Activation of Adriamycin by the pH-dependent Formaldehyde-releasing Prodrug Hexamethylenetetramine¹

Chemotherapeutic drugs are important in most types of current cancer treatment protocols. One of the most widely used agents, Adriamycin (doxorubicin), exhibits a wide activity against leukemias and breast, lung, thyroid, and ovarian carcinomas, as well as Hodgkin’s and non-Hodgkin’s lymphomas (1, 2). Despite the remarkable efficiency with which Adriamycin kills cells, treatment is complicated by many factors, including a dose-limiting cardiotoxicity and MDR³ (1, 2). Adriamycin undergoes metabolic reduction and in the presence of oxygen leads to the production of reactive oxygen species (3, 4), and these species are particularly damaging to cardiac tissue that do not possess enzymes to detoxify the resulting radicals (1). MDR occurs as a result of up-regulation of the membrane-associated P-glycoprotein efflux pump (5). MDR therefore results in a diminished response to anthracycline anticancer agents such as Adriamycin.

The mechanism of action of Adriamycin and other anthracyclines appears to involve impairment of topoisomerase II activity (6–10) as well as the formation of DNA adducts (11–14). Initial investigations of Adriamycin-DNA adducts using an in vitro transcription footprinting assay suggested that the formation of adducts was enhanced in the presence of iron (11). However, it was later shown that the Fe(III)/DTT/Tris buffer resulted in the production of formaldehyde that mediated the formation of the drug-DNA adducts (3).

Following the realization that formaldehyde mediated the formation of Adriamycin-DNA adducts, studies were undertaken to fully understand this drug-DNA interaction in vitro. It has been shown that formaldehyde supplies a methylene group that links the 3′ amino of Adriamycin to the 2-amino of deoxyguanosine residues of DNA via Schiff base chemistry (3, 13–16) and that the formaldehyde-conjugated complex is the active form of the drug (17). Although the adducts are attached covalently to only one strand of DNA and are therefore mono-adducts, they stabilize the local region of DNA sufficiently that they are resistant to thermal denaturation and can therefore be detected in denaturation-based cross-linking assays (12–14, 17). It has been demonstrated that the presence of the amino group on the sugar moiety is critically important for the formation and stabilization of the drug-DNA interaction (18–20). A preactivated form of Adriamycin (doxoform) has been shown to be taken up by cells at an accelerated rate, is retained longer in the nucleus, and is substantially more cytotoxic than Adriamycin (21).

An alternative approach to enhance the activity of Adriamycin is to increase the intracellular level of formaldehyde, leading to enhanced formation of Adriamycin-DNA lesions. This approach has been investigated recently using a combination of Adriamycin and the formaldehyde-releasing prodrug pivaloyloxymethyl butyrate (AN-9; Ref. 22). The prodrug is cleaved intracellularly by esterases, liberating formaldehyde, which reacts with Adriamycin and results in a dramatic increase of the number of Adriamycin-DNA adducts, together with a synergistic cytotoxic response. However, AN-9 also releases butyric acid, which inhibits histone deacetylase and, as such, potentially modulates the chromatin accessibility of Adriamycin, thus contributing to the observed synergistic response.

HMTA is a tertiary amine that hydrolyzes under acidic conditions to yield ammonia and formaldehyde (23). We have therefore examined it as a pH-dependent formaldehyde-releasing prodrug, thus enabling us to study the effects of formaldehyde release independently of butyric acid release. HMTA has been examined as an antiseptic for the treatment of urinary tract infections (24), as well as being studied in patients with maxillofacial phelegmons (25) and as a prophylactic agent against recurrent acute cystitis (26). It has also been tested in water systems at a concentration of 1 mg/liter (27) for its antiseptic properties (25). It is particularly well tolerated by humans, even at high doses of up to 5 g/kg/day (28). The degradation of HMTA produces six molecules of formaldehyde for every molecule of drug hydrolyzed, suggesting that it may be a more efficient prodrug than AN-9, which releases only one molecule of formaldehyde when degraded by esterases (29).

