Biological evaluation of cryptophycin 52 fragment A analogues: Effect of the multidrug resistance ATP binding cassette transporters on antitumor activity

Cryptophycin 1, isolated from terrestrial cyanobacterium (Nostoc sp.), was found to possess significant antitumor activity (1–5). Cryptophycin 52 (LY355703), a synthetic derivative of the naturally occurring cryptophycin 1, was selected from a wide range of synthetic analogues for clinical evaluation as a cancer therapeutic. It differs structurally from cryptophycin 1 by an additional methyl group in the C fragment of the molecule (Fig. 1). Like the natural product, cryptophycin 52 inhibits the polymerization of microtubules and suppresses their dynamic instability. Cryptophycin 52 was reported to be the most potent suppressor of microtubule dynamics thus far evaluated. It binds to microtubule ends in vitro with high affinity and significantly reduced the rate and extent of microtubule shortening and growing at concentrations that did not significantly reduce the polymer mass or mean microtubule length (6–9). It was shown to have significant antiproliferative and cytotoxic activity on cancer cells. Cultured tumor cells exposed to picomolar concentrations of cryptophycin 52 were prevented from progressing through the cell cycle. These cells accumulated in mitotic metaphase and subsequently underwent apoptosis (10).

The antiproliferative activity of cryptophycin 52 against tumor cells in vitro was significantly more potent (40–400 times) compared with the clinically relevant antimicrotubule agents paclitaxel or the Vinca alkaloids, vinblastine and vincristine. Moreover, whereas paclitaxel and the Vinca alkaloids are sensitive to the multidrug resistance (MDR) transporters P-glycoprotein (P-gp, MDR-1) and/or MDR-associated protein (MRP-1), cryptophycin 52 seems to be relatively insensitive to these drug resistance mechanisms (10). These results correlate well with its in vivo efficacy against a wide range of human xenografts and murine solid tumors including ones expressing P-gp.

The clinical potential, interesting mode of action, and synthetically challenging structure of cryptophycins have made them a focus of several synthetic studies (11–23) and structure-activity studies (24–29). The β-epoxide of fragment A (cryptophycins 1 and 52) or their corresponding chlorohydrins (cryptophycins 8 and 55; Fig. 2) were found to be essential for the antiproliferative activity of the cryptophycins (30, 31). The clinical evaluation of cryptophycin 52 was done in a Cremophor EL formulation, which, although clinically acceptable, is less desirable than other formulations due to potential hypersensitivity reactions and/or the need to premedicate patients with antihistamines. We have described previously our synthetic preparation of a series of cryptophycin 52 fragment A analogues aimed at exploring the structure-activity relationship and improving the aqueous solubility of the molecule allowing for the modification of this formulation. Substitutions on the phenyl ring adjacent to the epoxide in fragment A were found to be well tolerated and produced derivatives with comparable activity with the parent compound when evaluated in vitro in the CCRF-CEM cell line (32, 33). In this report, the antitumor activity of several of these analogues was compared with the activity of cryptophycins 52 and 55 in both in vivo and in vitro models expressing a MDR phenotype.

Vinblastine, vincristine, and paclitaxel used in the in vitro studies were purchased from Sigma Chemical Co. (St. Louis, MO). All compounds were dissolved in DMSO as 2 to 10 mmol/L stocks and stored at −70°C. For experimental use, aliquots were thawed at room temperature and dilutions were prepared in DMSO at 1,000 or 10,000 times concentrates prior to dilution into cell culture medium and addition to the cells.

1. BDF male inbred mice (C57BL/6, C3H/He) were implanted bilaterally s.c. on day 0 with 30 to 60 mg tumor fragments of Panc-03.

2. Chemotherapy, delivered i.v. bolus, was started either within 3 days after implantation (early stage disease) or after the tumors had grown to a palpable size (200–700 mg, upstaged disease).

3. Tumors were measured with caliper once or twice weekly until either the tumors exceeded 1,600 mg or cure was assured. Tumor weights were estimated from two dimensional measurements: tumor weight (mg) = (a × b²) / 2, where a and b are tumor length and width (mm), respectively.

