by wt-hCPA1, wt-hCPA2, and hCPA1-T268G Compared to Stability in Human Pancreatic Juice kcat/Km (s-1 M-1) prodruga

classb

wt-hCPA1

MTX-Phe GW1031-Phe GW1843-Phe

I I I

441000 335000 690

GW1031-Glu MTX-Glu MTX-Asp

II II II

120 170 80

GW1031-naphthylAla MTX-naphthylAla MTX-3-cyclobutylPhee MTX-2-cyclopentylPhe MTX-3-cyclopentylPhe MTX-3-tert-butylPhe MTX-3-cyclopentylTyr

III III III III III III III

260 1400 0 200 0 0 0

MTX-2-carboxyPhe GW1031-2-carboxyPhe

IV IV

3 0.2

wt-hCPA2

T268G-hCPA1

pancreatic juicec

89400 nd3 nd

7350000 nd 12600

1 2 630

nd 0 0

90 120 60

nd 830 1410

nd 1400000 260 771 25 79 0

171000 1360000 1800000 157000 92400 384000 158000

5 2 2100 680 21400 4170 50000

nd nd

680 180

>50000 7500

a All data from MTX prodrugs are reproduced from Smith et al. (19) and included here for comparison with in vivo data. b Class I: prodrugs with natural, nonpolar amino acids (substrates for wild-type CPA1). Class II: prodrugs with natural, acidic amino acids (substrates for wt carboxypeptidase G). Class III: prodrugs with nonnatural, bulky, aromatic amino acids (substrates for larger pocket mutants). Class IV: prodrugs with nonnatural, acidic, aromatic amino acids. c Pancreatic juice: numbers represent stability relative to MTX-Phe. d nd: not determined. e In Phe and Tyr prodrugs, position 2 ) ortho and position 3 ) meta.

aromatic amino acids and were designed to be substrates for larger-pocket hCPA mutants. Class IV prodrugs have nonnatural, acidic, aromatic amino acids and were designed as substrates for positively charged mutant carboxypeptidases. Kinetic Assays. kcat and Km values for prodrugs of MTX were reported previously (19). Here, kinetic constants for the prodrugs of GW1031 and GW1843 were measured using wild-type (hCPA1) and mutant human CPA1 (hCPA1-T268G). Values were compared to those of MTX-Phe, a prodrug suggested by Vitols et al. (26) to have good potential in ADEPT. An ADEPT prodrug should show resistance to hydrolysis by endogenous enzymes such as wild-type hCPA1 and be easily hydrolyzed by an activating enzyme such as hCPA1-T268G. Therefore, desirable substrates for ADEPT would have much lower kcat/Km ratios with hCPA1 than that with hCPA1-T268G. In addition, a poor kcat/Km for hCPA2 also is needed (19). Our data (Table 1) indicate that while MTX-Phe demonstrated a extremely high kcat/Km ratio for hCPA1-T268G, it was also a very good substrate for hCPA1 and hCPA2. GW1031-Phe also had a similarly high kcat/Km for the wild-type enzyme, but GW1843-Phe was not hydrolyzed by either enzyme as efficiently as the other class I prodrugs. These data indicate that the class I prodrugs MTX-Phe and GW1031-Phe would not likely be good candidates in ADEPT since premature prodrug hydrolysis would likely occur in vivo. The other class I prodrug, GW1843-Phe, and all class II prodrugs tested were poor substrates for either the hCPA1 or the hCPA1T268G enzyme. This poor hydrolysis with hCPA1-T268G in addition to the lack of specificity for the mutant enzyme over the wild-type enzyme indicates that these compounds are poor substrates in ADEPT with hCPA1T268G. Class III prodrugs gave kinetic profiles where the kcat/Km for the mutant enzyme was much better than that observed with the wild-type enzyme. GW1031-naphthylAla, like MTX-naphthylAla reported previously (19), is a poor hCPA1 substrate but is an excellent hCPA1-T268G substrate. However, MTX-naphthylAla is an excellent substrate for hCPA2 as well, and therefore would be a poor ADEPT prodrug. With the exception of the naphthylAla prodrugs and MTX-2-cyclopentylPhe, class III prodrugs had no observed turnover of prodrug with

hCPA1 (19). All class III MTX prodrugs showed some turnover with hCPA2, except MTX-3-cyclopentylTyr. However, observed turnover was modest with the exception of MTX-naphthylAla. The very high kcat/Km and specificity of class III prodrugs for hCPA1-T268G indicate that class III prodrugs show potential for ADEPT. The kinetic data from class IV prodrugs resembled that of class II prodrugs in that there was poor enzyme activity with both enzymes (19). Therefore, class IV prodrugs are not likely candidates for ADEPT with hCPA1-T268G. Based on the kinetic data, the top four prodrugs (all class III) are rank-ordered for stability versus the hCPA enzymes as MTX-3-cyclopentylTyr > MTX-3-cyclopentylPhe > MTX-3-tert-butylPhe > MTX-3-cyclobutylPhe. Pancreatic Fluid Assays. In an attempt to model their in vivo stability, selected prodrugs were analyzed for their stability in activated human pancreatic fluid (19). MTX-Phe, a prodrug previously reported in the literature as having potential in ADEPT, was chosen as the prodrug with which others were compared. Higher relative stabilities in pancreatic fluid should correlate to an increase in in vivo prodrug stability. Results (Table 1) from the pancreatic fluid assay correlated fairly well with the kinetic data shown in the same table. Higher relative stability in pancreatic fluid paralleled resistance to hydrolysis by wild-type hCPA1 and hCPA2 in our kinetic assays (19). Based on the pancreatic fluid assays, the three best prodrugs are rank-ordered for stability as MTX-3-cyclopentylTyr > MTX-3-cyclopentylPhe > GW1031-2-carboxyPhe. The latter, which belongs to class IV, would not be a good prodrug with T268G. If more weight is given to the pancreatic fluid assay, which should be a better indication of in vivo stability than the kinetic assays, then the data indicate that MTX-3cyclopentylTyr > MTX-3-cyclopentylPhe > MTX-3-cyclobutylPhe, MTX-3-tert-butylPhe. If a correlation of the pancreatic juice stability and the in vivo stability could be found, it would point to the former as a useful primary screen for in vivo stability of compounds. In vivo experiments were conducted to test this. In Vivo Stability and Biodistribution. Fourteen prodrugs were examined for their stability (Table 2) and biodistribution in mice (Table 3). The compound suggested by Vitols et al. (MTX-Phe) showed poor stability

