Effect of Nitroreduction on the Alkylating Reactivity and Cytotoxicity of

Mar 5, 2003 - Incubation of 8 in TES buffer provided polar products 20 and 21 (HPLC retention times 12.7 and 8.3 min, respectively, relative to 16.8 m...
0 downloads 5 Views 176KB Size
Chem. Res. Toxicol. 2003, 16, 469-478

469

Effect of Nitroreduction on the Alkylating Reactivity and Cytotoxicity of the 2,4-Dinitrobenzamide-5-aziridine CB 1954 and the Corresponding Nitrogen Mustard SN 23862: Distinct Mechanisms of Bioreductive Activation Nuala A. Helsby,* S. James Wheeler, Frederik B. Pruijn, Brian D. Palmer, Shangjin Yang, William A. Denny, and William R. Wilson Auckland Cancer Society Research Centre, The University of Auckland, Private Bag 92019, Auckland, New Zealand Received November 17, 2002

The dinitrobenzamide aziridine CB 1954 (1) and its nitrogen mustard analogue SN 23862 (6) are prodrugs that are activated by enzymatic nitroreduction in tumors. Bioactivation of 1 is considered to be due to reduction of its 4-nitro group to the hydroxylamine and subsequent formation of the N-acetoxy derivative; this acts as a reactive center, in concert with the aziridine moiety, to provide a bifunctional DNA cross-linking agent (Knox model). It is currently unclear whether bioactivation of 6 occurs by the same mechanism or results from the electronic effects of nitroreduction on reactivity of the nitrogen mustard moiety. To discriminate between these mechanisms, we have synthesized the hydroxylamine and amine derivatives of 1 and 6, plus related compounds, and determined their alkylating reactivities in aqueous solution, using LC/MS to identify reaction pathways. The relationships between substituent electronic effects, reactivity, and cytotoxicity were determined using the UV4 cell line, which is defective in nucleotide excision repair (thus avoiding differences in repair kinetics). Alkylating reactivity correlated with the electron-donating character of the ortho or para substituent in the case of the mustards, with a less marked electronic effect for the aziridines. Importantly, there was a highly significant linear relationship between cytotoxic potency and alkylating reactivity in both the aziridine and the mustard series, with the notable exception of 4, the 4-hydroxylamine of 1, which was 300-fold more toxic than predicted by this relationship. This demonstrates that the high potency of 4 does not result from activation of the aziridine ring, supporting the Knox model. The single-step bioactivation of 6, to amino or hydroxylamine metabolites with similar potency to 4, is a potential advantage in the use of dinitrobenzamide mustards as prodrugs for activation by nitroreductases.

Introduction CB 1954 (1) [5-(aziridin-1-yl)-2,4-dinitrobenzamide] (Scheme 1) originally came to our attention because of its excellent therapeutic activity against Walker rat 256 tumors (1). It was subsequently found to be a prodrug, activated by the two electron nitroreductase DT-diaphorase [NAD(P)H quinone oxidoreductase; NQO1; EC 1.6.99.2] in Walker 256 cells (2). The critical metabolite was shown to be the 4-hydroxylamine 4, which appears to be further activated by acetyl CoA to form a second reactive center that acts in concert with the aziridine moiety to form a cytotoxic bifunctional DNA interstrand cross-linking agent (the Knox model) (3). However, CB 1954 was not effective as an antitumor agent in humans, presumably because it is a poorer substrate for human than rat DT-diaphorase (4). Bioactivation of CB 1954 by rat DT-diaphorase led to suggestions of new therapeutic approaches involving introduction of the rat enzyme into tumors using antibody-enzyme conjugates (5) or gene therapy (6). It was subsequently found that the Escherichia coli aerobic nitroreductase (NTR, the product of the nfsB * To whom correspondence should be addressed. Tel: + 64-9-3737599 ext 86090. Fax: +64-9-373-7571. E-mail: [email protected].

gene) is a better candidate for such applications; while it reduces both the 2- and the 4-NO2 groups of CB 1954, it generates the critical 4-NHOH metabolite with faster kinetics than does rat DT-diaphorase (4, 5). CB 1954/ NTR is currently in clinical trials as a means of arming adenoviral vectors for gene therapy of cancer (7, 8). CB 1954 is also a substrate for oxygen sensitive one electron reductases (e.g., NADPH cytochrome P450 oxidoreductase, EC 1.6.2.4) (9) and consequently shows selective toxicity under hypoxic conditions in some tumor cell lines (10, 11). This selective bioactivation is of potential interest because hypoxia is an exploitable feature of tumors (12-16). We have reported the nitrogen mustard analogue of CB 1954, SN 23862 (6; Scheme 1) as an alternative dinitrobenzamide prodrug. The design of this compound was conceived as a way of exploiting the large change in electron density in a benzene ring on reduction of an electron-withdrawing nitro group to an electron-donating hydroxylamine or amine; this “electronic switch” is used to increase the reactivity of a prepositioned nitrogen mustard moiety (17-19). The mustard 6, and analogues with other benzamide side chains or mustard leaving groups, was shown to be selectively toxic to tumor cells

10.1021/tx025662b CCC: $25.00 © 2003 American Chemical Society Published on Web 03/05/2003

470

Chem. Res. Toxicol., Vol. 16, No. 4, 2003

Helsby et al.

Scheme 1. Proposed Mechanisms of Reductive Activation of the Dinitrobenzamide Aziridine Prodrug CB 1954 (1) and Its Nitrogen Mustard Analogue SN 23862 (6)a

a Bioactivation of 1 is due to reduction of the 4-nitro group to the hydroxylamine 4 followed by acetylation to form a second reactive center that can, in concert with the aziridine group, cross-link DNA (the Knox model). Bioactivation of 6 could occur via a similar mechanism (pathway B) or via an “electronic switch”, which increases the reactivity of the prepositioned mustard moiety and forms a DNA crosslinking moiety 7 (pathway A).

