Novel Acid Labile COL1 Trityl-Linked Difluoronucleoside

Bioconjugate Chem. 1996, 7, 497−510

497

Novel Acid Labile COL1 Trityl-Linked Difluoronucleoside Immunoconjugates: Synthesis, Characterization, and Biological Activity1 Vinod F. Patel,* Julie N. Hardin, John M. Mastro, Kevin L. Law, John L. Zimmermann, William J. Ehlhardt, Joseph M. Woodland, and James J. Starling Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, Indiana 46285. Received October 5, 1995X

LY207702 (1) is a difluorinated purine nucleoside that exhibits impressive antitumor activity in preclinical models. This agent, however, also possesses cardiotoxicity which limits the potential clinical utility of this novel drug candidate. We therefore developed linker chemistry whereby regioselective N6-tritylation of LY207702 (1) allowed this drug to be coupled to -lysine amino groups of mAb’s reactive with human tumor-associated antigens. The resulting immunoconjugates 3 possessed conjugation ratios ranging from 5 to 7 mol of LY207702/mol of mAb, minimal aggregate content (510%), and good immunoreactivity. The electronic nature of substituents on the aromatic rings of the trityl group dictated the degree of acid lability of the trityl linker. Increased electronic stabilization of the transient trityl carbocation led to increase in the release rate of free drug, i.e., m-DMT 10a ) p-DMT 10b > p-MMT 10d > p-T 10f. Consequently, the more acid labile DMT conjugates 3a and 3b proved to be the most potent cytotoxic agents, and the most stable p-T conjugate 3f exhibited the least antitumor activity when evaluated in vitro and in vivo. p-MeT-linked conjugate 3e, the most stable construct that retained excellent in vivo antitumor activity, was selected for more extensive evaluation. No detectable free drug or metabolite was observed in mouse plasma at a single intravenous dose of p-MeT conjugate 3e, which was consistent with its predicted stability under physiological conditions. This construct did, however, exhibit significant antigen-mediated antitumor activity in vivo. No cardiotoxicity was detected in mice dosed with conjugate 3e (6 mg/kg free drug content per day for 21 days) equivalent to ∼8 times the total dose required for complete regression of well-established (∼1 g) HC1 human colon tumor xenografts in nude mice. Cardiotoxicity was induced in 20% of free drug 1 treated group at the equivalent dose. Cardiomyopathy was, however, observed when the dose of conjugate 3e was increased to 8 mg/kg per day for 21 days. These data suggest that antitumor activity of LY207702 (1) was maintained and its cardiotoxic potential reduced when this agent was administered to human tumor xenograft bearing nude mice as COL1-N6-p-MeT-207702 conjugate 3e.

INTRODUCTION

A continuing challenge in the design of oncolytics for the treatment of human solid tumors is to enhance the cytotoxic effect and the selectivity of the agent toward tumor cells. A myriad of agents have been identified and shown to possess potent cytotoxicity against tumor cells; however, most of these agents are limited in their use due to the lack of selectivity toward tumor cells versus normal cells (1). Traditionally, structure-activity relationships (SAR)2 have been employed in an attempt to overcome this problem, but this approach is not always successful. Conversion of an active drug into an inactive prodrug (2), which selectively dissociates to the active component at the tumor site, provides an alternative method of enhancing selectivity and is more generally applicable to the structurally diverse classes of antitumor agents. An extension of this concept whereby the drug is covalently attached to a tumor specific monoclonal antibody (mAb) provides an alternative method of delivering drug to the tumor site (3). The ability to attach * Author to whom correspondence should be addressed: Lilly Research Laboratories (Drop Code 0540), Lilly Corporate Center, Indianapolis, IN 46285. Tel: (317) 276-9582. Fax: (317) 277-3652. E-mail PATEL VINOD [email protected]. X Abstract published in Advance ACS Abstracts, July 1, 1996. 1 An abstract of this work was published in the Proceedings of the 86th Annual Meeting of the American Association for Cancer Research, Toronto, Canada, March 1995.

S1043-1802(96)00038-9 CCC: $12.00

numerous drug molecules on each antibody also presents a unique opportunity for selective dose intensification of drug at the tumor. In general, drug immunoconjugates are comprised of three distinct components: (i) a cytotoxic drug, (ii) a tumor specific mAb for targeting, and (iii) the linker for attachment of the drug to the antibody. The role of each of these entities is critical in defining the overall properties of the drug immunoconjugate. Drug immunoconjugates have attracted much attention in recent years as agents for cancer therapy; however, much of the research has focused on optimization of the drug and the monoclonal antibody (4). The ability of the linker to release active drug in a timely and predictable fashion is also of paramount importance in the overall design of drug immunoconjugates but has received less attention (5). 2 Abbreviations: AcCl, acetyl chloride; AcOH, acetic acid; AST, aspartate transaminase; ALT, alanine transaminase; CEA, carcinoembryonic antigen; CR, conjugation ratio; DCC, dicyclohexylcarbodiimide; m-DMT, 3-carboxy-4′,4′′-dimethoxytrityl; pDMT, 4-carboxy-4′,4′′-dimethoxytrityl; p-MMeT, 4-carboxy-4′methoxy-4′′-methyltrityl; p-MMT, 4-carboxy-4′-methoxytrityl; p-MeT, 4-carboxy-4′-methyltrityl; p-T, 4-carboxytrityl; DMF, dimethylformamide; EtOAc, ethyl acetate; IgG, immunoglobulin G; iPrNH2, isopropylamine; iPA, isopropylamide; iPrOH, isopropyl alcohol; mAb, monoclonal antibody; MIR, molar input ratio; NHS, N-hydroxysuccinimide; PBS, phosphate-buffered saline; SAR, structure-activity relationship; TBAPC, tetra-nbutylammonium perchlorate; THF, tetrahydrofuran.

© 1996 American Chemical Society

498 Bioconjugate Chem., Vol. 7, No. 4, 1996

For drug immunoconjugates to selectively destroy tumor cells, it is essential that the release of cytotoxic drug from the conjugate should occur while the antibody is bound to its antigen; however, the construct should remain stable during systemic circulation. For noninternalizing antibodies the extracellular acidic tumor environment (6) and for internalizing antibodies translocation to the acidic lysosomal compartment (7) provide convenient trigger mechanisms for drug release from acid labile conjugates. This rationale has led to the development of several acid labile linkers, including cis-aconityl (8), hydrazone (9), hydrazide (10), and acetal (11) groups, for drug immunoconjugates. The trityl group is a wellknown acid labile protecting group widely used in organic synthesis (12). Variations in acid lability can be readily achieved by appropriately substituting the aromatic rings of the trityl moiety (12). The versatile triphenylmethyl group has also found applications in solid phase synthesis of oligonucleotides (13) and peptides (14). It was apparent to us that the trityl moiety would be a particularly attractive linker candidate for use in the construction of drug immunoconjugates as the aromatic rings provide sites for appending substituents that allow (a) the control of drug release and (b) attachment of the mAb. Furthermore, the ability to readily attach a trityl group to a variety of pendant nucleophilic functionalities suggested that drug immunoconjugates could be prepared with a range of structurally diverse antitumor agents. In this paper we report on the synthesis, characterization, and biological properties of novel acid labile COL1N6-trityl-207702 conjugates 3 for their potential use in cancer therapy. COL1 is a murine IgG2a mAb, isolated by Schlom and co-workers (15) that binds to an epitope of CEA that is highly expressed on tumor cell membranes and is only minimally present in normal tissue. COL1 did not appear to be internalized after binding to LS174T colon tumor cells as determined by acid dissociation of 125 I-labeled antibody (16). LY207702 (1) is a member of the 2′,2′-difluoro-2′-deoxyribofuranosylpurine class of nucleoside antimetabolites which exhibits exquisite and unique antitumor activity in preclinical human tumor xenograft models, but also possesses cardiotoxicity (17). Our objective, therefore, was to determine if COL1-N6trityl-207702 conjugates 3 could maintain the antitumor activity of LY207702 (1) while displaying less cardiotoxicity than the free drug. EXPERIMENTAL PROCEDURES

All reactions described herein were performed under an inert atmosphere of dry nitrogen in flame-dried glassware unless otherwise noted. Tetrahydrofuran was freshly distilled from sodium benzophenone ketyl for use. All other solvents and reagents were used as supplied unless stated otherwise. In those cases where solvents and reagents were dried and purified, literature procedures were followed. Organic extracts were dried over Na2SO4 and concentrated in vacuo at room temperature. Melting points were determined on a 6427-H10 ThomasHoover melting point apparatus and are uncorrected. NMR spectra were obtained on a General Electric QE300 instrument at 300 MHz for 1H and 74.48 Hz for 13C in various solvents (specified) and are reported as parts per million (δ) values downfield from tetramethylsilane. Multiplicities of resonances are described as broad (b), singlet (s), doublet (d), triplet (t), quartet (q), or multiplet (m). IR spectra were run as KBr pellets or CHCl3 solutions on a Nicolet 510P FT-IR. Optical rotations were measured on a Perkin-Elmer 241 spectrometer, and UV spectra were obtained on a Shimadzu 2101PC instrument. The mass spectral data were obtained on either a

Patel et al.

