The reactivity and sequence selectivity of 4a-d against ten different synthetic double-stranded DNA oligomers (ds-1-ds‑10) and against the synthetic single-stranded DNA oligo­nucleo­tide ON-1 of ds‑1 (Figure 3) was investigated using ESI-FTICR MS in order to rationalise differences in the compounds’ cytotoxicities.

An alkylation of the DNA oligonucleotides ON-1 and ON-2 by the drugs 5a-d results in the form­ation of covalent adducts denoted by ON-1*a-d and ON-2*a-d, respectively (Figure 4). In the absence of suitable nucleophiles like DNA, the drugs are partially hydrolysed to give the inactive alcohols 6a-d.

For the mass spectrometric investigations, the hydrochlorides of seco-drugs 4a-d were incubated with the DNA oligo­nucleotides in water for 24 h at 25 °C. The samples were then diluted with equivalent amounts of methanol and investigated directly by means of ESI-FTICR mass spectro­metry without preliminary purification or enrichment of the products as described previously [16,18,19]. Covalent adducts were identified by comparison of calculated and found masses and alkylation positions were determined by identification of the characteristic fragment ions formed from the alkylated oligonucleotides applying capillary skimmer dissociation (CSD) [16,18,19,29,30,31,32].

If only one of the two oligonucleotides of the double-stranded DNA was alkylated, the percentage of alkylation of ON-1 or ON-2 was calculated based on the relative peak intensities of the corresponding isotope peaks of the unreacted oligonucleotides of ON-1 and ON-2 after 24 h of incubation as compared to the same ratio after 0 h of incubation. In case both oligonucleotides were alkylated, the preferred alkylation of one of the two oligo­nucleo­tides was determined based on the relative peak intensities of the covalent adducts. Analogously, if an oligonucleotide was alkylated in more than one position, the preferred binding site was determined based on the relative peak intensities of the respective characteristic fragment ions. The percentage of alkylation of the single-stranded oligo­nucleotide ON-1 of ds-1 was determined by calculating the ratio of the peak intensity of the covalent adduct as compared to the unmodified oligonucleotide.

Figure 5 and Table 2 show the results of the mass spectrometric investigations (as described above) in terms of the main alkylation site and the percentage of alkylation. In Figure 5, alkylation positions are denoted by underlines and important differences in the base sequence in the 5'‑direction of potential competing binding sites of ON-1 as compared to ON-2 are marked in red. Notably, all drugs alkylated mainly the nucleobase adenine (A) in AT-rich DNA regions of at least four consecutive AT base pairs with the alkylated adenine situated at the 3'-end of this sequence. This sequence selectivity could also be observed for the natural products CC‑1065 and duocarmycin SA [5] and the alkylation of the oligonucleotide ON-1 of ds-4 by the drug 5d is the only exception hereof. Furthermore, all drugs prefer the nucleobase adenine (A) over thymine (T) in the first, second and third position in the 5'-direction of the binding site as is obvious from the preferred binding positions in the oligonucleotides ds-6, ds-7 and ds-8, respectively (Figure 5 and Table 3). Hence, the drugs preferably alkylate the oligo­nucleotides ON-2 of these double-stranded oligonucleotides instead of the oligonucleotides ON-1.

In addition, all drugs show a preference for purine bases (A or G) over pyrimidine bases (C or T) in the first position in the 3'-direction of the binding site. This results for example in the alkylation of A⁸ of ON-2 in ds‑6 additionally to A⁹ of the same DNA oligonucleotide (Figure 5, Table 3).

Interestingly, the seco-drugs 4a and 4c containing a right hand dimethylamino-ethoxyindole subunit prefer AT base pairs in the fourth position in the 5'-direction of the binding site with the AT-rich sequence being located in the middle of the double strand. This is obvious due to the lower alkylation efficiencies of the double-strands ds-2 and ds-3 as compared to ds‑1. In contrast, the seco-drugs 4b and 4d containing a right hand morpholinoindole subunit prefer CG base pairs in the fourth position in the 5'-direction of the binding site and AT-rich sequences located at the end of the double strand. This can be seen by the higher alkylation efficiencies of the double-strands ds-6 and ds-7 in comparison to ds-1. Furthermore, all drugs showed a much higher reactivity against the double-stranded DNA oligo­nucleotide ds‑1 than against the respective single-stranded oligonucleotide ON-1 of ds-1 (Table 2).

In summary, all drugs show a high selectivity for adenines in AT-rich DNA regions and differ only slightly in the preferred base sequence. Thus, differences in cytotoxicity are presumably not due to differences in sequence selectivity. Unexpectedly, also the alkylation efficiency regarding the same DNA oligonucleotide does not seem to correlate with the cytotoxicity of the drugs because, with the exception of ds-4, 4a and 4b alkylated the DNA with a significantly higher efficiency than the seco-drugs 4c and 4d even though they show a lower cytotoxicity as observed in the cell culture assays than the latter (Table 2).

