Half-intercalation stabilizes slipped mispairing and explains genome vulnerability to frameshift mutagenesis by endogenous “molecular bookmarks”

Abstract

Some sixty years ago chemicals that intercalate between base pairs of duplex DNA were found to amplify frameshift mutagenesis. Surprisingly, the robust induction of frameshifts by intercalators still lacks a mechanistic model, leaving this classic phenomenon annoyingly intractable. A promising idea of asymmetric half-intercalation stabilizing frameshift intermediates during DNA synthesis was never developed into a model. Instead, researchers of frameshift mutagenesis embraced the powerful slipped-mispairing concept that unexpectedly struggled with the role of intercalators in frameshifting. I propose that the slipped-mispairing and the half-intercalation ideas are two sides of the same coin. Further, I review existing findings to test predictions of the combined “half-intercalation into the slipped-mispairing intermediate” model against accumulated knowledge. The existence of potential endogenous intercalators and the phenomenon of “DNA bookmarks” reveal ample potential for natural frameshift mutagenisis in the cell. From this alarming perspective, I discuss how the cell could prevent genome deterioration from frameshift mutagenesis.

Graphical Abstract

DNA duplex is a stack of planar base pairs, surrounded by two spiral sugar-phosphate backbones (left). Planar polycyclic molecules are known to intercalate symmetrically between DNA duplex base pairs (right). This review proposes how asymmetric half-intercalation could promote one-nucleotide insertions or deletions at slipped-mispairing intermediates during DNA synthesis (center).

1. Introduction: Finding the information molecule and mutating it

The late 1940s and early 1950s were exciting times for DNA research, as several lines of powerful evidence came together pointing to DNA as the information molecule of the cell. Among them were the genetic transformation and bacteriophage infection studies in bacteria ^[1], the cellular DNA content and metabolic stability studies in vertebrates and flowering plants ^[2] and of course the crowning jewel — the crystal structure of the DNA double helix ^[3], framing a compelling model of information transfer based on base-pairing ^[4]. The “info-transfer” core of life turned out to be an elegant design, in which the infinitely “high” stack of base pairs is held together by the two spiral cables of the sugar-phosphate backbone.

One of the intriguing pieces of early evidence for the information role of nucleic acids was the finding that the action spectrum of cell-inactivating UV-irradiation coincides with that of nucleic acids, rather than that of any other cellular constituents ^[5], while UV also blocks DNA synthesis ^[6]. When UV-treatment was found not only to kill, but also to mutate ^[7], a fruitful concept of a direct targeting of the information molecule was born. Nitrogen mustard contributed a similar line of evidence: not only was it highly mutagenic ^[8], but it also specifically blocked DNA synthesis ^[9], while its interaction and the effect spectra were specific for DNA (in those times referred to as “thymonucleic acid”) ^[10]. As researchers realized that DNA is the molecule to target if mutations are desired, they attempted to disturb DNA metabolism, with immediate results ^[11] that, among other things, confirmed the info-carrier role of DNA. Soon, incorporation of bromodeoxyuridine (BrdU) into DNA in place of thymidine ushered the use of base analogs for mutagenesis via alternative base pairing (Fig. 1A) ^[12].

Fig. 1. The early findings and models.

A. Alternative base pairing explains transition mutations. Double arrows indicate base pairing. The two natural base pairs (A:T, G:C) are shown in black. A pyrimidine base analog bromodeoxyuridine (BrdU), that can pair with either adenine or guanine, is shown in blue. A purine base analog 2-aminopurine (2AP), that can pair with either thymine or cytosine, is shown in red. B. The structure of two intercalator mutagens: 9-aminoacridine and proflavine. C. The model of Lerman of acridine interaction within duplex DNA. DNA base pairs are shown as light blue “pebbles” stacked on each other, surrounded by the two darker blue spirals of the sugar-phosphate backbone. The normal DNA duplex is shown on the left. A duplex in which four orange molecules of acridine have intercalated is shown on the right. Note that the backbone is distorted to accommodate intercalation. Also, fewer base pairs per turn means that DNA accommodating intercalators becomes positively-supercoiled. The black “X” marks the position of crossing-over between the two DNAs, according to Lerman’s explanation of how intercalation induces frameshift mutagenesis.

2. Two types of point mutations identify two distinct ways to change DNA sequence 1 nt at a time

In the late 1950s, Freese studied the nature of mutants in bacteriophage T4 generated by two base analogs, BrdU and 2-aminopurine ^{[13, 14]}. He found that mutations were all changes of the original base for another base (base substitutions). Most of them were substitutions of one purine for the other purine (or one pyrimidine for the other pyrimidine), which he called “transitions” ^[13] (Fig. 1A). Less frequently, he had also seen substitution of a purine for one of the two pyrimidines (or vice versa, a pyrimidine for one of the two purines), — he called them “transversions” ^[13]. Importantly, Freese had also found that base substitution mutations could be reverted by the same base analog mutagens that caused them in the first place, thereby confirming their origin through alternative base-pairing (Fig. 1A) ^[13]. Thus, a method to generate point mutations (changes of one or a few nucleotides in the same location) was born.