ADEPT with T268G-hCPA and MTX Prodrugs

Bioconjugate Chem., Vol. 10, No. 1, 1999 43

Table 2. In Vivo Prodrug Stability (% of Prodrug to Prodrug + Drug)a Prodrugb MTX-Phe GW1031-Phe GW1843-Phe GW1031-Glu MTX-Asp GW1031-naphthylAla MTX-naphthylAla MTX-2-cyclopentylPhe MTX-3-cyclobutylPhe MTX-3-cyclopentylPhe MTX-3-tert-butylPhe MTX-3-cyclopentylTyr MTX-2-carboxyPhe GW1031-2-carboxyPhe

time (h)

plasma

liver

kidney

Lg. Intc

SIPd

spleen

0.5 2 0.5 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2

31 3 80 100 93 99 100 100 100 93 48 66 nd 95 90 97 nd 100 nd 100 100 100 100 100 100 100 99

95 8 45

100 ndf 75

65 22 67

17 0

81 nd 39

65 nd 63

100 100 70 0 100 100

96 76 82 59

98 97 97 93 87 90 59 nd 90 76 81 73 83 94 91 89 100 100 100 89 99 98

98 97 98 98 89 74

93 100 100 65 94 87

84 81 83 82 100 100 53 22 100 100 100 100 100 99

93 57 74 85 100 100 nd 67 100 100 100 100 100 100

91 47 83 68 86 54 10 1 2 0 77 35 87 68 97 93 91 67 100 100 100 97 99 98

98 100 96 nd 100 nd nd nd 100 100 100 100 99 nd

cecum

feces

63 48

21 52

75 56 100 100 100 99 100 81 100 100

71 55 92 68 97 94

tumore

100 88

100 100 100 100 100 100

a CD-1 nu/nu female mice were dosed ip or iv with 50 mg/kg of prodrug. After 0.5 or 2 h, plasma and tissues were removed. Plasma was frozen at -70 °C, and tissues were snap-frozen in liquid N2, and stored at -70 °C. Samples were diluted or homogenized in 0.1 N HCl. Prodrug and drug were extracted in acetonitrile, and supernatants were analyzed by HPLC as a percentage of prodrug to prodrug + drug. b Prodrug: MTX-Asp, MTX-2-cyclopentylPhe, MTX-3-cyclobutylPhe, MTX-3-cyclopentylPhe, MTX-3-tert-butylPhe, and MTX-3cyclopentylTyr dosed iv; all others dosed ip. c Lg.Int: large intestine. d SIP: small intestine and pancreas. e Tumor: LS174T for MTX-3cyclobutylPhe; WiDr for MTX-Phe; GW1031-Phe, GW1031-Glu, and Wein-133 for all others. f nd ) no drug or prodrug detected.

Table 3. In Vivo Prodrug Biodistribution (nmol of Prodrug/g of Tissue)a prodrugb MTX MTX-Phe GW1031-Phe GW1843-Phe GW1031-Glu MTX-Asp GW1031-naphthylAla MTX-naphthylAla MTX-2-cyclopentylPhe MTX-3-cyclobutylPhe MTX-3-cyclopentylPhe MTX-3-tert-butylPhe MTX-3-cyclopentylTyr MTX-2-carboxyPhe GW1031-2-carboxyPhe

time (h)

plasma

liver

kidney

Lg.Intc

SIPd

spleen

cecum

feces

0.5 2 0.5 2 0.5 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2 0.5 2

13.7 0.6 17.8 1.1 32.0 22.1 0.3 41.5 4.2 46.7 2.0 32.4 2.8 8.0 nd 28.5 1.1 22.2 nd 1.1 nd 2.6 0.7 3.3 0.6 51.4 0.2 71.4 1.4

634.5 29.0 335.1 0.6 79.7

15.2 8.5 26.9 0 6.7

6.8 55.0 19.3 1.1 17.9

199.8 87.2 17.3 ndf

0 0 19.9 nd 7.8

11.7 255.9

554.7 1380.0

770.3 579.8 287.8 161.6 135.3 371.2 49.8 nd 113.5 43.9 53.8 47.8 54.3 126.2 58.6 39.5 47.7 110.1 171.9 7.5 343.4 33.3

175.8 70.3 93.4 12.6 29.2 4.0

12.2 1.0 5.7 19.4 14.9 1.1

5.6 3.2 7.3 3.0 6.8 1.3 2.2 0.2 3.7 0.5 55.4 3.8 94.5 14.5

5.2 27.3 2.7 4.4 2.6 6.2 nd 14.7 7.9 12.5 12.1 91.8 25.3 45.4

170.2 17.6 48.2 55.6 21.3 14.2 27.7 1.8 7.5 nd 364.5 31.7 301.3 140.0 377.9 211.0 550.8 44.9 373.8 322.2 127.4 24.1 188.1 69.1

