`Ann. Rev. Biophys. Bioeng.
`
`© 1981 by Annual Reviews
`Copyright
`
`Inc. All rights reserved
`
`THE INTERACTION OF
`.9160
`INTERCALATING DRUGS WITH
`NUCLEIC ACIDS
`
`Helen M. Berman and Peter R. Young
`
`The Institute for Cancer Research, Fox Chase Cancer Center, Philadelphia,
`19111
`Pennsylvania
`
`INTRODUCTION
`
`In recent years increased attention has been focused on the ways in
`
`
`which drugs interact with biological systems, with the goal of under
`
`standing the toxic as well as chemotherapeutic effects of these small
`
`
`molecules. In the cell many drugs, particularly those with planar chro
`
`mophores, bind to nucleic acids. It is thought that this complex forma
`
`tion may be the first step in mutagenesis and possibly carcinogenesis.
`Nucleic acids with their evenly stacked base pairs and shallow and deep
`
`
`grooves are attractive targets for these molecules. Planar drugs can
`
`intercalate between base pairs, drawing them apart from their normal
`3.4 A spacing to 6.8 A while bulkier drugs can fit into either groove,
`
`
`
`sometimes with minimal distortion of the structure. Both procedures can
`
`profoundly influence the recognition properties of the nucleic acid.
`
`
`Some drugs such as actinomycin D, which is a powerful antibiotic, and
`
`daunomycin, a potent chemotherapeutic agent, exhibit both binding
`modes.
`This article is confined to a discussion of the binding of intercalating
`
`
`drugs to poly-and oligonucleotides, with emphasis on the results of
`
`solution and X-ray crystallographic analyses of some typical members
`of this class (Figure 1).
`
`THE INTERCALATION HYPOTHESIS
`
`In 1961 Lerman (60) published a classic account of the physical chemi
`cal behavior of DNA in the presence of small amounts of planar
`
`
`
`acridine molecules. The results of sedimentation and viscometry experi-
`
`0084-6589/81/0615-0087$01.00
`
`87
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 87
`
`Further
`
`ANNUAL
`REVIEWS
`Quick links to online content
`
`UC Ex-2002
`Thermo Fisher v. UC Regents
`IPR2018-01156
`
`
`
`88 BERMAN & YOUNG
`
`RI'.H2• �R3"�'H 9-anlnoocrldlne
`(f)
`3 •• H.(C2H5)2.�'CI. R3"H. �-oCH3 QuInacrine
`RI'.H.CH(CH31.(C'2)
`Rl'�'H. �'R3".H2 proflavIne
`aranq e
`Rl'�'H, �R3""(CH3)2 acrIdIne
`
`ethldlun
`R-C2"S
`R-(CH21.T. (C2"5)2 propldlun
`I GH3
`�
`° 0" CotH
`3
`YYYxt
`OH
`° vtH;- 0...J
`0" 0
`CH� °
`daunomycin H�
`l-AI" - Me l-N-IleVal 0-CN
`.
`0" ��-D-Ser l-fY' D-ser-�. J
`D
`/ "N-/ "
`"
`� � ° '" /DIIte,>-
`/
`N
`l-N-MeVaJ L-CyS l-Ala
`echlnanvctn
`
`3
`
`octlnonr;"ctn D
`
`or (tPHI
`�tertlYlPt(Hml+
`
`
`
`
`
`Irehdlanlne A ClDA)
`
`Figure 1 Some examples of intercalating agents most commonly studied.
`
`ments indicated that these drugs cause DNA to become longer and
`
`of 3.4 A spacing and the loss of the layer line stiffer. The retention
`
`
`
`by a model in
`
`pattern in the fiber diffraction photograph were explained
`
`
`which the planar chromophore is sandwiched in between the base pairs
`
`
`manner. The many experiments done
`of the double helix in an aperiodic
`
`
`
`
`since the proposal of this intercalation hypothesis have validated, re
`
`
`fined, and extended the model in attempts to answer the many ques
`tions it raised:
`
`1. What is the relationship
`between intercalation and the biological
`
`
`
`effects of acridines and other drugs with planar chromophores?
`
`
`2. What are the detailed conformational changes that occur upon
`intercalation?
`
`drug?
`preferences? 4. Are there base sequence or composition
`
`
`5. What is the maximum ratio of dye/DNA that can be accommo
`dated in the helix?