But right away a complication developed in the same rII cistron of T4, where other types of chemicals also stimulated mutations ^[15]. Brenner and Orgel, comparing the robust mutagenesis by acridines (Fig. 1B) to the one by BrdU found that 1) acridines also generated what looked like point mutations; 2) these mutations could not be reverted by base analogs like 5-BrdU or 2-AP (this was already reported by Freese ^[13]), suggesting a different nature; 3) however, the acridine-induced mutations were revertible by acridines, the same chemicals that produced them in the first place; 4) at the same time, acridines could not revert the base-substitution mutations, produced by base analogs ^[16]. Thus, DNA sequence changes affecting a single base pair turned out to comprise at least two kinds: the base substitutions and the new, mysterious kind. As it turned out, the mysterious kind represented the second dimension of the information variability of the DNA molecule, — the number of base pairs!

3. DNA structure tolerates intercalation

The clue to the mystery was soon provided by Lerman and colleagues who showed that a particular acridine, proflavine (Fig. 1B), interacts with DNA duplex strongly yet noncovalently, by sliding between adjacent base pairs using stacking interactions ^{[17, 18]}. Perhaps the most incredible aspect of the proposed scenario was the ability of the seemingly tight sugar-phosphate backbone of DNA to accommodate an additional planar molecule between adjacent base pairs (Fig. 1C), but the experimental evidence was compelling. Such “intercalation” mode of non-covalent DNA binding was consistent with a number of phenomena (^[19]; also see ^{[20, 21]} for later reviews): 1) proflavine binds strongly to DNA, but at a limited number of sites (~1 per 5 base pairs) and perpendicular to the main DNA axis; 2) the DNA binding reduces proflavine’s chemical reactivity; 3) the binding increases the length of the DNA molecules; 4) viscosity of DNA is also increased due to increased rigidity and stabilization of the internal structure; 5) proflavine-bound DNA duplexes acquire greater thermal stability.

Acridines (Fig. 1B) are representatives of a broad group of diverse chemicals, known as “intercalators”, that are able to insert themselves between adjacent base pairs in duplex DNA. Typical intercalators are polycyclic aromatic agents, most of them heterocycles with some positive charge, — in this respect they superficially resemble nucleic acid base pairs ^[22]. DNA bases are either single (pyrimidines) or double (purines) heterocycles themselves, and within DNA duplex the hydrogen bond distance between the bases of the base pair approximates another ring hexamer. Hence molecules with at least three connected rings, such as acridines, intercalate readily and form good stacking in which they are sandwiched between two base pairs ^{[21, 23, 24]}.

4. Intercalators cause frameshifts

The finding that mutagens like acridines intercalate between DNA base pairs prompted Brenner and colleagues to guess the nature of the new type of point mutations as one base pair insertions or deletions (“indels” in modern terms) and to propose a one-sentence idea of how these 1-bp-indels could form ^[25] (Fig. 2): “acridines are bound to DNA by sliding between adjacent base-pairs, thus forcing them 6.8 A apart, rather than 3.4 A. If this occasionally happened between the bases on one chain of the DNA, but not the other, during replication, it might easily lead to the addition or subtraction of a base.”

Fig. 2. Brenner’s idea of half-intercalation inducing frameshift mutagenesis.

DNA structure is shown in navy, the frameshift-affected base pair is in light blue, the intercalator molecule is in red. A. Normal DNA synthesis. B. Half-intercalation in the primer strand leads to 1-nt deletion. C. Half-intercalation in the template strand leads to 1-nt insertion.

Such 1-bp-indels, if happened within an open reading frame of a gene, would shift the translation frame one nucleotide, with severe genetic consequences for the gene ^[26] — therefore they became known as “frameshift mutations” ^[27]. Thus, point mutations turned out to belong to two major classes: they were either base substitutions or frameshifts. No further classes of 1 bp changes were ever identified (or in fact possible).

But the elegant idea of Brenner and colleagues (Fig. 2) was too speculative for 1961, because it made two (quite reasonable, but unsupported at that time) assumptions: (#1) a robust asymmetric, one-strand intercalation into duplex DNA (we will call it “half-intercalation” for clarity, after ^[28]); (#2) to cause frameshift, this half-intercalation should be near the 3’-end of the extending primer during DNA synthesis. Unfortunately, assumption #2 was generalized by others at that time to mean “chromosomal DNA replication”, which immediately made the idea unlikely. Not only did Lerman observe exclusively symmetric, “full” intercalation affecting equally both strands of duplex DNA ^{[17, 18]}, but also intercalators caused no mutagenesis in growing bacteria, being mutagens only for big lytic phages developing in these bacteria ^[29]. Since bacterial replication was obviously impervious to intercalators (that issue, due to permeability problems and active excretion of intrecalators from bacterial cells, was later solved ^{[30, 31]}), while the affected phages were all known to exercise extreme levels of homologous recombination, Lerman naturally proposed that intercalators promote frameshifts during homologous recombination, inducing something akin to unequal crossing-over, but at the nucleotide level ^[18] (as schematically represented by Waring ^[32]). It was a safe proposal, as the nature of homologous recombination was still shrouded in mystery at that time, while the principle of unequal crossing-over ^[33] was not only visually-appealing ^[34], but seemed relevant. It should be stressed that, in contrast to Brenner’s idea of half-intercalation during DNA synthesis (Fig. 2), Lerman’s idea was based on the symmetric, two-strand intercalation operating with non-replicating DNA duplexes (Fig. 1C), — reflecting what was considered at that time the optimal conditions for intercalator-induced mutagenesis.

