Doxorubicin-formaldehyde conjugates targeting α_vβ₃ integrin Free

Doxorubicin and its congener, epidoxorubicin, are among the most effective chemotherapeutics for the treatment of a variety of solid tumors including breast tumors (1). The mechanism of action of doxorubicin has been debated since its discovery in the late 1960s, but its interaction with DNA is undeniable. Evidence has accumulated over the past 10 years that suggests doxorubicin combines with formaldehyde to form covalent bonds with DNA (2–5). The source of formaldehyde in cells is unclear, but analysis of cancer cells treated with doxorubicin reveals elevated levels of formaldehyde (6). Doxorubicin induction of oxidative stress is a possible pathway to the formaldehyde necessary for drug-DNA virtual cross-linking (7). Synthetic derivatives of doxorubicin, which contain formaldehyde in the form of an N-Mannich base, are significantly more cytotoxic to both sensitive and resistant cancer cells than doxorubicin itself (8, 9).

Two resistance mechanisms to doxorubicin are the overexpression of the drug efflux pump P-170 glycoprotein and suppression of oxidative stress mechanisms that produce formaldehyde (7, 10, 11). Using mass spectrometry, our group has observed doxorubicin inducing formaldehyde production in MCF-7 breast cancer cells but not in doxorubicin-resistant MCF-7/Adr cells (6). To combat both resistance mechanisms, we developed a conjugate formed from doxorubicin, formaldehyde, and salicylamide, known as doxsaliform (9). Doxsaliform contains an N-Mannich base that, on time, hydrolyzes to release the doxorubicin active metabolite with formaldehyde already incorporated (Fig. 1). Doxsaliform exhibits greater toxicity to both sensitive and resistant cancer cells (MCF-7, MCF-7/Adr, Rtx-6, MDA-MB-231, MDA-MB-435, and PC-3; refs. 9, 12, 13).

Significant limitations for doxorubicin treatment of cancer are drug resistance and chronic cardiotoxicity (14). One of the most promising methods to reduce the side effects of a cytotoxin like doxorubicin is selective delivery to cancer cells and/or their associated angiogenesis. A protein complex that may be a good target for drug delivery is the α_vβ₃ integrin. α_vβ₃ is involved in many cell-matrix recognition and cell adhesion phenomena, giving it an important role in angiogenesis and tumor metastasis. The α_vβ₃ integrin is overexpressed on the surface of tumor and endothelial cells responsible for angiogenesis (15), and its expression correlates with tumor progression in glioma, melanoma, breast cancer, and ovarian cancer (16–21). α_vβ₃ exists in discrete activation states, and activation can be induced with manganous ion (22). Activated α_vβ₃ supports breast cancer cell arrest during blood flow and strongly promotes breast cancer metastasis (23, 24). In tumor-induced angiogenesis, invasive endothelial cells bind via this integrin to extracellular matrix components. The inhibition of this interaction induces apoptosis of the proliferative angiogenic vascular cells (25). These factors combined make α_vβ₃ an attractive target for antiangiogenic and antimetastatic therapies. Several RGD peptide and peptide mimetics developed over the last decade exhibit excellent binding affinity and selectivity for α_vβ₃ (26). The peptide cyclic-(N-Me-VRGDf) known as Cilengitide has proceeded as far as phase II clinical trials as a potent antagonist of α_vβ₃ (27). Small RGD-containing peptides have successfully been used to deliver cytotoxins, magnetic resonance imaging contrast agents, radionuclides, liposomes, and fluorescent agents to tumors that express α_vβ₃ (28–30).

Arap and coworkers′ (28) report that a doxorubicin-CDCRGDCFC (RGD-4C) conjugate that targets α_vβ₃ substantially inhibited tumor growth in mice relative to doxorubicin with fewer side effects prompted further exploration. de Groot and coworkers (31) reported that doxorubicin conjugated with RGD-4C via a plasmin-cleavable tether inhibited human umbilical vascular endothelial cell binding to plates coated with vitronectin with an IC₅₀ of ∼150 nmol/L and exhibited a cytotoxicity IC₅₀ of 750 nmol/L against the same cell line. The plasmin-activated prodrug failed to inhibit tumor growth in vivo better than doxorubicin alone but did exhibit less toxicity based on weight loss in a tumor-bearing mouse model (32). Here we report the synthesis and biological evaluation of doxsaliform conjugated to two different RGD-containing peptides, RGD-4C and cyclic-(N-Me-VRGDf).

The conjugation of doxsaliform to α_vβ₃-targeting peptides serves several purposes. The drug conjugate is a prodrug with little or no activity until the trigger (N-Mannich base hydrolysis) releases the cytotoxin from the peptide. RGD-4C and cyclic-(N-Me-VRGDf) have both been shown to accumulate in tumor relative to other tissue, with a peak accumulation point of ∼40 to 60 minutes (33). Based on this delivery schedule, a triggered release of doxorubicin active metabolite with a half-life of 60 minutes should localize a good portion of the drug in tumor relative to other tissue. We hypothesize that this design will reduce side effects such as cardiotoxicity and increase the amount of active drug in and around the tumor.

Design. Doxsaliform was conjugated to the RGD-containing peptides RGD-4C and cyclic-(N-Me-VRGDf) via a short hydroxylamine ether tether that forms an oxime bond with a formyl group added at the 5″ position of the salicylamide of doxsaliform. This oxime was found to be quite stable under a variety of aqueous conditions. The N-Mannich base that contains the formaldehyde equivalent necessary to produce the doxorubicin active metabolite hydrolyzes with a half-life of 60 minutes at physiologic temperature and pH (9). Hydrolysis of the N-Mannich base is also the trigger that releases the doxorubicin active metabolite from the targeting peptide. The synthesis of acyclic-RGD-4C-doxsaliform and cyclic-(N-Me-VRGDf-NH)-doxsaliform are detailed below and their structures are shown in Fig. 2. See Fig. 3 and Fig. 4, respectively, for synthetic schemes.