under hypoxic conditions (20) and was subsequently shown to be a superior substrate to 1 for E. coli NTR (21). NTR reduces only the 2-nitro group of 6 to the hydroxylamine 7, and this is presumed to be responsible for its marked cytotoxic bioactivation by NTR (21). Other advantages of the nitrogen mustards over the aziridines as prodrugs for hypoxia or NTR include their insensitivity to activation by DT-diaphorase (which has the potential to bioactivate CB 1954 in normal tissues) and their superior bystander effects as demonstrated by very efficient killing of NTR negative cells in mixtures with NTR-expressing cells in multicellular layer cultures or tumor xenografts (22). Although the suggested bioactivation of the nitrogen mustard moiety of 6 by nitroreduction is plausible, there has been no direct experimental demonstration that nitroreduction increases the alkylating reactivity of the mustard group or that this is responsible for bioactivity. In fact, it has been reported that acetyl CoA potentiates DNA damage by the 2-hydroxylamine 7, suggesting that bioactivation of 6 may depend on formation of a reactive hydroxylamine as for the aziridine 1 (21). Thus, the mechanism of bioactivation of 6 is controversial. To clarify the contribution of the two candidate mechanisms of bioactivation (i.e., electronic effects of nitroreduction on reactivity of the alkylating moiety vs formation of a new reactive center with acetyl Co A), we have synthesized the amine and hydroxylamine derivatives of SN 23862 and CB 1954 and related compounds. We have measured their alkylating reactivity, using LC/MS to identify the reaction products, and have examined the relationships between substituent electronic effects, re-

activity, and cytotoxicity. The latter studies have utilized a cell line, UV4, defective in nucleotide excision repair (23) as a result of mutation of the ERCC1 gene (24) in order to minimize any differences between compounds in the repair of DNA adducts.

Experimental Procedures Compounds. CB 1954 (1) was supplied by Prof. M. Jarman (Institute for Cancer Research, Sutton, U.K.), and 3 and 5 were supplied by Prof. Richard J. Knox (Enact Pharma plc, Porton Down Science Park, Salisbury, Wiltshire SP4 6BU, U.K.). Synthesis of 2 and 4 was as described previously (25). SN 23862 (6) and the reduction products 7-10 and 34 were synthesized as previously described (20, 26). Compound 12 was synthesized as described previously (27). Compound 11 was prepared as follows. A mixture of 5-[bis(2-chloroethyl)amino]-2-nitrobenzamide (13.1 mg, 0.043 mmol) (20) and SnCl2 (32.0 mg, 0.171 mmol) in 6 N HCl (5 mL) was stirred at room temperature for 6 h. The solvent was removed in vacuo, and the residue was partitioned between ice-cold diethyl ether and concentrated aqueous ammonia. The ether portion was washed well with cold water, dried over sodium sulfate, and added to methanol, which had been saturated with gaseous HCl. The solution was concentrated to dryness, and the residue was crystallized from methanol/ether to give the dihydrochloride salt of 11 as a white powder (12.4 mg, 83%); mp >300 °C. 1H NMR δ (D2O): 7.32 (br d, J ) 8.8 Hz, 1H), 7.19 (br s, 1H), 7.11 (br d, J ) 8.8 Hz, 1H), 3.85 (br t, J ) 6.0 Hz, 4H), 3.77 (br t, J ) 6.0 Hz, 4H). The unstable hydroxylamines 2, 4, and 7 were stored at -80 °C in DMSO purged with nitrogen containing 1 mM ascorbate. LC/MS prior to use routinely confirmed the purity of these compounds. Alkylating Reactivity. The rate of loss of each analogue was determined at 37 °C in sterile aqueous solutions buffered

Reductive Activation of Dinitrobenzamide Alkylators with 50 mM Bis-Tris [bis(2-hydroxyethyl)imino-tris(hydroxymethyl)methane] (pH 6.5) or 100 mM TES [Tris(hydroxymethyl)methylamino-1-ethanesulfonic acid] (pH 7.4), both containing 0.1 mM EDTA, with addition of ascorbate (1 mM) in the case of the hydroxylamines (2, 4, and 7). Compounds (final concentration 5-50 µM) were diluted into the buffer (4 mL) to start the reaction. Aliquots (250 µL) were removed at various times and assayed by HPLC or LC/MS. The reactions were quenched after sampling by addition of ice-cold ammonium formate buffer (45 mM; pH 3.5), and the samples were frozen in liquid nitrogen and stored at -80 °C until HPLC analysis. For the hydroxylamines (2, 4, and 7), reactions were performed in real time at 37 °C within the thermostated autosampler compartment of the Agilent 1100 HPLC using buffer and drug solutions preequilbrated at 37 °C. HPLC. HPLC analyses used an Agilent 1100 system with a Waters 8 mm × 100 mm µ-Bondapak C18 column or Alltech Altima (4.6 mm × 150 mm or 3.2 mm × 150 mm) C8 5 µm column. The mobile phase was comprised of ammonium formate (45 mM, pH 4.2, or 450 mM, pH 6.5) and 80% acetonitrile in water, using nonlinear gradients at flow rates of 0.5-1 mL/min. Detection was by diode array absorbance, using a signal giving maximal absorbance for the compound of interest (237-372 nm) with a bandwidth of 4 nm. The reference signal was 550 nm with a bandwidth of 50 nm. LC/MS. On-line mass spectra were obtained with an Agilent 1100 HPLC interfaced to a single quadrupole mass spectrometer (Agilent LC/MSD, model A). Products of 1, 2, 4, 6, and 7 were identified using negative mode atmospheric pressure chemical ionization (APCI), with nitrogen as the nebulizing and drying gas at a flow rate of 5 L/min and a nebulizing pressure of 55 psi. The gas temperature was 350 °C, the vaporizer temperature was 450 °C, the capillary voltage was 4000 V, and the corona current was 4 µA. The mass/charge (m/z) ratio was scanned from 100 to 1000. For products of 3 and 8, the mass detector was operated in electrospray (ES) positive mode, with a fragmentor voltage of 60 V. Kinetic Analysis. The pseudo first-order kinetics of loss of starting compound (kloss) was fitted by linear regression. When parallel reactions were observed (e.g., competing nucleophiles) or subsequent reactions were detected, a kinetic modeling program (Model Maker, version 4.0, Sherwell Scientific Publishing, Ltd.) was used to estimate rate constants for the individual reactions. Cytotoxicity Assays. We have previously reported the IC50 values of SN 23862 (6) and its reduction products 8, 9, 10, and 23 in the UV4 cell line (26) and the IC50 values of CB 1954 (1) and its reduction products 3 and 5 (28). Cytotoxicities of 2, 4, 7, and 11 were determined using the same method. Briefly, cells were grown in minimal essential medium (R-MEM) containing 5% fetal bovine serum in a 5% CO2 incubator (37 °C, 24 h) at a seeding density of 200 UV4 cells in 150 µL per well. Compounds were diluted from DMSO stock solutions into culture medium and then added to cells in five 2-fold dilutions. After 4 h, drugs were removed by washing cultures with fresh medium and the trays were incubated for a further 4 days. Cell density was determined by staining with sulforhodamine B as described previously (29), and the IC50 was calculated as the drug concentration providing 50% inhibition of growth relative to controls on the same plate.