End points used to assess antitumor activity were as follows:

1. Tumor growth delay [treated versus control (T/C) value]

2. Tumor cell kill

3. Tumor growth inhibition (% T/C value)

Most cryptophycin analogues were formulated in 2% propylene glycol, 8% Cremophor EL. Chlorohydrins were acidified with 0.05% citric acid. All injections were carried out within 20 minutes of aqueous preparation from stock solutions.

The effects of the cryptophycin analogues were evaluated in two murine solid tumors, a pancreatic adenocarcinoma (Panc-03) and an Adriamycin-resistant mammary adenocarcinoma (Mamm-17/Adr). Panc-03 is a tumor model that is sensitive to a variety of oncolytic agents, such as the Vinca alkaloids and paclitaxel, which are substrates for MDR transporters. Mamm-17/Adr is a resistant tumor model that overexpresses P-gp, the best characterized member of the ATP binding cassette transport proteins associated with the MDR phenotype.

Except for those instances wherein the supply of compound was very limited, cryptophycin 52 analogues listed in Table 1 were initially evaluated in the drug-sensitive Panc-03 and, on showing a good efficacy profile, were subsequently tested in the Mamm-17/Adr-resistant tumor. All chemotherapeutic agents were given by the i.v. bolus route at the maximum tolerated doses (MTD). The effectiveness of the drugs was determined with two variables: % T/C and log kill values. A T/C ≤ 30% is considered significant antitumor activity, where T and C are the median tumor burden in the treatment group and control group, respectively, multiplied by 100. Tumor cell kill is determined from the difference in tumor growth delay in days, between the treated and control animals, to reach 1,000 mg in tumor weight (34, 35).

We compared the antiproliferative effects of the cryptophycin analogues in a human promyelocytic leukemia cell line HL-60/S and in two MDR cell lines derived from the parental HL-60/S line by selection in anticancer drugs. HL-60/Vinc (sometimes cited as HL-60/VCR) was selected for the MDR phenotype by growth in the presence of vincristine and has been shown to overexpress P-gp (36). HL-60/Adr, which overexpresses a second ATP binding cassette transport protein associated with the MDR phenotype, MRP-1, was selected by growth in the presence of doxorubicin (37, 38). The cell lines were kindly provided to us by Dr. Melvin Center (Kansas State University, Manhattan, KS).

The cell lines were grown at 37°C in the presence of 5% CO₂ in RPMI 1640 containing 25 mmol/L HEPES buffer (purchased either from BioWhittaker, Walkersville, MD or Life Technologies, Grand Island, NY) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT) and 2 mmol/L l-glutamine (BioWhittaker). Subcultures of cells were plated into Corning 24-well tissue culture plates at a density of 2 × 10⁴ cells/well and incubated 24 hours prior to compound addition. The IC₅₀ for antiproliferative activity of test compounds in both sensitive and resistant cell lines was determined using metabolic reduction of Alamar blue (Sensititer-Alamar, Westlake, OH) as a surrogate measure for cell number as described previously (10). The ratio of the IC₅₀ for the resistant line to the IC₅₀ of the sensitive parental line was used to calculate a resistance factor (fold difference).

Drug resistance is a major challenge facing the oncologist and leads to the failure of many modern chemotherapy treatments (39, 40). Therefore, we chose the murine Mamm-17/Adr tumor model to evaluate the effectiveness of cryptophycin analogues in tumors that show the MDR phenotype. Cryptophycins 52 and 55 were active against both sensitive (non-MDR) and resistant solid tumors as reflected by their T/C and log kill values (Table 1; Fig. 3). Cryptophycin 52 at 32 or 40 mg/kg had a log kill of 1.9 and 2.0 against both the sensitive Panc-03 and the MDR Mamm-17/Adr tumors. Cryptophycin 55, a chlorohydrin, was given at a MTD of ∼400 mg/kg, which was 10-fold higher than that of cryptophycin 52 (MTD, ∼40 mg/kg), but at this regimen, it was the most efficacious agent against Panc-03 producing >4.5 log kill value. It had a log kill of 1.6 to 1.9 against the Mamm-17/Adr tumor at 158 to 220 mg/kg. Paclitaxel and Adriamycin, on the other hand, showed good activity against Panc-03 with log kill values of 4.2 and 1.5, respectively, but were both ineffective in the Mamm-17/Adr (log kills, 0.41 and 0.55). Clearly, these results show that the cryptophycins are more effective agents than Adriamycin or paclitaxel in treating the multidrug-resistant tumor.