Warburg (30) suggested that tumors were acidic because of the production of lactic acid, due to the metabolism of tumor cells in an anaerobic environment. However, more recently it has been shown that the intracellular pH in tumors tends more toward neutral or alkaline values (31, 32), whereas the extracellular pH was shown to be acidic (33–35). Due to the acidic extracellular pH of solid tumors, it would be expected that HMTA would be preferentially hydrolyzed to release formaldehyde in the tissues adjacent to such tumors.

It has been shown that the dose of Adriamycin required to induce a cytotoxic response can be diminished when coadministered with a formaldehyde-releasing prodrug (22). It was therefore expected that a similar response would also be achieved with the coadministration of HMTA with Adriamycin. Moreover, the fact that formaldehyde is released from HMTA in acidic environments (such as those associated with solid tumors) raises the possibility of a tumor-localized release of formaldehyde and hence a tumor-localized response to Adriamycin.

HMTA-mediated Adriamycin-DNA lesions were therefore investigated at various pH values in vitro to establish the potential for tumor-localized responses to formaldehyde-releasing prodrugs and Adriamycin. The cytotoxicity of Adriamycin and prodrug as single agents and in combination was investigated in tumor cells in culture, together with the ability of the drug combination to form drug-DNA adducts in cellular DNA.

Adriamycin, daunomycin, and epirubicin were gifts from Farmitalia Carlo Erba (Milan, Italy). Idarubicin hydrochloride was purchased from Pharmacia and Upjohn. HMTA (Aldrich, Milwaukee, WI) was freshly dissolved in Milli-Q water as a 50 mm stock solution. The anthracyclines were dissolved in Milli-Q water to a stock concentration of approximately 1 mm and stored at −20°C. Calf thymus DNA was purchased from Worthington Chemical Corp. Tris-saturated ultrapure phenol was purchased from Life Technologies, Inc., and formaldehyde was purchased from BDH. ProbeQuant G-50 columns, [α- ³²P]dCTP, [α- ³²P]dATP, [α- ³²P]UTP (3000 Ci/mmol) and [14-¹⁴C]Adriamycin hydrochloride were obtained from Amersham Pharmacia Biotech. The plasmid containing the DHFR probe, pBH31R1.8 (36), was a gift from Dr. V. A. Bohr (National Institute on Aging, NIH, Baltimore, MD), whereas the plasmid with a mitochondrial insert (pCRII-H1) was a gift from Dr. C. A. Filburn (National Institute on Aging, NIH). Qiagen Plasmid Maxi Kits and QIAamp blood kit were from Qiagen. The Klenow fragment of DNA polymerase, glycogen, and the random primed labeling kit were from Roche Molecular Biosciences. Lambda exonuclease was purchased from Life Technologies, Inc. All other enzymes were purchased from New England Biolabs.

The plasmid pCCI (37) was linearized with HindIII and end-labeled with [α-³²P]dCTP or [α-³²P]dATP in the presence of the Klenow fragment of Escherichia coli DNA polymerase I. Unincorporated label was removed using G-50 ProbeQuant columns. The labeled DNA was resuspended in calf thymus DNA to a final concentration of 400 μΜ bp in TE buffer.

End-labeled DNA (25 μm bp) was reacted with drugs at 37°C for defined times in PBS (adjusted to the desired pH). Unreacted drugs were extracted with phenol and chloroform, and the DNA was precipitated in ethanol (using glycogen as an inert carrier). The pellet was washed with 70% ethanol, dried, and resuspended in TE. Samples were denatured in a final concentration of 66% loading dye (60% formamide, 6.6 mm EDTA, 0.07% xylene cyanol, and 0.07% bromphenol blue) at 65°C for 5 min. Samples were quenched on ice and then loaded onto a 0.8% agarose gel (40 cm), and the DNA was separated electrophoretically in TAE buffer [40 mm Tris-acetate, 1 mm EDTA (pH 8.0)] at 45 V for 16 h. The gels were dried, and all image analysis was performed on a Molecular Dynamics model 400B PhosphorImager using ImageQuaNT software (Molecular Dynamics).