Materials and Instruments. All reactions were done under inert atmosphere. Fmoc amino acids and other peptide synthesis reagents were purchased from Novabiochem (San Diego, CA) and used without further purification. For amino acids with sensitive side chains the following were used: Fmoc-Asp-tBu, Fmoc-Cys-triphenylmethyl, Fmoc-Arg-2,2,4,6,7-pentamethyldihydrobenzofuran-5-sulfonyl, Fmoc-D-4-aminoPhe[t-butoxycarbonyl (Boc)]. Fmoc-2-(2-aminoethoxy)-ethylamine hydrochloride was obtained from Neosystem (Strasbourg, France), (Boc-aminoxy)-acetic acid from Fluka (Milwaukee,WI), and Oregon Green 488 N-hydroxysuccinimide from Molecular Probes (Eugene, OR). 5″-Formyldoxsaliform was synthesized as previously described (13). Melting points were uncorrected. The ¹H, COSY, HSQC, HMBC, and ¹³C nuclear magnetic resonance (NMR) high-resolution spectra were obtained with a Varian Inova 500 spectrometer (Palo Alto, CA). Electrospray mass spectra were measured with a Perkin-Elmer Sciex API III (Norwalk, CT), equipped with an ion-spray source, at atmospheric pressure. Analytic high-performance liquid chromatography (HPLC) was carried out on a Hewlett-Packard/Agilent 1050/1100 system (Palo Alto, CA) consisting of a Hewlett-Packard 1050 series auto injector and pumping system, Hewlett-Packard 1046A fluorescence detector, Agilent 1100 diode array UV-Vis detector, and Agilent ChemStation data system (Palo Alto, CA). A Vydac (Hesperia, CA) protein C-4 column (4.6 × 250 mm) was used for analytic HPLC with a flow rate of 0.5 mL/min and a gradient solvent system of 0.1% or trifluoroacetic acid (TFA)/acetonitrile: 0 to 15 minutes, 98% to 40% aqueous; 15 to 20 minutes, 40% to 15% aqueous; 20 to 25 minutes, 15% to 98% aqueous; detection at 220, 254, 280, and 480 nm. For preparative HPLC, a Vydac 214TP1022 C-4 column (22 × 250 mm) was used with the same solvent system on a Varian/Ranin (Palo Alto, CA) semipreparative HPLC consisting of Varian Pro Star 210 pumping system, Ranin Dynamax UV-1 detector, and Ranin/Varian Macintosh data system eluting at 15 mL/min. The gradient used for cyclic-(N-Me-VRGDf-NH)-tether was 0 to 11.5 minutes, 98% to 80% aqueous; 11.5 to 15 minutes, 80% to 30% aqueous; 15 to 16 minutes, 30% to 55% aqueous; and 16 to 20 minutes, 55% to 98% aqueous (detection at 254 nm). The gradient used for complete drug conjugates was 0 to 30 minutes, 90% to 60% aqueous, and 30 to 50 minutes, 60% to 90% aqueous (detection at 470 nm).

Synthesis of the Hydroxylamine Ether Tether, (2-{2-[2-(2,2-Dimethyl-Propionylaminooxy)-Acetylamino]-Ethoxy}-Ethyl)-Carbamic Acid 9H-Fluoren-9-ylmethyl Ester 1. Fmoc-2-(2-aminoethoxy)-ethylamine hydrochloride 1.30 g was weighed and placed in a dry 250-mL round-bottomed flask under an argon atmosphere. Anhydrous dimethylformamide (DMF, 10 mL) was added by syringe followed by 2.0 mL of pyridine with stirring. (Bocaminoxy)-acetic acid [1.04 g, 2 equivalents (eq)] and water-soluble carbodiimide (0.69 g, 2 eq) were measured out and added in one portion to the solution of amine. The reaction was monitored by analytic HPLC and 0.33 eq of (Boc-aminoxy)-acetic acid and water-soluble carbodiimide were added after 1 hour to drive the reaction to completion. The reaction was then diluted with ethyl acetate (100 mL) and washed with dilute acetic acid (3 × 50 mL) followed by sodium bicarbonate (pH 8.5). The organic phase was dried with sodium sulfate and concentrated under vacuum to yield 1.78 g (99%) of clear solid product 1. ¹H NMR in chloroform-d: 1.42 (s, 9H), 3.20 (m, 2H), 3.53 (m, 6H), 4.20 (t, J = 6.8 Hz, 1H), 4.27 (s,2H), 4.43 (d, J = 6.8 Hz, 2H), 5.75 (br, 1H), 7.26 (t, J = 7.6 Hz, 2H), 7.43 (dd, J = 4.8, 7.2 Hz, 2H), 7.62 (d, J = 7.2 Hz, 2H), 7.76 (d, J = 7.6 Hz, 2H); electrospray-mass spectrometry (ESI-MS) relative intensity (m/z): 500, calculated for (M + H⁺) m/z 500.23.

Partial Deprotection of 1 to {2-[2-(2-Aminooxy-Acetylamino)-Ethoxy]-Ethyl}-Carbamic Acid 9H-Fluoren-9-ylmethyl Ester 2 and Loading of 2 onto Resin. The Fmoc-2-(2-aminoethoxy)-ethyl-Boc-aminoxy-amide (1, 1.7 g) was dissolved in a solution containing 10 mL of TFA and 1.1 mL of thioanisole at 0°C. The solution was allowed to stir for 1 hour at room temperature and then concentrated (<5 mL) and the product precipitated into cold diethyl ether (100 mL). The precipitate was then collected as the TFA salt by filtration and washed with ether (3 × 20 mL). ¹H NMR in methanol-d₄ showed complete removal of the Boc protecting group, so the compound 2 was then loaded on trityl chloride resin as follows. To a dry 250-mL round-bottomed flask was added 50 mL of dry methylene chloride, 2.2 mL anhydrous pyridine, and 2. After the amine went into solution with stirring, 1.1 g of trityl chloride resin was added in one portion and the mixture allowed to stir for 22 hours. The resin was then collected by filtration, washed with 17:2:1 [volume for volume (v/v)] methylene chloride:methanol:diisopropylethylamine (2 × 25 mL), and with methylene chloride (2 × 30 mL) followed by methanol (3 × 50 mL). The resin was then dried under vacuum and the loading was determined by treatment of an aliquot (5 mg) with 0.5 mL of 20% piperidine/DMF for 15 minutes and dilution to 50 mL with DMF followed by UV absorbance measurement at 301 nm. Resin loading ranged from 0.5 to 0.84 mmol/g.