5. Streisinger explains frameshifts by slipped mispairing

However, the notion of frameshift formation due to asymmetric events during DNA synthesis, — the template and the primer strands shifting relative each other, — very soon resurfaced and in fact became extremely popular (ironically, the new idea specifically avoided involvement of intercalation). Perhaps the resistance to the original asymmetric frameshift mutagenesis idea of Brenner and colleagues ^[25] was due to the assumption that the slipped-mispairing intermediate, stabilized by the asymmetric intercalation near the extending 3’-end (Fig. 2), would be unstable without intercalator, — while spontaneous frameshift mutations form with relatively high frequencies, similar to base substitutions ^[35]. The expectation that a strand with an extra nucleotide would not be able to form a local duplex with the complementary strand lacking this nucleotide was in fact already tested and refuted: the two strands still formed a (less stable) duplex, accommodating the unpaired nucleotide(s) in an extra-helical loop ^[36].

Moreover, nucleic acid synthesis in vitro by both RNA polymerase ^[37] and DNA polymerase ^[38] on templates with repeats was found to be prone to enzyme slippage along the template, resulting in the products either much longer than, or uncharacteristic of, the original templates. Likely influenced by all these findings, Streisinger and colleagues proposed their famous slipped-mispairing model of spontaneous frameshift mutagenesis ^[39]. According to this model, small indels are formed during repair synthesis, when the 3’-end is extended across a run of the same base-pairs or short repeats. In case of 1 bp indels, if the primer “slips” 1 nt forward or backward on a run of the same nucleotide (Fig. 3A→B), the DNA polymerase may not notice the slip, as the 3’-end remains perfectly paired, — and will continue polymerizing as if nothing happened, creating a frameshift. The Streisinger model did not elaborate how the product of replication, the indel-mismatch, would survive the subsequent mismatch repair proofreading of the extra-helical loop, as mismatch repair of indels was described only some 20 years later ^[40].

Fig. 3. Streisinger’s model of slipped mispairing to explain frameshift mutagenesis and possible roles of intercalators in amplifying this mutagenesis (framed).

DNA structure is shown in navy, the frameshift-affected base pair (A, B and F only) is in light blue, the nucleotide that is or will be pushed extrahelical is in light purple, the intercalator molecule is in red. A. Normal DNA synthesis. B. The primer strand 3’-end slips one nucleotide backwards on a homonucleotide run, forming a compensatory bulge with the extra base bulging out. C-F: Various ideas on how intercalators could promote frameshifts. C. The slipped mispairing intermediate is stabilized by two symmetric intercalations into the DNA duplex on both sides of the bulge. D. The slipped mispairing is stabilized by stacking interactions with intercalators, sandwiching the extrahelical base. E. Slipped mispairing is caused by intercalation into the ssDNA template. F. (Not a part of the original model) Half-intercalation across the extra base absorbs the bulge into the duplex and stabilizes the slipped mispairing intermediate.

The slipped mispairing model (Fig. 3A→B), being naturally asymmetrical, complements the half-intercalation-promoted frameshift mutagenesis idea of Brenner (Fig. 2), explaining how the very 3’-end of the primer strand could move back or forth along the template strand without losing complementarity contact with it. An obvious prediction of stabilization of the intercalator-frameshift intermediate by slipped mispairing is that acridine-stimulated frameshifts should preferentially happen in runs of the same nucleotide. This was indeed found to be the case ^{[41, 42]}. Surprisingly, Streisinger and colleagues struggled to reconcile their model with the fact that intercalators stimulate frameshift mutagenesis, originally proposing that symmetric intercalation stabilizes the already formed slipped complex (Fig. 3C) ^[39] and later adding that stacking with aromatic polycycles (which all intercalators are) could also stabilize the extrahelical base without actual intercalation (Fig. 3D) ^[41]. Curiously, subsequent students of intercalation adopted without changes Streisinger’s slipped-mispairing model of frameshift mutagenesis, in which intercalators were relegated to the secondary roles, stabilizing either the slipped intermediate ^{[24, 43]} or the extra-helical base ^{[44, 45]}.

In fact, from the perspective of Brenner’s asymmetric, half-intercalation idea (Fig. 2), the slipped mispairing model of Streisinger is the “missing half”. Indeed, since both half-intercalation and slipped-mispairing generate asymmetric (therefore less stable) duplexes, combining the two together in a single complex stabilizes both by returning the lost symmetry to the resulting duplex (Fig. 3B→F), as was noted at least once before ^[46]. However, the complementary relationship between slipped mispairing and half-intercalation is not obvious, — in fact, neither of the two original formulations mentioned possible existence of a missing half!