`
`
`
`
`
`3. In what ways is the intercalation geometry affected by the particular
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 88
`
`
`
`NUCLEIC ACID-DRUG COMPLEXES 89
`
`In the descriptions that follow, it will become apparent that while our
`
`
`
`
`
`knowledge has advanced considerably over what it was 20 years ago,
`
`
`none of these questions are yet fully answered, and indeed the experi
`
`ments described below raise many more.
`
`BIOLOGICAL EFFECTS OF INTERCALATING
`
`DRUGS
`
`The wide range of biological effects of intercalating drugs has been
`
`
`(34, 75, 108) and is briefly
`extensively documented
`
`summarized here. At
`
`
`
`
`the macroscopic level, the most commonly cited effects are inhibition of
`
`
`cell growth, cell death, and cell transformation. Since these effects are
`
`
`
`
`particularly noted with rapidly proliferating cells (108), many of the
`
`
`
`commonly studied intercalating drugs find uses as antibacterial, anti
`
`
`parasitic, and antitumor agents.
`At the subcellular level, intercalating drugs produce substantial
`
`
`
`
`
`
`changes in chromatin structure (108) and can result in the "curing" or
`
`
`
`and mitochondrial selective loss of small circular DNAs such as plasmids
`
`
`DNA (8, 34). These observations suggest that the common site of action
`
`
`
`
`can be direct effects of these drugs is the DNA. Indeed, three possible
`drugs to DNA: (a) inhibition traced to the binding of intercalating of
`
`
`enzymes, (b) frameshift DNA dependent mutagenesis,
`and (c) damage
`
`
`
`
`to the DNA. It is likely that many of the observed cellular effects result
`
`
`from one or more of these direct activities (48). Furthermore, such
`
`
`
`
`activities are sufficiently defined to allow quantitative structure-activity
`
`
`relationships to be developed (30).
`
`
`
`Inhibition oj DNA-Dependent Enzymes
`
`
`
`Enzymes typically hindered by the intercalative drugs are DNA poly
`
`
`
`
`
`merases (16, 42), RNA polymerases (106, 125), and nuc1eases (29)
`
`
`
`
`
`resulting in inhibition of replication, translation, repair, and processing
`
`
`(108). In most cases these enzymes are inhibited by competition for
`
`DNA sites with the intercalating drug. However, there can be dif
`drugs. Thus, in a study of T7 RNA
`
`ferences in action between different
`
`
`
`polymerase activity (106) it was noted that 9-aminoacridine inhibited
`
`
`
`initiation of RNA synthesis, whereas actinomycin D was active only at
`
`
`
`the elongation step. On the other hand, di-acridine derivatives linked by
`
`
`alkyldiamine chains were found to be inhibitory at only some of the
`
`
`
`
`promoters and not at all active against elongation, suggesting some
`
`
`sequence specificity in their action.
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 89
`
`
`
`90 BERMAN & .YOUNG
`
`Frameshift Mutagenesis
`
`Frameshift mutagenesis is the deletion. or addition of base pairs to
`
`
`
`
`
`
`
`DNA, resulting in a shift of the codon-reading frame and usually loss of
`
`
`
`
`inducers of the active product. Intercalating agents are the archetypal
`
`
`
`such mutagenesis, and yet there are only limited examples of a correla
`
`
`tion between binding affinity and mutagenicity. Thus, Lerman (61)
`
`
`
`
`
`observed no correlation between acridine binding affinity and mutagen
`
`
`
`icity, although Pezzuto et al (88) did find a correlation for some
`
`
`3-amino-l-methyl-5H-pyrido[4,3-b]-indoles. Furthermore, the frameshift
`
`
`
`activity of the steroidal diamine irehdiamine, which cannot intercalate
`
`
`
`due to its nonplanarity (65), calls into question whether intercalation is
`
`
`required at all. Part of this lack of correlation may be due to differences
`
`
`
`in transport between different drugs or binding to other cellular recep
`tors such that some drugs do not reach the target DNA as effectively.
`
`
`
`Even allowing for this, there is strong evidence for frameshift activity
`
`
`
`being further dependent on both the exact sequence and the nature of
`
`the drug (104). A good example of this is the ability of 9-aminoacridine
`
`to revert only one of two Salmonella typhimurium
`
`deletion mutants that
`(i.e. GGGG is favored over
`
`differ only in the sequence to be reverted
`
`
`
`CGCGCG) (68). In contrast, ethidium bromide and proflavine are
`
`inactive towards both mutants (68).