DNA containing drug-DNA adducts stabilized the DNA sufficiently to resist denaturation at 65°C and therefore migrated as dsDNA, whereas DNA that lacked adducts was denatured under the same conditions and migrated as ssDNA. Relative drug-DNA adduct levels were determined as the percentage of DNA that migrated as dsDNA and calculated using ImageQuaNT software.

A 188-bp fragment was excised from pCC1 using EcoRI and PvuII. This fragment was separated electrophoretically and collected using a Biotrap electroeluter (Schleicher and Schuell). The DNA fragment was subjected to 3′ end-labeling with the Klenow fragment and [α-³²P]dATP. The labeled DNA was resuspended in calf thymus DNA to a final concentration of 400 μm bp and used for exonuclease studies.

End-labeled DNA was incubated in the presence of Adriamycin and either HMTA or formaldehyde to induce the formation of drug-DNA adducts. The samples were precipitated with ethanol and resuspended in λ exonuclease buffer [67 mm glycine-KOH (pH 9.4), 2.5 mm MgCl₂, and 50 μg/ml BSA] before being digested for 2 h with 5 units of λ exonuclease at 37°C. The digestion was terminated by the addition of an equal volume of 90% formamide (containing 0.1% xylene cyanol and 0.1% bromphenol blue) in TE buffer. Sequence identification was performed with a Maxam-Gilbert G sequencing lane (38).

IMR-32 neuroblastoma cells were maintained at 37°C, 5% CO₂ in DMEM (pH 6.8–7.4) obtained from Trace Scientific and supplemented with 10% fetal calf serum, 0.1 mg/ml streptomycin, and 100 U/ml penicillin.

IMR-32 cells (1 × 10⁴) were seeded into individual 96-well plates in 100 μl of DMEM (10% FCS) and allowed to adhere overnight. Drug treatment consisted of either HMTA, Adriamycin, or a combination of both. After a 3-day incubation period, the cells were treated with 300 μg/ml 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxy-phenyl)-2-(4-sulfonyl)-2H-tetrazolium (39, 40) and incubated for 4 h, and the absorbance was measured at 490 nm using a Molecular Devices Spectra Max 250 micro plate reader. The IC₅₀ was determined as the concentration at which only 50% of cells survived.

Specific probes were used to detect adducts in both mtDNA and nuclear DNA as outlined previously (41). Briefly, a 1.8-kb probe specific for the DHFR gene was isolated from pBH31R1.8 and used to assess nuclear DNA adducts. The 1.8-kb fragment (coding for exons 1 and 2 of the DHFR gene) was labeled using a random primed labeling kit, incorporating [α-³²P]dATP. To probe mtDNA, a strand-specific probe was made by generating run-off transcripts from the T7 promoter and incorporating [α-³²P]UTP. Cells (10⁶) were seeded in 10-cm Petri dishes approximately 14 h before the addition of drug. Cells were treated with HMTA and Adriamycin at varying concentrations and for the desired times. Cells were then harvested, and media were washed from the pellet with chilled PBS. DNA was extracted using the QIAamp blood kit, digested with BamH1 (to linearize the mitochondrial genome) or HindIII (to release the 1.8-kb DHFR fragment), and separated electrophoretically on a 0.5% agarose gel. DNA was transferred to nylon membranes and probed with the DHFR (nuclear DNA) and mtDNA probes. Adducts were quantitated, and the frequency was calculated as described previously using the Poisson distribution (42).