VG-70SE or a Varian MAT-731 (for FD) and a VGZAB2SE or VG-ZAB3F (for FAB) spectrometer. “Flash chromatography” was performed on silica gel 60 230400 mesh (40-63 µm) according to the procedure of Still et al. (18). TLC was carried out on E. Merck silica gel 60 F254 plates. Analytical HPLC separations were performed on a LDC system with a CM 4000 multiple solvent delivery system, spectroMonitor 5000 photodiode array detector set to record at 254 nm, a Hitachi D-2500 chromato-Integrator and a Gilson FC204 fraction collector. A µPorasil 10 µm column (3.9 × 150 mm) was used for normal phase (NP) and a Nova-Pak HR C18 6 µm column (3.9 × 150 mm) for reverse phase (RP). A mobile phase of EtOAc:n-hexanes for NP and acetonitrile:ammonium acetate(aq) for RP was used at a flow rate of 1.0 mL/min. All solvents were filtered through a 0.22 µm membrane and then thoroughly degassed under vacuum. Boric acid (12.37 g) was dissolved in Milli-Qwater (2 L) and the pH adjusted to 8.60 with 50% aqueous sodium hydroxide to give a 0.1 M borate buffer solution. Dulbecco’s PBS powder (GIBCO) was dissolved in Milli-Q-water (1 L) to give a 0.1 M solution of pH 7.40 PBS. Monoclonal antibodies were purified on a protein-A column, and fractions containing protein were concentrated to ∼15-25 mg/mL. mAb was dialyzed in a molecular porous membrane (Spectra/Por, MW cutoff ) 12000-14000, capacity ) 2 mL/cm) tubing against 0.1 M pH 8.60 borate buffer (×100 volume) at 4 °C for 2030 h and sterile filtered, and then the protein concentration was determined. Centrifugation was performed in a IEC Clinical centrifuge at 4000 rpm (2000G). Protein solutions were concentrated in an Amicon stirred-cell (series 8000) containing a disc membrane (MW cutoff 30000 and presoaked in water to remove azide) at 55 psi nitrogen pressure. Protein chromatography was performed using a FPLC Pharmacia LKB system equipped with a p-500 Pump, LCC-500 plus Controller, 100 Fraction collector, and a MII-UV monitor. Gel filtration purification was performed using a G-25 Sephadex medium (bead diameter 50-150 µm dry) column eluting with PBS. Analytical runs were performed on a Superose 12 HR10/30 column eluting with 15% acetonitrile:PBS at a flow rate of 0.8 mL/min. Peak integrations were determined by area under the curve, where AUC ) 1/2width × height. UV measurements were performed on a Beckmann DU-70 spectrophotometer. A 1 cm path length UV cell was used to measure absorbance of the product diluted in PBS. LY207702 (1) was prepared according to known procedures (19). All other chemicals were obtained from commercial sources. Physical chemical data were obtained at Lilly Research Laboratories. 4,4′-Methoxymethylbenzophenone (5c) (20). A solution of p-toluoyl chloride (1.55 g) in dry CH2Cl2 (3 mL) was added to a suspension of AlCl3 (4 g) in dry CH2Cl2 (20 mL) at 0 °C and the mixture stirred at 0 °C for 1 h. A solution of anisole (1.08 g) in dry CH2Cl2 (3 mL) was added dropwise at a rate such that the temperature was maintained 0 °C. The orange mixture was stirred for a further 2 h at 0 °C and quenched with cold 1 N aqueous HCl (10 mL). Organics were separated, washed with saturated aqueous NaHCO3 (50 mL) and H2O (50 mL), dried, concentrated and purified by column chromatography (10% EtOAc:hexanes) to provide benzophenone 5c as a white solid (1.85 g, 82%): mp 88.5-89.5 °C; IR (KBr) νmax 1645, 1596, 1260, 760 cm-1; 1H (CDCl3) δ 7.82 (d, J ) 8 Hz, 2 ArH), 7.68 (d, J ) 8 Hz, 2 ArH), 7.26 (d, J ) 8 Hz, 2 ArH), 6.97 (d, J ) 8 Hz, 2 ArH), 3.90 (s, OMe), 2.44 (s, Me) ppm. Anal. Calcd for C15H14O2 requires M 226; C, 79.62; H, 6.24. Found: m/z 226 (100); C, 79.82; H, 6.35.

COL1 Trityl Difluoronucleoside Immunoconjugates

Tetra-n-butylammonium Perchlorate (TBAPC). A solution of KClO4 (1.38 g) in H2O (40 mL) was added to a solution of tetra-n-butylammonium sulfate (3.39 g) in H2O (15 mL). A white floculent precipitate immediately formed and was filtered, washed with H2O, and dried over phosphorus pentoxide to yield the desired product as a white powder (1.61 g, 48%): IR (KBr) νmax 2967, 2879, 1475, 1093, 623 cm-1; 1H (CDCl3) δ 3.26-3.20 (m, 8H, 4 CH2N), 1.70-1.59 (m, 8H, 4 CH2), 1.44 (dq, J ) 6 and 8 Hz, 8H, 4 CH2), 1.01 (t, J ) 8 Hz, 6H, 2 Me) ppm. Anal. Calcd for C16H36NO4Cl requires M 341.5; C, 56.20; H, 10.61; N, 4.10; Cl, 10.37. Found MS-FAB m/z 242.4 (M - ClO4+, 100); C, 56.36; H, 10.43; N, 3.96; Cl, 10.25. General Procedure A. A suspension of flame-dried magnesium turnings (48 mmol) in dry THF (25 mL) was sonicated for 20 min, and neat aryl bromide 4 (40 mmol) was added, followed by addition of a catalytic amount of iodine (10 mg), and the mixture was sonicated for 30 min to form the Grignard reagent. A suspension of the benzophenone 5 (44 mmol) in dry THF (10 mL) was added to the Grignard solution, and the mixture was stirred at room temperature for 2.5 h. The reaction was quenched with 5% aqueous KHSO4 (50 mL) and extracted with EtOAc (3 × 20 mL). Combined organics were dried, concentrated, and purified by column chromatography (gradient elution 25-75% EtOAc:hexanes) to give carbinol 6. Racemic 2-[4-[(4-Methylphenyl)phenylhydroxymethyl]phenyl]-4,4-dimethyl-1,3-oxazoline (6e). 4Methylbenzophenone (5e) (10.53 g) was reacted with the Grignard derived from 2-(4-bromophenyl)-4,4-dimethyl1,3-oxazoline (4) (12.40 g), according to general procedure A, to give carbinol 6e as a white solid (18.20 g, 99%): mp 68-72 °C; IR (KBr) νmax 2971, 1647, 1274 cm-1; UV (EtOH) λmax 251.5 ( ) 19 295), 204.5 ( ) 59 991) nm; 1 H (CDCl3) δ 7.94 (br d, J ) 8 Hz, 2 ArH), 7.38 (d, J ) 8 Hz, 2 ArH), 7.36-7.24 (m, 5 ArH), 7.13 (s, 4 ArH), 4.15 (s, OCH2), 2.80 (br s, OH), 2.36 (s, Me), 1.42 (s, 6H, 2 Me) ppm. Anal. Calcd for C25H25NO2 requires M 372; C, 80.83; H, 6.78; N, 3.77. Found: m/z 372 (100); C, 80.55; H, 7.01; N, 3.73. General Procedure B. Oxazoline 6 (22.4 mmol) was dissolved in 80% aqueous glacial AcOH (80 mL) and heated at reflux for 21 h. The solution was concentrated (50 °C) to an oily residue which was redissolved in 20% (w/v) NaOH in ethanol/water (1:1, 70 mL) and reheated at reflux for 3 h. The reaction solution was concentrated, diluted with water (50 mL), and acidified with 1 N aqueous HCl to pH 3. Product was extracted with EtOAc (4 × 10 mL), and the combined organics were dried and concentrated to give carboxylic acid 7. Racemic 4-Carboxy-4′-methyltriphenylhydroxymethane (7e). Oxazoline 6e (18.0 g) was hydrolyzed according to general procedure B to give acid 7e as a yellow solid (13.35 g, 87%): mp 91-94 °C; IR (CHCl3) νmax 3380, 3020, 1692, 1282, 1009 cm-1; UV (EtOH) λmax 247 ( ) 12 558), 203.5 ( ) 45 569) nm; 1H (CDCl3) δ 8.03 (d, J ) 8 Hz, 2 ArH), 7.44 (d, J ) 8 Hz, 2 ArH), 7.35-7.23 (m, 5 ArH), 7.13 (s, 4 ArH), 2.36 (s, Me) ppm. Anal. Calcd for C21H18O3 requires M 318.4; C, 79.23; H, 5.70. Found: m/z 318 (100); C, 78.87; H, 6.03. General Procedure C. To a stirred suspension of arene carboxylic acid 7 (18.1 mmol) and NHS (18.1 mmol) in dry CH2Cl2 (100 mL) was added DCC (21.7 mmol) at room temperature and the mixture stirred for 4 h. The reaction mixture was filtered through Celite, washed with CH2Cl2, concentrated, and purified by column chromatography (gradient elution 25-50% EtOAc:hexanes) to give active ester 8.

