Enhanced Amsacrine-induced Mutagenesis in Plateau-Phase Chinese Hamster Ovary Cells, with Targeting of +1 Frameshifts to Free 3′ Ends of Topoisomerase II Cleavable Complexes¹ Free

The cleavable complex of topoisomerase II is an intermediate in strand passage wherein the two strands of a DNA duplex have been cleaved, on a four-base 5′ stagger. The 3′ termini of the breaks are free hydroxyls, but each 5′ terminus is covalently bound to a topoisomerase II subunit, and the two DNA ends are held together by interactions between two such subunits. Many topoisomerase-based cancer chemotherapeutic agents effect cell killing by trapping this cleavable complex intermediate and preventing religation of the DNA strands. This trapping can result in disruption of DNA metabolism, particularly the decatenation and separation of sister duplexes after replication, or can lead, by an unknown process, to irreversible double-strand breaks (1, 2, 3, 4).

At sublethal concentrations, topoisomerase inhibitors that stabilize the cleavable complex are potent mutagens, but the spectrum of the recovered mutations depends strongly on the nature of the genetic locus examined (5, 6, 7). At heterozygous loci in mammalian cells (as typified by the TK^± assay), these agents can produce specific-locus mutant frequencies as high as several percent, and most of the mutations are very large multilocus deletions (8, 9). However, at the hemizygous aprt³ locus of CHO-D422 cells, where such large deletions would be lethal, mutant frequencies are much lower, and smaller deletions and insertions (usually <20 bp), as well as base substitutions, become dominant. The small deletions induced by the nonintercalating topoisomerase II inhibitor teniposide in this system seemed to be targeted to potential sites of cleavable complex formation, but the exact positioning of the deletion end points with respect to the potential cleavage sites was highly variable, suggesting somewhat complex processing of the DNA ends before religation (10).

In T4 phage, intercalating topoisomerase II inhibitors such as 9-aminoacridine, proflavine, and the cancer chemotherapeutic agent amsacrine (m-AMSA), are potent frameshift mutagens, producing predominantly +1, +2, and −1 frameshifts that are targeted to hotspots of cleavable complex formation. These frameshifts are almost always positioned such that they can be explained by removal or templated addition of one to two nucleotides at the exposed 3′ DNA termini of the cleavable complexes (11), presumably by T4 DNA polymerase and its potent associated 3′→ 5′ exonuclease (12). The apparent low frequency of such events among mutations induced by the nonintercalating inhibitor teniposide in CHO cells (10) raises the question of whether such frameshifts might be induced only by intercalating inhibitors or, alternatively, might result from unique features of DNA processing in T4.

This and certain other proposed mechanisms of mutagenesis by these agents (most notably, misrepair of double-strand breaks; ref. 10) do not involve replicative DNA synthesis, raising the possibility that mutations could be induced even in nonproliferating cells. Indeed, we previously reported that two free radical-based DNA double-strand cleaving agents, bleomycin and neocarzinostatin, were significantly mutagenic only in confluence-arrested cells (13, 14).

To address both these questions, aprt mutations induced by treatment with m-AMSA in both log- and plateau-phase CHO cells were analyzed.

For plateau-phase treatment, 1000-cell inocula were grown to confluent monolayers in 6-well plates, as described (13), and treated with m-AMSA for 2 days, with fresh, drug-containing conditioned medium added every 12 h. The cells were then trypsinized, and an aliquot was plated at low density to assess clonogenic survival. The bulk of the cells were released into exponential growth for 6 days, and then a total of 2.5 × 10⁶ cells were plated in the presence of 0.4 mm 8-azaadenine to select aprt mutants (10).

For log-phase treatment, 5 × 10⁵ cells grown from a 1000-cell inoculum were seeded into a 75-cm² tissue culture flask. After incubation for 24 h, they were treated with m-AMSA for 16 h, and survival and mutagenesis were assessed as above (10).