General Procedure for the Synthesis of Linear Peptides. The linear peptides were synthesized by the solid-phase method using Fmoc strategy (for details see Applied Biosystems peptide synthesizer user's manual), starting with the preloaded Fmoc tether from above. The peptides were prepared on a 0.25-mmol scale by single amino acid couplings using a 4-fold excess of Fmoc amino acids and 2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (TBTU)/N-hydroxybenzotriazole (HOBT) activation on an Applied Biosystems 433A peptide synthesizer. Fmoc groups were removed by sequential treatment (3×) with 20% piperidine/DMF. Acyclic-RGD-4C was synthesized in the order Cys-Phe-Cys-Asp-Gly-Arg-Cys-Asp-Cys and final Fmoc deprotection of the peptide was done while still on the resin. The linear peptide was cleaved from the resin and deprotected by a 3-hour treatment with degassed reagent K. The resin was then filtered and the mother liquor concentrated under vacuum (<5 mL) and the product precipitated dropwise into cold ether (60 mL). The peptide was collected by filtration (#1 filter paper) and washed with ether (3 × 20 mL). The crude peptide was dried under vacuum overnight and analyzed by analytic HPLC and mass spectrometry. The analytic HPLC trace showed a single peak (retention time, 22.22 minutes), ESI-MS, m/z 1,180.6, calculated for (M+H⁺) 1,179.39.

Synthesis of the Acyclic-RGD-4C-Doxsaliform. To a 50-mL pear-shaped flask containing 3 mg of 5″-formyldoxsaliform was added 2 mL of 0.1% TFA and the solution was degassed with argon by bubbling for 5 minutes. Acyclic-RDG-4C-tether (3, 12 mg, 3 eq) was dissolved in 1.5 mL of degassed methanol and then added in one portion to the 5″-formyldoxsaliform solution by syringe. The reaction was allowed to stir at room temperature and was monitored by HPLC. After ∼3 hours, the reaction was found to be complete by HPLC analysis (new peak found retention time = 17.14 minutes, 480 nm). The reaction was purified directly by preparative HPLC and all major peaks analyzed by mass spectrometry. The product showed a mass spectral ion at m/z 1,883.2 (M + H⁺; calculated 1,883.6) and a base peak at m/z 942.2 [(M + 2H⁺)/2]. The yield of acyclic-RGD-4C-doxsaliform was 1.2 mg of compound, 98% pure by analytic HPLC. Drug was then formulated with 3 eq of citric acid and 6 eq of lactose and stored at −80°C.

Synthesis of Bicyclic-RGD-4C-Tether 4. Acyclic-RGD-4C-tether 3 (30 mg) was dissolved in a solution of 50 mL of TFA and 2.5 mL of DMSO. Anisole (0.5 mL) was then added by syringe with stirring and the solution stirred for 1 hour. The reaction was monitored by HPLC and stopped when complete (usually 1 hour). The solution was then concentrated under high vacuum to yield a mixture (∼50:50) of bicyclic isomers. HPLC gave two peaks at retention time, 27.05 and 27.31 minutes for the two isomers; ESI-MS, m/z 1,175.6, calculated for (M + H⁺) 1,175.4 for both isomers.

Synthesis of Protected Acyclic-N-Me-VRGDf-NH₂ 5. General Fmoc synthesis was done as for acyclic-RGD-4C-tether, but TBTU/HOBT hydrate coupling was found to be inefficient for coupling to N-methyl valine. Peptide still on the resin was treated with 2 eq bromo-tris-pyrrolidino-phosphonium hexafluorophosphate, Fmoc-d-4-aminophe(Boc) and 4 eq diisopropylethylamine in dry dichloromethane (5 mL per gram resin). The mixture was placed on a shaker for 16 hours, washed with 3 × 10 mL of dichloromethane, and checked by the chloranil test for coupling completion. If not complete, coupling was repeated for 3 hours. When coupling was finished, the resin was then treated with 3 × 10mL of 20% piperidine in DMF for a period of 10 minutes to complete deprotection. Resin was then returned to the ABI synthesizer to complete the peptide synthesis using standard Fmoc synthesis protocol. Cleavage of the linear peptide was effected with 1% TFA in dichloromethane (3× 10 mL) with shaking for 5 minutes each time. The solution was concentrated under high vacuum to give the linear peptide with protecting groups intact in 98% yield as determined by analytic HPLC (one peak with retention time 17.9 minutes); ESI-MS, m/z 1,029.6, calculated for (M + H⁺) 1,029.51.

Cyclization of Protected N-Me-VRGDf-NH₂ 5 to yield 6. Linear peptide (386 mg) with all protecting groups intact was dissolved in 50 mL of ethanol and 1.2 eq of 10% aqueous HCl (v/v) was added to displace the TFA salt. When this step was omitted, trifluoroacetylation of the peptide occurred during the cyclization reaction. The solution was concentrated under vacuum. Linear peptide was then dissolved in anhydrous DMF (125 mL) and 2 eq of water-soluble carbodiimide were added in one portion. Reaction was monitored by HPLC and typically complete within 3 hours. The solution was concentrated under vacuum and the residue was dissolved in 50 mL of ethyl acetate and washed with 10% HCl (v/v; 2 × 50 mL). The organic phase was dried with sodium sulfate and concentrated under vacuum to give pure cyclic-(N-Me-VRGDf-NH₂) with protecting groups intact 6. Analytic HPLC showed one peak (retention time, 17.2 minutes); ESI-MS m/z 1,012.5, calculated for (M + H⁺) 1,012.51.