Consequently other factors seem to influence the biological activity of these alkylating agents and it could be argued that differences in the rate of the cyclisation reaction of the seco-drugs 4a-d to give the drugs 5a-d or maybe differences in the rate of formation of the DNA adducts might play a crucial role. Therefore, the hydrochlorides of seco-drugs 4a-d were incubated with the double-stranded DNA oligo­nucleo­tides in buffer or water for up to 24 h at 25 °C and the kinetics of the reactions were investigated using HPLC. Since covalent and non-covalent adducts of the DNA and the drugs elute with the same retention time under the conditions of these measurements [16], mixtures of both covalent and non-covalent adducts are denoted ds‑x⁽*⁾a-d in the following whereas purely covalently bound adducts as detected by means of mass spectrometry are denoted ds‑x*a-d. For evaluation, the area under the curve (AuC) in the HPLC chromatograms of the respective seco-drugs and their derivatives was determined based on their absorption of light at λ = 350 nm because at this wavelength the absorption of unmodified DNA can be neglected. It should be noted that the results obtained are only semi-quantitative, since we did not correct the calculated concentrations according to the molar extinction coefficients of the seco-drugs and their derivatives; however, it can be assumed that the extinction coefficients are approximately the same. Indeed, the results allow a good and straight­forward comparison of the reactivity of the seco-drugs 4a-d and their drugs 5a-d, respectively.

Figure 6 shows a series of HPLC chromatograms obtained at 0-4 h after starting the incubation of the hydrochloride of seco-drug 4a with ds-1 in phosphate buffer (pH 7). As can clearly be seen, the seco-drug cyclises to give the corresponding drug 5a very rapidly and the formation of the respective DNA-adduct ds-1⁽*⁾a has finished already after 2 h. Table 4 shows the resulting AuCs determined after up to 6 hours of incubation of a 1:1 mixture of the hydrochlorides of the seco-drugs 4a-d with the DNA in phosphate buffer (pH 7).

The seco-drugs 4a and 4b cyclise rapidly to give the corresponding drugs 5a and 5b which subsequently form the respective DNA adducts ds-1⁽*⁾a and ds-1⁽*⁾b within two hours. Hence, no free seco-drugs or drugs can be observed after two hours of incubation. In contrast, the more cytotoxic seco-drugs 4c and 4d cyclise to the corresponding drugs 5c and 5d with a much lower reaction rate and, moreover, the DNA adduct formation proceeds quite slowly. Thus, after six hours of incubation, DNA adduct formation is not completed and therefore, seco-drugs 4c and 4d can still be detected in considerable amounts.

Since seco-drugs such as 4a-d cyclise to the corresponding drugs nearly quantitatively in less than 90 minutes in buffer without DNA [16,17], 4c and 4d obviously are stabilised by an interaction with the DNA. This interaction is relatively weak because it can be disrupted under conditions of chromato­graphy to give back the free seco-drugs 4c and 4d as well as unchanged DNA. In addition, the formation of stable non-covalent and covalent complexes of the respective drugs 5c and 5d with the DNA oligonucleotides seems to be disfavoured because the amounts of ds-1⁽*⁾c ds-1⁽*⁾d increase only slowly and after prolonged incubation times, free drugs 5c (up to 4 h) and 5d (up to 6 h) can still be detected. Interestingly, the products of hydrolysis 6a-d are not observed, indicating that the rate of nucleophilic attack by water is reduced in the presence of DNA.

Further, we investigated whether there might be differences in seco-drug stabilisation or DNA adduct formation correlating with the base sequence. Using again HPLC, the interaction of the four seco-drugs 4a-d with the DNA oligonucleotides ds-1-ds-10 was analysed. Table 5 and Table 6 show the respective results obtained after 24 hours of incubation of a 1:1 mixture of DNA with the hydrochlorides of 4a-d in water at 25 °C.

After 24 hours of incubation, only traces (≤9%) of the seco-drugs 4a and 4b were detected. Furthermore, with the exception of ds-4,ds-6 and ds-10, nearly all detectable amounts of the drugs 5a and 5b which had been formed during the time of incubation, were converted to the corresponding DNA adducts ds‑x⁽*⁾a and ds‑x⁽*⁾b, respectively. Additionally, even in those cases where only small amounts of DNA adduct were formed (ds-4,ds-6 undds-10), no hydrolysis of the drugs 5a and 5b to the hydroxylated derivatives 6a and 6b (Figure 4) occurred in contrast to a notable hydrolysis of these drugs in the absence of DNA [16]. This indicates a weak interaction of the drugs with the DNA oligo­nucleo­tides that stabilises the drugs against hydrolysis on the one hand, but disfavours the alkylation reaction on the other hand. Consistent with this observation, the formation of covalent adducts as determined by the mass spectrometric investigations is low regarding ds-4,ds-6 and ds-10.