Individual mutant colonies were recovered by trypsinization in cloning rings, and expanded to ∼10⁷ cells for automated DNA extraction (program 163; Autogen Instruments, Framingham, MA). Coding portions of aprt were amplified by PCR in two segments comprising either exons 1 and 2 or exons 3–5, and each exon was sequenced (10). Mutants lacking one or both products were subjected to PCR with additional primers within and flanking aprt to localize the sequence alteration (14). Mutants with no alterations in aprt exons were, likewise, subjected to additional PCR to screen for possible rearrangements in the large intron 2. No more than two mutants were selected from each treated culture; in cases where the two clones had identical mutations, they were assumed to have arisen from a single mutational event, and that mutation was counted only once in the data analysis.

Cloned, 5′-end-labeled fragments containing each of the aprt exons were prepared as described previously (14). Labeled fragments (∼ 0.2 μg) were treated with 4 units of human topoisomerase IIα (p170 form; Topogen, Columbus, OH) at 37°C for 10 min in 20 μl of the buffer provided by the vendor [50 mm Tris-Cl (pH 8.0), 120 mm KCl, 10 mm MgCl₂, 0.5 mm ATP, 0.5 mm DTT, and 30 μg/ml BSA] plus 10 μm or 20 μm m-AMSA. (Previous experiments with other fragments have indicated that the cleavage specificity of the human enzyme is indistinguishable from that of the rodent enzyme, purified from murine L1210 cells and provided by Y. Pommier, National Cancer Institute). Cleavable complexes were trapped with SDS plus proteinase K, and the DNA was phenol-extracted, precipitated, and analyzed on 7% denaturing polyacrylamide gels (10).

Site-specific cleavage was quantitated by phosphorimage analysis using ImageQuant 2.0 software (Molecular Dynamics, Sunnyvale, CA) and was normalized as follows. First, the phosphorimage intensity in each lane was transformed into a line graph. For each sample treated with topoisomerase plus m-AMSA, the region of the graph containing fragments resulting from cleavage within the exon was determined by reference to Maxam-Gilbert sequencing markers. The total integrated phosphorimage intensity in this region was calculated, and after subtraction of the intensity in the same area of the control lanes (average of the m-AMSA-only and topoisomerase-only samples), it was divided by the number of bp in the exon to give the average frequency of m-AMSA- and topoisomerase-dependent cleavage per base within that exon, expressed as phosphorimage intensity. The intensity of each individual band that seemed significantly stronger in the topoisomerase+ m-AMSA-treated samples was then determined and, after subtraction of the intensity of any detectable band at the same position in the control lanes, was divided by the average cleavage per bp, to give a dimensionless quantity representing the relative cleavage frequency at each site, as compared with an “average” base position in that exon (14). Any base position with a relative cleavage frequency >1.0 was considered to be a cleavage site.

The statistical significance of the apparent targeting of mutations to certain sequence positions was determined from the goodness-of-fit statistic (15) for 2 × 2 tables in which the classification of the observed mutation sites according to a given criterion was compared with the classification of all sequence positions in aprt exons according to the same criterion. In the specific case of +1 frameshifts, each observed frameshift was classified on the basis of whether it was consistent with duplication of a base at the 5′ terminus of a site for cleavage by m-AMSA plus topoisomerase II in vitro; for comparison, each base in all aprt exons was classified on the basis of whether duplication of that base would be indistinguishable from duplication of a 5′ terminal base at a cleavage site. For example, in the sequence … TCC↓CCA… (where “↓” represents a cleavage site), a +1 duplication to give … TCCCCCA… would be considered to be targeted to that cleavage site, but all four Cs in the sequence would have to be classified as potential target sites in the calculation.

The significance of differences between spectra was assessed by Monte Carlo simulations of the appropriate 2 × N tables, where N is the number of mutant categories (16). To assess differences in a particular category of mutant, all other categories were combined and the P for the resulting 2 × 2 table was multiplied by the number of hypotheses tested, in this case equal to the number of categories N.