Results and Discussion Alkylating Reactivity of Dinitrobenzamide Mustard Derivatives. The reactivities of the nitro compounds and their reduction products were compared by monitoring their kinetics of loss due to reaction with nucleophiles (OH- and buffer constituents) in aqueous solution at 37 °C. Product analysis (by mass spectrometry and comparison of HPLC retention and absorbance spectra with synthetic standards) and kinetic modeling

Chem. Res. Toxicol., Vol. 16, No. 4, 2003 471

Figure 1. Loss of the 2-amino-4-nitrobenzamide mustard 8 (b) and formation of the corresponding mono-ol 20 (2), diol 21 (9), TES adduct 30 (∆), and the hydrolysis product of the latter 31 (0) during incubation in TES buffer (pH 7.4) containing 0.1 mM EDTA at 37 °C, with quantitation by HPLC. Solid lines are fits to the kinetic model shown in Scheme 2.

were used to delineate the competing and sequential reactions in order to assign individual rate constants. The general approach is illustrated with 8, the 2-amine reduction product of chloromustard SN 23862 (6), in Figure 1 and Scheme 2. Incubation of 8 in TES buffer provided polar products 20 and 21 (HPLC retention times 12.7 and 8.3 min, respectively, relative to 16.8 min for 8), which were identified by LC/MS as the mono-ol (20, [M + H]+ ) 303) and the diol (21, [M + H]+ ) 285) hydrolysis products. The consecutive loss of the characteristic chlorine satellite peaks in the mass spectra confirmed identification of these compounds. The TES sulfonate monoconjugate 30 (retention time 9.4 min) was identified by LC/MS ([M + H]+ ) 514), and its hydrolysis product 31 was tentatively assigned by comparison of HPLC retention time (5 min) and similarity of absorbance spectra with 30. The kinetics (Figure 1) were well-fitted assuming the relationships shown in Scheme 2, with 20 and 30 behaving as intermediates, confirming the structural assignments. The fitted first-order rate constants are shown in Scheme 2. The overall loss of 8 (kloss ) 1.8 h-1) is slightly faster than formation of the mono-ol 20 (k1 ) 1.49 h-1) because of competing formation of the sulfonate conjugate 30 (k2 ) 0.35 h-1). The hydrolysis reactions (k1 and k3) proceed faster than the sulfonate conjugation reactions (k2 and k5); given that the sulfonate anion (44 mM) is present in higher concentration than the hydroxide ion at pH 7.4 (0.25 µM), the product distribution must reflect the greater nucleophilicity of the hydroxide ion. Further experiments examining hydrolysis of 8 showed that there was no significant difference in kloss over the pH range of 4.2-8.5 (data not shown); the insensitivity to OH- concentration is consistent with an SN1 mechanism confirming earlier findings (27, 30-32) that the rate-limiting step for aromatic mustard hydrolysis is formation of an aziridinium ion intermediate. Because OH- and TES- compete for a common intermediate, kloss is the appropriate measure of overall alkylating reactivity (kalk) for this compound. In contrast to the 2-amino derivative 8, the dinitro parent compound SN 23862 (6) was very stable in aqueous solution, as noted previously (28), demonstrating the strong activating effect of resonant amino groups on

472

Chem. Res. Toxicol., Vol. 16, No. 4, 2003

Helsby et al.

Scheme 2. Proposed Reaction Scheme of the Products Formed during Incubation of 8, the 2-Amine Derivative of SN 23862, in TES Buffer, pH 7.4, at 37 °C

the alkylating reactivity of the mustard moiety. Traces of three more polar products were detected, with retention times, absorbance spectra, and mass spectra consistent with the mono-ol 16 ([M - H]- ) 331) and diol 17 ([M - H]- ) 313) hydrolysis products and the TES monoadduct 29. The product analysis again implies that kloss ) kalk, determined as 0.00083 h-1. The proximate reduction product of SN 23862 (6) via nitro reduction by NTR is the 2-hydroxylamine derivative 7 (21). Measurement of the reactivity of 7 toward nucleophiles was complicated by its instability in the presence of oxygen, forming the nonpolar 2,2′-azoxy dimer ([M - H]- ) 653 with the characteristic fragment [M - H - O]- ) 637), as observed previously during synthesis of 7 (26). This may arise through disproportionation of the hydroxylamine 7 to the amine 8 and the corresponding nitroso compound, followed by condensation of the two products. Ascorbate is known to reduce aromatic nitroso groups to hydroxylamines (33, 34); we therefore added ascorbate to 1 mM, which completely suppressed formation of the azoxy conjugate for up to 400 min, allowing investigation of the hydrolysis of 7 during this period (kloss ) 0.424 h-1). Under these conditions, the initial reaction products included the amine mustard 8, the hydroxylamine mono-ol 18, the amine mono-ol 20, and the diol reaction product 19 (Figure 2). The stable end product of the reaction was the amine diol 21, not the hydroxylamine diol 19. It is proposed that the hydroxylamine 7 and its transient hydrolysis products 18 and 19 all undergo disproportionation reactions to form the corresponding amine products 8, 20, and 21 (Scheme 3). Because of the competing disproportionation reaction to 8, kloss is not a direct measure of alkylating reactivity of 7. The rate constant k3 for formation of 8 from 7, by disproportionation, was fitted utilizing the known value of k4 (kalk for 8, above). This enabled the determination of the rate constants for hydrolysis of the hydroxylamine (k1 ) 0.22 h-1 and k2 ) 0.072 h-1). In this case, k1 rather than kloss, is equal to kalk. Other competing reactions that occur and prevent the use of kloss as a direct measure of kalk include intramo-

Figure 2. LC/MS analysis of products of the 2-hydroxylamine derivative of SN 23862, 7 (10 µM), following incubation in TES buffer (pH 7.4) containing 0.1 mM EDTA and ascorbate (1 mM) at 37 °C. A typical chromatogram (diode array detection) is shown at a reaction time of 100 min, and numbers refer to structures (Table 1), with parent molecular ion m/z values in italics.