23.4 1.8 3.4 0 30.9 1.5 1.4 1.1 2.7 nd 8.4 nd nd nd 0.6 0.2 12.4 1.0 30.9 nd

tumore

3.0 nd 2.6

8.5 35.8

4.2 207.0

6.8 235.1 4.5 0.8 4.1 108.3 16.3 402.3 14.4 4.8

150.2 646.2 426.8 389.8 1104.0 1506.0 880.0 1226.0 665.9 738.4

6.0 1.2 7.5 0.8

4.8 2.4

6.9 1.5 6.7 1.3

a Mice were dosed and samples processed as described in Table 2. Plasma and tissue prodrug levels were determined. Values are in nmol/g for tissue and µM for plasma. b Prodrug:MTX, MTX-Asp, MTX-2-cyclopentylPhe, MTX-3-cyclobutylPhe, MTX-3-cyclopentylPhe, MTX-3-tert-butylPhe, and MTX-3-cyclopentylTyr dosed iv; all others dosed ip. c Lg.Int: large intestine. d SIP: small intestine and pancreas. e Tumor: LS174T for MTX-3-cyclobutylPhe; WiDr for MTX-Phe; GW1031-Phe, GW1031-Glu, and Wein-133 for all others. f nd ) no drug or prodrug detected.

in vivo, as expected, based on the kinetic and pancreatic fluid assay data. MTX-Phe was completely converted to drug in the small intestine and pancreas by 2 h (19). One

other class I prodrug, GW1031-Phe, was similarly unstable, and the other, GW1843-Phe, was somewhat more stable than MTX-Phe in vivo, consistent with their CPA

44 Bioconjugate Chem., Vol. 10, No. 1, 1999

Wolfe et al.

Figure 2. Toxicity of MTX and prodrugs in CD-1 nu/nu mice. CD-1 nu/nu female mice were dosed iv for 5 days with varying concentrations of MTX or selected class III prodrugs. A prodrug was deemed toxic if a 15% weight loss occurred. Doses shown for MTX and MTX-2-cyclopentylPhe were at the limit of toxicity. Doses for MTX-3-cyclopentylPhe and MTX-3-tert-butylPhe were the highest doses tested. Data are represented as percent weight change as a function of time.

substrate activity and pancreatic fluid stability. The two prodrugs tested in vivo from class II, GW1031-Glu and MTX-Asp, were more stable, again consistent with their stability from hCPA1 hydrolysis. Hydrolysis of these compounds was observed, however, particularly in tissues taken from the gut, where concentrations of CPA are the highest. Class III prodrugs generally gave better in vivo stability results than class I and class II. The exception was MTX-naphthylAla which is a substrate for wild-type hCPA2 and gave a profile similar to those of class I. The rest of the class III prodrugs are rank-ordered for stability as MTX-3-cyclopentylTyr > MTX-3-cyclopentylPhe > MTX-3-cyclobutylPhe > MTX-3-tert-butylPhe > MTX-2-cyclopentylPhe. Class IV prodrugs had similar stability profiles to MTX-3-cyclopentylTyr, the most stable class III prodrug. The biodistributions of class I, class II, and class IV prodrugs and the naphthylAla class III prodrugs indicate that the liver is the primary site of initial prodrug accumulation. The other class III prodrugs were found more in the small intestine and pancreas than the liver at the earliest time point tested. In all studies where feces were analyzed, the prodrug concentration in the feces after 2 h was higher than any tissue studied. This generally was the case after 30 min as well. All prodrug classes demonstrated rapid plasma clearance. This rapid biliary excretion was similar to that observed for MTX (27). Based upon kinetic data, pancreatic juice assays, and the in vivo stability and biodistribution results in mice, five prodrugs (MTX-Phe, MTX-2-cyclopentylPhe, MTX-3-cyclopentylPhe, MTX-3-tert-butylPhe, and MTX-3-cyclopentylTyr) were selected for mouse toxicity studies and compared to MTX. Toxicity of MTX and Prodrugs in Mice. Selected prodrugs were evaluated in vivo for toxicity and compared to MTX (Figures 2 and 3). Toxicity was defined as death and/or a greater than 15% weight loss. In Figure 2, doses shown for MTX (4 mg/kg), MTX-Phe (15 mg/kg), and MTX-2-cyclopentylPhe (20 mg/kg) were at the limit of toxicity, and consistent with this, each produced a significant delay in weight increase at this dose. MTXPhe resulted in 4 of 4 deaths at the dose above this limit (45 mg/kg). Doses for MTX-3-cyclopentylPhe (40 mg/kg) and MTX-3-tert-butylPhe (40 mg/kg) were the highest

Figure 3. Toxicity of MTX and prodrugs in Swiss nude mice. Swiss nude female mice were dosed as described in Figure 2. The dose shown for MTX-3-cyclopentylTyr was the highest dose tested. The dose for MTX-3-cyclobutylPhe was the lowest dose tested and resulted in 2 of 3 deaths. The dose for MTX was the highest dose tested without toxicity. Data are represented as described in Figure 2.

doses tested; the latter compound produced a delay in weight increase. In Figure 3, the dose shown for MTX3-cyclobutylPhe (30 mg/kg) was the lowest dose tested and resulted in 2 of 3 deaths. The dose for MTX (2.5 mg/ kg) was the highest dose tested without toxicity. The dose for MTX-3-cyclopentylTyr (50 mg/kg) was the highest dose tested and had no toxicity. The data indicate that the in vivo stable prodrugs MTX-3-cyclopentylPhe, MTX3-tert-butylPhe, and MTX-3-cyclopentylTyr could be safely dosed at 10-20 times the MTX maximum tolerated dose while the less stable prodrugs MTX-Phe, MTX-2-cyclopentylPhe, and MTX-3-cyclobutylPhe were toxic. These unstable and toxic prodrugs therefore would be less useful ADEPT prodrugs. In Vivo Targeting Studies. Targeting studies were done to determine if an antibody-enzyme conjugate, ING1-hCPA1-T268G, which targets Ep-CAM-1 (28) on LS174T tumors, could be shown to target the tumor in vivo and if the enzyme moiety of the conjugate retained its catalytic activity in vivo. Desired results would be an accumulation of antibody-enzyme conjugate only at the tumor with a concurrent drop of plasma levels of conjugate. In vivo targeting studies were first done with 90Ylabeled antibody to verify that ING1 targets LS174T. The data indicated that dosing with 10 µg of ING1 does target