Bioconjugate Chem., Vol. 7, No. 4, 1996 499

Racemic N-Succinimidyl 4-[(4-Methylphenyl)phenylhydroxymethyl]benzoate (8e). Carboxylic acid 7e (13.20 g) was coupled to NHS (4.77 g) using DCC (10.27 g), according to general procedure C, to give active ester 8e as a solid (15.65 g, 91%): mp 156-158 °C; IR (KBr) νmax 3467, 1771, 1731, 1212, 996 cm-1; UV (EtOH) λmax 249 ( ) 18 105), 206 ( ) 48 790) nm; 1H (DMSOd6) δ 8.03 (d, J ) 9 Hz, 2 ArH), 7.50 (d, J ) 9 Hz, 2 ArH), 7.36-7.26 (m, 3 ArH), 7.23-7.19 (m, 2 ArH), 7.16-7.06 (br q, J ) 8 Hz, 4 ArH), 2.89 (s, 4H, 2 CH2), 2.30 (s, Me) ppm. Anal. Calcd for C25H21NO5 requires M 415; C, 72.28; H, 5.10; N, 3.37. Found: m/z 415 (100); C, 71.95; H, 5.26; N, 4.06. General Procedure D. Hydroxy ester 8 (4.98 mmol) was dissolved in freshly distilled AcCl (25 mL) and the solution heated at reflux for 5-24 h, cooled to room temperature, and concentrated to approximately onethird of its original volume. Dry Et2O (20 mL) was added to form a precipitate, which was filtered, washed with further dry Et2O, and dried under vacuum to give trityl chloride 9. In those cases where product could not be precipitated, the reaction solution was concentrated to dryness and the product 9 azeotroped with dry CH2Cl2 (5 × 10 mL). Racemic N-Succinimidyl 4-[(4-Methylphenyl)phenylchloromethyl]benzoate (9e). Hydroxy ester 8e (2.50 g) was heated in AcCl according to general procedure D to provide trityl chloride 9e as a yellow solid (2.35 g, 90%): IR (CHCl3) νmax 1775, 1744, 1257, 1201, 1071, 996 cm-1; 1H (DMSO-d6) δ 8.04 (d, J ) 8 Hz, 2 ArH), 7.50 (d, J ) 8 Hz, 2 ArH), 7.36-7.25 (m, 2 ArH), 7.257.17 (m, 3 ArH), 7.12 (q, J ) 8 Hz, 4 ArH), 2.88 (s, 4 H, 2 CH2), 2.28 (s, Me) ppm. Anal. Calcd for C25H20NO4Cl requires M 433.5; C, 69.21; H, 4.65; N, 3.23. Found: m/z 415 (100); C, 69.50; H, 4.79; N, 3.47. General Procedure E. To a solution of 2,6-diamino9-(2′-deoxy-2′,2′-difluoro-β-D-ribofuaranosyl)purine (1) (0.331 mmol) in a solvent mixture of dry DMF:CH2Cl2 (1:1; 6 mL) was added TBAPC (0.397 mmol) and dry collidine (0.497 mmol) at room temperature. The solution was stirred for 5 min and trityl chloride 9 (0.331 mmol) added portionwise over ∼3 min. The resulting solution was stirred at room temperature until all the starting material was consumed (monitored by TLC). The crude reaction was diluted with EtOAc (1 volume) and the solution directly applied onto a silica column and purified by chromatography (gradient elution EtOAc then 5-20% iPrOH/EtOAc) to give trityl nucleoside 2. N-Succinimidyl 3-[Bis(4-methoxyphenyl)[[2-amino9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzoate (2a). Nucleoside 1 (100 mg) was reacted with m-DMT chloride 9a (159 mg) in the presence of TBAPC (136 mg) and collidine (70 µL) for 20 min, according to general procedure E, to give product 2a as a white solid (190 mg; 77%): TLC Rf (10% MeOH/ CH2Cl2) 0.42; IR (KBr) νmax 3379, 1740, 1606, 1253, 1072 cm-1; UV (EtOH) λmax 278 ( ) 9970.5), 225.5 ( ) 32 651), 203.5 ( ) 50 143) nm; 1H (DMSO-d6) δ (8.07 (s, 1 ArH), 7.90-7.77 (m, 2 ArH), 7.88 (s, H8), 7.50 (t, J ) 9 Hz, 1 ArH), 7.27 (dd, J ) 4 and 10 Hz, 4 ArH), 6.95 (s, NH), 6.80 (dd, J ) 6 and 10 Hz, 4 ArH), 6.74 (br s, NH2), 6.24 (d, J ) 6 Hz, 3′OH), 5.60-5.44 (m, H1′), 5.19 (t, J ) 4 Hz, 5′OH), 4.41-4.23 (m, H3′), 3.82-3.54 (m, 3H, H4′5′5′), 3.70 (s, 6H, 2 OMe), 2.84 (s, 4 H, 2 CH2) ppm. Anal. Calcd for C36H33N7O9F2 requires M 746; C, 57.99; H, 4.40; N, 13.15. Found: m/z 746 (100), 630 (10), 444 (15), 302 (10); C, 57.81; H, 4.73; N, 12.89. N-Succinimidyl 4-[Bis(4-methoxyphenyl)[[2-amino9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzoate (2b). Nucleoside 1 (100 mg)