`Given the known activity of proflavine in T4 frameshift mutagenesis
`
`
`
`as compared to Salmonella,
`
`it is clear that the available repair systems
`
`
`
`
`are also important determinants of activity (27). Unfortunately, in no
`
`
`
`system are the intermediate steps involved in frameshift mutagenesis
`
`
`
`
`known in detail. Moreover, the nonlinear concentration dependence
`
`
`
`
`generally observed for mutagenesis implicates more than one bound
`
`mutagen per frameshift (27, 137).
`
`Damage to DNA In Vivo
`
`
`Several intercalating drugs have been shown to produce single strand
`
`breaks in the DNA of mammalian cells (48, 102, 103, 108), although the
`
`
`
`
`mechanism underlying this phenomenon is unknown. It has been estab
`
`
`lished that ethidium bromide can induce nicks photochemically in the
`
`
`can absence of enzymes (24, 67), but it is likely that drug intercalation
`
`
`
`
`also cause nicks by inducing nuclease activity (83) and by inhibition of
`
`
`
`repair processes subsequent to excision ( l08).
`
`SOLUTION STUDIES OF INTERCALATING
`
`DRUGS
`
`Most of our ideas about the various ways in which intercalating dyes
`
`
`
`
`
`
`can interact with nucleic acids derive from solution studies. These not
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 90
`
`
`
`NUCLEIC ACID-DRUG COMPLEXES 91
`
`only can discern extent and heterogeneity of binding but they also
`
`
`
`
`provide us with a means for comparing behavior of different drugs,
`
`
`
`
`
`different DNA sequences, and different solution environments. Further
`
`
`
`more, it is possible to characterize the effects of such binding on DNA
`
`
`structure and dynamics, and to a more limited extent to learn about
`
`
`
`details of the binding site stereochemistry. Solution techniques are
`
`
`
`
`
`therefore important in setting limits to the features that are likely to be
`
`
`important in determining in vivo activity.
`
`Binding of Intercalating Drugs to Nucleic A cids
`
`
`
`Any discussion of the consequences of ligand binding to DNA or RNA
`
`
`must begin with an understanding of the extent of such binding. Both
`
`
`
`direct and indirect analyses of binding (33) indicate at least two distinct
`
`
`
`
`kinds of binding process. These are designated here as binding processes
`
`
`
`I and II, following the description used by Blake & Peacocke (15).
`
`BINDING PROCESS I Binding Process I has been associated with inter
`
`
`
`
`calative binding by criteria to be discussed below. It is strong (K
`
`
`association = 104 _106 M -1) and characterized by a limit in the maxi
`
`mum number of sites available on the DNA. For drugs without side
`
`
`
`chains, such as ethidium bromide and the acridines, the limit in sites (as
`
`
`defined by Scatchard analysis) is one per 2-2.5 base pairs (126),
`
`
`
`whereas for actinomycin D with its polypeptide side chains, the limit is
`
`
`one per 6 base pairs of DNA ( 72). These values have given rise to the
`
`
`concept of neighboring site exclusion whereby a bound intercalator
`
`
`
`
`
`inhibits binding of further drug at adjacent site(s) (10, 23). Indeed, such
`
`
`behavior is predicted to not only reduce the total number of strong
`
`
`binding sites available but also to lead to an apparent anticooperativity
`
`
`
`
`
`in the binding. This becomes particularly evident as saturating levels of
`
`
`
`
`drug are bound, resulting in distinct curvature of the Scatchard plot (10,
`
`
`
`23, 69, 72). Detailed analysis of binding isotherms of the simple interca
`
`
`lator proflavine has suggested that some anticooperative ligand-ligand
`
`
`
`interaction may occur even at the second nearest neighbor sites (100).
`
`
`
`This is especially true for quinacrine, which has a cationic side chain in
`
`
`
`
`
`addition to its positively charged acridine ring (135), suggesting that the
`
`
`
`anticooperativity between bound ligands involves a large electrostatic
`component.