In exponentially growing CHO cells, the survival and mutagenesis responses were qualitatively similar to those reported previously for teniposide, a nonintercalating but otherwise functionally similar topoisomerase II poison (10). Significant mutagenesis was seen even at >90% survival, but the mutant frequency quickly leveled off at about 7 × 10^-6/survivor (eight times background), and seemed even to decrease slightly at the highest dose (Fig. 1, A and B).

As expected from the known down-regulation of topoisomerase II in plateau phase (17), m-AMSA was about 10-fold less cytotoxic to plateau-phase cells, although the treatment time was 3-fold longer. At very low m-AMSA concentrations, the mutant frequencies in log- and plateau-phase cells were similar (e.g., 4.5 versus. 4.1 × 10^-6 at 0.01 μm). However, in plateau-phase cells, the mutant frequency continued to increase with drug concentration, reaching nearly 3 × 10^-5 at the highest dose where mutagenesis could still be assayed (Fig. 1, C– E). This response was more similar to that seen for the radiomimetic DNA-cleaving agents bleomycin and neocarzinostatin, in plateau-phase cells (13, 14).

Except for the presence of several large-scale deletions/rearrangements, the spectrum of m-AMSA-induced mutations in log-phase cells was similar to that of spontaneous mutations and included base substitutions as well as small deletions, insertions, and duplications (Table 1). However, in addition to large-scale rearrangements, m-AMSA treatment of plateau-phase cells resulted in a relative decrease in the proportion of base substitutions (P < 0.002) and an increase in +1 frameshifts (P < 0.05). Also, among the base substitutions from cells treated in plateau phase, an unusually high proportion were G•C→C•G transversions (11 of 22 as compared with 3 of 17 spontaneous substitutions). However, there was no apparent targeting of the base substitutions to sites of cleavable complexes. Contrary to the spectrum of base substitutions in lymphocytes of patients treated with the nonintercalating topoisomerase II inhibitor etoposide (18), there was no apparent increase in A•T→T•A transversions as a result of m-AMSA treatment of CHO cells. To assess possible differences between mutants induced by treatment with different concentrations of m-AMSA, each spectrum was partitioned into high- and low-dose spectra with approximately equal numbers of mutants in each (data not shown); no significant dose-dependent differences were found (P > 0.25 in both cases).

To evaluate specific hypotheses concerning the role of cleavable complexes in mediating various m-AMSA-induced mutations, it was necessary to determine the sites in aprt at which cleavable complexes would form. Therefore, cleavable complexes in end-labeled aprt DNA fragments treated with topoisomerase plus m-AMSA were trapped with detergent, and the cleavage sites were mapped at single-nucleotide resolution (Fig. 2). Except for a few of the weakest sites, nearly all of the detected cleavage sites were paired in opposite strands on a four-base stagger (Fig. 3), as expected for cleavage of both strands in a single cleavable complex (1, 2, 3). Furthermore, the two sites in a given pair usually showed similar cleavage intensities. There were two predominant cleavage hotspots, one at bp 1797–1800 in exon 3 and one at bp 2039–2042 in exon 4 (Fig. 2). Cleavage at these sites was ∼20-fold greater than the average cleavage frequency/bp and about twice as great as at any other sites in the aprt exons (Fig. 3).

Frameshifts corresponding to duplications of a single bp accounted for 5% of log-phase and 16% of plateau-phase m-AMSA-induced mutations, a total of 14 independent mutations. Nearly all of these (4 of 4 log-phase and 9 of 10 plateau-phase mutants; ▿ in Fig. 3) occurred precisely at the cleavage sites of cleavable complexes, as detected in aprt DNA in vitro. In every case, the frameshift could be described as a duplication of the base at the 5′ terminus of the transient, topoisomerase II-mediated break, consistent with a model involving polymerase-catalyzed, templated addition of a single nucleotide to the free 3′ end of the break, followed by resealing of the break (11). One of the two predominant hotspots for m-AMSA-induced cleavage in aprt (at bp 1800; see Fig. 2,A) was also a frameshift hotspot, accounting for one log-phase and three plateau-phase +1 frameshifts. Because there are only 116 sequence positions in aprt exons at which a single-base duplication would be indistinguishable from duplication of a 5′-terminal base at a cleavage site, the fraction of +1 frameshifts positioned at potential cleavage sites is much greater than would be expected on the basis of chance (13 of 14 versus 116 of 543; P < 10^-7). Several of the slightly larger duplications (e.g., +3 at bp 2037–2042 and +5 at bp 2272–2277; see Fig. 3 B) were also consistent with templated addition at free 3′-ends of topoisomerase II cleavable complexes. However, there were also several duplications (including the +3 duplication hotspot at bp 2276–2279) that did not conform to the templated extension model.