Selective Removal of Boc Protecting Group from d-4-Amino-Phe of Fully Protected Cyclic-(N-Me-VRGDf-NH₂) 6 to Yield 7. Fully protected cyclic-(N-Me-VRGDf-NH₂; 6, 319 mg) was dissolved in 10 mL of dry ethyl acetate, and 1 mol/L anhydrous HCl in ethyl acetate (1.5 mL) was added while the mixture was maintained at 0°C with an ice bath. The mixture was allowed to stir for 3 hours at 0°C and then concentrated under vacuum. The product was then lyophilized from water to give a clear solid. This method completely removed the Boc group from the d-4-amino-Phe, but a small amount of peptide also experienced hydrolysis of the Asp tert-butyl protecting group to release the acid. This mixture was carried forward because the deprotected Asp was not deemed problematic. HPLC analysis shows two peaks (retention time, 12.2 and 13.9 minutes); ESI-MS for these two peaks, m/z 856.4 and 912.6, respectively. Calculated for deprotection of both d-4-amino-Phe(Boc) and Asp(tBu) (M + H⁺) 856.39; calculated for deprotection of only d-4-amino-Phe(Boc) (M + H⁺) 912.46.

Addition of Boc-Aminoxyacetic Acid Tether to Partially Protected Cyclic-(N-Me-VRGDf-NH₂) 7. Clear solid (288 mg) from the above reaction was dissolved in 50 mL anhydrous DMF and 3 eq of Boc-aminooxyacetic acid were added followed by 1.5 eq of water-soluble carbodiimide. After stirring for 1.5 hours, the reaction was complete based on analytic HPLC. The mixture was concentrated under vacuum, the residue dissolved in ethyl acetate (50 mL), and washed with water (2 × 20 mL) and then 10% HCl (v/v; 2×50 mL). The organic phase was separated, dried with sodium sulfate, and concentrated under vacuum. Two peaks were observed by HPLC (retention time, 12.7 and 14.4 minutes); ESI-MS for the two products, m/z 1,028.8 and 1,084.5, respectively; calculated for (M + H⁺) 1,028.47 and 1,084.53.

Removal of All Protecting Groups from Cyclic-(N-Me-VRGDf-NH)-Tether to Yield 9. Peptide from the above reaction was added to a dry 50 mL round-bottomed flask and cooled to 0°C under an argon atmosphere. Reagent K (5 mL) was added and the solution allowed to stir for 3 hours at room temperature. Solution was then added dropwise with vigorous stirring to 100 mL of anhydrous ether that had been cooled with an ice bath. The white precipitate was collected by filtration and washed thrice with ether (15 mL) and dried under vacuum. Pure product cyclic-(N-Me-VRGDf-NH)-tether (9, 171 mg) was obtained as determined by HPLC (retention time, 7.2 minutes); ESI-MS, m/z 677.2, calculated for (M + H⁺) 677.33. To assign the ¹H NMR spectrum unequivocally, the following spectra were run: ¹H NMR, COSY, HSQC, and HMBC all in D₂O. ¹H NMR: 0.47 (3H, d, J = 6 Hz, CH₃, Val), 0.80 (3H, d, J = 6 Hz, CH₃, Val), 1.48 (2H, m, CH₂, Arg), 1.83 (2H, m, CH₂, Arg), 1.85 (1H, m, CH, Val), 2.62 (1H, dd, J = 17 and 6 Hz, CH₂, d-Phe) 2.80 (3H, s, CH₃, N-Me Val), 2.6-2.9 (3H, m, CH₂, d-Phe, and Asp), 3.07-3.13 (2H, m, CH₂, Arg), 3.45 (1H, d, J = 14 Hz, Gly), 3.83 (1H, m, CH, Arg), 4.04 (1H, d, J = 14 Hz, Gly), 4.23 (1H, d, J = 11 Hz, CH, Val), 4.47 (1H, t, J = 6 Hz, CH, d-Phe), 4.6-4.8 (under HOD peak, CH₂, hydroxylamine ether tether), 5.09 (1H, t, J = 7Hz, CH, Asp),7.18 (2H, d, J = 8 Hz, CH, Phe), 7.29 (1H, d, J = 8 Hz, CH, Phe).