In summary, by the mid-1960s, the three ideas of how intercalation could promote frameshift mutagenesis were ^[32]: 1) the popular model of Lerman, according to which, with no connection to DNA synthesis, symmetric duplex intercalation stimulates unequal crossing-over ^[18] (Fig. 1C); 2) the speculative idea of Brenner that postulated half-intercalation to cause shifting of the primer strand relative the template strand ^[25] (Fig. 2); 3) the powerful slipped mispairing concept of Streisinger, based on a stochastic asymmetry of the DNA synthesis intermediate, where intercalators were not required, but could play an auxiliary stabilization role ^[39] (Fig. 3).

6. DNA polymerases slipped-mispair without intercalators

Unfortunately, there was no further theoretical elaboration of how intercalation could promote frameshifts, likely due to two subsequent developments. One was that by the mid-1980s, the favorite proposal of Lerman that intercalators promote unequal crossing-over ^[18] (Fig. 1C) had lost its relevance, as the molecular mechanisms of homologous recombination were found to be clearly incompatible with this scenario ^[47]. In stark contrast, the slipped mispairing idea of Streisinger (Fig. 3A→B) transformed into a mini-cult, completely overshadowing competing models. Although specifically proposed to explain 1 or 2 bp indels ^{[39, 41]}, the slipped-mispairing explanation grew broadly popular with students of genome rearrangements, becoming almost synonymous with a related concept of “template switching” and accounting for all kinds of deletions or additions promoted by the presence of short direct repeats in DNA ^[48].

In addition, in vitro studies with pure DNA polymerases confirmed their unexpected propensity for making spontaneous 1 nt indels, especially 1-nt-deletions — and the slipped-mispairing idea offered a testable explanation for these frameshifts ^[49]. Further interest in slipped mispairing was sparked by emergence of an alternative model to explain the paradoxical stimulation of DNA polymerase frameshifting in vitro by huge excess of one DNA precursor over the other three. Basically, in addition to the expected increased misincorporation of the dN in excess, certain frameshifts were also elevated in what could be described as “misincorporation-enforced template-primer dislocation”. The misincorporation could be either pending (called “dNTP-stabilized misalignment” ^[50]) or already accomplished (the “misincorporation” proper ^[51]), but regardless of the details, the new model offered a testable distinction from Streisinger’s. Indeed, slipped mispairing envisioned template-primer misalignment happening only in runs of same base pairs, the extrahelical base popping ≥2 nucleotides away from the 3’-end (Fig. 3). In contrast, the misincorporation model had no sequence requirements for frameshifts and positioned the extrahelical base right next to the (mispaired) 3’-end ^[50].

Subsequently, polymerases with proofreading capabilities turned out to be suboptimal in vitro experimental systems for the spontaneous frameshift mutagenesis, as they were able to either prevent or correct most of the template-primer misalignments ^[52]. At the same time, translesion DNA polymerases were found to be uniquely susceptible to slipped mispairing ^{[53, 54]}, — so much so, in fact, that physical analysis or determination of atomic structure of the misaligned template-primer pairs in their active sites became possible. The underlying property of translesion polymerases that would allow such complexes to form is the unusual tolerance of their active sites for extrahelical or additional nucleotides right near the primer 3’-end. Remarkably, while both physical experiments ^{[53, 55]} and crystal structures ^[56] were consistent with the misincorporation mechanism of 1 nt indels by translesion polymerases, the classic slipped mispairing intermediate in the runs of the same nucleotides was also reported ^[57]. One particular advantage of the misincorporation model of frameshifts was that it predicts formation of 1 nt-indels in random sequences, explaining significant frameshift mutagenesis by translesion polymerases outside same-nucleotide runs ^{[53, 54, 58]}.

The importance of translesion polymerases for the spontaneous frameshift mutagenesis in eukaryotes ^[59] and in stressed bacteria ^[60] further promoted the concept of template-primer misalignment during DNA synthesis (in both the slipped mispairing and the misincorporation versions), — while the interest in intercalators further amplifying frameshift mutagenesis fell on the wayside. Even in the obvious cases of the powerful stimulation of frameshifts by intercalators, the results would now be explained by Streisinger model in its pure form, that specifically excludes intercalation (Fig. 3A→B) ^{[44, 61]}. Indeed, why does one need intercalators if translesion polymerases make high enough levels of frameshifts by themselves? Therefore, even though advanced ideas on how intercalators could stabilize frameshift intermediates kept surfacing ^{[29, 41]}, a detailed model of how intercalation could promote frameshift mutagenesis was never formulated.

