Gene-specific Formation and Repair of DNA Monoadducts and Interstrand Cross-links after Therapeutic Exposure to Nitrogen Mustards

Nitrogen mustards are bifunctional alkylating agents, and one of the main classes of clinically used anticancer drugs. These drugs react extensively with cellular macromolecules (DNA, RNA, and proteins), thus inducing multiple kinds of molecular lesions. The critical cellular target of these agents is DNA, which is alkylated primarily at the N-7 position of guanine with lesser reaction at the N-3 position of adenine (1). Besides monofunctional binding of the drug to a single site in the DNA molecule (monoadducts), drug-induced cross-linking between bases in the complementary strands of a DNA molecule (DNA interstrand cross-links), as well as DNA-protein cross-links, also occur (2, 3). Nucleotide excision repair and base excision repair play a crucial role in the repair of monoadducts (4, 5). Furthermore, a number of multistep DNA repair pathways including nucleotide excision repair, homologous recombination, and postreplication/translesion repair all contribute to the DNA interstrand cross-link repair (6).

Cyclophosphamide is a nitrogen mustard widely used, in combination with other agents, for the treatment of lymphoma, multiple myeloma, breast cancer, and systemic rheumatic diseases (7, 8). It is a known carcinogen associated with therapy-related leukemia and bladder cancer (9, 10). Cyclophosphamide requires metabolic activation by the cytochrome P-450 enzymes to become an alkylating agent (11). It has been reported that the drug induces 67% phosphotriester monoadducts, 26% N-7-guanine monoadducts and 6.7% N-7-guanine-N-7-guanine diadducts (12).

The nitrogen mustard melphalan does not require metabolic activation to become an alkylating agent and forms ∼38% N-7-guanine monoadducts, 20% N-3-adenine monoadducts, 20% N-7-guanine-N-7-guanine diadducts and 13% N-3-adenine-N-7-guanine diadducts (13). Melphalan can induce cancer: it gives rise to a variety of tumors in mice and rats; and there is some evidence that it can cause leukemia in humans (14). The role of melphalan today is essentially reserved for the management of multiple myeloma. Oral melphalan at low doses, in combination with prednisone, is still being considered as an acceptable primary treatment for elderly patients with multiple myeloma. Melphalan i.v. at high doses (200 mg/m²), with autologous stem cell support, is the standard high-dose treatment for many patients with multiple myeloma. Despite its remarkable activity, complete responses are noted in up to 40% of patients, long-term disease-free survival is seen in [l]10% of patients, and the majority of patients experience disease recurrence (15).

Evidence has accumulated that the biological role of DNA lesions must be evaluated in terms of the fine structure of DNA repair (16). Selective DNA damage and repair appear to affect cellular sensitivity to DNA-amaging agents, cancer susceptibility, aging,and other biological endpoints (17). It was shown previously that in Chinese hamster ovary cells there was heterogeneity of mechlorethamine-induced DNA damage, with active genes being damaged more than the bulk genomic DNA (18). Also, repair of mechlorethamine amage was faster in the active genes than in the genome as a whole or in noncoding sequences (18). Bartosova et al. (19) found that the gene-specific repair activity of UV irradiation-induced cyclobutane pyrimidine dimers in CD4⁺ lymphocytes can reflect the proliferative state of the cells as well as the transcriptional state of the gene. Finally, similar relative rates and extent of cisplatin-induced interstrand cross-links repair were observed in active or silent genes, in human G₀ or proliferating CD4⁺ lymphocytes (20).

This is the first report on the kinetics of gene-specific monoadducts and DNA interstrand cross-link formation and repair in a readily accessible tissue of cancer patients after chemotherapeutic exposure to melphalan or cyclophosphamide. Results of in vitro studies with lymphocytes from healthy volunteers after melphalan treatment are also given.

Blood samples were collected after informed consent from seven consecutive patients with multiple myeloma and receiving (200 mg/m²; Table 1) and one polyarteritis nodosa patient treated with cyclophosphamide (1 g/m²). Among the seven multiple myeloma patients, only one had been previously exposed to standard dose, oral melphalan. This was patient 3 who received melphalan and prednisone for 12 months as primary treatment, 5 years before the high-dose melphalan, and had achieved a partial response at that time. The remaining six patients have not received alkylating agents with the exception of a single dose of cyclophosphamide 4 g/m² which was given 8–12 days before high-dose melphalan as part of the regimen administered to mobilize peripheral blood stem cells. Samples of venous blood were collected just before and various times after completion of drug administration. To avoid the decrease in DNA yield from blood samples stored at −70°C, 0.67 ml of 0.48 g/100 ml citric acid, 1.32 g/100 ml sodium citrate, 1.47 g/100 ml glucose solution per 4 ml of blood was added. The sample was immediately transferred to −70°C freezer and used for the extraction of DNA from whole blood.