IMR-32 cells were seeded at 7.5 × 10⁵ cells/3.5-cm Petri dish. The cells were incubated with varying concentrations of HMTA and [¹⁴C]Adriamycin for the desired times and then harvested. The genomic DNA was isolated as described for the gene-specific detection of adducts and subjected to two phenol extractions and one chloroform extraction before being precipitated in ammonium acetate. Pellets were resuspended in 100 μm TE buffer, and DNA concentration was calculated at 260 nm. A 50-μl aliquot was added to 1 ml of Optiphase Hisafe scintillation mixture, and the incorporation of [¹⁴C]Adriamycin into DNA was quantitated in a Wallac 1410 Liquid Scintillation Counter and calculated as Adriamycin adducts/10 kb.

Because it has been shown that formaldehyde activates Adriamycin, leading to the formation of Adriamycin-DNA adducts, and that this process is also facilitated intracellularly by formaldehyde-releasing prodrugs, it was likely that HMTA, which is known to dissociate to formaldehyde and ammonia, may also facilitate the formation of Adriamycin-DNA adducts.

To establish the optimal pH for the release of formaldehyde from HMTA and hence the optimal pH for the formation of drug-DNA adducts, linearized end-labeled pCC1 DNA was reacted with Adriamycin (1 μm) and HMTA (1 mm) for 6 h at a range of pH values from 5.3 to 7.7 (Fig. 1). After a cleanup and denaturation procedure, DNA adducts (migrating as dsDNA) were separated from DNA devoid of adducts (migrating as ssDNA) by gel electrophoresis.

It was seen that the more acidic the reaction conditions, the greater the number of adducts detected. This was most pronounced at pH values <6.3 (Fig. 1). At pH 7.7, the percentage of DNA containing adducts was approximately 30%, compared with essentially 100% at pH 5.3. Because of the pronounced dependence of adduct formation on pH, the optimal HMTA concentration and rates of formation were examined at physiologically relevant pH extremes of pH 6.4 and 7.4 to reflect typical scenarios for tumor and normal cell environments, respectively.

The extent of formation of adducts at the two pH extremes was investigated to establish the optimal HMTA concentration required for additional studies of this reaction. Concentrations that resulted in moderate to high levels of adducts (50–80% dsDNA) were optimal. At very low levels, quantitation of adducts is less reliable, whereas at concentrations that resulted in >90% lesions, multiple adducts can contribute to one “cross-linked” DNA duplex, also making quantitation less reliable. DNA was incubated with Adriamycin (0.5 μm) and HMTA (0–2 mm) for 6 h at 37°C. HMTA-mediated Adriamycin-DNA adduct formation was clearly concentration dependent (Fig. 2), with increasing HMTA resulting in a near linear increase of adduct formation. At pH 6.4, the formation of adducts was more pronounced, with 3–4-fold more lesions formed compared with pH 7.4. The maximal enhancement (4-fold) of adducts (pH 6.4 compared with pH 7.4) was at a HMTA concentration of 1 mm, and this level was therefore used for all subsequent analysis of adduct formation at these pH values, unless stated otherwise.

Because maximal drug-DNA adduct formation is dependent on the availability of formaldehyde, it was useful to establish the time frame for complete hydrolysis of HMTA at the two pH extremes of interest. DNA was incubated with 0.5 μm HMTA and 0.5 μm Adriamycin over 8 h at pH 6.4 and 7.4. It can be seen that up to 5-fold more adducts were formed at pH 6.4 than at pH 7.4 (Fig. 3), but only after an extensive reaction time of 8 h. This indicates that at pH 6.4, HMTA is degraded much more than at pH 7.4, albeit slowly, leading to greater release of formaldehyde. For convenience, an incubation time of 6 h was selected for all additional studies.

It was important to assess the requirement for HMTA in drug-DNA adduct formation and to confirm that Adriamycin formed few adducts with DNA in the absence of HMTA at pH 6.4. Varying concentrations of Adriamycin (0–1 μm) were incubated with DNA in the presence and absence of 2 mm HMTA for 6 h (Fig. 4). There was a concentration-dependent increase in DNA adducts formed in the presence of HMTA at pH 6.4, but no lesions were detected in the absence of HMTA under these conditions. A concentration of approximately 0.6 μm Adriamycin was required to form 50% dsDNA under these conditions, and lesions were detected at drug levels above 0.2 μm.