Conjugation of Cyclic-(N-Me-VRGDf-NH)-Tether 9 to Doxsaliform to Yield Cyclic-(N-Me-VRGDf-NH)-Doxsaliform 10. 5″-Formyldoxsaliform (4 mg) was dissolved in a 3:1 mixture (2 mL) of water:ethanol (v/v, pH 2.0 TFA), and 8 mg cyclic-(N-Me-VRGDf-NH)-tether 9 was added. The solution was stirred at room temperature for 5 hours until one peak with absorbance at 480 nm was observed by analytic HPLC. The product was purified by preparative HPLC to yield 4.2 mg of pure cyclic-(N-Me-VRGDf-NH)-doxsaliform 10. The conjugate was concentrated under vacuum at room temperature and stored as a red solid at −80°C. HPLC analysis showed one predominate peak (retention time, 12.7 minutes, 96%); ESI-MS, m/z 1,379.6, calculated for (M + H⁺) 1,379.54. To assign the ¹H NMR spectrum the following spectra were obtained in DMF-d₇; ¹H NMR, COSY, HSQC, HMBC, ROESY. ¹H NMR: 0.41 (3H, d, J = 6 Hz, CH₃, Val), 0.77 (3H, d, J = 6 Hz, CH₃, Val), 1.14 (3H, d, J = 7 Hz, CH₃, 5′), 1.45–1.51 (2H, m, CH₂, Arg), 1.87–1.93 (2H, m, CH₂, Arg), 1.93–1.95 (2H, m, CH₂, 2′), 2.03–2.05 (1H, m, CH, Val), 2.12 (1H, dd, J = 6 and 15 Hz, CH₂, 8), 2.22 (1H, m, CH₂, 8), 2.46 (1H, dd, J = 6 and 17 Hz,CH₂, d-Phe), 2.74 (3H, s, CH₃, N-Me Val), 2.80 (2H, m,under DMF peak, CH₂, 10), 2.9–2.99 (2H, m, CH₂, Asp),3.13–3.19 (3H, m, CH₂ and CH, Arg, and 3′), 3.32 (1H, d,J = 14 Hz, CH₂, Gly), 3.74 (2H, m, CH, 9 and Arg), 3.95 (3H, s, CH₃, 4, O-Me), 3.97 (1H, m, CH₂, Gly), 4.20 (1H, q, J = 7 Hz, CH, 5′), 4.34 (1H, d, J = 11 Hz, CH, Val), 4.51 (1H, t, J = 7 Hz, CH, d-Phe), 4.60 (4H, two singlets, CH₂, 14,and hydroxylamine ether tether), 4.73 (1H, d, J = 13 Hz, CH₂, N-Mannich base), 4.83 (1H, d, J = 13 Hz, CH₂, N-Mannich base), 4.92 (1H, dd, J = 6 and 9 Hz, CH, Asp), 5.0 (1H, bs, CH, 7), 5.34 (1H, bs, CH, 1′), 6.92 (1H, d, J = 9 Hz, CH, 3′), 7.04 (2H, d, J = 9 Hz, CH, d-Phe), 7.50 (2H, d, J = 9 Hz, CH, d-Phe), 7.55 (1H, dd, J = 3 and 9 Hz, CH, 4′), 7.61 (1H, dd, J = 2 and 7 Hz, CH, 3), 7.85 (1H, under DMF peak, CH, 2), 7.91 (1H, under DMF peak, CH, 1), 8.03 (1H, d, J = 3 Hz, CH, 6′), 8.14 (1H, s, NH, 4-Phe).

Conjugation of Cyclic-(Me-VRGDf-NH)-Tether 9 to Oregon Green to Yield Cyclic-(N-Me-VRGDf-NH)-Oregon Green 11. To a dry-25 mL round-bottomed flask was added 2 mg of 5′-Oregon Green 488 N-hydroxysuccinimide, 1 eq of cyclic-(Me-VRGDf-NH)-tether (2.7 mg) and 5 mL of anhydrous DMF. The solution was allowed to stir for 5.5 hours while monitored by HPLC. Solution was then concentrated under vacuum and the residue resuspended in methanol for purification by preparative HPLC, which yielded 1.1 mg (26%) of a solid yellow product. Analytic HPLC showed a single peak (retention time, 16.9 minutes); ESI-MS, m/z 1,071.5, calculated for (M + H⁺) 1,071.36.

Cell Culture. Human breast carcinoma cell line MDA-MB-435 (34) was maintained in DMEM medium supplemented with 10% fetal bovine serum, penicillin (100 units/mL), streptomycin (0.1 mg/mL), l-glutamine (2 mmol/L), sodium pyruvate (1 mmol/L), nonessential amino acids, and vitamins.

Purified Proteins. Human vitronectin and bovine serum albumin (BSA; A-7030) were purchased from Sigma (St. Louis, MO). α_vβ₃-Specific monoclonal antibody LM609 was purchased from Chemicon (Temecula, CA).

Cell Adhesion Assay. Cell adhesion was determined by coating wells of 96-well plates (Corning, New York, NY) with 100 μL of 5 μg vitronectin/mL in Dulbecco's PBS from 2.0 hours to overnight at room temperature. Wells were washed twice with deionized water and nonspecific binding sites were blocked with 200 μL heat-inactivated (20 minutes at 60°C) 1.0% BSA in Dulbecco's PBS from 2.0 hours to overnight at 37°C. Wells were washed five times with deionized water and allowed to dry for 30 minutes at room temperature or stored at 4°C for extended periods. Cells were harvested from a subconfluent T-175 tissue culture flask by rinsing with 35 mL Dulbecco's PBS and incubating with 2 mL of 4 mmol/L EDTA for 3 minutes at 37°C. The EDTA solution was neutralized by adding 48 mL of DMEM containing penicillin-streptomycin. Cells were washed once with 50 mL DMEM + penicillin-streptomycin and resuspended in DMEM + penicillin-streptomycin at a final concentration of 8.5 × 10⁵ cells/mL. MnCl₂ was added, resulting in a final concentration of 500 μM. Peptides or antibody was added prior to adding cells (100 μL) to 96-well plates. Cells were allowed to adhere for 90 to 100 minutes at 37°C. Nonadherent cells were removed by aspiration and washing twice with Dulbecco's PBS containing 900 μM Ca²⁺ and 500 μM Mg²⁺. Adherent cells were quantified via measuring cellular metabolism of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Promega, Madison, WI) at 37°C. For every experiment, each condition was done in triplicate; experiments were done at least twice.

α_vβ₃ Binding Assessed by Flow Cytometry with Cyclic-(N-Me-VRGDf-NH)-Oregon Green. Cells were harvested from five subconfluent T-175 cell culture flasks by rinsing with 35 mL Dulbecco's PBS and incubating with 2 mL of 4 mmol/L EDTA for 3 minutes at 37°C. The EDTA solution was neutralized by adding 40 mL Dulbecco's PBS. Cells were washed once with 9 mL Dulbecco's PBS and resuspended in Dulbecco's PBS + 0.5% BSA or Dulbecco's PBS + 0.5% BSA, 500 μM MnCl₂, and 500 μM MgCl₂. Various concentrations of cyclic-(N-Me-VRGDf-NH)-Oregon Green were added and allowed to incubate for 90 minutes at 37°C. Cells were washed twice, resuspended in 500 μL Dulbecco's PBS + 0.5% BSA and analyzed by flow cytometry on a MoFlo (Ft. Collins, CO) flow cytometer. Cells were analyzed with excitation at 488 nm (Ar ion laser), with emission monitored between 510 and 550 nm. Ten thousand cells were analyzed per condition. The data are presented as the mean fluorescence for each condition with the background, drug-free cell fluorescence subtracted.