Peripheral blood mononuclear cells were isolated from freshly drawn peripheral blood from healthy volunteers using standard methods (21). Peripheral blood mononuclear cells (3 × 10⁶ cells/ml) were treated with 5, 10, or 25 μg/ml melphalan for 1 h at 37°C in RPMI supplemented with 10% FCS. In all experiments, cell viability, assessed by trypan blue exclusion, was >95%. After incubation, the cell suspensions were cooled in ice and washed twice. For time course experiments, cells were resuspended in drug-free medium and incubated at 37°C for 2, 4, 8, 24, and 48 h. Cells were harvested and stored at −70°C. All experiments were performed in duplicates.

Isolation of genomic DNA was performed as described by Sunters et al. (22) with minor modifications. After DNA isolation, genomic DNA was digested to completion with HindIII or EcoRI for the analysis of p53 or N-ras alkylation, respectively. Then, DNA samples, dissolved in sterile deionized H₂O, were heated at 70°C for 30 min to depurinate N-alkylated bases. Apurinic sites were converted to single-strand breaks by the addition of freshly prepared NaOH for 30 min at 37°C. The DNA samples were size-fractionated using a 0.6% agarose gel in Tris-borate EDTA at 2 V/cm for 16 h. The gel was transferred onto Hybond N⁺ (Amersham Pharmacia Biotech) as described by Southern (23).

Hybond N⁺ membranes were hybridized to a gene-specific labeled probe according to manufacturer’s instructions. The p53-specific probe was a 610-bp fragment of the human p53 gene, encompassing exons 7 and 8 and intron 7, and was amplified as described previously (24). The following primer pair was used: forward, 5′-AGG-TTG-GCT-CTG-ACT-GTA-CC-3′; reverse, 5′-ATT-GTC-CTG-CTT-GCT-TAC-CTC-3′. The N-ras-specific probe was a 112-bp fragment of the human N-ras gene and was amplified as described previously (25). The following primer pair was used: forward, 5′-GTT-ATA-GAT-GGT-GAA-ACC-TG-3′; reverse, 5′-ATA-CAC-AGA-GGA-AGC-CTT-CG-3′.

After exposure, films were scanned using a Molecular Dynamics Personal Densitometer, and band volumes were determined. The level of N-alkylpurines per fragment (L/F) were calculated using the equation L/F = log_e(fraction of undamaged DNA) (20). Autoradiographs used for these calculations were all from the linear range of the film.

For the detection of DNA interstrand cross-links, cells were treated as above, except that the pH of the lysis buffer was 10.4 (22). DNA samples were digested to completion with HindIII or EcoRI for the analysis of p53 or N-ras alkylation, respectively. In this case, alkylations were not converted to strand breaks, and DNA was denatured before gel electrophoresis and Southern blotting (22). The frequency of DNA interstrand cross-links was derived directly from the denatured samples; the frequency of DNA monoadducts was derived by subtracting that value from the value obtained from the analysis of N-alkylpurines.

The group t test was used to determine differences in the mean values between groups of samples. Statistical significance of a test was assumed when P < 0.05.

Human lymphocytes were treated with 5, 10, or 25 μg/ml melphalan for 1 h at 37°C and lysed immediately (for zero time points), or after reincubation in drug-free medium for the desired period (0–48 h). The drug doses used here were similar to the plasma levels of melphalan detected in multiple myeloma patients after chemotherapeutic treatment (26). Genomic DNA was prepared and N-alkylpurines were converted to strand breaks before Southern analysis. The blotted DNA was hybridized with probes specific for the human p53 or N-ras genes. Fig. 1 A shows the melphalan-dependent loss of signal from the full-length, single-stranded band as a result of N-alkylpurine formation in the p53 gene. Similar results were obtained by studying the N-ras gene (data not shown). Cyclophosphamide was not used in such experiments in vitro because this drug requires metabolic activation by the cytochrome P-450 enzymes to become an alkylating agent (11).