was reacted with p-DMT chloride 9b (159 mg) in the presence of TBAPC (136 mg) and collidine (70 µL) for 30 min, according to general procedure E, to give product 2b as a white powder (232 mg; 94%): TLC Rf (10% MeOH/CH2Cl2) 0.22; IR (KBr) νmax 3372, 2933, 1741, 1608, 1510, 1072 cm-1; UV (EtOH) λmax 261 ( ) 22 597), 226.5 ( ) 43 631) nm; 1H (DMSO-d6) δ 7.95 (d, J ) 10 Hz, 2 ArH), 7.87 (s, H8), 7.62 (d, J ) 8 Hz, 2 ArH), 7.27 (d, J ) 8 Hz, 4 ArH), 6.91 (s, NH), 6.80 (dd, J ) 3 and 10 Hz, 4 ArH), 6.72 (br s, NH2), 6.25 (d, J ) 6 Hz, 3′OH), 5.65-5.44 (m, H1′), 5.19 (t, J ) 6 Hz, 5′OH), 4.44-4.26 (m, H3′), 3.83-3.63 (m, 3H, H4′5′5′), 3.70 (s, 6H, 2 OMe), 2.85 (s, 4H, 2 CH2) ppm. Anal. Calcd for C36H33N7O9F2 requires M 746; C, 57.99; H, 4.46; N, 13.15. Found: m/z M 746 (100), 444 (12), 302 (3); C, 58.05; H, 4.82; N, 12.80. N-Succinimidyl 4-[(4-Methoxyphenyl)(4′-methylphenyl)[[2-amino-9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzoate (2c). Nucleoside 1 (200 mg) was reacted with p-MMeT chloride 9c (307 mg) in the presence of TBAPC (271 mg) and collidine (132 µL) for 2 h, according to general procedure E, to give product 2c as a yellow solid (404 mg; 84%): diastereomeric mixture; TLC Rf (10% MeOH/CH2Cl2) 0.49; IR (KBr) νmax 3376, 1772, 1740, 1606, 1256, 1073 cm-1; UV (EtOH) λmax 259 ( ) 26 254) nm; 1H (DMSOd6) δ 7.95 (d, J ) 8 Hz, 2 ArH), 7.88 (s, H8), 7.62 (d, J ) 10 Hz, 2 ArH), 7.30-7.22 (m, 4 ArH), 7.05 (dd, J ) 2 and 8 Hz, 2 ArH), 6.92 (s, NH), 6.79 (dd, J ) 8 and 10 Hz, 2 ArH), 6.74 (br s, NH2), 6.25 (d, J ) 6 Hz, 3′OH), 5.64-5.43 (m, H1′), 5.20 (t, J ) 6 Hz, 5′OH), 4.44-4.24 (m, H3′), 3.84-3.54 (m, 3H, H4′5′5′), 3.72 (s, OMe), 2.87 (s, 4H, 2 CH2), 2.23 (s, Me) ppm. Anal. Calcd for C36H33N7O8F2 requires M 730; C, 59.26; H, 4.56; N, 13.44. Found: m/z 730 (100), 428 (5), 302 (2); C, 59.20; H, 4.84; N, 13.48. N-Succinimidyl 4-[(4-Methoxyphenyl)phenyl[[2amino-9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzoate (2d). Nucleoside 1 (302 mg) was reacted with p-MMT chloride 9d (450 mg) in the presence of TBAPC (271 mg) and collidine (160 µL) for 2 h, according to general procedure E, to give product 2d as a pale yellow powder (493 mg; 70%): diastereomeric mixture; TLC Rf (10% MeOH/CH2Cl2) 0.34; IR (KBr) νmax 3374, 1739, 1607, 1511, 1256, 1207, 1073 cm-1; 1H (DMSO-d6) δ 7.95 (d, J ) 8 Hz, 2 ArH), 7.87 (s, H8), 7.64 (d, J ) 1 Hz, 2 ArH), 7.38 (d, J ) 8 Hz, 2 ArH), 7.30-7.13 (m, 5 ArH), 7.02 (s, NH), 6.80 (dd, J ) 2 and 7 Hz, 2 ArH), 6.75 (br s, NH2), 6.25 (d, J ) 6 Hz, 3′OH), 5.60-5.44 (m, H1′), 5.20 (t, J ) 6 Hz, 5′OH), 4.4.2-4.26 (m, H3′), 3.82-3.52 (m, 3H, H4′5′5′), 3.70 (s, OMe), 2.87 (s, 4H, 2 CH2) ppm; C35H31N7O8F2 requires M 716, found m/z 716 (100). N-Succinimidyl 4-[(4-Methylphenyl)phenyl[[2amino-9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzoate (2e). Nucleoside 1 (200 mg) was reacted with p-MeT chloride 9e (287 mg) in the presence of TBAPC (271 mg) and collidine (132 µL) for 3 h, according to general procedure E, to give product 2e as a white solid (604 mg; 65%): diastereomeric mixture; TLC Rf (10% MeOH/CH2Cl2) 0.50; IR (KBr) νmax 3379, 1773, 1740, 1606, 1467, 1205, 1073 cm-1; 1 H (DMSO-d6) δ 7.96 (br d, J ) 6 Hz, 2 ArH), 7.86 (s, H8), 7.64 (d, J ) 10 Hz, 2 ArH), 7.42-7.35 (m, 2 ArH), 7.30-7.12 (m, 5 ArH), 7.10-7.02 (m, 2 ArH), 7.14 (s, NH), 6.76 (br s, NH2), 6.26 (d, J ) 7 Hz, 3′OH), 5.58-5.45 (m, 3H, H4′5′5′), 2.87 (br s, 4H, 2 CH2), 2.24 (s, Me) ppm. Anal. Calcd for C35H31N7O7F2 requires M 699; C, 60.08; H, 4.47; N, 14.01; F, 5.43. Found: m/z 699 (100); C, 59.89; H, 4.30; N, 13.82; F, 4.99.

Patel et al.

N-Succinimidyl 4-[Bisphenyl[[2-amino-9-(2′-deoxy2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzoate (2f). Nucleoside 1 (302 mg) was reacted with p-T chloride 9f (402 mg) in the presence of TBAPC (401 mg) and collidine (160 µL) for 1 h, according to general procedure E, to give product 2f as a white solid (391 mg; 57%): TLC Rf (10% MeOH/CH2Cl2) 0.27; IR (CHCl3) νmax 3379, 1772, 1739, 1607, 1467, 1207, 1073 cm-1; UV (EtOH) λmax 259 ( ) 24 723) nm; 1H (DMSOd6) δ 7.97 (d, J ) 9 Hz, 2 ArH), 7.88 (s, H8), 7.66 (d, J ) 9 Hz, 2 ArH), 7.42-7.36 (m, 4 ArH), 7.28-7.10 (m, 6 ArH), 7.04 (s, NH), 6.76 (br s, NH2), 6.15 (d, J ) 6 Hz, 3′OH), 5.60-5.44 (m, H1′), 5.20 (t, J ) 6 Hz, 5′OH), 4.404.24 (m, H3′), 3.80-3.54 (m, 3H, H4′5′5′), 2.88 (s, 4H, 2 CH2) ppm; C34H29N7O7F2 requires M 685, found m/z 685 (100). General Procedure F. To a stirred solution of active ester 2 (0.147 mmol) in dry CH2Cl2 (3 mL) was added neat iPrNH2 (0.734 mmol) at room temperature, and a precipitate formed almost immediately. The reaction mixture was stirred for a further 1 h, filtered, concentrated, and purified by column chromatography (gradient elution CH2Cl2-10% MeOH/CH2Cl2) to provide amide 10. N-Isopropyl-3-[bis(4-methoxyphenyl)[[2-amino-9(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzamide (10a). Active ester 2a (30 mg) was reacted with iPrNH2, according to general procedure F, to give amide 10a as a white solid (20.6 mg; 74%): TLC Rf (10% MeOH/CH2Cl2) 0.35; IR (CHCl3) νmax 3365, 2931, 1608, 1509, 1465, 1251, 1178, 1034, 830, 592 cm-1; UV (EtOH) λmax 277.5 ( ) 14 006), 224.5 ( ) 47 152), 205 ( ) 70 300) nm; 1H NMR (acetone-d6) δ 8.03 (s, ArH), 7.83 (s, H8), 7.65 (d, J ) 8 Hz, ArH), 7.52 (d, J ) 8 Hz, ArH), 7.48 (d, J ) 9 Hz, ArH), 7.26 (d, J ) 8 Hz, 4 ArH), 6.80 (d, J ) 8 Hz, 4 ArH), 6.35 (s, NH), 6.20 (br s, NH2), 5.76-5.71 (m, H1′), 5.55 (br s, 3′OH), 4.82 (br s, 5′OH), 4.62-4.57 (m, H3′), 4.20-4.12 (m, CH), 3.93-3.74 (m, 3H, H4′5′5′), 3.74 (s, 6H, 2 OMe), 1.14 (J ) 6.6 Hz, 6H, 2 Me) ppm; C35H37N7O6F2 requires M 690, found m/z 690 (100). N-Isopropyl-4-[bis(4-methoxyphenyl)[[2-amino-9(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzamide (10b). Active ester 2b (60 mg) was reacted with iPrNH2, according to general procedure F, to give amide 10b as a white solid (38 mg; 69%): TLC Rf (10% MeOH/CH2Cl2) 0.41; IR (KBr) νmax 3356, 1608, 1509, 1464, 1251, 1178, 1033 cm-1; UV (EtOH) λmax 227.5 ( ) 47 461), 204 ( ) 79 092) nm; 1H (CDCl3/MeOD) δ 7.83 (s, H8), 7.62 (d, J ) 8 Hz, 2 ArH), 7.40 (d, J ) 8 Hz, 2 ArH), 7.19 (d, J ) 8 Hz, 4 ArH), 6.74 (d, J ) 8 Hz, 4 ArH), 5.53-5.45 (m, H1′), 4.65-4.55 (m, 2H, CH, H3′), 4.35-4.15 (m, 3H, H4′5′5′), 3.74 (s, 6H, 2 OMe), 1.19 (d, J ) 7 Hz, 6H, 2 Me) ppm; C35H37N7O6F2 requires M 689, found m/z 689 (10), 389 (30), 302 (25). N-Isopropyl-4-[(4-methoxyphenyl)(4′-methylphenyl)[[2-amino-9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzamide (10c). Active ester 2c (50 mg) was reacted with iPrNH2, according to general procedure F, to give amide 10c as a white solid (43.6 mg; 95%): TLC Rf (10% MeOH/CH2Cl2) 0.41; IR (KBr) νmax 3360, 1607, 1510, 1465, 1252, 1076, 1036 cm-1; UV (EtOH) λmax 205 ( ) 72 094) nm; 1H (MeOD-d4) δ 7.95 (s, H8), 7.67 (d, J ) 8 Hz, 2 ArH), 7.45 (d, J ) 8 Hz, 2 ArH), 7.22 (t, J ) 8 Hz, 4 ArH), 7.07 (d, J ) 6 Hz, 2 ArH), 6.79 (dd, J ) 2 and 8 Hz, 2 ArH), 5.605.47 (m, H1′), 4.40-4.27 (m, H3′), 4.23-4.10 (m, CH), 3.90-3.70 (m, 3H, H4′5′5′), 3.75 (s, OMe), 2.27 (s, Me), 1.21 (d, J ) 7 Hz, 6H, 2 Me) ppm; C35H37N7O5F2 requires M 673, found m/z 673 (100), 372 (20), 302 (40).