ADEPT with T268G-hCPA and MTX Prodrugs

Bioconjugate Chem., Vol. 10, No. 1, 1999 45 Table 4. Activity of ING1-T268G-CPA1 after Direct Injection into LS174T Tumors (% Conversion of MTX-3-cyclopentylTyr)a timeb plasma tumor (h) [MTX]c [prodrug] % conv [MTX] [prodrug] % conv CTLd 0.5 6 24 48 96

Figure 4. In vivo targeting of ING1-hCPA1-T268G in mice. Swiss nude female mice bearing LS174T tumors sc were dosed iv with 300 µg of ING1-hCPA1-T268G conjugate. At 6, 24, 48, or 72 h following dosing, plasma and tissues were collected and snap-frozen in liquid N2 and stored -80 °C. Plasma and tissues were diluted or homogenized in assay buffer and incubated (without centrifugation) in 20 µM MTX-3-cyclopentylTyr for varying times. Reactions were stopped with the addition of perchloric acid, and supernatants were analyzed by HPLC. Data are reported as % injected dose/g of tissue as a function of time.

Figure 5. In vivo therapy study. Swiss nu/nu female mice bearing LS174T sc implants were dosed with 300 µg of ING1hCPA1-T268G. After 24, 48, or 72 h, mice were dosed iv for 5 days with 30 mg/kg MTX-3-cyclopentylTyr. Mice were monitored for tumor growth using electronic caliper measurements.

LS174T with levels approaching 30% ID/g of tissue at 48 h while antibody levels in the liver were only 7% (data not shown). All other tissues studied (plasma, kidney, bone, and spleen) had less accumulation of antibody than tumor. Tumor-to-blood ratios were 1.0, 1.9, 4.4, and 19.6 at 24 h, 48 h, 72 h, and 144 h, respectively. Thus, ING1 targeted tumor and was cleared from circulation. A second targeting study was then performed by dosing with 300 µg of the ING1-hCPA1-T268G antibodyenzyme conjugate. Targeting studies with ING1-hCPA1T268G (Figure 4) gave tumor-to-blood ratios of 2.7 at 24 h. As expected, conjugate levels in the blood dropped rapidly while levels in the tumor showed a much slower rate of decrease. The level of active conjugate found in tumor was approximately 1% ID/g of tissue. These data were encouraging, and an in vivo therapy study was designed. In Vivo Therapy Study. An in vivo therapy study in Swiss nude mice bearing LS174T was performed using a single 300 µg dose of ING1-hCPA1-T268G followed by five daily doses of 30 mg/kg MTX-3-cyclopentylTyr. Prodrug dosing was commenced at 24, 48, or 72 h postconjugate dose. Mice were monitored for tumor growth inhibition and compared to untreated mice. Results (Figure 5) indicate that under these conditions, the ADEPT was not efficacious. Conjugate Activity (Post-Tumor Injection). In an attempt to determine why the therapy studied failed, the activity of ING1-hCPA1-T268G conjugate was studied as a function of time after direct injection into LS174T tumors. In this experiment, conjugate was injected into the tumor. Then immediately or 6, 24, 48, or 96 h later MTX-3-cyclopentylTyr was dosed systemically as in the

0 0 0 0 0 0

18.6 18.5 25.9 5.83 23.2 24.3

0 0 0 0 0 0

0 2.2 2.0 0 0 0

5.2 2.8 5.0 9.6 7.7 9.8

0 44.0 28.6 0 0 0

a Swiss nude female mice bearing LS174T tumors were dosed with 300 µg of ING1-hCPA1-T268G conjugate at 3 sites directly into the tumor. Mice were then dosed iv with 100 mg/kg MTX-3cyclopentylTyr at 0.5, 6, 24, 48, or 96 h post-conjugate administration. Mice were sacrificed 30 min following prodrug administration. Plasma and tumor were extracted and snap-frozen in liquid N2 and stored at -80 °C. Samples were processed and analyzed by HPLC for prodrug concentration and stability (% conversion of MTX-3-cyclopentylTyr). b Time: time between conjugate dose and tissue extraction. c Concentration in µM (plasma) or nmol/g of tissue (tumor). d CTL: no conjugate control.

therapy experiment. After another 30 min, animals were killed, and MTX and MTX-3-cyclopentylTyr levels were determined in tumor and plasma as described in the in vivo drug stability section. This protocol allowed determination of hCPA1-T268G activity. The results (Table 4) show absolute levels of these compounds as well as the inferred percent conversion of prodrug to drug at that site. It can be seen that prodrug was found in both the tumor and plasma at all time points. In addition, drug (MTX) was also found in the tumors at the early time points. The targeted enzyme converted 42.9, 30.1, and 0% of MTX-3-cyclopentylTyr to MTX at 0.5, 6, and 24 h, respectively. No MTX was found in the plasma at any time point. This indicates that the MTX found in the tumors was generated exclusively via the tumor-localized enzyme. In addition, since MTX was not found in tumors from (control) animals that were not dosed with conjugate, the prodrug was synthesized specifically by the targeted enzyme. This experiment then clearly shows that active enzyme was localized to the tumor by the ING1-hCPA1-T268G conjugate, and the activity was capable of producing MTX in situ. However, enzyme activity targeted to the tumor was lost by 24 h. Therefore, the result explains the lack of response in the in vivo therapy study since the earliest prodrug was dosed for therapy 24 h after conjugate administration. By then, conjugate activity would of been lost. In Vitro Conjugate Stability Studies. We have reported that hCPA1-T268G is relatively stable at 37 °C (19). Therefore the complete loss of activity by 24 h was surprising. One possible explanation for the loss of conjugate activity in the therapy study is that serum factors may inactivate the enzyme. To test this, an in vitro 37 °C stability study was performed on both hCPA1 and hCPA1-T268G in PBS and in human serum. Results in Figure 6 show that, as we reported previously, the halflife for the wt hCPA1 was significantly longer in PBS than was that of hCPA1-T268G. This loss of stability upon mutation may derive from the destabilizing effect that cavity-creating mutations can produce (29). Surprisingly, however, the half-life of hCPA1 was reduced dramatically from 140 h in PBS to ∼24 h in human serum. The half-life of the hCPA1-T268G was also reduced by human serum, but not as dramatically, from 24 to ∼15 h. This inactivation of the enzyme in human serum resulted in 70% loss of hCPA1-T268G activity by 24 h and 90% loss of activity by 72 h. Thus, serum factors do inactivate the enzyme and can probably account for