A dose-dependent rapid accumulation of monoadducts, reaching maximal levels within 2 h was found in both genes studied (Figs. 2,A and 3A Fig. 3.). Thereafter, adduct levels declined slowly to ∼10 or 20% of maximal levels after 48 h in the p53 or N-ras gene, respectively. As shown in Figs. 2,B and 3,B, the semilogarithmic plots of adduct concentration versus time obtained (2–48 h) are linear, indicating that repair is first-order in monoadducts with an apparent t_1/2 of 14.5 ± 0.3 h (mean ± SD) in the p53 gene and 18.8 ± 1.5 h (mean ± SD) in the N-ras gene. These results show consistent patterns of repair of monoadducts formed in both p53 and N-ras genes of peripheral blood lymphocytes after melphalan treatment in vitro. Furthermore, similar slopes are observed after different doses of melphalan, suggesting no change in the system available for repair of monoadducts during the whole experiment (Figs. 2,B and 3 B). Finally, no monoadducts could be detected before the drug treatment.

The extent of DNA interstrand cross-linking in the p53 and N-ras genes was also measured in human lymphocytes treated with 5, 10, or 25 μg/ml melphalan. In this case, alkylations are not converted to strand breaks, and DNA is denatured before gel electrophoresis and Southern blotting. Under the conditions used, gene fragments containing DNA interstrand cross-links reanneal in the gel and migrate as double-stranded DNA. Uncross-linked material remains denatured and migrates as single-stranded DNA. Fig. 1 B shows the formation and repair of DNA interstrand cross-links in the p53 gene. Similar results were obtained by studying the N-ras gene (data not shown). Control, nondenatured, double-stranded DNA is electrophoresed in parallel and serves as a marker for the migration of cross-linked DNA. The disappearance of the DNA interstrand cross-links, as measured by the assay used in the present study, reflects the excision of one of the cross-linked bases by a glycosylase or endonuclease (“unhooking” step) and not necessarily the complete removal of the damage (27). If inhibition of strand separation is the critical cytotoxic lesion, then the “unhooking” step may be the critical first step in the repair process. The resultant monoadduct may then be observed as persisting damage in the assay that measures total damage produced by the nitrogen mustards.

The accumulation of DNA interstrand cross-links was slower than that of the monoadducts, reaching maximal levels within 8 h in both genes studied (Figs. 2,C and 3,C). Thereafter, adduct levels declined slowly to reach ∼10 or 20% of maximal levels after 48 h in the p53 or N-ras gene, respectively. As can be seen in Figs. 2,D and 3,D, the semilogarithmic plots of adduct concentration versus time obtained (8–48 h) are linear, indicating that repair of DNA interstrand cross-links is first-order, with an apparent t_1/2 of 12.4 ± 0.8 h (mean ± SD) in the p53 gene and 14.1 ± 2.2 h (mean ± SD) in the N-ras gene. Again, these data show consistent patterns of repair of interstrand cross-links formed in both p53 and N-ras genes of peripheral blood lymphocytes after melphalan treatment in vitro. These results are in agreement with those of Hansson et al. (28) who found that the rate constant for the overall removal of DNA interstrand cross-links after melphalan exposure is 12.9 h. Also, Ross et al. (29) found a rapid removal of mechlorethamine-induced DNA interstrand cross-links (t_1/2: 7 h) in L1210 cells, whereas melphalan-induced cross-links were repaired more slowly. After different doses of melphalan, similar slopes are observed, suggesting no change in the cross-link repair system during the whole experiment (Figs. 2,D and 3 D). Furthermore, the data presented here indicate that the amounts of DNA interstrand cross-links in lymphocyte DNA were proportional to the melphalan doses, at all time points and in both genes analyzed. Finally, no DNA interstrand cross-links could be detected before the drug treatment.

As can be seen in Table 2, the ratio of interstrand cross-links to monoadducts depends on the time point analyzed and the dose of melphalan used. This ratio, at the end of the melphalan treatment, was close to those found by previous investigators (22). However, at the following time points, the ratio increased because of the protracted formation of DNA interstrand cross-links and the parallel repair of monoadducts, reached maximal levels within 8 h, and decreased thereafter.

Blood samples were collected from seven multiple myeloma patients (patients 1–7; Table 1) receiving single i.v. injections of melphalan at a dose of 200 mg/m² and from one polyarteritis nodosa patient (patient 8) after a single i.v. injection of cyclophosphamide at a dose of 1 g/m². Monoadducts and interstrand cross-links were detected in leukocyte DNA 2–48 h after exposure to these drugs.