N-Isopropyl-4-[(4-methoxyphenyl)phenyl[[2-amino9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzamide (10d). Active ester 2d (50 mg) was reacted with iPrNH2, according to general procedure F, to give amide 10d as a white solid (36 mg; 78%): TLC Rf (20% MeOH/CH2Cl2) 0.68; IR (KBr) νmax 3373, 1609, 1511, 1466, 1253, 1077 cm-1; UV (EtOH) λmax 224 ( ) 42 444), 204.5 ( ) 68 723) nm; 1H (acetone-d6) δ 7.80 (s, H8), 7.72 (d, J ) 6 Hz, 2 ArH), 7.46 (d, J ) 6 Hz, 2 ArH), 7.42-7.36 (m, 2 ArH), 7.32-7.14 (m, 5 ArH), 6.82 (d, J ) 8 Hz, 2 ArH), 6.28 (s, NH), 6.10 (br s, NH2), 5.80-5.65 (m, H1′), 5.42 (d, J ) 6 Hz, OH), 4.72-4.50 (m, 2H, OH, H3′), 4.28-4.10 (m, CH), 3.86-3.70 (m, 3H, H4′5′5′), 3.76 (s, OMe), 1.21 (d, J ) 6 Hz, 6H, 2 Me) ppm; C34H35N7O5F2 requires M 659, found m/z 659 (100), 358 (15). N-Isopropyl-4-[(4-methylphenyl)phenyl[[2-amino9-(2′-deoxy-2′,2′-difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzamide (10e). Active ester 2e (50 mg) was reacted with iPrNH2, according to general procedure F, to give amide 10e as a white solid (36.1 mg; 78%): TLC Rf (10% MeOH/CH2Cl2) 0.51; IR (CHCl3) νmax 3358, 1633, 1427, 1077 cm-1; UV (EtOH) λmax 204 ( ) 62 080) nm; 1H (MeOD-d4) δ 7.88 (s, H8), 7.67 (d, J ) 8 Hz, 2 ArH), 7.46 (d, J ) 10 Hz, 2 ArH), 7.34 (d, J ) 8 Hz, 2 ArH), 7.26-7.12 (m, 5 ArH), 7.05 (d, J ) 7 Hz, 2 ArH), 5.60-5.46 (m, H1′), 4.41-4.24 (m, H3′), 4.23-4.10 (m, CH), 3.92-3.64 (m, 3H, H4′5′5′), 2.28 (s, Me), 1.22 (d, J ) 6 Hz, 2 Me) ppm; C34H35N7O4F2 requires M 644, found m/z 644 (65). N-Isopropyl-4-[bisphenyl[[2-amino-9-(2′-deoxy-2′,2′difluoro-β-ribofuranosyl)purin-6-yl]amino]methyl]benzamide (10f). Active ester 2f (50 mg) was reacted with iPrNH2, according to general procedure F, to give amide 10f as a white solid (32 mg; 70%): TLC Rf (10% MeOH/CH2Cl2) 0.58; mp 190 °C dec; IR (KBr) νmax 3362, 1633, 1467, 1077 cm-1; UV (EtOH) λmax 205 ( ) 65 359) nm; 1H (acetone-d6) δ 7.80 (s, H8), 7.73 (d, J ) 8 Hz, 2 ArH), 7.48 (d, J ) 8 Hz, 2 ArH), 7.42-7.37 (m, 4 ArH), 7.32-7.15 (m, 6 ArH), 6.36 (s, NH), 6.12 (br s, NH2), 5.78-5.64 (m, H1′), 5.43 (d, J ) 8 Hz, OH), 4.74-4.52 (m, 2H, OH, H3′), 4.26-4.16 (m, CH), 3.96-3.72 (m, 3H, H4′5′5′), 1.20 (d, J ) 6 Hz, 6H, 2 Me) ppm; C33H33N7O4F2 requires M 630, found m/z 630 (70), 328 (20). Antibody Purification. Ascites fluid, derived from the hybridoma cells, was used as the antibody source for the studies described in this paper. The mAbs were purified from ascites fluid using Prosep-A protein A chromatography (Bioprocessing, Inc., Princeton, NJ). Briefly, ascites fluid (containing about 5 mg/mL IgG) was diluted (50:50 v/v) with 1 M glycine/0.3 M NaCl, pH 8.6 buffer and filtered. This antibody-containing solution was applied to the Prosep-A column which had been previously equilibrated in 1 M glycine/0.15 M NaCl, pH 8.6 buffer. After all unbound material was washed out, the protein A bound IgG was eluted with 0.1 M citrate, pH 3.5 buffer and immediately neutralized with 1 M TrisHCl, pH 8.6 buffer. Protein was diluted with PBS (dilution factor (df) ) 20) and UV absorbance measured at 279 nm. Protein concentrations were determined using the Beer-Lambert Law (where A279 ) 1.43 at 1 mg/ mL for monoclonal antibody). Purified IgG was subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (Phastgel, Pharmacia, Piscataway, NJ) using reducing (12.5% gel in the presence of β-mercaptoethanol) and nonreducing (7.5% gel in the absence of β-mercaptoethanol) conditions. General Conjugation Procedure. To a stirred 0.1 M borate buffer (pH ) 8.64) solution of protein (1.442 mL of 10.4 mg/mL ) 15 mg; 1.00 × 10-7 mol) was slowly


(over 3-5 min) added to a solution of N6-trityl-207702 active ester 2 (0.55 mg; 8.00 × 10-7 mol) in redistilled dimethylformamide (117 µL, 7.5% of total volume) at room temperature. The reaction mixture turned cloudy and was allowed to stir for a further 1 h. The clear solution was centrifuged at 2000g for 10 min and the supernatant then purified on a G-25 Sephadex (medium) HR 16/50 column (pre-equilibrated with PBS). The product was eluted with PBS, and the fractions containing product were combined and concentrated. The final product was sterile filtered and then analyzed by Superose 12 column for aggregate content and free drug content. The conjugation ratio (CR), protein and drug concentrations, and protein yield were determined from the UV spectrum. The protein concentration [mAb] in mg/mL was calculated using the formula

[Ab] ) A279df/1.43 (where A279 ) 1.43 at 1 mg/mL for monoclonal antibody) and the drug concentration [drug] in mg/mL from the Beer-Lambert law:

[drug] ) A/l amide (10)

/104

(a) m-DMT (b) p-DMT (c) p-MMeT (d) p-MMT (e) p-MeT (f) p-T

0.993 2.06 1.47 1.71 1.87 2.22

where is the molar absorptivity (molar extinction coefficient), l is the path length of the sample (cm), [drug] is the concentration of the drug (mol dm-3), and A is the absorbance of the solution (optical density, OD) at λmax ) 254 nm for 3b-f and 263 nm for 3a. Conjugation ratio (CR) is defined as the number of moles of drug per mole of antibody. For LY207702 (1) MW ) 302 and mAb MW ) 150 000:

CR ) {[drug]/302}/{[Ab]/150 000} Evaluation of mAb and Conjugates Binding Activity. Competitive binding analyses were performed as follows: serial dilutions (250 µg/mL, 125 µg/mL, ...) of mAb or conjugate 3 were analyzed for their ability to displace the binding of 50 µg/mL COL1-FITC, prepared by the method described by Harlow and Lane (21), to LS174T human colon carcinoma cells (obtained from the American Type Culture Collection, Rockville, MD). Cell surface fluorescence was quantitated by flow cytometry. Indirect immunofluorescence was accomplished by incubating half log dilutions of mAbs or conjugates (starting at 50 µg/mL) with LS174T tumor cells. After unbound antibody was washed away, FITC-conjugated rabbit antimouse IgG serum (Jackson Immunoresearch Laboratories, Inc., West Grove, PA) was added and cell surface binding was also quantitated by flow cytometry. The level of immunofluorescence observed in both assays is given as linear equivalent relative fluorescence intensity (LERFI) units. In Vitro Cytotoxicity. In vitro cytotoxicity was determined using a [3H]leucine incorporation assay as previously described (22). Nude Mouse Xenografts. The CEA-expressing human colon carcinoma cell line, HC1 (23), was used in the in vivo evaluation of free and conjugated LY207702. Approximately 8 mm3 fragments were trochar implanted


Patel et al.