Uptake of Doxorubicin, Doxsaliform, and Acyclic-RGD-4C-Doxsaliform. A flow cytometry method of measuring uptake of doxorubicin, doxsaliform, and acyclic-RGD-4C-doxsaliform in breast cancer cells was done as previously described, with modifications (35, 36). MDA-MB-435 breast cancer cells in log phase growth were dissociated with trypsin-EDTA, counted, resuspended in medium at 2× 10⁵ cells/mL, and plated into six-well plates (5 × 10⁵ cells per well) and allowed to adhere overnight. Drug solutions of doxorubicin, doxsaliform, and ayclic-RGD-4C-doxsaliform were prepared in DMSO with 1% acetic acid at 50 μM. Before treatment with drug, cell medium was removed, cells were washed with HBSS (0.5 mL), and then fresh cell medium, with or without 500 μM Mn²⁺, was placed into the wells (2 mL). Drug treatments of 0.5 μM were accomplished by the addition of 20 μL from the drug solutions to the desired wells and incubation for various amounts of time (20, 40, and 60 minutes). For each time point, the cell medium was removed, cells were washed with HBSS and trypsinized, and trypsinization was quenched with 5 mL of cell medium at 4°C. Cells were pelleted by centrifugation at 200 × g for 5 minutes at 10°C. The supernatant was decanted, and the cells were resuspended in 5 mL of Dulbecco's PBS at 4°C, repelleted, resuspended in 2 mL of Dulbecco's PBS at 4°C, and placed on ice. Drug treatments were done in such a manner that all cell treatment times would end at approximately the same time to ensure comparable measurements with the FACScan instrument. The amount of drug uptake was measured by flow cytometry on a Becton Dickinson Biosciences FACScan (San Jose, CA) flow cytometer using Becton Dickinson Biosciences CellQuest software (version 3.3). Cells were analyzed with excitation at 488 nm (15 mW Ar ion laser), with emission monitored between 570 and 600 nm. Instrument settings were optimized for the cell line and held constant for all experiments; for anthracycline fluorescence analysis, 10,000 cells were analyzed for each sample. The data are presented as the mean fluorescence for each condition divided by the background, drug-free mean fluorescence.

Growth Inhibition Assay. Growth inhibition was determined as previously described (9) with minor modifications. Cells were treated for 4 hours, then allowed to grow until control wells reached ∼80% confluence (4–5 days). Cells were quantified by measuring crystal violet staining or cellular metabolism of MTT. For every experiment, each condition was done in hexaplicate; experiments were done at least twice.

We began our search for an α_vβ₃-targeting group for doxsaliform with RGD-4C because of the success of Arap and coworkers with bicyclic-RGD-4C conjugated to doxorubicin (28). The two predominant bicyclic structures are formed by oxidation of the four thiols to two disulfide bridges (37). We observed that on formation of the disulfide bridges, RGD-4C became poorly water soluble over a range of pH. Strong evidence for the molecular structure of doxorubicin-RGD-4C prepared by Arap and coworkers was not reported (28). Based on the observed change in solubility on formation of the disulfide bridges, we hypothesized that acyclic-RGD-4C was the actual peptide that targeted Arap and coworkers' phage to MDA-MB-435 tumors in mice. Linear RGD-containing peptides are known to have a short circulation time in the bloodstream due to the activity of proteases, but because targeted delivery to tumor is relatively rapid, we sought to test this peptide-drug conjugate. The synthetic strategy for acyclic-RGD-4C-doxsaliform outlined in Fig. 3, used an oximation reaction of a formyl group placed at the 5″ position of the salicylamide group of doxsaliform and a hydroxylamine ether tether at the carboxyl terminus of the peptide. The oximation reaction was regioselective for the aryl aldehyde and produced a robust connection between the targeting group and the salicylamide trigger, time-release group. Both acyclic-RGD-4C-tether and acyclic-RGD-4C-doxsaliform have good water solubility.

An attractive alternative to RGD-4C is the cyclic peptide, cyclic-(N-Me-VRGDf), developed by Merck (Rahway,NJ) as a selective and potent α_vβ₃ antagonist. The X-ray crystal structure of cyclic-(N-Me-VRGDf) bound to α_vβ₃ shows the d-phenylalanine and N-methyl valine directed toward solvent, making these residues attractive attachment points for conjugation of cytotoxin or other molecular probe (see Fig. 5; ref. 38). We chose to attach a short tether to the 4 position of d-phenylalanine because this would take advantage of the rigid nature of the aromatic ring, essentially creating a short linear tether and projecting the steric bulk of doxsaliform toward solvent. Cyclic-(N-Me-VRGDf-NH)-tether with the hydroxylamine functional group at the terminus of the tether was synthesized from start to finish in high yield with no chromatography as shown in Fig. 4. Again, the targeting group was connected to 5″-formyldoxsaliform via the oximation reaction. The conjugate, cyclic-(N-Me-VRGDf-NH)-doxsaliform, obtained in pure form after preparative HPLC, was stable for months while stored at −80°C as the TFA salt of its N-Mannich base.