7. Propagation of the confusion — or why should we care?

The instrumental role of intercalator-induced frameshift mutations in figuring out the formality of the genetic code ^[26] made the intercalation phenomenon an important part of the fundamental narrative of Molecular Biology (visualizing DNA in gels with ethidium bromide (EtBr) staining also helped). Inevitably, therefore, the fact that intercalators cause frameshift mutagenesis became a fixture of the early molecular biology textbooks (for example, ^[62]). The modern schemes typically feature a depiction of a planar molecule symmetrically intercalated in the DNA duplex (Fig. 1C) (often a reproduction of the classic illustration by Waring ^[32]), accompanied by a separate panel showing the two types of frameshift mutations, 1 bp deletion or 1 bp insertion, that such symmetric intercalation mysteriously causes (Fig. 4). Perhaps such disjointed schemes reflect the fact that the primary effect of the symmetric DNA intercalation is a powerful inhibition of DNA synthesis, — the outcome both obvious from the nature of duplex DNA intercalation (generation of positive supercoiling in DNA) (Fig. 1C) and easily detectable ^{[20, 21]}. Remarkably, such schemes are mute on how strong DNA synthesis inhibition could lead specifically to 1-bp-indels, propagating the confusion.

Fig. 4. A scheme of a typical textbook “explanation” of how symmetric intercalation causes 1-bp-indels (frameshift mutations).

The two strands of the DNA duplex are shown with structural details, with bases represented by blue rectangles and deoxyriboses by yellow pentagons. The base pair affected in the frameshift mutations is shown in a darker blue, the intercalator is in red.

Yet, no matter how esoteric frameshift mutagenesis may sound today, almost 60 years since it contributed to figuring out the genetic code, development of a mechanistic model of intercalation-induced frameshifting is urged by practical reasons: 1) although originally believed to be rare, spontaneous frameshifts happen at frequencies similar to those of base-substitutions ^[35]; 2) because of its gene-inactivating nature, spontaneous or induced frameshift mutagenesis is a serious contributor to the overall genome instability; 3) obviously, we cannot explore the mechanisms of this important process without formulating its model and testing derived predictions; 4) less obviously, but more importantly, if we do not appreciate the critical contribution of intercalators to frameshift mutagenesis, we remain oblivious to the various classes of endogenous intercalators that are likely contributors to the relatively frequent spontaneous frameshifts; 5) last but not least, since the phenomenon of intercalation is widely used in the design of modern DNA-targeting drugs ^[63], the risk assessment of the associated frameshift mutagenesis is currently impossible without the knowledge of its mechanism. And exploration of such a mechanism should start with formulation of a model of the intercalation-induced frameshifts.

8. All intercalators inhibit DNA synthesis, but only some of them cause frameshifts

But before unifying the original half-intercalation idea ^[25] and the slipped-mispairing model ^[39] into a single scheme, let us first briefly consider the biological properties of various intercalators in relation to their physical capacity for DNA intercalation. The main biological readout for intercalators is their ability to inhibit DNA synthesis, both in vitro and in vivo. In fact, intercalators were known to block phage DNA replication when the DNA structure was not even known ^[64]. Moreover, due to this DNA synthesis-blocking capacity, modern-day intercalating drugs featuring minor-grove anchors or bis-intercalator capabilities offer potent anti-cancer treatments ^{[45, 63]}. The ability of intercalator drugs to kill human cells in culture (“anti-cancer activity”) is mechanistically related to their inhibition of topoisomerase II ^[44].

Similar to other treatments that interfere with the progress of replication forks, intercalators are also known as general mutagens (inducers of mostly base-substitutions and DNA rearrangements). At the same time, and quite remarkably, only some of the intercalators are specifically active as frameshift mutagens, as was first reported by Lerman himself ^[19]. These unexpected differences were further documented by Drake in his famous early book on the subject, with a necessary caveat about the importance of knowing the actual intracellular concentrations of the compared compounds ^[43]. Even in the simple tricyclic compounds of the same core structure and the same DNA intercalation capacity that vary only in the position of a solubilizing side chain, mutagenicity varies from almost as high as in the positive control (9-aminoacridine, Fig. 1B) to zero, with no apparent explanation ^[65].

Indeed, the ability of intercalators to bind duplex DNA correlates with their ability to inhibit DNA synthesis ^{[66, 67]}, but the correlation with their frameshift mutagenicity is perplexing. The best duplex DNA binders are usually poor mutagens, even though they inhibit DNA synthesis deeply and are potent anti-cancer agents ^[68], — the lucky combination making them safer therapeutics. Even the notoriously carcinogenic polycyclic aromatic hydrocarbons (PAHs), that readily intercalate into duplex DNA (^[69], and citations therein), induce no frameshift mutagenesis in the standard Ames test, unless “metabolically activated” ^[70] (oxidized into reactive species ^[71]).

The best known DNA-intercalator, EtBr, inhibits nucleic acid synthesis in vivo ^[24]. Ames and colleagues famously reported EtBr as a strong inducer of frameshifts after metabolic activation ^[72], but in fact without such activation, EtBr is a non-mutagen in the same indicator bacteria, although it still inhibits their growth ^[73], — presumably by inhibiting DNA synthesis. Moreover, the enhanced modern EtBr analogs, the SYBR family of super-intercalators (whose intercalation is additionally stabilized by their positively-charged side arms embracing the sugar-phosphate DNA backbone) ^[74] are non-mutagens in the standard Ames assay, whether activated or not ^[73]. At the same time, just like EtBr, SYBR green is a strong inhibitor of DNA synthesis, at least in vitro ^[75].