Small deletions were similar in log- and plateau-phase cells and ranged from 1–40 bp. Slightly less than half of these were deletions of one repeat unit in tandem singlet, doublet, or triplet repeats (e.g., shortening of the tandem triplet repeat GTTGTTGT at bp 2023–2030 to GTTGT; Fig. 3). The tandem deletions did not seem to be targeted to cleavable complexes. There were also several small insertions that were duplications of one repeat unit in similar tandem repeats (double lines above the sequence in Fig. 3). Nearly all (8 of 10) m-AMSA-induced duplications were associated with the prominent potential cleavage site at bp 2273–2276, which also lies within the stem of a large potential cruciform structure (19) spanning bp 2269–2326. Two additional insertions could be described as imperfect duplications of adjacent sequences [i.e., insertion of GTA immediately preceding CAGTA at bp 2272 (Fig. 3,A) and insertion of ACAGT preceding TCAGT at bp 2271 (Fig. 3 B)].

The other, nontandem deletions were slightly larger, predominantly between 9 and 21 bp, and were usually marked by 1–6 bp repeats at the deletion end points. Although these deletions often encompassed prominent potential cleavable complex sites (Figs. 3 and 4) and the average cleavage frequency within the deleted sequences was 2.5 times higher than that in sequences not involved in any deletion (data not shown), this apparent correspondence was not statistically significant.

Attempts to sequence the large-scale rearrangements are in progress. As reported previously, one of these was an apparent topoisomerase II subunit exchange event resulting in a precise reciprocal chromosomal translocation, without loss of even a single bp of sequence (20). However, mapping data obtained thus far (data not shown) indicate that most or all of the remaining rearrangements are accompanied by loss of large segments of aprt, although none of them seem to have lost the entire gene.

For log-phase cells, the proportion of 8-azaadenine-resistant clones having no detectable alteration in the aprt gene was comparable with that reported for other mutagens, but for plateau-phase cells, the proportion of such clones was unusually high, about 16%. Although these clones were not further characterized, they are reminiscent of the 25% of m-AMSA-induced 6-thioguanine-resistant clones of AS52 cells (putative gpt mutants) in which the gpt coding sequence and promoter remained unaltered, but in which there were large-scale rearrangements in the vicinity of gpt, as well as reduced gpt expression (7). The results in both systems could be explained by rearrangements that translocate the entire, intact aprt or gpt gene into a genomic region with a chromatin structure that suppresses transcription.

At most genetic loci, including the tk locus in mouse lymphoblasts (9), the integrated Escherichia coli gpt gene in CHO-AS52 cells (7), and the functionally hemizygous X-linked HPRT locus in several cell lines (5, 6, 21), m-AMSA and other topoisomerase inhibitors induce primarily large-scale deletion and rearrangement mutations. Although these mutations presumably result from cellular processing of cleavable complexes, the exact mechanisms of their formation are not known. Very few of these large-scale mutations have been analyzed at the sequence level, and it is, thus, difficult to distinguish between topoisomerase subunit exchange mechanisms and models involving more complex processing and joining of broken DNA ends.