`The nature of neighboring site exclusion has been further explored
`
`
`
`in
`
`
`solution by studying the binding of drugs that have two potentially
`
`
`
`
`intercalating chromophores connected by chains of various lengths. It is
`
`
`
`
`found that the nearest neighbor site exclusion principle holds true for
`
`
`
`
`chromophores connected by spermidine or peptide chains, in that a
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 91
`
`
`
`92 BERMAN & YOUNG
`
`of 10.2 A is necessary minimum interchromophoric distance for bis
`
`
`
`
`
`
`intercalation (58, Ill, 120). However, in a series of diacridine deriva
`
`
`tives linked by a simple methylene chain, the shortest interchromo
`
`
`
`phoric distance that still allows intercalation of both acridines is only
`8.8 A, implying that neighbor exclusion
`
`
`is inoperative for these com
`
`
`
`pounds (118). The apparent contradiction here with the observed bind
`
`
`
`ing behavior of the monomeric chromophores has yet to be resolved.
`
`The topological state of the DNA can also affect the shape of the
`
`isotherm for binding process I. This is because of the "unwinding"
`
`
`
`capability of intercalating drugs (see below) that enables them to reverse
`
`
`
`
`the supercoiling of naturally occurring closed circular DNAs. Thus,
`
`
`addition of drug to such DNAs is at first more favorable than for
`
`nicked or linear DNA, since it helps relax the (negative) supercoils
`
`
`
`
`present. However, after the DNA is fully relaxed, the affinity decreases
`
`
`below that of nicked DNA because binding now results in the forma
`
`
`
`tion of (positive) supercoils. These changes in affinity are continuous,
`(10, 41).
`
`
`
`resulting in a smoothly curved Scatchard binding isotherm
`
`
`
`
`Thus a higher affinity is predicted for supercoiled, compared to relaxed,
`
`DNAs at low levels of drug. This may be important in the particular
`
`
`
`activity of the intercalating drugs toward small, closed circular DNAs
`(34).
`In order to develop a full understanding of the factors affecting DNA
`
`
`
`
`
`
`
`
`affinity, one must have some knowledge of the thermodynamic terms
`
`
`
`involved. A limited set of values have been determined from binding,
`
`
`
`kinetic, and microcalorimetric data. In general, these tend to agree on a
`
`
`negative enthalpy of between 5 and 8 kcal mol -\ for intercalation of
`
`
`
`singly charged cationic drugs, whether the charge is on the intercalated
`
`chromophore (proflavine and ethidium bromide) (19, 62, 92) or on a
`
`
`
`
`
`
`
`side chain (daunomycin) (91). A particular exception is actinomycin D,
`
`which has an enthalpy of between 0 and +2 kcal mol (35, 91). In this
`
`
`case, the binding process is marked by a positive entropy change (35,
`
`
`
`
`120) in contrast to the negative entropy changes for ethidium bromide
`(19, 62).
`and proflavine
`The electrostatic aspects of the binding have been treated according
`
`
`
`
`
`
`to counterion condensation theory (66, 94). The release of bound
`
`
`
`
`counterions (M +) on binding of a cationic ligand results in a depen
`
`dence of the observed binding constant (KOBS) on the counterion
`
`concentration
`810gKoBs
`---=::--= ===- = m' \f;
`<5log[M +]
`where m' is the number of ion pairs formed by the ligand with DNA
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 92
`
`
`
`NUCLEIC ACID-DRUG COMPLEXES
`
`93
`
`and � is the fraction of counterions associated with each phosphate.
`
`
`
`
`
`This latter value depends on the interanionic distance and is 0.88 for
`
`normal double-stranded DNA. As would be expected, ethidium bromide
`charge gives an m' value of 1 (40, 59). However,
`with its single positive
`
`
`other intercalators examined do not give integer values of m' (40, 135).
`
`
`The discrepancy arises partly from a reduction in � because of the
`
`
`
`alteration of the interphosphate distances at the site of intercalation
`
`
`
`(135). Hence, counterions are partially released upon intercalation, and
`
`
`
`this in itself will lead to an ionic strength binding dependence as
`
`
`
`evidenced for actinomycin D, which is uncharged (72).
`
`
`All the studies mentioned to this point have assumed a single intrinsic
`
`
`
`binding constant to DNA, independent of the sequence of nucleotide
`
`bases at each site. However, this is an oversimplification, since both
`
`
`
`
`compositional and sequence variations in binding constants have been
`
`
`
`observed. For example, from studies of the binding of planar tricylic
`
`drugs to DNAs of varying G-C content it has been possible to show a
`
`
`good correlation between the polarizability of the ring system and a
`for binding next to G· C base pairs (73). This has been
`preference
`
`
`
`rationalized in terms of the greater polarity of G· C base pairs over A· T
`
`
`base pairs (73). The preference can also be affected by side chains (71)
`
`and different ring systems (59).