DNA synthesis inhibition should cause replication fork instability, dependence on recombinational repair and induction of the SOS-mutagenesis ^[76]. Thus, in addition to frameshifts, intercalators should also cause chromosomal rearrangements due to template switching and base substitutions due to induction of translesion polymerases ^[77]. Indeed, once the permeability issue of E. coli and Salmonella is solved, proflavine becomes a strong mutagen in both, — yet, interestingly, only 20% of the induced mutations are frameshifts, while 50% are base substitutions and 30% are “stable and multisite” (= deletions) ^[30], revealing significant replication fork problems. Cell’s reaction to the replication stress due to intercalation is important for frameshift mutagenesis, as proflavine induces no frameshifts in recombinational repair-minus mutants ^[30], — a reflection of the critical role of SOS-induced translesion polymerases in bacterial frameshift mutagenesis, revealed in the studies of adaptive mutagenesis in E. coli ^[60].

Thus, intercalators that can penetrate the cell envelope and stay inside in significant concentrations, mostly inhibit DNA synthesis (by symmetric intercalation into template DNA), causing general mutagenesis (point mutations of both types and DNA rearrangements). In addition, some of them also amplify frameshift mutagenesis. The natural suspicion is that the frameshift mutagens are those intercalators that are capable of something else in addition to DNA synthesis inhibition via symmetric intercalation. We speculate that this something else is the capacity for asymmetric, half-intercalation.

9. The half-intercalation model of frameshift mutagenesis

Guided by this realization that only chemicals capable of asymmetric intercalation could be frameshift mutagens, the slipped-mispairing scheme of Streisinger and the half-intercalation idea of Brenner fuse into a testable model of intercalation-promoted frameshift mutagenesis. It starts with a DNA synthesis substrate: the full-length template strand and the shorter complementary primer strand that is extended over the template strand (Fig. 5 top). The template-primer misalignment at a run of the same nucleotide shifts the primer 3’-end one nucleotide position forward over the template, pushing one template base out of the helix (Fig. 5 middle left), while half-intercalation across the extrahelical bulge returns the symmetry to this 1 nt-deletion intermediate (Fig. 5 middle right). Alternatively, the misaligned primer is extended, leaving behind a duplex with an extra-helical 1 nt bulge (Fig. 5 bottom left), — in this case, half-intercalation into the bulge not only restores the symmetry, but also masks the frameshift intermediate from subsequent mismatch repair (Fig. 5 bottom right). A similar scheme, where the same actions happen in the opposite DNA strands of the DNA duplex, leads to 1-nt-insertion (Fig. 6). It should be noted that the idea of asymmetric intercalation stabilizing frameshift intermediates is so obvious, that sometimes it is simply depicted without formal introduction ^{[46, 78]}.

Fig. 5. A model of 1nt-deletion caused by slipped-back template strand with subsequent compensatory half-intercalation into the primer strand.

The color designation is like in Fig. 4; in addition, the base of the extra-helical nucleotide is shown in lavender. After the template strand slips back, the generic scenario demands removal of the 3’-end of the primer if it does not match the template. The three frameshift-promoting intercalation possibilities are: 1) direct intercalation into the primer strand causing the slippage; 2) right after the slippage (still within the polymerase), potentially inhibiting proofreading; 3) after the polymerase has left, potentially inhibiting mismatch repair.

Fig. 6. A model of 1nt-insertion caused by slipped-back primer strand with subsequent compensatory half-intercalation into the template strand.

The color designation is like in Fig. 5. In contrast to the first frameshift-promoting intercalation scenario of Fig. 5, here intercalation without prior slippage is into the single-stranded template.

The strength of any model is in the testable predictions it generates. An obvious general prediction of the model in Figs. 5 and 6 is that the processes that intercalators compromise to induce frameshifts should be associated with DNA synthesis of any kind — which was shown to be the case in vivo ^[79] and is especially well-documented in vitro ^[80]. A particular genetic prediction of the DNA synthesis nature of frameshifts is that intercalator-induced mutations should pass through a heterozygote stage, — this was indeed shown for acridines ^[81]. Finally, a more specific prediction of the half-intercalation idea is that such asymmetric intercalation close to the extending 3’-end should destabilize template-primer interaction during DNA synthesis. Indeed, 9-aminoacridine was shown to destabilize the primer-template interaction of the synthesizing DNA polymerase in vitro ^[67]. In fact, the author argued that intercalators promote slipped mispairing by this destabilization ^[67] (Fig. 3E and F), rather than by stabilization of the slipped intermediate, as Streisinger originally proposed ^[39] (Fig. 3C). But this highly-suggestive evidence for half-intercalation is still indirect. How do the structural predictions of the model fare against the physical evidence?

10. Physical evidence for half-intercalation

The critical structural prediction of the model is that, in addition to the major, symmetric mode of intercalation into fully duplex DNA, the frameshift-causing intercalators should be also capable of a minor, asymmetric mode of intercalation (“half-intercalation”) into one of the DNA strands of a duplex, that could be envisioned as “wedging” into the side of the duplex (Fig. 7A). Evidence for such half-intercalation comprises several groups of facts.