lecular alkylation, as observed with the 4-amine analogue 9, which forms the tetrahydroquinoxaline half mustard 23 (Table 1) (26). This was confirmed as the major product from 9 (k ) 1.62 h-1) during incubation in TES buffer by comparison with authentic 23 (Table 1). The overall kinetics of loss (kloss ) 1.78 h-1) was similar to

Reductive Activation of Dinitrobenzamide Alkylators

Chem. Res. Toxicol., Vol. 16, No. 4, 2003 473

Scheme 3. Proposed Reaction Scheme of the Products Formed during Incubation of 7, the 2-Hydroxylamine Derivative of SN 23862, in TES Buffer, pH 7.4, Containing 1 mM Ascorbate at 37 °C

Table 1. Structures of the Compounds Studied and Their Reaction Productsa

1 2 3 4 5 6

I I I I I III

X ) Y ) NO2 X ) NHOH, Y ) NO2 X ) NH2, Y ) NO2 X ) NO2, Y ) NHOH X ) NO2, Y ) NH2 X ) Y ) NO2, R1 ) R2 ) Cl

7 8

III III

X ) NHOH, Y ) NO2, R1 ) R2 ) Cl X ) NH2, Y ) NO2, R1 ) R2 ) Cl

9

III

X ) NO2, Y ) NH2, R1 ) R2) Cl

10

III

X ) NH2, Y ) NH2, R1 ) R2 ) Cl

11 12

III III

34

IV

X ) NH2, Y ) H, R1 ) R2 ) Cl X ) NH2, Y ) H, R1 ) R2 ) Cl (no carboxamide) X ) NH2, R 3 ) CH2CH2Cl

a

reaction products

kalk (h-1)

13 (II R ) OH), 27 (II R ) Cl) 3, 15 15 (II R ) OH), 28 (II R ) Cl) 5, 14 14 (II R ) OH), 29 (II R ) Cl) 16 (III R1 ) OH, R2 ) Cl), 17 (III R1 ) OH, R2 ) OH), 29 (III R1 ) TES, R2 ) Cl) 8, 18 (III R1 ) OH, R2 ) Cl), 19 (III R1 ) OH, R2 ) OH), 20, 21 20 (III R1 ) OH, R2 ) Cl), 21 (III R1 ) OH, R2 ) OH), 30 (III R1 ) TES, R2 ) Cl), 31 (III R1 ) TES, R2 ) OH) 22 (III R1 ) OH, R 2) Cl), 23 (IV R3 ) CH2CH2Cl), 24 (IV R3 ) CH2CH2OH) 25 (III R1 ) OH, R2 ) Cl), 26 (III R1 ) R2 ) OH), 32 (III R1 ) TES, R2 ) Cl), 33 (III R1 ) TES, R2 ) OH) ND ND

0.0023 0.016 0.039 0.00117 0.0037 0.00083

7.8a 15.6a

ND

10.2a

0.216 1.85 1.78 11.4a

ND, not determined; kloss is an estimate of kalk.

that for the 2-amine 8. Two minor products were tentatively identified by their absorption spectra and retention times as the mono-ol hydrolysis product 22 (k1 ) 0.162 h-1) and the tetrahydroquinoxaline hydrolysis product 24 (k2 ) 1.26 h-1). k1 is the only reaction involving an external nucleophile, but overall kalk is equivalent to kloss. The ratio of intramolecular to external nucleophilic attack will depend on the external nucleophile concentration (assuming that this competes with the 4-amino group for the electrophilic reaction center in the common aziridinium ion). The 2,4-diamino analogue 10, in con-

trast, hydrolyzed to the corresponding diol 26 as the major stable end product, confirming an earlier report that this compound does not undergo intramolecular alkylation (26). Three minor hydrolysis/alkylation products 25, 32, and 33 were tentatively assigned as the mono-ol and TES conjugates (Table 1), and kalk was estimated as kloss (11.4 h-1). Nucleophilic reactivity of other amino analogues was also measured from kloss (Table 1). The 2-amino, 4-H mustard 11 was approximately 4-fold more reactive than the corresponding 4-nitro derivative 8, demonstrating the deactivating effect

474

Chem. Res. Toxicol., Vol. 16, No. 4, 2003

Figure 3. (a) Rates of loss of aziridines 1 (O), 3 (∆), and 5 (2) in Bis-Tris buffer (pH 6.5) with 0.1 mM EDTA at 37 °C and (b) the hydroxylamines 2 (0) and 4 (9) in the same buffer containing 1 mM ascorbate.

of a single, unreduced nitro group. The p-phenylenediamine mustard 12, lacking the carboxamide group, was more reactive (kloss ) 15.6 h-1) than 11 (kloss ) 7.8 h-1), reflecting the slightly electron-withdrawing nature of the carboxamide group (σmeta ) 0.28). Over all of the substituted benzamide mustards (n ) 7), there was a strong correlation between kalk and substituent electronic effects (summing the σ effects for the nitro, amino, and carboxamide substituents), as described by the equation:

log kalk ) aσ + b Linear regression gave a ) -1.44 and b ) 0.25 (r2 ) 0.88, p < 0.005, and F ) 36.0). The compound with the largest residual in this regression was the diamino 10, reflecting its anomalously low alkylating reactivity noted above. This is consistent with the anomalously low basicity of ortho amino anilines (35). Excluding 10, the coefficient a increased to -1.69 (r2 ) 0.93, p < 0.005, and F ) 54.5). This is similar to the correlation between electronic properties of substituted N,N-bis(2-chloroethyl)anilines and chemical stability in culture medium (as a measure of hydrolytic reactivity) observed previously (18) as well as the relationship between electronic effects and chemical hydrolysis of substituted aniline chloro mustards (36). The present study clearly demonstrates that the alkylating reactivity of the mustard moiety is directly related to electronic substituent effects in the aromatic ring, with electron-donating groups increasing alkylating reactivity. Alkylating Reactivity of CB 1954 Derivatives. The alkylating reactivity of CB 1954 (1) and its reduction products was assessed by measuring their kinetics of loss and products of nucleophilic attack on the aziridine ring, in Bis-Tris buffer at pH 6.4 (Figure 3a). Compound 1 and the corresponding 4-amine 5 were relatively stable while the 2-amine 3 was much more reactive with a more than

Helsby et al.