46 Bioconjugate Chem., Vol. 10, No. 1, 1999

Figure 6. Stability of hCPA1 and hCPA1-T268G at 37 °C in vitro in PBS or human serum (HS). Wild-type (hCPA1) and mutant (hCPA1-T268G) enzymes were studied for their stability at 0.1 mg/mL in PBS or human serum. Enzyme was diluted directly into PBS or human serum. At specified times, samples were removed and assayed for CPA activity with hippurylphenylalanine. Data are shown as percent enzyme activity as a function of time.

some of the activity loss in vivo. However, since 30% activity remained at 24 h, the serum denaturation studies may not account for loss of all detectable activity in vivo at the tumor by 24 h. The stability of the enzyme under in vivo relevant conditions clearly warrants further study. DISCUSSION

Our approach to ADEPT is to overcome the problem of enzyme selectivity vs immunogenicity by mutating a human enzyme to provide it with a nonendogenous activity, thereby minimizing immunogenicity and maximizing selectivity. The objectives of the current study were to examine the stability of four classes of prodrugs which would be suitable substrates for the novel mutant human enzyme hCPA1-T268G but would not be substrates for the wild-type hCPA1. The short-term biodistribution of these prodrugs was also determined in this effort. Selected prodrugs were then evaluated in vivo for their toxicity, elimination, and pharmacokinetic profiles. Finally targeting studies were done with antibody and antibody-enzyme conjugate in preparation for an in vivo therapy study. Three assays were developed to examine prodrug stability. For a primary screen, kinetic assays were performed to generate kcat and Km values. The in vivo stabilities of prodrugs were then determined in mice. Compounds with both high and low kcat/Km ratios were studied in vivo to demonstrate a correlation between the kinetic and in vivo data. A third assay, human pancreatic fluid stability, was later developed as a simpler and perhaps more relevant reflection of prodrug stability in humans. This last assay was eventually used as a secondary screen prior to the more time-consuming in vivo study. Promising MTX prodrugs were dosed in mice up to 20 times the MTD of MTX to determine if a MTD of prodrug could be obtained. Targeting studies were done to demonstrate that our antibody was selective by targeting tumor better than normal tissue and that our antibody-enzyme conjugate was both selective and active. Finally, an in vivo therapy study was done. The data indicate that we were able to generate very stable prodrugs of both the DHFR inhibitor MTX and the TS inhibitor GW1031 (MTX-3-cyclopentylTyr, MTX-carboxyPhe, GW1031-2-carboxyPhe, and MTX-3-cyclo-

Wolfe et al.

pentylPhe). The pancreatic fluid and in vivo screens produced the same trends for stability. It was expected that the stability of prodrugs in vivo or in pancreatic fluid would be predicted from the kcat/Km for the prodrug with hCPA1 and hCPA2. Generally, the hCPA1 kinetics alone were enough to predict in vivo and pancreatic fluid stabilities. The naphthylalanyl prodrugs of GW1031 and MTX, however, were much less stable in both the in vivo and pancreatic fluid assays than predicted from the kinetic data of hCPA1 alone. The lack of stability of these compounds in vivo, particularly in the small intestine and pancreas, seemed to suggest the presence of carboxypeptidases other than CPA1 in both mice and humans. Another CPA, CPA2, has been identified in rat (30) and human (31) but has yet to be reported in the cow (32). This enzyme prefers bulkier amino acids than CPA1. Indeed, we have shown previously that the MTX-naphthylalanine prodrug is hydrolyzed 1000 times more efficiently by hCPA2 than hCPA1 (19, 31). We assume that CPA2 is the activity in mice and human pancreatic juice responsible for the hydrolysis of the naphthylalanyl prodrugs of GW1031 and MTX (19), and when it was included in the analysis, the stabilities of all compounds in vivo or in pancreatic fluid were readily predicted (19). The highest in vivo stability, along with high hCPA1T268G activity, was observed with the nonnatural, bulky, aromatic amino acid prodrug MTX-3-cyclopentylTyr, a class III prodrug. With the exception of the naphthylalanyl prodrugs as discussed above, this class in general showed the most promise for ADEPT with good in vivo stability and high hCPA1-T268G activity. Prodrugs of class IV, those with a nonnatural, acidic, aromatic amino acid substituent, and which were designed as substrates for mutant carboxypeptidases with positively charged active sites, also showed excellent in vivo stability. These compounds were, however, not good substrates for hCPA1T268G, and efforts to produce an enzyme with an active site positive charge have thus far been unsuccessful (Table 1 and ref 19). Nonetheless, their in vivo properties make them of continued interest. Compounds of class II, those prodrugs with an aspartate or glutamate substituent, were less stable in vivo, but considerably more stable than those of class I. As such they too may have utility if an appropriate enzyme can be produced. Finally, compounds of class I, those with a natural, nonpolar amino acid attached and which are substrates for wildtype CPA1, had low stability in vivo. These compounds would appear to have the lowest probable utility for ADEPT. It is possible that semistable prodrugs could be used in ADEPT. Obviously, very stable prodrugs such as MTX3-cyclopentylTyr would be best; however, prodrugs with partial degradation may still be useful if systemic toxicity can be avoided. For example, Senter et al. (14) using an alkaline phosphatase catalyzed system observed better antitumor activity with etoposide phosphate than etoposide, even though they observed prodrug hydrolysis in the absence of antibody-enzyme conjugate. In our case, the maximum tolerated dose of stable and semistable methotrexate prodrugs in mice was determined to ascertain how much endogenous degradation would be acceptable. Toxicity studies indicated that stable prodrugs such as MTX-3-cyclopentylTyr, MTX-3-cyclopentylPhe, and MTX-3-tert-butylPhe could be dosed at least 10-20 times the maximum tolerated dose of MTX. Semistable prodrugs such as MTX-2-cyclopentylPhe, as expected, were more toxic than our most stable prodrugs but less toxic than MTX alone. Vitols et al. (33) reported the class I prodrug metho-