Notably, the effect of stabilisation of the seco-drugs and drugs by the DNA is much more pro­nounced in the case of 4c and 4d as well as 5c and 5d, in comparison to that of 4a and 4b as well as 5a and 5b. Though in the absence of DNA all seco-drugs 4a-d cyclise quantitatively within 90 min in buffer or cell culture media [16,17], in the presence of DNA, the seco-drugs 4c and 4d cyclise to give the corresponding drugs 5c and 5d to a much lower extent than their methylated analogues 4a and 4b. Since the formation of the drugs is the prerequisite for the alkylation reaction, this might explain why the DNA alkylation is lower in case of 4c and 4d as compared to 4a and 4b. Interestingly, the extent of drug formation depends on the kind of double-stranded DNA oligonucleotide used, indicating a specific interaction of the seco-drugs with the DNA. The most pronounced stabilisation of the seco-drugs can be observed using ds-5, in the presence of which only about 40% of 4c and 4d, respectively, are converted to the corresponding drugs in 24 hours. Nevertheless, with the exception of ds-4,ds-6, ds-10 and partially also ds‑7, most of 5c and 5d reacted with the DNA oligonucleotides under formation of the non-covalent and covalent DNA adducts ds-x⁽*⁾c and ds-x⁽*⁾d as was already found for 5a and 5b. However, additional DNA fragments indicative of strand cleavage could be observed in case of ds-6,ds-7 and ds-9 after incubation with 4c and 4d in contrast to the respective incubations with 4a and 4b. Figure 7 displays HPLC chromatograms showing the DNA fragmentation exemplarily for the reaction of seco-drugs 4c and 4d with the oligonucleotide ds-6. DNA was detected using a wavelength of λ = 260 nm and the seco-drugs 4c and 4d as well as all their derivatives including DNA adducts were detected using a wavelength of λ = 350 nm. As can clearly be seen, the drugs 5c and 5d induce cleavage of the intact DNA to form DNA fragments a part of which is still covalently bound to the drug.

The high cytotoxicity of 4c and 4d might thus also be modulated by DNA strand cleavage. DNA lesions such as single- and double-strand breaks have been reported previously to be caused by analogues of CC-1065 and the duocarmycins such as adozelesin and bizelesin [33]. These damages can lead to cell death in case they cannot be repaired by the cellular repair machinery [34].

The hydrochlorides of seco-drugs 4a-d were synthesised according to previously published pro­cedures and stock solutions in DMSO were prepared [6,7,8,9]. The synthetic duplex-DNA oligomers ds-1-ds-10 and the single-stranded oligonucleotide ON-1 of ds-1 were purchased from IBA (Göttingen, Ger­many) as aqueous solutions (0.1 mm) of the sodium and ammonium salts, respectively. The phosphate buffer (pH 7.0) was composed of Na₂HPO₄/NaH₂PO₄ (10 mm) and NaCl (0.1 m) in bidistilled water.

Incubations of the hydrochlorides of seco-drugs 4a-d with the double-stranded DNA oligo­nucleo­tides ds-1 -10 and the single-stranded DNA oligonucleotide ON-1 of ds-1 were carried out in a 1:1 ratio of seco-drug to DNA. For each experi­ment, an aliquot (1 µL, 5 nmol) of the stock solution of the respective seco-drug hydrochlorides in DMSO (5 mmol L^-14a-d) was mixed with an aliquot of DNA in water (50 µL, 5 nmol, 0.1 mmol L^-1 DNA). Either, the reaction mixture was incubated at 25 °C for 24 h as such or phosphate buffer (pH 7, 50 µL) was added before starting the incubation. Samples were taken at different incubation times and analysed by means of mass spectrometry and chromatography.

For mass spectrometric investigations of the reaction mixtures, samples were taken at 0 h and 24 h, diluted with an equivalent amount of methanol and the resulting solution introduced directly into the ion source of the ESI-FTICR mass spectrometer. High-resolution mass spectrometry was performed using a 7 T-FTICR-MS instrument (APEX IV, Bruker Daltonics, Billerica, USA) equipped with an APOLLO electrospray ion source and a syringe pump (74900 series, Cole-Parmer, Vernon Hills, USA) with a flow rate of 2 µL min^‑1 for sample injection. The ions were accumulated in the hexapole region for 0.8 s and transferred subsequently into the ICR cell. For gentle desolvatisation the drying gas temperature was set to 100 °C and the capillary exit voltage to -100 V. Enhanced fragmentation of alkylated oligonucleotides was achieved by capillary-skimmer dissociation (CSD) with a capillary exit voltage of -150 V. Ions were generated in the negative ion mode.

HPLC separations were performed with an Agilent 1200 with DAD from Agilent Technologies, an Aquapore OD-300 Column (220 × 4.6 mm, 7 µm) from Perkin Elmer and a Bondapak^® C18 Column (300 × 3.9 mm, particle size 10 µm, pore size 125 Å) from Waters. Samples were eluted within 45 min with a flow rate of 1 mL min^-1 (Aquapore OD-300) or 2 mL min^-1 (Bondapak^® C18) at 28 °C by applying a two-stage gradient (0-2 min: 5% B, 2-22 min: 5→20% B, 55-45 min: 20→80% B, 45-50 min: 80% B, 50-60 min: 80→5% B). Eluent A: 0.1 mol L^-1 triethylammonium acetate buffer (H₂O, pH 7.0). Eluent B: 0.1 mol L^-1 triethylammonium acetate buffer (80% acetonitrile and 20% water, pH 7.0). The absorption of the drugs and their derivatives at λ = 350 nm was used to calculate the AuCs.