Bicyclic-RGD-4C-tether (as a 50:50 mixture of the 1-4;2-3 and 1-3;2-4 isomers), acyclic-RGD-4C-tether, and cyclic-(N-Me-VRGDf-NH)-tether were assayed for their ability to bind the α_vβ₃ integrin present on viable MDA-MB-435 cells using a vitronectin competition assay (39). Vitronectin is the endogenous ligand for the α_vβ₃ integrin. Conditions for inhibition of MDA-MB-435 cell adhesion to vitronectin, including the requirement of Mn²⁺, were established by using an α_vβ₃-specific monoclonal antibody (LM609). Targeting compounds or targeted doxsaliform were added to cell suspensions created by release of cells from cell culture flasks with EDTA as opposed to trypsin to preserve the integrity of α_vβ₃. Drug-treated cells were then added to cell culture plates coated with BSA with or without vitronectin. Cells were allowed to adhere, wells were washed with Dulbecco's PBS, and cells were quantified by measuring cellular metabolism of MTT. Nonspecific binding (cells bound to BSA-coated wells) was subtracted from total binding (cells bound to BSA-and vitronectin-coated wells) to determine specific binding to vitronectin. The concentrations of compound required to inhibit binding of 50% of the cells to vitronectin (IC₅₀ values) are shown in Table 1. The acyclic-RGD-4C isomer was chosen over the bicyclic isomer for further experiments due to better water solubility and higher binding affinity for the α_vβ₃ integrin. Next, acyclic-RGD-4C-doxsaliform and cyclic-(N-Me-VRGDf-NH)-doxsaliform compounds were assayed for their ability to bind the α_vβ₃ integrin (Table 1). The binding affinities of both acyclic-RGD-4C-tether and cyclic-(N-Me-VRGDf-NH)-tether decreased by only one order of magnitude on addition of doxsaliform, indicating that the tethering system does not preclude binding. IC₅₀ values for both acyclic-RGD-4C-doxsaliform and cyclic-(N-Me-VRGDf-NH)-doxsaliform in the vitronectin assay are significantly lower than those for the RGD-4C-doxorubicin conjugate with the plasmin-cleavable tether, pioneered by de Groot and coworkers (ref. 40; 10 and 5 nmol/L versus 150 nmol/L).

Cyclic-(N-Me-VRGDf-NH)-tether was also analyzed for binding to α_vβ₃ on MDA-MB-435 cells as a function of Mn²⁺ activation with cyclic-(N-Me-VRGDf-NH)-tether bound to Oregon Green fluorescent dye (see Figs. 4 and 6). Binding as a function of cyclic-(N-Me-VRGDf-NH)-Oregon Green concentration in the presence and absence of Mn²⁺ was measured by flow cytometry, as shown in Fig. 6. The experiment was done with cells in suspension, released from the growth flask with EDTA. In the presence of Mn²⁺, binding of dye to cells increased with concentration of dye and plateaued at about 100 nmol/L. In the absence of Mn²⁺, little binding of dye was observed even at 100 nmol/L cyclic-(N-Me-VRGDf-NH)-Oregon Green, consistent with targeted dye binding to activated α_vβ₃. As reported in Table 1, the IC₅₀ for targeted dye binding to cells is 2 nmol/L, approximately midway between the values for cyclic-(N-Me-VRGDf-NH)-tether and cyclic-(N-Me-VRGDf-NH)-doxsaliform.

Uptake of acyclic-RGD-4C-doxsaliform by MDA-MB-435 cells was measured by flow cytometry after drug treatment for various periods of time in the presence and absence of additional Mn²⁺ beyond that present in fetal bovine serum. Concentration of targeted drug relative to doxorubicin and doxsaliform in cells was determined from emission of the doxorubicin fluorophore. After a 1-hour drug treatment time, doxorubicin was taken up 3-fold more than acyclic-RGD-4C-doxsaliform, as shown in Fig. 7, and uptake of acyclic-RGD-4C-doxsaliform was independent of additional Mn²⁺. In Table 2, uptake of acyclic-RGD-4C-doxsaliform is compared with uptake of doxorubicin and doxsaliform at two time points (30 minutes and 4 hours) and three drug concentrations (100, 500, and 1,000 nmol/L) in the absence of additional Mn²⁺. At the 30-minute time point, ∼30% of the time-release trigger of acyclic-RGD-4C-doxsaliform or doxsaliform had fired, and at the 4-hour time point, more than 90% of the trigger had fired based on the known half-life for the trigger (9). After treatment for 30 minutes with 500 nmol/L drug, uptake of fluorophore from acyclic-RGD-4C-doxsaliform was 60% of doxorubicin and 20% of doxsaliform. However, after treatment for 4 hours with 500 nmol/L drug, uptake of fluorophore from acyclic-RGD-4C-doxsaliform and doxorubicin was comparable and uptake of fluorophore from doxsaliform was only 2-fold higher. These results suggest that acyclic-RGD-4C-doxsaliform does not significantly penetrate the cell membrane and that the doxorubicin fluorophore only enters after the trigger releases the doxorubicin-formaldehyde conjugate.

Acyclic-RGD-4C-doxsaliform and cyclic-(N-Me-VRGDf-NH)-doxsaliform were also assayed for their ability to inhibit growth of MDA-MB-435 cells relative to doxorubicin and doxsaliform. Cells treated in cell culture plates and nontreated (control) cells were allowed to grow to near confluency and then quantified via measuring crystal violet staining or cellular metabolism of MTT. The concentrations of drug required to inhibit growth of cells by 50% (IC₅₀ values) are shown in Table 1 as a function of drug treatment time. The data in Table 1 were obtained in the absence of additional Mn²⁺ because control experiments showed no effect from Mn²⁺ on cytotoxicity. Both acyclic RGD-4C-doxsaliform and cyclic-(N-Me-VRGDf)-doxsaliform are more cytotoxic than clinical doxorubicin and comparable in cytotoxicity to doxsaliform with a drug treatment time of 4 hours. With shorter drug treatment times, the cytotoxicities of targeted drugs and doxorubicin are comparable. The slight decrease in cytotoxicity observed for cyclic-(N-Me-VRGDf-NH)-doxsaliform relative to doxsaliform is comparable to the loss relative to parent drug observed by de Groot and coworkers (31, 41). Earlier control experiments established that the miniscule amounts of formaldehyde that would be released even from complete hydrolysis of the conjugate would contribute nothing to the observed growth inhibition (8). Conjugation of cytotoxic drugs to triggers and targeting groups often causes a drop in cytotoxicity (42).