ADEPT with T268G-hCPA and MTX Prodrugs

trexate-phenylalanine and showed in vitro that UCLAP3 growth inhibition (ID50 values) of the prodrug in the presence of their CPA-mAb conjugate approached that of methotrexate. From these data, they inferred that methotrexate-phenylalanine would be an optimal ADEPT prodrug. However, our data indicate that class I prodrugs including methotrexate-phenylalanine are highly susceptible to in vivo degradation by endogenous enzymes, like CPA, and that using class I prodrugs designed for these enzymes would give unsatisfactory results. It is worthy of note, though, that although MTX-Phe was rapidly hydolyzed in vivo and as a result more toxic than the more in vivo stable prodrugs, its toxicity was still less than that of MTX, which may provide it some utility in ADEPT with hCPA1 or hCPA1-T268G. Pharmacokinetic analysis showed that for all compounds, prodrug levels decreased rapidly in the plasma between 30 min and 2 h and followed this trend in almost all the tissues studied. The exceptions to this trend can probably be attributed to the variations typically seen between animals. Initial prodrug levels were high in the liver. This might be expected since the liver functions as a drug detoxification organ. The rapid biliary excretions of MTX and MTX prodrugs are responsible for the high prodrug levels in feces and small intestine and pancreas. This rapid excretion may be species-specific since the principle route of MTX clearance in humans is via the urine (27). The levels of prodrug found in tumor, while low, are consistent with the levels of methotrexate found in murine leukemias following ip dosage with the maximum tolerated dose of 3 mg/kg of methotrexate (34). Metabolic cage studies were employed to monitor biliary excretion of class III prodrugs (data not shown). These studies indicated that class III prodrugs exhibit rapid fecal excretion with a small percentage of prodrug found in the urine. This fecal excretion is consistent with that of MTX (27). We have previously demonstrated that ADEPT using a mutated human enzyme and novel prodrugs works in cell culture (19). In that study, HT-29 cells were exposed to varying levels of antibody-enzyme conjugate followed by varying concentrations of MTX-3-cyclopentylTyr. IC50’s were determined and compared to MTX alone. The data demonstrated the efficacy of this variation on ADEPT in vitro. The current work extends those data further by including targeting studies and an in vivo therapy study. ING1, an antibody that recognizes an epitope on EpCAM1 (28) on human colorectal carcinomas, was dosed into mice to determine antibody specificity. Data indicated a time-dependent drop in percent ID in all tissues studied. After plasma clearance, the levels were highest in the tumor, followed by spleen, liver, kidney, and bone. At 144 h, the latest time point tested, the tumor-to-blood ratio was 19. The in vivo targeting of the antibodyenzyme conjugate also showed an encouraging tumorto-plasma ratio. At 24 h, a ratio of 2.7 was observed with ∼1% ID/g of tumor found. A possible explanation for the lack of a tumor response in vivo to therapy was instability of the conjugate. To test this hypothesis, a study was set up where conjugate was directly injected into the tumor, followed by a bolus dose of MTX-3-cyclopentylTyr. Then enzyme activity as a function of time postinjection was determined by monitoring percent prodrug hydrolysis 30 min after each prodrug injection time. The data suggested that the conjugate does suffer a time-dependent drop in activity in vivo. While there was in vivo enzyme activity present at up to 6 h post-conjugate injection, there was no detectable activity in vivo 24 h after injection. This