To establish whether the HMTA-mediated drug-DNA adducts were similar to the formaldehyde-mediated Adriamycin-DNA adducts, the half-life of these lesions was determined. The loss of adducts was measured over 36 h at 37°C, and from the first-order kinetic decay (Fig. 5, inset), the half-lives were found to be 12.3 ± 1.3 h for the HMTA-mediated lesion, compared with 13.2 ± 0.5 h for the formaldehyde-mediated lesion (Fig. 5). The stabilities are therefore essentially identical and correlate with the previously determined half-life of 14.4 ± 1.9 h for formaldehyde-mediated DNA adducts (19), and this suggests that the HMTA-mediated lesion is the same as the formaldehyde-mediated lesion.

To investigate the potential of other widely used anthracyclines to form DNA adducts, varying concentrations of these anthracyclines (0–1 μm Adriamycin, daunomycin, and idarubicin and 0–50 μm epirubicin) were examined at a HMTA concentration of 1 mm. Although each of these drugs has a similar basic structure, the small changes in chemical composition resulted in significant clinical differences. Adriamycin and daunomycin induced adducts to a greater degree at pH 7.4 compared with idarubicin, whereas at pH 6.4, Adriamycin, daunomycin, and idarubicin all formed approximately the same amount of adducts (±5%) at all drug concentrations and at a much greater level that that detected at pH 7.4 (Fig. 6). Idarubicin (1 μm) formed approximately 5-fold more adducts at pH 6.4 compared with pH 7.4, whereas daunomycin formed approximately 3-fold more adducts at pH 6.4. In contrast, epirubicin exhibited a dramatically reduced ability to form lesions at both pH values.

A 188-bp fragment of DNA was 3′ end-labeled and reacted with Adriamycin and either HMTA or formaldehyde, resulting in the formation of drug-DNA adducts. This DNA was then subjected to 5′ exonuclease digestion using λ exonuclease, and the residual fragments were resolved electrophoretically (Fig. 7A). The location of blockages was compared with a Maxam-Gilbert G sequencing lane. The blockages were calculated as the mole fraction of fragments, with the DNA fragment length indicating the site of blockage (Fig. 7B). All blockages were observed 1–2 bases before a GpC sequence. The frequency of blockages was as high as 0.052 mole fraction (i.e., 5.2% drug occupancy) for the lesion associated with the blockage at the cytosine residue at position 111. For clarity, only lesions with a fractional occupancy of ≥1% have been shown. It was found that blockages for the HMTA-mediated adduct were the same as those observed for the formaldehyde-mediated adducts (Fig. 7A). In both cases, the lesions were specific to GpC regions. Because it has been shown previously that formaldehyde-mediated Adriamycin-DNA adducts are specific to GpC sites (43), this result confirms that HMTA and formaldehyde induce the same drug-DNA lesion and are in fact the same lesion.

Once the relationship between formaldehyde release from HMTA and the formation of Adriamycin-DNA adducts had been established in vitro, it was of interest to investigate this phenomenon in cells. IMR-32 neuroblastoma cells were used for this study because they represent a model of solid tumors and are one of the tumors that exhibit good clinical responses to Adriamycin (1). The IC₅₀ values of Adriamycin and HMTA were established alone and in combination (Table 1) and revealed a progressive increase of activity of Adriamycin with increasing levels of HMTA.