10-fold higher first-order rate constant (Table 1). The reaction products of 1 were identified by LC/MS and coelution with authentic standards, as the ring-opened alcohol 13 ([M - H]- ) 269) and chloroethyl product 27 ([M - H]- ) 287) resulting from nucleophilic attack by OH- and Cl-, respectively. Similarly, the products of incubation of the 2-amine (3) were also identified by LC/ MS (positive mode electrospray ionization) as the alcohol 15 ([M + H]+ ) 241) or the chloroethyl product 28 ([M + H]+ ) 259). The reaction products of the 4-amine 5 were also tentatively identified as the ring-opened hydrolysis 14 or chlorination 29 products based on spectral similarity and retention times relative to the products of 1 and 3. The product analysis is consistent with a common protonated aziridinium ion intermediate (or the subsequent carbocation (37)), again indicating that kloss is a measure of overall alkylating reactivity (kalk). The 2- and 4-hydroxylamines (2 and 4), synthesized by Zn/NH4OAc reduction of CB 1954 (identities confirmed by NMR spectrometry), showed overall kinetics of loss (2, kloss ) 0.0203 h-1; 4, kloss ) 0.0013 h-1), in ascorbate-containing buffer, similar to that of the corresponding amines (Figure 3b), but gave different product profiles. The end reaction product of 2, identified by LC/MS, was the 2-amino 5-hydroxyethyl derivative 15 (Figure 4); the identity of this product was confirmed by coelution with the product from hydrolysis of 3 in perchloric acid. The mass spectrum of 2 in Bis-Tris buffer contains a molecular ion at 237 m/z; however, under the ionization conditions used, the hydroxylamine 2 appears to disproportionate to the nitroso (235 m/z) and the amino (221 m/z) derivatives and these ions and their fragments dominate the mass spectrum of 2. The starting material also contained 4% of the 2-amino 5-aziridine 3, which slowly increased in concentration. The formation of the amine 15 indicated disproportionation of the hydroxylamine to the corresponding nitroso (not detected because this will be rapidly reduced by ascorbate) and amine. Formation of the 4-nitroso derivative from 4 has been reported previously (34). Kinetic models were therefore fitted to clarify whether disproportionation occurs primarily before or after hydrolysis of the aziridine, using the rate constant determined above for hydrolysis of the potential intermediate 3. The concentration profiles were not consistent with disproportionation followed by hydrolysis of 3 but were well-fitted by the model of Scheme 4 with loss of 2 being mainly via hydrolysis followed by disproportionation of the hydroxyethyl intermediate (which was not detected). The estimated value of kalk () k1) was 0.016 h-1, which is intermediate between the value for 1 and its 2-amine 3. Thus, electron-donating characteristics of the substituent para to the aziridine ring increase its alkylating reactivity. Lewis (37) also noted that the presence of electron-withdrawing groups will decrease the acid-catalyzed reactivity of the aziridine moiety for a series of CB 1954 analogues due to a decrease in protonation of the aziridine nitrogen atom. However, the dependence on σpara for the aziridine derivatives in the present study was much less pronounced than for the mustards, with a differential of only 17-fold for 1 vs 3 (cf. 2200-fold for the corresponding mustards, 6 and 8). In the case of the 4-hydroxylamine 4, the kinetics were even slower than for the 2-hydroxylamine 2 (Figure 3b), with traces of two products observed; these were again identified by LC/MS as the amine 5 ([M - H]- ) 221)

Reductive Activation of Dinitrobenzamide Alkylators

Chem. Res. Toxicol., Vol. 16, No. 4, 2003 475

Figure 4. HPLC analysis of reaction products of the 2-hydroxylamine derivative of CB 1954, 2 (10 µM), following incubation in Bis-Tris buffer (pH 6.5), containing 0.1 mM EDTA and ascorbate (1 mM) at 37 °C. A typical HPLC chromatogram is shown at a reaction time of 200 min (diode array absorbance detection) with mass spectra (negative mode APCI; fragmentor voltage, 100 V).

Scheme 4. Proposed Reaction Scheme of the Products Formed during Incubation of 2, the 2-Hydroxylamine Derivative of CB 1954, in Bis-Tris Buffer, pH 6.5, Containing 1 mM Ascorbate at 37 °C

and its hydrolysis product 14 ([M - H]- ) 239). Fitting the concentration profiles as above provided an estimate of 0.00117 h-1 for kalk (Table 1). Thus, the rate constants for 4 and 5 were of the same order as CB 1954 (1). Therefore, there is little activation of alkylating reactivity by electron donation at the ortho position.

Cytotoxicity. The cytotoxicity of the analogues was determined following a 4 h drug exposure of an ERCC1 mutant (UV4) derived from AA8 cells. The cytotoxic potency of the 2-hydroxylamine derivative of SN 23862 (7) has not been reported previously; its IC50 was 0.23 ( 0.005 µM (n ) 3), similar to that of the 2-amine 8 (0.22 ( 0.01µM) and 3783-fold more potent than the parent dinitro compound 6. The similarity in potency of the amine 8 and hydroxylamine 7 is, at first, surprising given their 8.3-fold difference in alkylating reactivity. From the kinetic model, it is expected that most of the 7 will react during the 4 h period of exposure of the cells, whether by alkylation of nucleophiles or reduction/disproportionation to 8, which would account for the similar cytotoxic potencies of 7 and 8. Given that both alkylation and disproportionation of 7 are expected to contribute to its overall ability to form DNA cross-links, kloss is used rather than kalk in the regression analysis of reactivity vs cytotoxicity below. However, there is no need to invoke mechanisms other than nitroreduction, which acts as an electronic switch activating the alkylating moiety, to account for the increased cytotoxicity of 7. The major product of enzymatic nitroreduction of SN 23862 (6) in the NTR-gene-directed enzyme prodrug therapy (GDEPT) system is the 2-hydroxylamine 7 (21), which we show undergoes rapid nonenzymatic conversion to the highly cytotoxic amine 8. Recent data indicate that the bioactive bystander metabolite of this prodrug is indeed the amine derivative 8 (38).