The formation and loss of monoadducts in leukocyte DNA of drug-treated individuals are shown in Figs. 4, A and B, and 5, A and B. All eight individuals examined showed a maximum of the adduct levels 2 h after the drug injection in both genes studied. The mean values of peak monoadduct levels were very similar in the two genes analyzed (p53: 13.02 ± 3.68 adducts/10⁶ nucleotides, range 8.37–18.41; N-ras: 12.29 ± 6.27 adducts/10⁶ nucleotides, range 0–18.97). Thereafter, melphalan-induced adduct levels decreased by 34.7 ± 6.6% (range 26.3–41.2%) in the p53 gene and 29.0 ± 4.5% (range 24.1–34.2%) in the N-ras gene during the first 24 h, whereas adduct levels in the cyclophosphamide-treated individual decreased by 50% after 48 h. No monoadducts could be detected before the drug injection.

The formation and repair of DNA interstrand cross-links in leukocyte DNA of melphalan-treated patients are shown in Figs. 4, C and D, and 5, C and D. All seven individuals examined showed a relatively slow accumulation of DNA interstrand cross-links with a maximum 8 h after the drug injection in both genes. The mean values of peak interstrand cross-link levels were very similar in the two genes analyzed (p53: 1.42 ± 0.62 adducts/10⁶ nucleotides, range 0.80–2.48; N-ras: 1.37 ± 1.05 adducts/10⁶ nucleotides, range 0–2.81). Thereafter, adduct levels decreased by 29.5 ± 9.7% (range 21.0–41.7%) in the p53 gene and 30.0 ± 12.3% (range 17.9–47.8%) in the N-ras gene during the first 24 h. These results, in accordance with the in vitro data, show consistent patterns of repair in both p53 and N-ras genes in blood leukocytes DNA of melphalan-treated patients. After cyclophosphamide administration, DNA interstrand cross-links were detected in leukocytes within 2 h, whereas maximal adduct levels occurred after 24 h. This may not be a real peak level of the interstrand cross-links, because no blood samples were collected from this patient at the 8-h time point. No DNA interstrand cross-links could be detected before the drug injection. Moreover, the ratio of DNA interstrand cross-links to monoadducts after the chemotherapeutic treatment increased rapidly and reached its maximum value after 8 h in patients 1–7 (p53: 0.12 ± 0.02, range 0.08–0.15; N-ras: 0.13 ± 0.02, range 0.11–0.16) or 24 h in patient 8 (p53: 0.26; N-ras: 0.26). Thereafter, the ratio remained constant for up to 24 h in patients 1–7 and up to 48 h in patient 8.

Furthermore, we observed that among the seven patients with multiple myeloma, the three (patients 3, 6, and 7) who showed the least levels of DNA adducts over time, as represented by the AUC² for DNA adducts during the whole experiment (0–48 h), did not respond to high-dose melphalan (Table 1). Four patients responded to high-dose melphalan: one patient had primary refractory disease; and three patients had responded to conventional therapy (Table 1). Two of the three latter patients (patients 1 and 4) achieved a complete response, i.e., negative serum and urine immunofixation and absence of clonal plasma cells from the marrow. Patient 5 had the highest number of monoadducts and interstrand cross-links in both genes studied (Table 1).

Finally, marked differences were found between responders (patients 1, 2, 4, and 5) and nonresponders (patients 3, 6, and 7) to chemotherapy in the mean AUC values of both DNA adducts and in both genes analyzed (296.8 ± 57.9 versus 137.4 ± 11.0, respectively, and P = 0.006 for p53 monoadducts; 31.4 ± 13.4 versus 9.8 ± 1.5, respectively, and P = 0.04 for p53 interstrand cross-links; 314.0 ± 58.8 versus 109.2 ± 138.7, respectively, P = 0.04 for N-ras monoadducts; 36.4 ± 16.0 versus 8.8 ± 15.2 adducts/10⁶ nucleotides × h, respectively, P = 0.07 for N-ras interstrand cross-links; Table 1; Fig. 6).

The measurement of DNA adducts formed in patients treated with alkylating drugs is of interest, because such adducts may indicate the biological dose received at the level of the individual patient, which may be correlated with therapeutic response or long-term complications (mainly therapy-induced carcinogenesis). The presence of total adducts, monoadducts, and DNA interstrand cross-links as a result of in vitro and in vivo exposure to therapeutic drugs has been reported in the past (26, 28, 30, 31, 32, 33). However, the present study is the first dosimetry report of the gene-specific formation and repair of these adducts after in vitro exposure of human lymphocytes to melphalan or after therapeutic exposure of humans to melphalan or cyclophosphamide.