Cells were incubated with Adriamycin and HMTA at varying concentrations, and drug-DNA adducts were detected using a gene-specific assay (41). It was seen that drug-DNA lesions increased with increasing levels of HMTA. This relationship was evident in both nuclear and mtDNA, with approximately the same amount of lesions being formed in both genomes (Fig. 8). The control reactions (absence of prodrug) confirm the absolute necessity for HMTA to be present to induce detectable levels of drug-DNA adducts. At 15 μm Adriamycin and 1 mm HMTA, 1.9 and 1.8 lesions/10 kb were formed in mtDNA and nuclear DNA, respectively, after exposure to these agents for 8 h.

IMR-32 cells were incubated with 10 μm Adriamycin and 1 mm HMTA for 0–8 h. The longer the incubation period, the greater the extent of formaldehyde released and hence the more drug-DNA adducts that were formed. Adduct levels were approximately the same for both nuclear and mtDNA, with slightly more lesions observed in nuclear DNA (Fig. 9). At 10 μm Adriamycin and 1 mm HMTA, there were 0.48 and 0.40 lesions formed/10 kb in the nuclear DNA and mtDNA, respectively. To ensure that the concentrations of drug-prodrug used in the gene-specific assays did not result in high levels of nonviable cells in the time frame investigated, the viability of the cells (at maximal time points of Figs. 8 and 9) was assessed and found to be in excess of 95%.

The use of [¹⁴C]Adriamycin enabled the detection of lesions at decreased drug concentrations compared with that required for the gene-specific method described above. The formation of Adriamycin-DNA adducts was concentration and time dependent (Fig. 10). At 2 μm [¹⁴C]Adriamycin and 0–2.5 mm HMTA, an exponential increase in lesions was observed, with 23 lesions formed/10 kb at 2.5 mm HMTA. After an 8-h exposure to 2 μm [¹⁴C]Adriamycin and 1 mm HMTA, approximately 4 lesions/10 kb were detected. These levels of lesions are substantially greater than those detected using the gene-specific assay, due to the additional procedures required to isolate and prepare the genomic DNA in the gene-specific assay, and the resulting effects these procedures have on the labile Adriamycin-DNA adducts has been reported previously (22).

Current treatment of tumors with Adriamycin results in many dose-limiting side effects. The ability to increase the cytotoxicity of Adriamycin by using a formaldehyde-releasing prodrug lends to the possibility of reducing the concentration of Adriamycin required for similar levels of cell kill. HMTA is a potential solid tumor-specific activator of Adriamycin because it degrades preferentially in acidic environments to produce six molecules of formaldehyde and four molecules of ammonia (44).

This hydrolysis is also catalyzed by the presence of an active formaldehyde acceptor (44) such as Adriamycin.

The potential for HMTA to release formaldehyde suggests that it would serve as an efficient formaldehyde-releasing prodrug when coadministered with Adriamycin and that it would function similar to other formaldehyde-releasing prodrugs such as AN-9 (22). The hydrolysis constant of HMTA is favored by acidic conditions (44), and this indicates possible selective release of formaldehyde in low pH necrotic areas, hence providing the potential for selective enhancement of Adriamycin activity in solid tumors. Increased release of formaldehyde was clearly observed within the physiological pH range of 6.4–7.4, with up to 4-fold more drug-DNA adducts forming at pH 6.4 (Fig. 1). This suggests that a pH as low as 6.4 may result in selective hydrolysis of HMTA near tumor cells (compared with normal cells), indicating that HMTA may serve as a useful prodrug for the site-specific delivery of formaldehyde. This acid-dependent release of formaldehyde from HMTA has been reported in humans (45), where up to 9% of the total HMTA dose was detected as formaldehyde in the urine (pH range, 5.5–6.8) with even greater release (up to 30%) in the acidic stomach environment (46).