Figure 1. Purine difluoronucleoside, LY207702 (1), its trityl derivatives 2, and immunoconjugates 3. Table 1. Trityl Derivatives (a) (b) (c) (d) (e) (f)

m-DMT p-DMT p-MMeT p-MMT p-MeT p-T

R

R′

OMe OMe OMe OMe Me H

OMe OMe Me H H H

into recipient mice (female, 20-25 g, Charles River Breeding Laboratories, Boston, MA). All animal experiments were done in accordance with the rules and regulations within the AAALAC approved animal research facility at Lilly Research Laboratories. RESULTS AND DISCUSSION

Chemistry. The synthesis of drug conjugates 3 was achieved by initial tritylation of LY207702 (1) followed

by coupling of the resulting drug linker 2 to a monoclonal antibody (Figure 1) (24). Trityl linkers used in this study included dimethoxytrityl (DMT), monomethoxymethyltrityl (MMeT), monomethoxytrityl (MMT), methyltrityl (MeT), and trityl (T) each of which possess a carboxy substituent in the para (p) or meta (m) position on one of the aromatic rings (Table 1). Assuming that the acid mediated dissociation of trityl derivatives proceeds via initial protonation of the heteroatom attached to the trityl group followed by cleavage to give a transient trityl carbocation, the rate of dissociation should be proportional to the stability of the trityl carbocation (25). On the basis of this simplified electronic argument, linkers a, b, d, and f were expected to provide a range of acid stabilities and therefore were proposed for initial investigation. Trityl chloride derivative 9, previously reported by Gildea (26), was chosen as the bis, differentially reactive electrophile for the sequential introduction of LY207702 (1) and mAb in the preparation of conjugate 3. Thus, addition of the Grignard reagent, derived from aryl bromide 4, to benzophenone 5 gave carbinol 6 (Scheme 1). Hydrolysis of the 2-oxazoline ring in 6 led to the corresponding carboxylic acid 7, which was then converted to the corresponding N-hydroxysuccinimide active ester 8 under standard DCC coupling conditions. Treatment of alcohol 8 with refluxing acetyl chloride provided the desired trityl chloride 9. N-Tritylation, for protection during synthesis, is commonly achieved by reacting an amine with a trityl chloride at room temperature usually in pyridine as solvent (27). The less reactive aromatic amines present in purine heterocycles can also be alkylated under similar conditions (28). In the presence of primary hydroxyl groups, however, excess trityl chloride and elevated temperatures lead to the formation of bisN,O-tritylated product (29). In one instance, the use of 1.1 equiv of trityl chloride led to preferential O-tritylation in the presence of a purine N6-amino group (30). It has been shown that even the extremely unreactive guanosine N2-amino group can also be tritylated under standard conditions (31). It was, therefore, remarkable that reaction of unprotected difluoronucleoside, LY207702 (1), with 1.1 equiv of trityl chloride 9, under conditions developed by Reddy and co-workers (32), cleanly produced the desired N6-tritylated nucleoside 2 as the sole product in 54-93% isolated yield (Scheme 2). It is worthy to note

Scheme 1a

a (i) (a) Mg/THF, room temperature, sonication; (b) benzophenone 5, THF, room temperature; (ii) (a) 80% aqueous AcOH, 55-60 °C; (b) 20% NaOH in aqueous EtOH; (c) 6 M HCl; (iii) DCC, NHS, CH2Cl2; (iv) AcCl, 60 °C.



Scheme 2

Scheme 3a

a

(i) iPrNH2, CH2Cl2, room temperature; (ii) 1H NMR, ∼6 mM in 0.1 M KH2PO4 in D2O/MeOH-d4.

Scheme 4a

a (i) 2, molar equivalents ) 8, COL1 (20-30 mg/mL) in 0.1 M borate buffer, pH ∼8.6, 7.5% DMF, room temperature, 1 h; (ii) centrifuge, purify on Sephadex G-25 column (eluting with pH 7.4 PBS), and concentration/sterile filter.

that exclusive regioselective tritylation of N6 amino group of 1 and selective alkylation versus acylation was achieved under these conditions (24). Subsequent acylation of isopropylamine with active ester 2 was found to be rapid and quantitative, at ambient temperatures, to provide the corresponding trityl amide 10 (Scheme 3). This encouraging result indicated that the acylation of -amino lysine groups on the mAb with NHS ester 2 should also proceed with ease under mild conditions that would be necessary in the final step of the synthesis of conjugate 3. Release Rates of LY207702 1 from Trityl-207702 Amides 10. To determine the effect of acid-mediated release of drug on biological activity of trityl-linked conjugates, it was necessary to first identify a range of trityl linkers with differing acid labilities. Trityl amides

10a-f were, therefore, subjected to hydrolysis under various pH conditions and the rate of consumption of amide 10 measured (33). In these experiments, a solution of amide 10 in deuterated methanol was diluted with a solution of 0.1 M KH2PO4 in D2O buffered at pH 5.40, 6.40, and 7.40 to achieve a final concentration of ca. 6 mM (Scheme 3). The solutions were heated at 37 °C in an NMR tube and hydrolysis conveniently monitored by 1H NMR spectroscopy. The starting trityl nucleoside derivative 10 and released free drug 1 were readily identified by the presence of a distinct 8 H singlet at 7.83 and 7.95 ppm, respectively. Integration of the respective peaks enabled the rate of consumption of amide 10 and appearance of 1 to be determined. It was found that the rate of formation of free drug 1 was equal to the rate of consumption of drug linker 10. The formation of the


Patel et al.

Table 2. Half-Lives (h) of N6-Trityl Amides under Various pH Conditions pH

m-DMT (10a)

p-DMT (10b)

p-MMeT (10c)

p-MMT (10d)

p-MeT (10e)

p-T (10f)

5.40 6.40 7.40

2.35 23.3 133.3a

3.71 14.8 146.3

2.40 43 229

16.7 253 800b

26.5 368c stabled

stablee stablee stablee

a

Extrapolated figures. b 14% dissociation after 100 h. c See footnote 5.

expected hydrolysis products, LY207702 (1) and carbinol 11, was also detected by TLC analysis and confirmed by isolation and characterization. The data from the release studies is represented in graphical form (amount of 10 (%) vs time (h)) for drug linkers 10a-f at (a) pH 5.40, (b) pH 6.40, and (c) pH 7.40 (Figure 2a-c). In all cases, a linear relationship was observed, suggesting that release of drug 1 from drug linker 10 proceeds in a pseudo-first-order manner under these conditions. In order to gain a better understanding of the relative acid lability of these trityl linkers, the release rates were determined from these plots and expressed as half-lives, t1/2 (h), the results of which are summarized in Table 2. Indeed, as predicted earlier, a wide range (e.g. t1/2 ) 2.35 h for 10a and >800 h for 10f at pH 5.40) in release rates was observed for these trityl linkers. Moreover, for a particular trityl linker, the release rates were found to be pH dependent, where 5.40 > 6.40 > 7.40, suggesting that the rate of initial protonation of 10 significantly influences the rate of hydrolysis. Qualitative analysis of the data indicate that an increased stabilization of the transient trityl carbocation leads to increased rates of hydrolysis (i.e. 10a ) 10b > 10c > 10d > 10f . 10f). Since substituents R and R′ on the aromatic rings of the trityl group play a major role in stabilizing the trityl cation, it is, therefore, reasonable to conclude that the substituents also govern the release rates. The half-lives do not, however, completely correlate with the calculated charge4 on the presumed intermediate trityl cation, 3 Some cytotoxicity determinations differ from those in ref 23 due to the availability of further data. In those cases where a conjugate was prepared multiple times, the average cytotoxicity of all the conjugates is quoted. 4 Calculation of the charge on the methyl carbon atom of substituted trityl carbocation was determined by quantum chemistry performed on Silicon Graphics Indiago 2 Workstation with MOPAC program using the AM1 Hamiltonian through the interface provided by the program CERIUS2. All of the structures for MOPAC calculations were geometry preoptimized using the molecular mechanics program with CERIUS2. The optimized geometry was used by MOPAC on the Workstation for single point calculation of the partial atomic charges as the carbocation after the removal of the methine hydrogen (keyword: charge)+1). The charge values are listed below:

R

NH2 R′ O

trityl cation

charge calcd

m-DMT p-DMT p-MMeT p-MMT p-MeT p-T

0.3948 0.4135 0.4213 0.4351 0.4469 0.5163

5 This figure was redetermined using more data points to give an accurate half-life for amide 10e. This figure significantly differs from that which was determined in ref 24 by extrapolation from a few early time points. It is apparent that extrapolation, in this manner, leads to overestimation of the half-lives.

d

5% dissociation after 631 h. e No dissociation up to day 33.