The 3-methyl substituted epoxide 1 and chlorohydrin 2 (T/C, 20% and 0%) were more active than the 3,4-dimethyl substituted chlorohydrin 3 (T/C, 60%). Similar to cryptophycin 55, chlorohydrin 2, at 3-fold higher doses than its corresponding β-epoxide 1, was better tolerated and could have been dosed even higher since no animal weight loss was observed at the 162 mg/kg dose. Although esters 5 and 6 were completely inactive, the methoxy derivative 4 was marginally active (T/C, 17%). The hydroxymethyl derivative 7 has proven to be quite efficacious and 5-fold more potent than cryptophycin 52 against Panc-03, with a MTD of 6 mg/kg and a T/C of 2%. Its efficacy and that of its chlorohydrin derivative 8 were unfortunately poor in the resistant Mamm-17/Adr tumor line, producing T/C values of only 20% to 30% at toxic doses. The glycine ester 9, also prepared from epoxide 7, was a more effective antitumor agent than either 7 or 8 in both tumor lines. Log kill values of 3.7 and 1.5 in Panc-03 and Mamm-17/Adr, respectively, were obtained with only ∼18 mg/kg total dose of this analogue.

Several analogues containing an amine functionality, prepared to impart additional aqueous solubility through their corresponding hydrochloride salts, have proven to be more potent in vivo than initially anticipated from their in vitro IC₅₀ values (32, 33). The MTDs for amines 10 to 13 ranged between 0.5 and 3 mg/kg compared with 30 to 400 mg/kg for cryptophycins 52 and 55. The diethyl amine epoxide 10, its corresponding chlorohydrin 11, and the diaminoethyl analogue 13 were all highly toxic and hence could not be properly evaluated. The piperizinyl dihydrochloride salt 12 and the sarcosine amide 14 had similar chemotherapeutic profiles. When given on days 3 and 5 at a total dose of 1.5 mg/kg, both agents effectively inhibited the growth of the pancreatic adenocarcinoma to produce log kill values of 4.4 and 4.7, respectively, but were both inactive in the Mamm-17/Adr tumor. Another approach for increasing the hydrophilicity of the molecule would have been through the sodium salt of an appropriate acid. With this goal in mind, acid 15 was prepared by a two-step oxidation of the corresponding alcohol, but it has proven to be marginally active. Finally, the contributions of the fragment A phenyl ring of cryptophycin 52 to the activity of these molecules was determined by its replacement with a heterocyclic ring as in derivative 16. The thiazole ring of analogue 16 again resulted in a loss of antitumor activity in the Panc-03 tumor line when compared with cryptophycin 52. These in vivo results clearly identified several cryptophycin 52 fragment A derivatives that are effective chemotherapeutic agents, but their great potency and activity in the murine pancreatic adenocarcinoma did not always transfer to the Mamm-17/Adr-resistant tumor.