The essentially identical stability of the formaldehyde-mediated adduct with that of HMTA-mediated adducts indicates that these lesions are the same. The half-lives of the formaldehyde-mediated lesion and HMTA lesions were essentially identical (Fig. 5). HMTA-mediated adducts were specific for GpC sequences, also identical to that known for the formaldehyde-mediated Adriamycin lesion (Ref. 47; Fig. 6, A and B). Collectively, these two independent results confirm that both lesions (formaldehyde-mediated and HMTA-mediated lesions) are actually the same lesion. Formaldehyde directly activates Adriamycin, resulting in drug-DNA adducts, whereas release of formaldehyde from the hydrolysis of HMTA activates Adriamycin, resulting in formation of the same formaldehyde-mediated Adriamycin-DNA adducts.

The preferential release of formaldehyde at pH 6.4 (compared with pH 7.4) resulted in an increase in adduct formation of Adriamycin, daunomycin, and idarubicin at the lower pH value. The methoxy group of Adriamycin and daunomycin therefore appears to enhance drug activation at pH 7.4 compared with idarubicin. The addition of the hydroxyl group on Adriamycin (absent on daunomycin) does not significantly contribute to formation of drug-DNA adducts at either pH or to the stability of the adducts (the half-lives of the daunomycin and Adriamycin-DNA adducts were essentially identical; Ref. 19). Idarubicin exhibited an approximate 9-fold increase in adduct formation at pH 6.4 compared with pH 7.4, but it is not clear why the methoxy group should have any influence on Schiff base chemistry or formation of the drug-DNA aminal (-N-CH₂-NH-) linkage. Epirubicin does not appear to form a significant level of drug-DNA adducts, and this relates to the fact that in the presence of formaldehyde, an initial cyclization to a five-membered oxazoline ring forms, similar to that present in doxoform (21), which subsequently undergoes a nucleophilic ring opening, leading to the aminal bridge found in the anthracycline-DNA adduct (21). The formation of the oxazoline does not take place in epirubicin because the NH₂ and OH groups are found in equatorial positions, whereas in Adriamycin, daumomycin, and idarubicin, the NH₂ group is found in an equatorial position, and the OH is in an axial position, favorable for five-membered ring cyclization.

The potential use of HMTA is not compromised by any known toxicity, and it has been used previously as an antibacterial agent for the treatment of urinary infections (48). The IC₅₀ of HMTA in IMR-32 cells is 504.9 ± 5.6 μm, whereas the IC₅₀ of Adriamycin alone is 9.1 ± 0.3 nm. When fixed concentrations of HMTA were used in combination with Adriamycin, the IC₅₀ of Adriamycin was decreased, and this was more pronounced at higher HMTA concentrations (Table 1). This suggests that by combining Adriamycin treatment of cells with HMTA, the dose of Adriamycin required to achieve sufficient cytotoxicity could be reduced and may lead to reduced side effects. These data also suggest that HMTA is an efficient prodrug for the activation of Adriamycin.

The use of cell culture experiments established that the formation of Adriamycin-DNA adducts was enhanced by HMTA in both nuclear DNA and mtDNA (Fig. 8). At 1 mm HMTA and 15 μm Adriamycin, 0.28 and 0.24 lesions/10 kb were formed in mtDNA and nuclear DNA (measured in the DHFR gene), respectively. The similar amounts of lesions formed in both of these genomes suggest that the uptake of both drug and prodrug is not impeded by either the mitochondrial or nuclear membranes.

The release of formaldehyde from HMTA was also seen to be time dependent. With increasing incubation times, cells treated with both Adriamycin and HMTA formed increasing levels of drug-DNA adducts (Fig. 9). The relative time of addition of Adriamycin and HMTA did not significantly affect the amount of lesions formed over the time interval examined (IMR-32 cells were exposed to 2 μm Adriamycin for 8 h, and 1 mm HMTA was added up to 4 h before or after the addition of Adriamycin; data not shown). This is in contrast to that observed with the Adriamycin/AN-9 combination, where adduct levels were maximal when Adriamycin was administered 2 h before AN-9 (22). The present results suggest that formaldehyde may be released much more slowly from HMTA than from AN-9, thereby possibly avoiding the initiation of formaldehyde detoxification mechanisms.