Of the chemotherapeutic alkylating agents known to be carcinogenic in humans, the nitrogen mustards are among the most potent (34). Acute myelogenous (nonlymphocytic) leukemia is the predominant mustard-associated cancer (35). In some case studies as many as 10–20% of long term survivors were found to contract this disease, which is frequently preceded by a preleukemic condition known as myelodysplastic syndrome (MDS). Although not a single gene has yet been identified which is consistently mutated in mustard-associated acute myelogenous (nonlymphocytic) leukemia, mutations in p53 occur in a small fraction of cases, whereas N-ras mutations occur somewhat more frequently (25–50% of patients) (24, 36). Thus, in the present study, the p53 and the N-ras genes were chosen as experimental tools.

In all experiments described here, the highest monoadduct levels were found within 2 h, whereas DNA interstrand cross-links accumulated slowly and reached maximal levels within 8 h in both genes analyzed. The relatively long time between monoadduct and DNA interstrand cross-link formation observed in the present study confirms previous reports (37, 38). Kohn (3) proposed a two-step model for the protracted formation of DNA interstrand cross-links by melphalan. According to this model, rapid binding of the drug to a nucleophilic site in one DNA strand as a monoadduct is followed by a delayed formation of a DNA interstrand cross-link through the binding of a second chloroethyl group to a site in the cDNA strand. Thus, repair of monoadducts before cross-link formation may play an important role in protecting cells from melphalan and cyclophosphamide cytotoxicity and may be a significant factor leading to chemotherapy failure. DNA interstrand cross-links represent only a small fraction of the total adducts formed. However, they are thought to be the main determinant of the toxicity of the nitrogen mustards (28, 39, 40). The exact mechanism by which DNA interstrand cross-links cause cell death is not known. Mu et al. (41) found that DNA interstrand cross-links induce a futile repair synthesis which is not accompanied by damage removal. It is possible that this futile repair cycle and the potentially preapoptotic signals resulting from this cycle, rather than the replication block per se, are the main causes of the lethality of cross-link-inducing anticancer drugs. Interestingly, the data from both genes analyzed in the present study show that smaller doses give greater ratios of DNA interstrand cross-links to monoadducts, indicating that smaller doses are more effective in inducing toxic DNA interstrand cross-links (Table 2). This finding has possible implications for the future design of more efficient chemotherapeutic protocols.

Individual differences in the response of tumor cells to antineoplastic agents in vitro as well as correlations between drug sensitivity in vitro and individual responses to therapy have been observed as well (42, 43). However, to perform these assays, invasive surgical treatment is necessary to obtain a sufficient amount of tumor cells, and they are rarely used in clinical practice. Determining DNA adduct formation and repair in peripheral blood leukocytes of patients may provide a noninvasive method for evaluating the effectiveness of some antineoplastic drugs. Previous investigations found a relatively small interindividual variation in Hodgkin’s disease patients treated with procarbazine and lymphoma patients exposed to dacarbazine, because adduct levels in different individuals fell within a range of three (44, 45). Studies on cisplatin adducts have indicated large interpatient variation in adduct formation apparently related to therapy (46, 47). In contrast, measurement of DNA adduct levels after melphalan administration showed a generally low degree of interpatient variation (26). Recently, Spanswick et al. (48) found that similar levels of high-dose melphalan-induced DNA interstrand cross-links were observed in plasma cells from multiple myeloma patients who had not received melphalan therapy (melphalan naive) and those previously treated with melphalan. However, significant differences in DNA interstrand cross-links repair were observed: cells from naive patients showed no repair; whereas those from previously treated patients exhibited between 42 and 100% repair. In the present study, marked differences were found in the individual amounts of each type of DNA adducts over time, as represented by the AUC for DNA adducts during the whole experiment (0–48 h; Table 1; Fig. 6). Among the seven patients with multiple myeloma, the three (patients 3, 6, and 7) who showed the lowest levels of DNA damage over time did not respond to the melphalan therapy (Table 1).

The results reported here show that it is feasible to measure monoadducts and DNA interstrand cross-links in a readily accessible tissue of humans exposed to bifunctional alkylating drugs and indicate a possibility of examining at the level of the individual patient the relationships between the induction/repair of cytotoxic DNA damage and clinical response or long-term complications. In this way, we expect to gain insight into the causes of individual variability in response to chemotherapy and to access the usefulness of DNA damage and repair as predictive molecular markers for the design of more effective and individualized therapeutic protocols.

We thank Dr. S. A. Kyrtopoulos for helpful discussions.