The ATP binding cassette class of transporters is the best characterized family of proteins implicated in causing resistance to many known oncolytic agents, including the widely used antimicrotubule agents such as paclitaxel and the Vinca alkaloids (39, 40). Cryptophycins 1 and 52 have shown relatively little susceptibility to the MRP-1 or P-gp efflux pumps when evaluated in tumor lines overexpressing either transporter (10). We suspected that drug resistance due to the ATP binding cassette transporters was causing the wide discrepancies observed with the antitumor activities of the fragment A analogues in the sensitive versus the resistant murine solid tumor lines. To confirm this hypothesis, we tested several analogues for their antiproliferative activity in cells overexpressing either P-gp or MRP-1 to determine their sensitivities to the drug resistance mechanisms. A set of leukemia cell lines (HL-60) was used for the evaluation, wherein HL-60/S is the parental sensitive line and HL-60/Adr and HL-60/Vinc each overexpress MRP-1 and P-gp, respectively. IC₅₀ values were used to calculate a resistance factor (ratio of the IC₅₀ for the resistant line to the sensitive parental line) that was used to compare the sensitivity of the analogues, cryptophycins 52/55, and the antimitotic agents paclitaxel, vincristine, and vinblastine to the MDR transporters.

With the exceptions of analogues 1, 13, and 14, all fragment A analogues that were evaluated in the HL-60/S were found to be more potent than cryptophycins 52 and 55 (Table 2; Fig. 3). Similar relative potencies were observed in the in vivo tumor model studies (Table 1; Fig. 3) wherein the highest given nonlethal doses of analogues 1 and 7, for example, were 64 and 3.5 mg/kg, respectively, whereas the MTD for cryptophycin 52 was ∼30 mg/kg. With in vitro activity in the picomolar range across the three cell lines, all the cryptophycins were more potent than paclitaxel, vincristine, or vinblastine (Table 2; Fig. 3). Among the oncolytics tested, vincristine was most sensitive to the MRP-1 pump with a ∼10-fold difference in activity between HL-60/Adr and HL-60/S. Paclitaxel, vincristine, and vinblastine were much more susceptible to P-gp-mediated drug resistance and were 505-, 357-, and 106-fold less potent in the HL-60/Vinc than in the HL-60/S, respectively. Compared with these antimicrotubule agents, cryptophycins 52 and 55 were only 6-fold less active in the P-gp overexpressing line than in the parental cell line. Unfortunately, the resistance factors for analogues 12,13 and 14 were more within the range of paclitaxel's and vincristine's. With the exception of derivative 1, all other fragment A analogues fell in the range of 10- to 30-fold resistance.

The evaluation of cryptophycin 52 fragment A analogues using the murine tumors Panc-03 and Mamm-17/Adr was effective at differentiating the antitumor activity of these compounds. In these studies, several analogues compared favorably with cryptophycins 52 and 55 in the sensitive pancreatic murine solid tumor (the exceptions being thiazole 16 and the phenolic derivatives 4 and 6). Although cryptophycin 55 could be given at doses 7- to 13-fold higher than cryptophycin 52, some fragment A chlorohydrin analogues such as 12 and 14, containing a terminal secondary amine, were 20-fold more potent than cryptophycin 52. On the other hand, some such as 11 and 13 possessed a very narrow therapeutic window and were very toxic even at doses <2 mg/kg. Although several analogues improved the aqueous solubility of the molecule via their hydrochloride salts, they were therapeutically ineffective in the resistant Mamm-17/Adr tumor line. The in vitro IC₅₀ values obtained for these analogues in the HL-60 cell lines clearly show that a simple addition of a lipophilic methyl group such as in analogue 1 did not change its relative activity in the different cell lines. However, the incorporation of hydrophilic groups such as amines and hydroxy groups as in analogues 7, 9, 11 to 14, and 17 significantly affected their sensitivities to the drug resistance mechanisms. The same functional groups that rendered these analogues more potent and more water soluble also contributed to making them better substrates of the P-gp efflux pump.

This study illustrates the multifactorial forces that drove the structure-activity relationship around the A fragment of the cryptophycin molecule. The initial driving force in this structure-activity relationship was to increase the aqueous solubility and potency of the molecule to facilitate formulation issues. These goals were accomplished through the various amine substitutions on the phenyl ring of fragment A and their corresponding hydrochloride salts. However, we discovered that these substitutions resulted in greater sensitivity of the resulting molecules to the MDR transporter P-gp, which is an unacceptable feature in the development of a clinically relevant antitumor agent.