476

Chem. Res. Toxicol., Vol. 16, No. 4, 2003

There was a statistically significant (p < 0.0005, F ) 94.2, and r2 ) 0.95) inverse linear relationship between log alkylating reactivity and log IC50 for all of the mustards investigated with a slope of -1.07 indicating an approximately linear relationship between kalk and IC50. The cytotoxicity of two analogues (9 and 34) is less than expected from their measured kalk. The 4-amine analogue 9 undergoes intramolecular alkylation and cyclization to form a tetrahydroquinoxaline hydrolysis product 23, and this reaction will lower the amount of 9 available to form a DNA cross-link. However, the cytotoxicity was related to the rate of hydrolysis (k1 ) 0.162 h-1) and this value was used in the regression analysis. Analogue 34 cannot form a DNA cross-link and would also be expected to be less cytotoxic than compounds with two mustard “arms”. However, the electronic effects of amine substitution para to the mustard moiety on cytotoxicity are apparent when 34 is compared with its nitro analogue 23 (IC50 0.87 ( 0.1 vs 49 ( 1 µM) (26). Thus, the activation of an alkylating moiety by increasing the electron-donating characteristics of the substituent para to the mustard group directly correlates with the cytotoxicity of the dinitrobenzamide mustard derivatives. Previous studies with substituted aniline mustards (18) have also indicated that cytotoxicity in cell culture is highly correlated with the electronic properties of the substituents. Overall, these data clearly demonstrate the ability of a “nitro switch”(17, 19) to activate the alkylating reactivity of a prepositioned mustard moiety, which directly results in cytotoxicity. The cytotoxic potency of the hydroxylamine derivatives, 2 and 4, of CB 1954 against UV4 cells has not been reported previously, although 4 is known to be more toxic than 2 against Chinese hamster V79 cells (39). The IC50 of 2 was 4.85 ( 0.71 µM (n ) 3), similar to that of the 2-amine 3 (4 ( 1 µM) again presumably due to formation of the amine 3 from 2 over the 4 h period of exposure of the cells. As with the mustard analogues, there was also a statistically significant (p < 0.05, F ) 24.5, and r2 ) 0.925) linear relationship (slope ) -1.18) between alkylating reactivity and cytotoxicity for the aziridine analogues (Figure 5b). Reduction of the 2-nitro moiety to either the hydroxylamine 2 or the amine 3 and reduction of the 4-nitro group to the amine 5 increase the cytotoxicity of the compounds, and this can be ascribed to an increase in alkylating reactivity. The cytotoxicity of 2 is 1.8-fold greater than predicted by the alkylating reactivity (kalk) of this hydroxylamine. However, the apparent cytotoxicity of 2 will be influenced by formation of the amine 3; therefore, compound 2 would be expected to have a similar cytotoxicity to 3. Again as with 7, the total reactivity of 2 (kloss), which includes rate of formation of the amine 3 by disproportionation, is related to its cytotoxicity, and these data are included in the regression analysis (Figure 5b). Thus, although nitroreduction has a relatively small effect on alkylating reactivity in the aziridine series, there is a similar relationship between alkylating reactivity and cytotoxicity to that seen for the dinitrobenzamide mustards. The only outlier in this correlation is the 4-hydroxylamine of CB 1954, 4, which was 300-fold more cytotoxic (IC50 0.60 ( 0.13 µM) than predicted by reactivity of the aziridine ring alone. This analogue is relatively stable and does not readily form the amine 5. Unlike the mustard hydroxylamine 7, 4 is relatively stable, and the corresponding amine 5 is less potent, so the anomalous

Helsby et al.

Figure 5. Correlation between alkylating reactivity (kalk) and cytotoxicity in UV4 cells for (a) dinitrobenzamide mustard analogues and (b) dinitrobenzamide aziridine analogues. Solid lines are regressions, with dashed lines showing 95% confidence limits, excluding the open symbols (see text). (a) r2 ) 0.95, p < 0.0005, and F ) 94.24. (b) r2 ) 0.925, p < 0.05, and F ) 24.5. For 9, the value for k1, which excludes the rate of intramolecular alkylation, was used. For the hydroxylamines 2 and 7, the value of kloss was used rather than kalk because the competing loss pathway (disproportionation) yields the amine metabolites, which are similarly bioactive.

cytotoxicity of 4 cannot be due to disproportionation to 5. Therefore, the mechanism of cytotoxicity of the 4-hydroxylamine 4 of CB 1954 differs from the related aziridines and mustards. Previous reports (3) have demonstrated that incubation of 4 with acetyl CoA increases the formation of DNA cross-links in a cell-free system. It was therefore suggested that 4 is bioactivated in vivo via N-acetylation to form an unstable N-acetoxy derivative. This then may form a reactive nitrenium intermediate that alkylates DNA in a manner similar to other hydroxylamines (40), with the aziridine ring forming the second reactive center. The anomalous stability of the ortho-substituted hydroxylamine 4 as compared with 2 may facilitate N-acetylation of the hydroxylamine 4 in cells enabling formation of this potent DNA crosslink.

Conclusions We have demonstrated that the alkylating reactivity of dinitrobenzamide mustards is highly sensitive to substituent electronic effects as expected. In particular, we show that reduction of either nitro group of the dinitrobenzamide mustard 6 provides a large increase in alkylating reactivity, confirming that nitroreduction can be used as an electronic switch to activate a prepositioned nitrogen mustard (19, 41). In contrast, electronic effects are much less pronounced for the corresponding aziridines. This difference is illustrated by the alkylating reactivities of the 2-amino reduction products, which are 2200-fold and 17-fold more reactive than the dinitro compounds for the mustard and aziridine, respectively. Notably, the alkylating reactivity of the 4-hydroxylamine derivative of CB 1954 (1) is no greater than the parent compound. We have also demonstrated that cytotoxicity in UV4 cells is directly related to the alkylating reactivity of the nitrogen mustard or aziridine moiety, with the notable exception of 4, which is 300-fold more potent than expected on the basis of its reactivity. This is consistent

Reductive Activation of Dinitrobenzamide Alkylators

with the special significance of the 4-hydroxylamine moiety in providing a second reactive center as originally proposed by Knox et al. (3). The present study shows that the 4-hydroxylamine is more stable than the 2-hydroxylamine; this may facilitate more efficient conversion to the putative N-acetoxy derivative and thus contribute to its anomalous cytotoxicity. The E. coli nitroreductase NTR reduces both the 2- and the 4-nitro groups of CB 1954 (1), but only the 2-nitro group of the mustard SN 23862 (6), to the corresponding hydroxylamines (6, 21). Given that the mustard 2-hydroxylamine 7 is readily reduced to the 2-amine 8, the critical cytotoxic metabolites are the 4-hydroxylamine (4) of CB 1954 and the 2-amine (8) of SN 23862. The latter is at least as cytotoxic as the former (UV4 IC50 value 0.22 vs 0.60 µM, respectively). The simple, one step bioactivation of the dinitrobenzamide mustard on nitroreduction is a valuable advantage in their use as prodrugs for GDEPT. This avoids the requirement for a stable hydroxylamine intermediate, simplifies drug design requirements, and eliminates any potential variability in acetylation of the hydroxylamine in different cell lines. Combining these features with the higher rate of nitroreduction of SN 23862 than CB 1954 by NTR and the specificity of the enzyme for critical 2-nitro group of the mustard (21), the dinitrobenzamide mustards can be expected to be superior prodrugs for NTR GDEPT as recently demonstrated in tumor xenograft models (22).