To ensure that adduct levels detected in the gene-specific assay (requiring short-term exposure to 15 μm Adriamycin) were representative of more clinical concentrations, the absolute level of lesions formed was also determined at lower drug levels (2 μm) using [¹⁴C]Adriamycin, and the trends observed mimicked that observed using higher drug levels in the Southern-based gene-specific assay (compare Figs. 8 and 9 with Fig. 10). However, 8-fold more adducts were detected by scintillation analysis of ¹⁴C-labeled adducts (2 μm Adriamycin) than by gene-specific Southern-based analysis (10 μm Adriamycin) after exposure to drug for 8 h. This difference reflects the loss of the unstable adducts during the additional processing steps required for the electrophoretic procedure and has been discussed previously (41).

The above-mentioned data suggest that the total time of exposure with HMTA is important. It is not known whether this is due to the formaldehyde that is released from HMTA being more susceptible to time-dependent detoxification processes or whether this is due to other mechanisms. Because formaldehyde is toxic to cells, it is not surprising that it has a limited availability once released within the cell. For clinical purposes, further investigation would be required in animal models to define the toxicity and optimal times and doses for this drug/prodrug combination.

Although HMTA can potentially release up to six times as much formaldehyde as compared with AN-9, it is surprising that the total number of HMTA-mediated adducts formed is lower than that detected with equimolar levels of AN-9 (data not shown). This may be due to the slower release of formaldehyde from HMTA, to other components released from AN-9 (butyric acid and pivalic acid), or to differential rates of uptake of the two formaldehyde-releasing prodrugs.

HMTA releases formaldehyde in a time- and pH-dependent manner and a decrease from pH 7.4 to pH 6.4 resulted in a dramatic increase of activation of Adriamycin to form drug-DNA adducts in vitro. HMTA releases formaldehyde, which then activates Adriamycin, resulting in drug-DNA adducts, and these adducts are the same as those formed by direct activation by formaldehyde. HMTA-mediated Adriamycin-DNA adducts form specifically at GpC sites, as do formaldehyde-mediated lesions, and HMTA can activate anthracyclines other than Adriamycin, including daunorubicin, idarubicin, and, to a lesser extent, epirubicin. These anthracyclines also form more adducts at pH 6.4 compared with pH 7.4. The idarubicin (1 μm) and HMTA (1 mm) combination yielded approximately 5-fold more lesions at pH 6.4 compared with pH 7.4. A small change in pH can therefore greatly increase the hydrolysis of HMTA to release formaldehyde. HMTA also releases formaldehyde in IMR-32 cells, resulting in formation of Adriamycin-DNA adducts, and this relationship was also time and concentration dependent.

The potential for HMTA to release formaldehyde in a pH-specific manner therefore may provide for selective localization and treatment of tumors with Adriamycin. It has been shown that HMTA releases formaldehyde in tumor cells; however, the localization of this release has yet to be investigated. This localization could enable a reduction in drug concentrations required to achieve effective tumor cell kill. Overall, the selective activation of Adriamycin (and other anthracyclines) by formaldehyde-releasing prodrugs may yield benefits such as tumor-specific cell kill and reduced drug levels, leading to a reduction of toxic side effects. Given the acidic nature of solid tumors, the acid-dependent release of formaldehyde from HMTA in solid tumors may provide a mechanism for a degree of localized activation of anthracyclines in these tumors. An additional specific example is stomach cancer, which could be treated p.o. with an Adriamycin/prodrug combination formulated at alkaline pH, and in the acidic stomach environment, HMTA is known to release formaldehyde, which in turn would activate Adriamycin. Other tumors that might be suited to this form of treatment are those with a tissue pH of <6.4 (e.g., squamous cell carcinoma, astocytomas, meningiomas, and sarcomas), but this treatment could also be applicable to those tumors with a pH of <6.8 (e.g., glioblastomas, mammary carcinomas, and adenocarcinomas; Ref. 31).