Doxsaliform prodrug-RGD conjugates were synthesized and evaluated for binding to α_vβ₃ in the vitronectin cell adhesion assay and for inhibition of MDA-MB-435 cancer cell growth. We hypothesized that a prodrug with this design would bind α_vβ₃ and localize in/or near the tumor and vascular endothelial cells of the developing blood supply. Upon hydrolysis of the N-Mannich base, the conjugate would release the doxorubicin active metabolite locally. Local delivery is required because the active metabolite has a short lifetime with respect to further hydrolysis to doxorubicin (half-life, ∼5 minutes; ref. 8). The advantage of delivering the doxorubicin active metabolite is that it is more cytotoxic to both sensitive and resistant tumor cells.

RGD-4C as a targeting group was explored first because of significant activity in tumor-bearing mice reported for RGD-4C-doxorubicin conjugates with the peptide in its oxidized form (28). The structures for the conjugates, however, were not well defined by the synthetic strategy or from spectroscopic data. A later report established that 1-4;2-3-bicyclic-RGD-4C has an order of magnitude better affinity for α_vβ₃ than the other major regioisomer, 1-3;2-4-bicyclic-RGD-4C (37). Oxidation of acyclic-RGD-4C-tether gave roughly a 50:50 mixture of the 1-4;2-3- and 1,3;2,4-bicyclic isomers. We found that acyclic-RGD-4C-tether (2) had better affinity for α_vβ₃ (Table 1) and much better aqueous solubility than a 50:50 mixture of the two regioisomers of bicyclic-RGD-4C-tether. The result that acyclic-RGD-4C bound with higher affinity than the mixture of bicyclic isomers is surprising because formation of the disulfide bridges makes the structure more rigid. Based on this result, we selected acyclic-RGD-4C-tether for conjugation to doxsaliform. Acyclic-RGD-4C-doxsaliform conjugate exhibited a decrease in affinity for α_vβ₃ relative to the peptide alone (10 nmol/L versus 1 nmol/L), but cytotoxicity against MDA-MB-435 cells (IC₅₀ = 50 nmol/L) was comparable to that of doxsaliform. Comparison of uptake of doxorubicin fluorophore by MDA-MB-435 cells treated with either acyclic-RGD-4C-doxsaliform, doxorubicin, or doxsaliform as a function of treatment time suggests that targeted drug does not penetrate the plasma membrane. Appearance of doxorubicin fluorophore from targeted drug in cells requires release of the doxorubicin-formaldehyde conjugate by the salicylamide trigger.

Because acyclic-RGD-4C-doxsaliform has the potential for instability due to 4 sulfhydryl groups, we also explored a cyclic RGD peptide with the cycle created via a peptide linkage between the amino and carboxyl termini, cyclic-(N-Me-VRGDf). Although a variant, cyclic-(KRGDf), has been used to attach various molecules to the α_vβ₃-targeting peptide at the ε-amino group of the Lys, we sought to use the linear d-phenylalanine of cyclic-(N-Me-VRGDf) as an attachment point guided by the cocrystal structure of the ligand-binding domain of α_vβ₃ bound to cyclic-(Me-VRGDf; Fig. 5). Based on molecular modeling of our conjugate bound to α_vβ₃, we hypothesized this linear tether would permit attachment of a large molecule without a significant decrease in binding affinity to α_vβ₃. Indeed, cyclic-(N-Me-VRGDf-NH)-doxsaliform exhibited an IC₅₀ in the vitronectin binding assay of 5 nmol/L. Furthermore, a conjugate of cyclic-(N-Me-VRGDf-NH) with Oregon Green showed dose- and Mn²⁺-dependent binding to MDA-MB-435 cells by flow cytometry. The cancer cell growth inhibition by cyclic-(N-Me-VRGDf-NH)-doxsaliform is better than doxorubicin but reduced two times relative to doxsaliform. A higher IC₅₀ relative to doxsaliform is attributed to a reduced rate of uptake because the targeted drugs do not seem to penetrate the plasma membrane, and for uptake, the salicylamide trigger must first release the doxorubicin-formaldehyde conjugate.

The likely scenario for prodrug activity in vivo based on these experiments would be binding to α_vβ₃ overexpressed by tumor and/or tumor vascular endothelial cells during circulation followed by hydrolysis of the N-Mannich base releasing the doxorubicin active metabolite extracellularly. The active metabolite should enter the cell more rapidly than free doxorubicin due to its lack of charge, then induce apoptosis via the formation of covalent cross-links in cellular DNA (43–45). Possibly, some conjugate could be internalized via receptor-mediated or fluid-phase endocytosis and hydrolyzed to the active metabolite intracellularly. Because the active metabolite is not cationic, as opposed to doxorubicin, the P-170 drug efflux pump resistance mechanism would likely have less effect (11). Similarly, resistance mechanisms that suppress oxidative stress and the production of formaldehyde will have little effect because the active metabolite released by the trigger already has formaldehyde incorporated.

Both RGD-doxsaliform conjugates have good affinity for α_vβ₃ and are more cytotoxic than clinical doxorubicin. The salicylamide N-Mannich base trigger hydrolyzes with a half-life of 60 minutes (9), which is appropriate for the rate of targeted drug delivery to tumor (28). Both RGD-targeted drug designs show good water solubility and are promising candidates for in vivo testing in tumor-bearing nude mice.

We thank Prof. Renata Pasqualini (M.D. Anderson Cancer Center) for a sample of MDA-MB-435 cells, Dr. Richard Shoemaker for help with the NMR experiments, Prof. Katheryn Resing for help with MS experiments, and Theresa Nahreini for help with the flow cytometry experiments.