Table 3. Physical Chemical and Cytotoxicity Data for COL1-N6-Trityl-207702 Conjugates CR (1) LY207702 (3) COL1-N6-trityl-207702 (a) m-DMT (b) p-DMT (c) p-MMeT (d) p-MMT (e) p-MeT (f) p-T COL1

6.44 7.44 7.49 5.73 5.20 4.44 -

protein yield (%)

IC50 (µg/mL)

-

0.302

52 65 44 61 47 50 -

0.352 0.288 2.71 0.966 6.04 >10 >330

suggesting that the hydrolysis of trityl groups is also influenced by other factors such as ionic strength (34) and counterion of the media, concentration, temperature as well as added nucleophiles (35). The presence of these and other factors at the site of release in a relevant biological system (e.g. extracellular environment of tumor cells in a cytotoxicity assay) is unknown, and therefore, precise conditions are difficult to reproduce. Nevertheless, the qualitative correlation between the electronic nature of R and R′ and the relative release rates of trityl linked drug provides a useful and predictive tool for linker design. Preparation and Characterization of Conjugates 3. The final step in the synthesis of drug immunoconjugate 3 required the formation of a stable amide bond between -amino group of lysine residues on mAb COL1 and the carboxy group of drug linker 2. Accordingly, a solution of 2 in dimethylformamide was added to an aqueous pH 8.6 buffered solution of mAb at room temperature and stirred for 1-2 h. The initial cloudy reaction mixture became a clear, homogeneous solution, indicating that the reaction was complete. TLC analysis of the reaction mixture for the consumption of active ester 2, however, provided a more accurate method of judging reaction progress. Final conjugates 3 were isolated by size exclusion chromatography and then analyzed by UV spectroscopy to determine the conjugation ratio (CR), protein and drug concentrations, and overall yield. These mild conjugation conditions led to high levels of drug (55-94%) incorporation and yielded conjugates 3 with conjugation ratio of 4-7 (Table 3). Total protein recovery was satisfactory (44-65%) at final concentrations of 2030 mg/mL; however, at higher concentrations of conjugate, precipitation was observed leading to significant reduction in yield. It is very likely that the hydrophobic nature of the trityl group renders these conjugates less soluble in aqueous media. Protein aggregation usually leads to nonspecific binding and precipitation, both of which are undesirable properties for conjugates. The aggregation content in conjugates 3 was readily assessed using an analytical Superose-12 column. As an example, a chromatogram of COL1-N6-p-MeT-207702 (3e), which represents a typical profile for these conjugates, is shown in Figure 3. In general, conjugates 3a-f consisted of 9296% of the desired monomeric form of the protein with the remainder comprised of low molecular weight (36%) and high molecular weight (1-2%) aggregates. This method of analysis was also able to show that no free drug 1 was present in the final preparations. Furthermore, no significant aggregates or degradation products


Figure 2. Release of LY207702 1 from Trityl-207702 amides 10a (~), 10b ([), 10c (O), 10d (]), 10e (9) and 10f (0) with time of incubation at (a) pH 5.40, (b) pH 6.40, and (c) pH 7.40 (37 °C).


were detected when conjugates 3a-f were subjected to analysis by SDS-PAGE under reducing and nonreducing conditions (data not shown). Conjugate 3, which was isolated in 0.1 M phosphate-buffered solution (PBS) and stored under sterile conditions at 4 °C, remained unchanged for several months with respect to levels of free drug 1 and aggregate. The quality of the conjugates, prepared on scales ranging from 10 mg to 10 g was consistent and sufficient for biological studies and did not require further purification. Noncompetitive and competitive assays using CEA positive LS174T cells showed that construct 3 retained good antigen binding characteristics in comparison to unconjugated mAb COL1, examples of which are shown in parts a and b of Figure 4, respectively. Correspondingly, conjugate 3 prepared with an irrelevant IgG mAb did not bind to LS174T or HC1 tumor human colon carcinoma cells (data not shown), indicating that components in the construct other than the mAb do not contribute to antigen binding. In order to minimize immunogenic response during human therapy with murine-based drug conjugates, it is desirable to deliver maximal quantities of drug using minimal amounts of antibody. One method of producing conjugates with higher drug loading (CR) is to increase the molar equivalents of drug. Coupling of 2 (16-fold molar excess) to mAb D612 (36), indeed, gave a conjugate with a higher drug load (CR ) 11.3), however, a 10% yield and ∼40% high MW aggregate content deemed the conjugate unacceptable for use. Furthermore, fractionation of this conjugate, using a hydrophobic interaction column, gave species with even higher CR (∼20) but unfortunately also of diminished quality. These results reflect the limitations of lysine coupling methodology and probably also the physical properties of mAb COL1. The procedure described using a drug-linker molar equivalents of 8 was deemed optimal and gave COL1-N6Trityl-207702 conjugate 3 with consistent and acceptable properties suitable for biological evaluation. Biological Properties. Cytotoxicity3 of conjugates 3a-f, LY207702 (1), and unconjugated mAb COL1 was determined by measuring [3H]leucine uptake in CEA positive LS174T cells after incubation for 48 h and expressed as IC50 values (Table 3). The most active conjugates, 3a and 3b, which were equipotent to free drug 1, corresponded to drug linkers 10a and 10b which released drug 1 at the highest rate under acidic conditions (see Table 2). The p-MeT conjugate 3e was found to be less active than p-MMT 3d, but significantly more active than p-T conjugate 3f. With the exception of conjugate 3c, the order of cytotoxicity of the conjugates, 3a ) 3b > 3d > 3e . 3f, correlated well with the rate of release of drug 1 from amides 10. The higher than expected IC50 for conjugate 3c could be due to factors other than electronics which influence release of drug from the conjugate; however, these were not identified in these studies. A quantitative analysis of the cytotoxicity data (Table 3), however, indicates that incubation of conjugates 3a and 3c at 37 °C for 48 h led to 100% dissociation to free drug 1. Likewise, conjugates 3c, 3d, 3e, and 3f were hydrolyzed 71%, 33%, 5%, and 0%, respectively. In contrast, incubation of prodrugs 10a-f for 48 h at 37 °C in methanolic aqueous phosphate buffer, for example, at pH 6.40 gave 78%, 88%, 53%, 8%, 8%, and 0% free drug, respectively (see Figure 2b). This comparison assumes, firstly, that cytotoxicity of conjugate 3 is solely derived from the release and action of LY207702 (1) and, secondly, that the pH experienced by conjugate 3 is 6.40 in the cellular assay. Since both unconjugated COL1 and tritylated COL1, a byproduct from the hydrolysis of conjugate 3, did not exhibit


Patel et al.

Figure 3. Profile of COL1-N6-p-MeT-207702 conjugate 3e on an analytical Superose 12 column. Conjugate (100 µL, df ) 10 in PBS), with CR ) 4.6, [COL1] ) 24.9 mg/mL, [LY207702] ) 0.231 mg/mL, was eluted with 15% CH3CN/PBS at flow rate ) 0.8 mL/min and detected at λ ) 279 nm. High MW ) >1×106, low MW ) ∼3 × 105, conjugate MW ) 150 000.

Figure 4. Immunoreactivity of COL1-N6-trityl-207702 conjugates 3 ([) versus mAb COL1 (~) to LS174T cells determined by (a) indirect immunofluorescence of COL1-N6-p-MeT-207702 3e and (b) competitive binding analysis of COL1-N6-p-DMT-207702 3b.

cytotoxicity, it is believed that cytotoxicity of conjugate 3 is, in fact, due to release of LY207702 (1). This is consistent with the knowledge that COL1 is a noninternalizing antibody (16), and therefore, the cytotoxicity demonstrated by conjugate 3 most likely results from the release of drug in the extracellular environment of the tumor cells and that LY207702 (1) is rapidly internalized and subsequently phosphorylated to its active metabolites (37). Although there is convincing literature evidence that the tumor pH is on average lower than normal tissue pH, the exact difference is dependent on tumor type and method of measurement (6). Gerweck (6) recently reported the extracellular pH of tumors in several patients and found that the pH ranged from 6.4 to 7.0. It is, therefore, difficult to design accurate release studies without knowing the pH of the tumor environment. It should be noted that the release studies were not performed on the actual conjugates 3, primarily due to difficulties encountered in assaying for free drug 1. The release of drug from the conjugate may possibly be influenced by additional factors, such as steric and ionic effects of amino acid residues at the site of the drug-linker attachment on the mAb. The above comparison between release rates of 1 from prodrug amide 10 and the cytotoxicity of the corresponding conjugates 3 does, however, reveals a qualitative relationship between the electronic nature of the substituents on the trityl linker a-f and cytotoxicity of the trityl-linked conjugates 3af. This correlation provides a useful means of ranking the in vitro cytotoxicity for a series of trityl-linked

conjugates, a property which should allow a more convergent approach to the design of an optimal trityl-linked conjugate. The sensitivity of LY207702 (1) to the HC1 human colon carcinoma and expression of high levels of CEA on the tumor cells made this model suitable for determining the antitumor activity and targeting properties of conjugate 3. The nature of the antigen, for example shedding which would lead to circulating levels of CEA, in the HC1 tumor model was not examined in these studies. Selective targeting, however, has previously been achieved using a different mAb drug conjugate that also recognizes CEA (38). The release studies indicate that conjugate 3 requires an acidic environment to release free drug 1; however, the pH of HC1 tumors was not determined due to the complexity encountered in measuring tumor tissue versus adjacent normal tissue as well as extracellular (pHe) versus intracellular (pHi). Assumptions were, therefore, made that the extracellular tumor pH in the HC1 model would be more acidic than normal tissue (6) and that this pH would be sufficient to cleave acid labile conjugate 3 while bound to the target antigen. This hypotheses was tested using antitumor efficacy as an end point. To assess antitumor activity, nude mice were, first, trocar implanted subcutaneously with HC1, and the tumor was allowed to establish (∼0.5 g). Mice were separated into groups of eight and treated iv with drug 1 or conjugate 3 twice weekly for 2 weeks, and tumors were measured weekly postimplantation. Initial studies with p-DMT-linked conjugate 3b showed a good dose



Figure 5. Effect of LY207702 (1) and COL1-N6-p-DMT-207702 conjugate 3b on HC1 human colon carcinoma. Tumor was trocar implanted sc in nude mice and allowed to establish to ∼0.5 g. Mice were separated into treatment groups of eight and dosed iv with LY207702 (1) (0), conjugate 3b (b), and saline control (O) twice weekly for 2 weeks (indicated by arrows) at (a) 1 mg/kg (drug equivalents), (b) 2 mg/kg, and (c) 4 mg/kg. Tumors were measured weekly postimplantation.