Acknowledgment. This study was supported by a grant from the Health Research Council of New Zealand. We thank Jane Botting and Susan Pullen for IC50 assays.

References (1) Cobb, L. M., Connors, T. A., Elson, L. A., Khan, A. H., Mitchley, B. C. V., Ross, W. C. J., and Whisson, M. E. (1969) 2,4-dinitro5-ethyleneiminobenzamide (CB 1954): A potent and selective inhibitor of the growth of the Walker carcinoma 256. Biochem. Pharmacol. 18, 1519-1527. (2) Knox, R. J., Boland, M. P., Friedlos, F., Coles, B., Southan, C., and Roberts, J. J. (1988) The nitroreductase enzyme in Walker cells that activates 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB 1954) to 5-(aziridin-1-yl)-4-hydroxylamino-2-nitrobenzamide is a form of NAD(P)H dehydrogenase (quinone) (EC 1.6.99.2). Biochem. Pharmacol. 37, 4671-4677. (3) Knox, R. J., Friedlos, F., Marchbank, T., and Roberts, J. J. (1991) Bioactivation of CB 1954: reaction of the active 4-hydroxylamino derivative with thioesters to form the ultimate DNA-DNA interstrand cross-linking species. Biochem. Pharmacol. 42, 16911697. (4) Boland, M. P., Knox, R. J., and Roberts, J. J. (1991) The differences in kinetics of rat and human DT diaphorase result in a differential sensitivity of derived cell lines to CB 1954 (5(aziridin-1-yl)-2,4-dinitrobenzamide). Biochem. Pharmacol. 41, 867-875. (5) Anlezark, G. M., Melton, R. G., Sherwood, R. F., Coles, B., Friedlos, F., and Knox, R. J. (1992) The bioactivation of 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB1954)-I. Purification and properties of a nitroreductase enzyme from Escherichia coli- -a potential enzyme for antibody-directed enzyme prodrug therapy (ADEPT). Biochem. Pharmacol. 44, 2289-2295. (6) Knox, R. J., Friedlos, F., Sherwood, R. F., Melton, R. G., and Anlezark, G. M. (1992) The bioactivation of 5-(aziridin-1-yl)-2,4dinitrobenzamide (CB1954)-II. A comparison of an Eschericia coli nitroreductase and Walker DT diaphorase. Biochem. Pharmacol. 44, 2297-2301. (7) Chung-Faye, G., Palmer, D., Anderson, D., Clark, J., Downes, M., Baddeley, J., Hussain, S., Murray, P. I., Searle, P., Seymour, L., Harris, P. A., Ferry, D., and Kerr, D. J. (2001) Virus-directed, Enzyme Prodrug Therapy with Nitroimidazole Reductase: A Phase I and Pharmacokinetic Study of its Prodrug, CB1954. Clin. Cancer. Res. 7, 2662-2668.