Bioconjugate Chem., Vol. 10, No. 1, 1999 47

explains the lack of antitumor activity in the in vivo therapy study since the earliest prodrug was dosed 24 h after conjugate administration. As shown in Figure 6, both wt- and T268G-hCPA1 were inactivated by human serum. Thus, a partial explanation for the loss of conjugate activity in vivo is inactivation by serum elements. (Preliminary experiments indicate that 5 mM mercaptoethanolamine in PBS can also accelerate denaturation of both wt- and T268G-hCPA1, resulting in loss of 50 and 75% activity by 18 h, respectively, which suggests that the serum factors may include thiols.) However, there appear to be additional features as well since at 37 °C in human serum, 30% of T268G-hCPA1 remained at 24 h and 10% remained at 72 h. Continued studies of this enzyme and its properties are warranted to enhance this promising approach to ADEPT. Further study using a different tumor model also may aid in determining if ADEPT works in vivo. In the current study, the human colon adenocarcinoma cell line LS174T was chosen based on the targeting ability of our ING1 antibody to it. Previous studies indicate that MTX, at nontoxic doses, only delays growth of LS174T tumors. Since one of the goals of ADEPT is to avoid the systemic toxicity seen during therapy while maintaining or improving efficacy, studies using a tumor model that responds better to MTX would help in determining the utility of ADEPT. The development of mAbs represents an additional mode of cancer therapy and has led to the emergence of ADEPT. The novel approach we have taken, in which a mutant human enzyme is used for the conversion of stable prodrugs to their toxic form, appears to offer a high potential of success. This approach addresses two of the hurdles facing ADEPT. The use of a human enzyme should minimize the immunogenicity. The active site mutation will allow for the development of an exogenous enzymatic activity allowing for increased prodrug stability. The development of an antibody with better specificity in combination with a more stable antibody-enzyme conjugate would be beneficial and should give better antitumor results with MTX prodrugs in mice. Antitumor studies in mice using the other two antifolates, GW1031 and GW1843, are difficult because they exert their effect by inhibiting thymidylate synthase. This inhibition cannot take place in the presence of high circulating thymidine levels as observed in mice (35). For this reason, mouse antitumor studies were not done with prodrugs of these two antifolates. LITERATURE CITED (1) Zubrod, C. G. (1982) Principles of chemotherapy. in Cancer Medicine (Holland, J. F., and Frei, E., Eds.) pp 627-632, Lea & Febiger, Philadelphia. (2) Mihich, E., and Creaven, P. J. (1982) Principles of pharmacology and toxicology. in Cancer Medicine (Holland, J. F., and Frei, E., Eds.) pp 685-715, Lea & Febiger, Philadelphia. (3) Hellstrom, K. E., Hellstrom, I., and Goodman, G. E. (1987) Antibodies for drug delivery. in Controlled Drug Delivery, Fundamentals and Applications (Robinson, J. R., and Lee, V. H. L., Eds.) 2nd ed., pp 623-653, Marcel Dekker, New York. (4) Smyth, M. J., Pietersz, G. A., Classon, B. J., and McKenzie, I. F. C. (1986) Specific targeting of chlorambucil to tumors with the use of monoclonal antibodies. J. Natl. Cancer Inst. 76(3), 503-510. (5) Senter, P. D. (1990) Activation of prodrugs by antibodyenzyme conjugates: a new approach to cancer therapy. FASEB J. 4, 188-193. (6) Bagshawe, K. D., Springer, C. J., Searle, F., Antoniw, P., Sharma, S. K., Melton, R. G., and Sherwood, R. F. (1988) A

48 Bioconjugate Chem., Vol. 10, No. 1, 1999 cytotoxic agent can be generated selectively at cancer sites. Br. J. Cancer 58, 700-703. (7) Bagshawe, K. D. (1989) Towards generating cytotoxic agents at cancer sites. Br. J. Cancer 60, 275-281. (8) Hellstrom, K. E., and Senter, P. D. (1991) Activation of prodrugs by targeted enzymes. Eur. J. Cancer 27(11), 13421343. (9) Philpott, G. W., Bower, R. J., Parker, K. L., Shearer, W. T., and Parker, C. W. (1974) Affinity cytotoxicity of tumor cells with antibody-glucose oxidase conjugates, peroxidase, and arsphenamine. Cancer Res. 34, 2159-2164. (10) Philpott, G. W., Bower, R. J., and Parker, C. (1973) Selective iodination and cytotoxicity of tumor cells with an antibody-enzyme conjugate. Surgery (St. Louis) 74(1), 5158. (11) Senter, P. D., Wallace, P. M., Svensson, H. P., Kerr, D. E., Hellstrom, I., and Hellstrom, K. E. (1991) Activation of prodrugs by antibody-enzyme conjugates. in Immunobiology of proteins and peptides VI (Atassi, M. Z., Ed.) pp 97-105, Plenum Press, New York. (12) Wallace, P. M., and Senter, P. D. (1991) In vitro and in vivo activities of monoclonal antibody-alkaline phosphatase conjugates in combination with phenol mustard phosphate. Bioconjugate Chem. 2(5), 349-352. (13) Haisma, H. J., Boven, E., van Muijen, M., De Vries, R., and Pinedo, H. M. (1992) Analysis of a conjugate between anti-carcinoembryonic antigen monoclonal antibody and alkaline phosphatase for specific activation of the prodrug etoposide phosphate. Cancer Immunol. Immunother. 34, 343348. (14) Senter, P. D., Schreiber, G. J., Hirschberg, D. L., Ashe, S. A., Hellstrom, K. E., and Hellstrom, I. (1989) Enhancement of the in vitro and in vivo antitumor activities of phosphorylated mitomycin C and etoposide derivatives by monoclonal antibody-alkaline phosphatase conjugates. Cancer Res. 49, 5789-5792. (15) Senter, P. D., Saulnier, M. G., Schreiber, G. J., Hirschberg, D. L., Brown, J. P., Hellstrom, I., and Hellstrom, K. E., (1988) Anti-tumor effects of antibody-alkaline phosphatase conjugates in combination with etoposide phosphate. Proc. Natl. Acad. Sci. U.S.A. 85, 4842-4846. (16) Haenseler, E., Esswein, A., Vitols, K. S., Montejano, Y., Mueller, B. M., Reisfeld, R. A., and Huennekens, F. M. (1992) Activation of methotrexate-R-alanine by carboxypeptidase A-monoclonal antibody conjugate. Biochemistry 31, 891-897. (17) Senter, P. D., Su, P. C. D., Katsuragi, T., Sakai, T., Cosand, W. L., Hellstrom, I., and Hellstrom, K. E., (1991) Generation of 5-fluorouracil from 5-fluorocytosine by monoclonal antibodycytosine deaminase conjugates. Bioconjugate Chem. 2, 447451. (18) Springer, C. J., Bagshawe, K. D., Sharma, S. K., Searle, F., Boden, J., Antoniw, P., et al. (1991) Ablation of human choriocarcinoma xenografts in nude mice by antibody-directed enzyme prodrug therapy (ADEPT) with three novel compounds. Eur. J. Cancer 27(11), 1361-1366. (19) Smith, G. K., Banks, S., Blumenkopf, T. A., Cory, M., Humphreys, J., Laethem, R. M., Miller, J., Moxham, C. P., Mullin, R., Ray, P. H., Walton, L. M., and Wolfe, L. A. (1997) Toward Antibody-directed enzyme prodrug therapy with the T268G mutant of human carboxypeptidase A1 and novel in vivo stable prodrugs of MTX. J. Biol. Chem. 272(25), 1580415816. (20) Pendergast, W., Dickeerson, S. H., Dev, I. K., Ferone, R., Duch, D. S., and Smith, G. K. (1994) Benzo[f]quinazoline inhibitors of thymidylate synthase: Methyleneamino-linked aroylglutamate derivitives. J. Med. Chem. 37(6), 838-844. (21) Duch, D. S., Banks, S., Dev, I. K., Dickerson, S. H., Ferone, R., Heath, L. S., Humphreys, J., Knick, V., Pendergast, W., Singer, S., Smith, G. K., Waters, K., and Wilson, H. R. (1993) Biochemical and cellular pharmacology of 1843U89, a novel