Figure 6. Comparison of antitumor activity of COL1-N6trityl-207702 conjugates 3 and LY207702 (1) against HC1 human colon carcinoma. Tumors were allowed to establish and nude mice (groups of eight) were treated iv with saline control (4), LY207702 1 (0) or conjugate 3b (b), 3d (9), 3e (2), 3f (O) at 4 mg/kg (drug equivalents) twice weekly for 2 weeks (indicated by arrows). Tumors were measured weekly postimplantation.

response whereby conjugates 3b and free drug 1 exhibited no activity at 1 mg/kg (drug equivalent), inhibition at 2 mg/kg, and complete and sustained regression at 4 mg/kg (Figure 5). It is interesting to note that the similar in vivo activity of conjugate 3b and free drug 1 is consistent with the in vitro cytotoxicity data (Table 3). Antitumor activity of conjugates 3b, 3d, 3e, and 3f was then assessed in a higher tumor burden (0.5-1.2 g) HC1 model at the 4 mg/kg dose schedule. In these experiments, the least in vitro cytotoxic p-T conjugate 3f displayed the least in vivo antitumor activity, exhibiting only moderate inhibition of tumor growth, while the more cytotoxic conjugates 3b, 3d, and 3e all regressed the large tumor volumes (Figure 6). The complete tumor regression displayed by conjugates 3b, 3d, and 3e, and free drug 1 reflects the remarkably high level of antitumor activity of this novel purine nucleoside in this resistant tumor xenograft (17). Particularly impressive was the sustained tumor regression observed several weeks after the last dose of conjugates 3b, 3d, and 3e was administered. p-MeT conjugate 3e treated group did, however, show signs of tumor regrowth at 5-6 weeks after initiation of therapy. In contrast, p-DMT 3b and p-MMT 3d conjugates, which were equally efficacious, main-

tained complete tumor regression until the end of the experiment (6 weeks). The larger initial tumor size in the conjugate 3e experiment makes it difficult to draw a true comparison between the antitumor activities of conjugates 3b, 3d, and 3e. The in vivo activities of free drug 1 and conjugates 3b and 3d were essentially indistinguishable; however, the faster tumor regrowth of conjugate 3e indicates that this may be the least active of these. The lack of a good correlation between the in vitro activity and in vivo activity of conjugate 3 is probably attributed to several factors, including (i) the use of a more drug sensitive HC1 tumor cell line used in vivo versus the less responsive LS174T cells used in vitro, (ii) the unoptimized in vivo dosing schedule, and (iii) the unknown tumor pH of LS174T (in vitro) and HC1 (in vivo) which, if different, would give rise to different release rates of the same conjugate 3. A hallmark of drug conjugates is the ability to selectively deliver drug to the tumor and thereby reduce systemic toxicity. For this to be effective, a drug conjugate possessing a linker which releases drug while bound to the tumor antigen is desirable. A series of experiments were, therefore, conducted comparing antitumor activity of antigen-positive (COL1) and antigen-negative (IgG) binding conjugates 3 to identify a trityl-linked conjugate 3 which demonstrated activity primarily through antigen binding. Accordingly, conjugates 3b, 3d, 3e, and 3f were administered at a dose of 4 mg/kg twice weekly for 2 weeks to HC1 tumor-bearing mice, and antitumor activity was measured. Surprisingly, in all cases, p-DMT 3b, p-MMT 3d, and p-MeT 3e linked conjugates, there were no significant differences in antitumor activity between the irrelevant (IgG) and relevant (COL1) conjugates in this large (∼1 g) tumor burden model (data not shown). Immunofluorescent analysis of enzymatically dissociated tumor cells obtained from mice treated with COL1-pMeT-207702 or irrelevant IgG-p-MeT-207702, however, demonstrated that HC1 tumor cell binding occurred only in the COL1 conjugate 3e treated animals (data not shown) (22). It is conceivable, however, that the more labile irrelevant IgG conjugate 3b may be rapidly releasing drug systemically prior to antigen binding and essentially acting in the same manner as administering free drug 1. However, the origin of activity displayed by the supposedly more stable (see Table 2) irrelevant IgG conjugates 3d and 3e is unclear. Since the concept of targeting drugs using mAbs relies on specific binding of the conjugate to the target antigen with subsequent release of drug, the “mistiming” of these two events would lead to loss of the desired effect. It is also possible that


Figure 7. Comparison of antitumor activity of mAb-N6-pMeT-207702 conjugates 3e in HC1 tumors: COL1 (relevant) versus IgG (irrelevant) implanted sc with HC1 tumors on day 1. Mice with palpable tumors were selected on day 14 and separated into treatment groups of seven. (a) Tumor mass measured weekly for groups dosed with saline control (O), LY207702 (1) (4), relevant COL1 conjugate 3e (b), and irrelevant IgG conjugate 3e (0) at 4 mg/kg drug equivalents iv on days 15, 19, and 22. (b) Tumor mass measured on day 40 for each of the treatment groups.

the lipophilic trityl constructs 3 may bind nonspecifically to tumor cells exposed to an acidic environment which causes drug to be released at the tumor site. These and other unknown factors, which have not been investigated, may influence drug release in vivo and account for the observed results with irrelevant conjugate. Another possibility is that low levels of irrelevant conjugate were localized to the large tumor masses via Fc receptormediated binding. Once localized, the exquisite activity of released LY207702 (1) against HC1 may have been sufficient to cause tumor regression. In order to determine if tumor size was a contributing factor to the apparent cause of antigen-mediated activity of the mAbtrityl-207702 conjugates 3, we evaluated these constructs in a lower tumor burden model. Consequently, a similar study was performed in which the HC1 tumor was implanted, and 14 days later mice bearing palpable tumors were separated and treated with conjugates 3b, 3d, 3e, and 3f under identical dosing conditions and schedule as described above. In this model, a clear difference in antitumor activity between irrelevant (IgG) and relevant (COL1) p-MeT-linked conjugates 3e was observed (Figure 7a). A repeat of this experiment, with a different batch of conjugate 3e, gave the same results.

Patel et al.

The significantly higher level of activity observed with COL1 conjugate 3e compared to the irrelevant conjugate, most pronounced on day 40 of the experiment (Figure 7b), suggested that antigen-mediated binding significantly contributes to its antitumor activity. It is also interesting to note that in this experiment COL1 conjugate 3e was more active than LY207702 (1). Conjugates 3b, 3d, and 3f did not, however, demonstrate this level of specificity, suggesting that p-MeT is the optimal linkage in this series of conjugates evaluated under the low tumor model conditions. These data provided the basis for selecting p-MeT-linked conjugate 3e for further evaluation. In an effort to establish the in vivo metabolic stability of p-MeT-linked conjugate 3e, it was administered to nontumor-bearing nude mice and plasma levels of LY207702 (1) measured. It is, however, known that LY207702 (1) is an excellent substrate for adenosine deaminase which rapidly converts it to the corresponding difluoroguanosine analogue (LY223592) (37). Thus, when a single iv dose of 8 mg/kg of LY207702 (1) was administered to female nude mice, a peak concentration of ∼1 µg/mL of metabolite LY223592 was observed 30 min after dosing, followed by a decrease in levels (t1/2 ) ∼30 min) as a result of excretion of drug. In contrast, no detectable levels (detection limit ) 200 ng/mL) of LY207702 (1) or its metabolite LY223592 were observed 30, 60, or 120 min after a single, 8 mg/kg (drug equivalents) iv dose of conjugate 3e, suggesting that direct release of drug is not rapid (t1/2 > 1 h). It is possible that antitumor activity observed with irrelevant p-MeT conjugate (Figure 7a) could be due to release of drug at levels not detected (i.e.

Novel Acid Labile COL1 Trityl-Linked Difluoronucleoside

Recommend Documents