Fig. 7. Half-intercalation via wedging and bookmarks.

DNA structure is shown in navy, the intercalator is in red. A. Asymmetric intercalation should lead to “wedging” into the DNA structure. B. DNA and the bookmark tripeptide Lys-Trp-Lys. C. Due to its positive charge, the tripeptide interacts tightly with the negatively-charged DNA backbone. D. The tryptophanyl residue wedges into the DNA duplex.

Intercalation into one strand of a DNA duplex assumes some ability to intercalate into single-stranded (ss) nucleic acids. Back in 1966, a remarkable “partial intercalation” model was proposed to describe intercalation into one strand of the duplex ^[82]. It explains the phenomenon of intercalation into ssDNA/ssRNA, which although significantly weaker than intercalation into duplex DNA, is readily detectable ^{[83, 84]}. On the other hand, the robust intercalation into ssDNA/ssRNA of tryptophan within the context of positively-charged oligopeptides is well-known ^[85].

Half-intercalation, if real, should be detectable even in the presence of the dominant symmetric intercalation, especially in the context of specific sequences that might favor asymmetric interactions. A crystallographic study distinguished the symmetric intercalation of 9-aminoacridine into fully duplex DNA from an asymmetric one-strand intercalation driven by stacking with guanines, — the authors argued that it is the latter that promotes frameshift mutagenesis ^{[86, 87]}. Of note, these authors propose yet another way intercalators could promote frameshifting, — by stimulating slipped mispairing during DNA synthesis via pre-binding to single-strand template DNA (Fig. 3E and Fig. 6 top right) ^[87].

An NMR study also found several modes of 9-aminoacridine intercalation into duplex DNA oligos, with some intercalation modes purely one-stranded (acridine again stacking with two consecutive guanines), and a detectable single-strand DNA intercalation ^[46]. The authors propose the half-intercalation model of frameshift mutagenesis, with the one-strand intercalation in the “shorter” DNA strand stabilizing the extra-helical nucleotide within the helix ^[46]. Sometimes bulky side chains prevent symmetric intercalation altogether, emphasizing half-intercalation. A derivative of a phenazinium, aposafranine, is capable of only partial, wedge-type intercalation ^[88], — likely because of the steric interference from its forth phenyl cycle, oriented perpendicular to the phenazinium tricycle.

The critical prediction of how half-intercalation promotes frameshifts concerns intercalation into the bulged DNA duplexes, in which one strand has extra nucleotide(s) expelled into the extrahelical bulge ^[36], — such bulged DNA essentially modeling the intermediates of slipped-mispairing (Fig. 3B). If frameshifts are indeed due to promotion or stabilization of slipped mispairing by half-intercalation (Figs 5 and 6), then intercalators should: 1) preferentially intercalate at the DNA bulge, even though 2) this intercalation could be only one-stranded, and therefore stacking will be decreased relative to the two-stranded, symmetric intercalation into regular DNA duplex. Indeed, intercalators EtBr and 9-aminoacridine bind bulged DNA duplex much stronger than regular helices of the same sequence, with binding sites close to or at the extrahelical base ^[89]. Moreover, NMR measurements of the degree of stacking of EtBr molecule within di- or trinucleotide-based minihelices revealed that the extra nucleotide in the dinucleotide-trinucleotide minihelix causes substantial decrease of EtBr stacking, arguing for half-intercalation of EtBr in those cases ^[84] (Fig. 3F).

In summary, there is significant physical evidence for half-intercalation as a minor occurrence that accompanies the dominant symmetric intercalation, perhaps even as an intermediate ^[90], and promising evidence for half-intercalation into the bulged DNA duplexes.

11. Do natural intercalators cause frameshifting?

In contrast to base substitutions, frameshift mutations tend to have severe consequences, inactivating the affected gene by truncating the corresponding protein due to frequent stop-codons outside open reading frames ^{[25, 26]}. Because of this, one expects the cell to ensure that spontaneous frameshifts are rare relative to base substitutions, but they are not ^[35], comprising on average ~10% of all point mutations across kingdoms ^[91]. Since frameshifts are efficiently intercepted at two levels: by DNA polymerase proofreading and by mismatch repair ^{[40, 92]}, the remaining significant background of these dangerous mutations suggests either the presence of endogenous intercalators in the cell or existence of cellular processes requiring intercalation, or both. In fact, several classes of potential endogenous intercalators are obvious, instantly flipping the question from “why so many frameshifts?” to “why so few?” Fortunately, the nature of endogenous intercalators suggests the likely metabolic tactics to keep “spontaneous” frameshift mutagenesis in check.

The first class of potential endogenous intercalators that comes to mind comprises the DNA bases themselves and especially their nucleosides that preposition the bases for a perfect half-intercalation into DNA duplex ^[93]. However, phosphorylation of nucleosides to nucleotides efficiently controls this intercalation threat, as the strong mutual repulsion of phosphates keeps nucleotides away from DNA. This logic is confirmed by the opposite example of facilitated intercalation of neocarzinostatin chromophore ^[94], or a plain cytosine ^[93], achieved by positively charging their sugar platforms, to promote electrostatic attraction to the negatively-charged DNA backbone. This is likely why natural sugars of nucleosides are never positively charged in the cell.