Wolfe et al. benzoquinazoline inhibitor of thymidylate synthase. Cancer Res. 53, 810-818. (22) Smith, G. K., Amyx, H., Boytos, C. M., Duch, D. S., Ferone, R., and Wilson, H. R. (1995) Enhanced antitumor activity for the thymidylate synthase inhibitor for 1843U89 through decreased host toxicity with oral folic acid. Cancer Res. 55, 6117-6125. (23) Pendergast, W., Dickerson, S., Dev, I., Ferone, R., Duch, D., and Smith, G. (1992) Benzoquinazoline inhibitors of thymidylate synthase: effect on cytotoxicity and thymidylate synthase activity of variation of side chain structure and C3 substitution. Proc. Am. Assoc. Cancer Res. 33, 407. (24) Banks, S. D., Waters, K. A., Barrett, L. L., and Smith, G. K. (1992) Destruction of WiDr multicellular tumor spheroids with novel thymidylate synthase inhibitor at physiological thymidine concentrations. Proc. Am. Assoc. Cancer Res. 33, 407. (25) Wilson, H. R., Heath, L. S., Knick, V. C., Koszalka, G. W., and Ferone, R. (1992) In vivo antitumor activity of GW1843, a new antifolate thymidylate synthase inhibitor. Proc. Am. Assoc. Cancer Res. 33, 407. (26) Vitols, K. s., Haag-Zeino, B., Baer, T., Montejano, Y. D., and Huennekens, F. M. (1995) Methotrexate-R-phenylalanine: Optimization of methotrexate prodrug for activation by carboxypeptidase A-monoclonal antibody conjugate. Cancer Res. 55, 478-481. (27) Rosenberg, I. H., and Selhub, J. (1986) Intestinal Absorption of Folates. in Folates and Pterins (Blakley, R. L., and Whitehead, V. M., Eds.) Vol. 3, pp 148-176, Wiley-Interscience, New York. (28) Lituinov, S. V., Velders, M. P., Bakker, H. A., Fleuren, G. J., and Warnaar, S. O. (1994) Ep-CAM: a human epithelial antigen is a homophilic cell-cell adhesion molecule. J. Cell Biol. 125(2), 437-446. (29) Kidera, A., Inaka, K., Matsushima, M., Go, N. (1994) Response of dynamic structure to removal of a disulfide bond: normal mode refinement of C77A/C95A mutant of human lysozyme. Protein Sci. 3 (1), 92-102. (30) Gardell, S. J., Craik, C. S., Clauser, E., Goldsmith, E. J., Stewart, C. B., Graf, M., and Rutter, W. J. (1988) A novel rat carboxypeptidase, CPA2: characterization, molecular cloning, and evolutionary implications on substrate specificity in the carboxypeptidase gene family. J. Biol. Chem. 263(33), 1782817836. (31) Laethem, R. M., Blumenkopf, T. A., Cory, M., Elwell, L., Moxham, C. P., Ray, P. H., Walton, L. M., and Smith, G. (1996) Expression and characterization of human pancreatic preprocarboxypeptidase A1 and preprocarboxypeptidase A2. Arch. Biochem. Biophys. 332, 8-18. (32) Cueni, L. B., Bazzone, T. J., Riordan, J. F., and Vallee, B. L. (1980) Affinity chromatographic sorting of carboxypeptidase A and its chemically modified derivatives. Anal. Biochem. 107, 341-349. (33) Vitols, K. S., Hanseler, E., Montejano, Y., Muller, B. M., Reisfeld, R. A., and Huennekens, F. M. (1992) Enhanced cytotoxicity of methotrexate alpha-peptide prodrugs for UCLAP3 cells treated with enzyme-conjugated monoclonal antibody KS1/4. Antibody, Immunoconjugates, Radiopharm. 5(1), 140. (34) Sirotnak, F. M., DeGraw, J. I. (1984) Selective antitumor action of folate analogues. in Folate Antagonist as Therapeutic Agents (Sirotnak, F. M., Burchall, J. J., Ensminger, W. B., and Montgomery, J. A., Eds.) Vol 2, pp 47-54, Academic Press, Inc., Orlando. (35) Jackman, A. L., Taylor, G. A., Calvert, A. H., and Harrap, K. R. (1984) Modulation of anti-metabolite effects: Effects of thymidine on the efficacy of the quinazoline-based thymidylate synthetase inhibitor, CB3717. Biochem. Pharmacol. 33, 3269-3275.

BC980057Z