The hemizygous aprt locus in CHO-D422 cells is a rather special case in that the gene is unusually small (∼2 kb) and is flanked downstream by an apparently essential intracisternal-A particle gene (22). This feature dramatically decreases the incidence of viable deletion mutants by constraining the downstream breakpoint to a region of a few thousand bp. Thus, point mutations and smaller deletions/insertions that might be rare in other systems (7) tend to dominate aprt mutation spectra, even for highly clastogenic agents such as X-rays (23), radiomimetic drugs (13, 14), and topoisomerase inhibitors (10), making aprt ideally suited for elucidating the mechanisms of these small-scale genetic alterations.

The most unusual feature of the spectrum of m-AMSA-induced aprt mutations in CHO-D422 cells is the presence of a number of +1 frameshifts, invariably reflecting duplication of a single bp in the sequence. Among spontaneous aprt mutations, +1 frameshifts are extremely rare, accounting for only one of over 200 such mutations that have been sequenced by various laboratories (10, 14, 19, 24, 25). The bases duplicated in most of the m-AMSA-induced +1 frameshifts were flanked on both sides by nonidentical bases, suggesting that they could not be accounted for by the replication slippage mechanisms often invoked to explain frameshift mutations (26). Most strikingly, 13 of the 14 m-AMSA-induced +1 frameshifts were consistent with duplication of a base at the 5′ terminus of a cleavage site in a potential topoisomerase II cleavable complex, a correlation far greater than would be expected from random positioning of these mutations (P < 10^-7). This is the same pattern reported previously for acridine-induced +1 frameshifts in T4 phage and is precisely the result predicted by a model involving polymerase-catalyzed extension of the exposed 3′ terminus, followed by religation of the break (11). Thus, it seems highly likely that the m-AMSA-induced frameshifts in CHO cells arise by an identical mechanism (Fig. 4). As has been noted previously (11), inhibitor data suggest that the topoisomerase II of T4 may be quite similar to the mammalian enzyme; certainly more similar than is, for example, E. coli gyrase.

However, in the T4 system, an almost equal number of acridine-induced −1 frameshifts, invariably reflecting loss of a base at the 3′ terminus of a cleavable complex, were also detected, and these have been attributed to removal of a single base by the potent 3′→5′ exonuclease of T4 polymerase, followed by ligation (11). In further support of this model, the prevalence of acridine-induced −1 frameshifts is increased in T4 strains having enhanced polymerase-associated exonuclease activity, and is reduced in strains having reduced exonuclease activity (12). Thus, the near absence of m-AMSA-induced −1 frameshifts in CHO aprt is most easily explained by the mammalian DNA polymerase having a less potent associated exonuclease.

A critical question with regard to the model in Fig. 4 is the exact mechanism by which religation is effected, whether by topoisomerase II, by DNA ligase (after removal of topoisomerase II), or by some more complex pathway. In vitro studies have shown that Drosophila topoisomerase II can form relatively stable cleavable complexes by reaction with single-stranded circular DNA and that these complexes can then react with oligomeric DNA duplexes bearing either blunt or staggered ends, resulting in ligation of the 5′ end of the linearized circle to the oligomer (27). This result indicates that a continuous DNA duplex is not absolutely required either to maintain reversibility of the cleavage/religation reaction, or to allow the religation. Thus, these results support at least the plausibility of the scheme shown in Fig. 4, with religation effected by topoisomerase.

A second question is whether m-AMSA intercalation, in addition to stabilizing the cleavable complex, might also act to stabilize a DNA duplex with a frameshift mismatch (presumably by intercalation into one strand only) and, thus, facilitate religation of the mismatched duplex. In certain systems that lack an acridine-sensitive type II topoisomerase (28), acridines induce +1 and −1 frameshifts in monotonous base runs, a result that tends to confirm the possibility that a frameshift-mismatched duplex might be stabilized by acridine intercalation into one strand. Likewise, the finding (10) that not a single +1 frameshift consistent with the model in Fig. 4 was detected among aprt mutations generated in log-phase CHO-D422 cells by teniposide (a nonintercalating but otherwise functionally similar topoisomerase II inhibitor) also suggests that m-AMSA intercalation may play some role in addition to merely stabilizing cleavable complexes. Admittedly, however, the number of +1 frameshifts involved is small (4 of 88 m-AMSA-induced versus 0 of 68 teniposide-induced mutations, in log-phase cells; P ∼ 0.13). (There are no data available on this question from the T4 system because T4 topoisomerase is insensitive to teniposide.)