The second class of potential endogenous intercalators comprises aromatic polycycles, like pteridine and flavin heterocycles, that serve as active parts of important enzyme cofactors. Again, the obvious neutralization strategy for the cell against these compounds is to charge them negatively, to keep them away from DNA. This may be also why, when these potential intercalators, as well as the nucleotides in the example above, are synthesized by the cell de novo, they are always built on the platform of phosphorylated sugars. In general, the phosphorylated sugar platforms seem critical for keeping various potential endogenous intercalators from amplifying frameshift mutagenesis.

The third class of endogenous intercalators are parts of various DNA-interacting proteins that use their side chain intercalation to change DNA conformation for various purposes ^[95]. Perhaps the simplest and the best known is intercalation into ssDNA by SSB-type proteins ^[96]; at the same time, various dsDNA-binding proteins intercalate into duplex DNA. The intercalating part, consisting of several side chains, is shaped like a wedge and inserts into DNA from one side, kinking the duplex (Fig. 7A). The side chains making the wedge group vary, but there is usually one or two aromatic aminoacids (Phe, Tyr, Trp) to intercalate and one positively-charged side chain to secure attachment to the sugar-phosphate DNA backbone ^[95]. To prevent this protein side-chain intercalation from promoting accidental frameshifts during DNA synthesis, the cell would need to make sure that similar combinations of aromatic and positively-charged side chains in DNA polymerases stay away from their active sites.

Finally, the fourth class of endogenous intercalators, and perhaps the least regulated and thus the most dangerous of all, is derived from the third class and comprises oligopeptides combining both positive and aromatic amino acids, the simplest being dipeptides of the Lys-Trp and Arg-Tyr type ^{[97, 98]}. Studies of interaction of such oligopeptides with duplex DNA gave rise to the concept of “bookmarks”, — side chains that proteins can intercalate into DNA duplex in order to stay with DNA or to change its conformation ^{[97, 99]}, as outlined in the preceding paragraph. The once active field of oligopeptide-DNA intercalation was thoroughly reviewed back in 1981, providing an illustration of the classic “wedging” ^[100] (Fig. 7B→D). The most remarkable aspect of the oligopeptide bookmarks is that they are only capable of half-intercalation (because of the small size of their aromatic side chains) and therefore, according to the half-intercalation model of frameshift mutagenesis, should be extremely mutagenic. The obvious way for the cell to address this oligopeptide intercalation threat is to expeditiously degrade oligopeptides.

Overall, the challenge from endogenous intercalators appears substantial, as there are several potential classes — moreover, all of them are of the frameshift-promoting half-intercalator type. Yet the cell has obvious ways to curb this unwanted half-intercalation, — hence the observed tolerable levels of spontaneous frameshifts. Characterization of potential endogenous intercalators, as well as cellular tactics to prevent them from contributing to frameshift mutagenesis, should be addressed experimentally.

12. Conclusion and outlook

Typical mutagens either chemically modify DNA bases and compromise their coding capacity (MNNG, hydroxylamine) or integrate into DNA themselves through ambiguous base pairing (base analogs (Fig. 1A)). In contrast, intercalators comprise a distinct class of mutagens that bind duplex DNA structure specifically, yet non-covalently, to change the overall DNA structure (Fig. 1C), promoting extra base incorporation or deletion at the template-primer complex, by as yet unclear mechanism (Fig. 4) ^[44].

The proposed model of half-intercalation causing and/or stabilizing the slipped-mispairing intermediate of frameshift mutagenesis during DNA synthesis generates testable predictions. It envisions that intercalators could promote frameshifting via interfering with either DNA polymerase proofreading of the misaligned template-primer pair or with subsequent mismatch repair of the bulged duplex. Hence its obvious genetic predictions are that inactivation of either proofreading or mismatch repair should affect modestly intercalator-amplified frameshifting. In fact, intercalators continue inducing frameshift mutagenesis in mismatch-repair mutants in E. coli ^[101], making it unlikely that intercalation simply masks the extrahelical bulges from mismatch repair (Figs 5 and 6, the bottom rows).

There are also predictions amenable to structural studies. First, the model can be directly tested with oligonucleotides, especially with bulged DNA duplexes. Second, intercalators could be designed with enhanced half-intercalation over symmetric intercalation — such “molecular bookmarks” are expected to be especially strong frameshift mutagens. Third, such “designer” intercalators could be made to intercalate between specific nucleotides (“selective bookmarks”), elevating frameshifts in specific sequence context, and causing predominantly deletions or insertions. As a result of these tests, a detailed structural mechanism of intercalation-promoted frameshift mutagenesis will be available to guide intelligent design of future DNA-acting drugs. As an educational benefit, textbooks will be able to present both logical and experimentally-tested model of intercalation-promoted frameshift mutagenesis.