Chem. Res. Toxicol., Vol. 16, No. 4, 2003 477 (8) Young, J. G., Leung, H. Y., Searle, P. F., Mautner, V., Howie, A., Horne, M., Ellis, J., Gilbert, J., Neal, D. E., Harris, P., Doherty, A. P., Wallace, D. M. A., and James, N. D. (2002) Direct injection of adenovirus for enzyme prodrug therapy of prostate cancer. Br. J. Cancer 86, s30. (9) Walton, M. I., Wolf, C. R., and Workman, P. (1989) Molecular enzymology of the reductive bioactivation of hypoxic cell cytotoxins. Int. J. Radiat. Oncol. Biol. Phys. 16, 983-986. (10) Stratford, I. J., Williamson, C., Hoe, S., and Adams, G. E. (1981) Radiosensitizing and cytotoxicity studies with CB 1954 (2,4dinitro-5-aziridinylbenzamide). Radiat. Res. 88, 502-509. (11) Chapman, J. D., Raleigh, J. A., Pedersen, J. E., Ngan, J., Shum, F. Y., Meeker, B. E., and Urtasun, R. C. (1979) Potentially three distinct roles for hypoxic cell sensitizers in the clinic. Proc. 6th Int. Congr. Radiat. Res. 885-893. (12) Rockwell, S. (1992) Use of hypoxia-directed drugs in therapy of solid tumors. Semin. Oncol. 19, 29-40. (13) Workman, P., and Stratford, I. J. (1993) The experimental development of bioreductive drugs and their role in cancer therapy. Cancer Metastasis Rev. 12, 73-82. (14) Denny, W. A., Wilson, W. R., and Hay, M. P. (1996) Recent developments in the design of bioreductive drugs. Br. J. Cancer 27, S32-S38. (15) Brown, J. M., and Siim, B. G. (1996) Hypoxia-Specific Cytotoxins in Cancer Therapy. Semin. Radiat. Oncol. 6, 22-36. (16) Wardman, P. (2001) Electron transfer and oxidative stress as key factors in the design of drugs selectively active in hypoxia. Curr. Med. Chem. 8, 739-761. (17) Denny, W. A., and Wilson, W. R. (1986) Considerations for the design of nitrophenyl mustards as agents with selective toxicity for hypoxic tumor cells. J. Med. Chem. 29, 879-887. (18) Palmer, B. D., Wilson, W. R., Pullen, S. M., and Denny, W. A. (1990) Hypoxia-selective antitumor agents. 3. Relationships between structure and cytotoxicity against cultured tumor cells for substituted N,N-bis(2-chloroethyl)anilines. J. Med. Chem. 33, 112-121. (19) Siim, B. G., Denny, W. A., and Wilson, W. R. (1997) Nitro reduction as an electronic switch for bioreductive drug activation. Oncol. Res. 9, 357-369. (20) Palmer, B. D., Wilson, W. R., Cliffe, S., and Denny, W. A. (1992) Hypoxia-selective antitumor agents. 5. Synthesis of water-soluble nitroaniline mustards with selective cytotoxicity for hypoxic mammalian cells. J. Med. Chem. 35, 3214-3222. (21) Anlezark, G. M., Melton, R. G., Sherwood, R. F., Wilson, W. R., Denny, W. A., Palmer, B. D., Knox, R. J., Friedlos, F., and Williams, A. (1995) Bioactivation of dinitrobenzamide mustards by an E. coli B nitroreductase. Biochem. Pharmacol. 50, 609618. (22) Wilson, W. R., Pullen, S. M., Hogg, A., Helsby, N. A., Hicks, K. O., and Denny, W. A. (2002) Quantitation of bystander effects in nitroreductase suicide gene therapy using three-dimensional cell cultures. Cancer Res. 62, 1425-1432. (23) Hoy, C. A., Thompson, L. H., Mooney, C. L., and Salazar, E. P. (1985) Defective DNA cross-link removal in Chinese hamster cell mutants hypersensitive to bifunctional alkylating agents. Cancer Res. 45, 1737-1743. (24) Thompson, L. H., Busch, D. B., Brookman, K., Mooney, C. L., and Glaser, D. A. (1981) Genetic diversity of UV-sensitive DNA repair mutants of Chinese hamster ovary cells. Proc. Natl. Acad. Sci. U.S.A. 78, 3734-3737. (25) Jarman, M., Melzack, D. H., and Ross, W. C. J. (1976) The metabolism of the anti-tumour agent 5-(1-aziridinyl)-2,4-dinitrobenzamide (CB 1954). Biochem. Pharmacol. 25, 24752478. (26) Palmer, B. D., van Zijl, P., Denny, W. A., and Wilson, W. R. (1995) Reductive chemistry of the novel hypoxia-selective cytotoxin 5-[N,N-bis(2-chloroethyl)amino]-2,4-dinitrobenzamide. J. Med. Chem. 38, 1229-1241. (27) Everett, J. L., and Ross, W. C. (1949) Aryl-2-halogenoalkylamines. Part II. J. Chem. Soc. Perkin Trans. 1, 1972-1983. (28) Kestell, P., Pruijn, F. B., Siim, B. G., Palmer, B. D., and Wilson, W. R. (2000) Pharmacokinetics and metabolism of the nitrogen mustard bioreductive drug 5-[N,N-bis(2-chloroethyl)amino]-2,4dinitrobenzamide (SN 23862) and the corresponding aziridine (CB 1954) in KHT tumour-bearing mice. Cancer Chemother. Pharmacol. 46, 365-374. (29) Skehan, P., Storeng, R., Scudiero, D., Monks, A., McMahon, J., Vistica, D., Warren, J. T., Bokesch, H., Kenney, S., and Boyd, M. R. (1990) New colorimetric cytotoxicity assay for anticancer-drug screening. J. Natl. Cancer Inst. 82, 1107-1112. (30) Ross, W. C. (1949) Aryl-2-halogenoalkylamines. Part I. J. Chem. Soc. Perkin Trans. 1, 183-191.

478

Chem. Res. Toxicol., Vol. 16, No. 4, 2003

(31) O’Connor, C. J., Denny, W. A., and Fan, J. Y. (1991) Alkylation of nucleic acids by DNA-targeted 4-anilinoquinolinium aniline mustards: kinetic studies. Chem.-Biol. Interact. 77, 223-241. (32) Bolton, M. G., Hilton, J., Robertson, K. D., Streeper, R. T., Colvin, O. M., and Noe, D. A. (1993) Kinetic analysis of the reaction of melphalan with water, phosphate, and glutathione. Drug Metab. Dispos. 21, 986-996. (33) Coleman, M. D., Breckenridge, A. M., and Park B. K. (1989) Bioactivation of dapsone to a cytotoxic metabolite by human hepatic microsomal enzymes. Br. J. Clin. Pharmacol. 28, 389395. (34) Knox, R. J., Friedlos, F., Biggs, P. J., Flitter, W. D., Gaskell, M., Goddard, P., Davies, L., and Jarman, M. (1993) Identification, synthesis and properties of 5-(aziridin-1-yl)-2-nitro-4-nitrosobenzamide, a novel DNA cross-linking agent derived from CB1954. Biochem. Pharmacol. 46, 797-803. (35) Perrin, D. D. (1965) Dissociation Constants of Organic Bases in Aqueous Solution, Butterworths, London. (36) Panthananickal, A., Hansch, C., and Leo, A. (1978) Structureactivity relationships in antitumor aniline mustards. J. Med. Chem. 21, 16-26.

Helsby et al. (37) Lewis, D. F. V. (1989) Molecular orbital calculations on tumorinhibitory phenyl aziridines: QSARs. Xenobiotica 19, 341-356. (38) Wilson, W. R., Ferry, D. M., Patterson, A. V., Pullen, S. M., Helsby, N. A., Pruijn, F. B., and Hicks, K. O. (2002) Extravascular transport of GDEPT prodrugs in multicellular layer cultures: Dinitrobenzamide mustard substrates for the E. coli nitroreductase NTR. Proc. Am. Assoc. Cancer Res. 43, 84-85. (39) Knox, R. J., Friedlos, F., Jarman, M., and Roberts, J. J. (1988) A new cytotoxic, DNA interstrand cross-linking agent, 5-(aziridin1-yl)-4-hydroxylamino-2-nitrobenzamide, is formed from 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB 1954) by a nitroreductase enzyme in Walker carcinoma cells. Biochem. Pharmacol. 37, 4661-4669. (40) Weisburger, J. H., and Weisburger, K. W. (1973) Biochemical formation and pharmacological, toxicological and pathological properties of hydroxylamines and hydroxamic acids. Pharmacol. Rev. 25, 1-66. (41) Denny, W. A., and Wilson, W. R. (1993) Bioreducible mustards: a paradigm for hypoxia-selective prodrugs of diffusible cytotoxins (HPDCs). Cancer Metastasis Rev. 12, 135-151.

TX025662B