The origins of other types of m-AMSA-induced aprt mutations are less clear. The present comparison of log- versus plateau-phase mutants was conducted partly as an attempt to determine whether certain types of mutations might be formed by replication-independent mechanisms. Although cells treated in plateau phase must eventually proliferate to allow detection of the mutants, it is expected that very few cleavable complexes will remain by the time the cells progress through S phase because m-AMSA-induced complexes are known to reverse rapidly on drug removal (29). Thus, the finding that mutant frequencies (with the possible exception of base substitutions) are much higher for cells treated in plateau phase suggests that, at least for mutations targeted to cleavable complexes, the conversion of the complexes to permanent DNA sequence changes must occur in G₁/G₀ phase, presumably by mechanisms other than attempted replication of a damaged template. Possible such mechanisms may include, in addition to that shown in Fig. 1, the conversion of cleavable complexes to frank double-strand breaks, followed by error-prone repair of those breaks by nonhomologous end-joining, leading to small deletions and insertions, as well as large-scale rearrangements. We previously invoked this mechanism to explain the apparent targeting of certain teniposide-induced small deletions and duplications in the aprt gene to sites of potential cleavable complex sites. Alternatively, frank breaks, once formed, may persist into S phase and promote deletions, insertions, and rearrangements during replication. Finally, Ripley (30) has proposed that essentially all small deletions and duplication induced in mammalian cells by topoisomerase inhibitors could be accounted for by a single mechanism in which one to several bases are added at the 3′ terminus of a break and/or deleted from the 5′ terminus. Whatever mechanism(s) is responsible, it seems clear that m-AMSA is capable of inducing mutations (including, presumably, oncogenic mutations) even in cells that are in a strictly nonproliferative state. Indeed, it may be that in the absence of replication or mitosis, cleavable complexes are still quite mutagenic but much less cytotoxic, thus accounting for the higher levels of mutagenesis achievable in plateau phase. The marked increase in the ratio of topoisomerase IIβ to topoisomerase IIα in plateau phase (31, 32) could conceivably be a factor, as well.

A substantial proportion of the m-AMSA-induced aprt mutations, in particular the base substitutions and the deletions at tandem repeats, show little, if any, evidence of being targeted to cleavable complexes. These types of mutations are not peculiar to topoisomerase inhibitors but are found in spontaneous mutation spectra, as well (10, 14, 19, 24). Nevertheless, the absolute frequencies of these mutations must be increased by m-AMSA treatment; otherwise they would be much less prevalent in spectra of treated cells, given the 6- or 27-fold increase in overall mutant frequency. One possibility is that these apparently untargeted mutations may result from a global decrease in replication fidelity arising, by some unknown mechanism, as a result of drug treatment. It is conceivable that such a loss of fidelity would persist even after the vast majority of drug-stabilized cleavable complexes have religated (33), as would be required to explain the prevalence of these types of mutations in cells treated in plateau phase.

The demonstration of m-AMSA-induced +1 insertions targeted to 3′ termini of cleavable complexes may have implications for m-AMSA-induced large-scale rearrangements, as well. Specifically, the finding that these 3′ termini can apparently be extended and then religated by topoisomerase II despite the frameshift mismatch in the usual four-base overlap suggests such extensions could also be tolerated in cases where subunits of two cleavable complexes exchange. Thus, contrary to the single apparent m-AMSA-mediated reciprocal exchange we have thus far sequenced (20), some such exchanges might show duplication of a base at the newly formed joint, even if the joint was, in fact, formed by subunit exchange, with topoisomerase II effecting the final religation.