`
`
`Varying the overall composition of the DNA may not discriminate
`
`
`
`between drugs that are sensitive to a precise sequence of bases at the
`
`
`
`intercalation site. That such differences can exist has been indicated by
`
`
`
`competitive binding studies of drugs to DNA. The competitive behavior
`
`
`
`
`of quinacrine and ethidium bromide (59) indicates similar preferences,
`
`
`
`and the noncompetitive interference with ethidium bromide binding by
`
`
`
`
`actinomycin D (59, 96) indicates different site preferences. A more
`
`
`
`
`systematic method for uncovering preferred binding sites is to study the
`
`
`binding of intercalating drugs to double-stranded synthetic polynucleo
`
`
`tides (7, 89, 120, 133) or to self-complementary or mixtures of comple
`
`
`mentary dinucleotides that can then form miniature intercalated
`
`
`
`duplexes. Using the latter approach, it has been shown that actinomycin
`
`
`
`D prefers the site GpC considerably more than its sequence isomer CpG
`
`
`(51, 52, 53) and that ethidium bromide shows a preference for pyrimi
`
`dine (3' - 5') purine sites over the isomeric purine (3' -5') pyrimidine
`
`
`site (55, 56, 96). More quantitative analysis of binding data for 4-
`
`
`
`nitroquinoline oxide (136) and 9-aminoacridine (137) has indicated only
`
`
`weak preferences in these cases. To relate the preferences seen at the
`
`
`
`dinucleotide level to those at the polymer level, and in particular to take
`
`into account the effect of adjacent sites, the binding of drugs to
`
`
`
`self-complementary oligonucleotides has been studied in a few cases (52,
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 93
`
`
`
`94 BERMAN
`
`& YOUNG
`
`85, 86, 139). These oligomers appear to display the site exclusion
`
`
`
`
`
`behavior of the polymers, but analysis of binding data in terms of
`
`
`
`
`sequence preferences is complicated by the presence of more than one
`
`
`
`intercalation site, each one of which may contribute differently to the
`
`
`
`
`
`
`experimental isotherm. Therefore, observed binding stoichiometries may
`
`
`
`
`(139). Further binding site(s) not necessarily represent the preferred
`
`studies will have to contend with this question.
`Apart from sequence specificity or preference, a further constraint
`
`
`
`
`
`
`
`may come from the chemical nature of the nucleic acid itself. This is
`
`
`
`most aptly illustrated by the relative binding of drugs to DNA and
`
`
`
`RNA. Some drugs, such as the acridines and ethidium bromide (59, 107,
`
`126, 127), bind to both DNA and RNA in a similar manner, whereas
`
`
`daunomycin, actinomycin D, and echinomycin do not bind to RNA at
`all (25, 72, 120).
`
`BINDING PROCESS II A second binding mode has been detected by both
`
`
`
`
`
`
`kinetic and equilibrium methods. Despite being weaker than intercala
`
`
`
`tion, this process may still have a biological role as evidenced by kinetic
`
`
`
`studies of proflavine binding to DNA, which indicate that even at low
`
`ratios of drug to DNA, where most of the drug is expected to be fully
`
`
`intercalated, as much as 7% of proflavine is found in an intermediate
`
`
`
`nonintercalative binding mode at physiological ionic strengths (62).
`
`
`
`
`Equilibrium binding studies, which detect secondary binding only after
`
`
`
`
`all intercalative sites are occupied, indicate a sensitivity to the salt
`
`
`
`
`concentration (15, 59, 126), which implies an electrostatic interaction of
`
`
`
`
`the drug with the negatively charged phosphate backbone. The extent to
`
`
`
`which this occurs appears to be related to the strength and salt sensitiv
`
`
`
`ity of the self-aggregation of the drug, suggesting that the negative DNA
`
`
`
`
`phosphate backbone acts as a template for aggregation of drug, particu
`
`
`
`larly after saturation of the intercalative sites (90). Analysis of the
`
`
`
`number of DNA sites available for this binding mode indicates dif
`
`
`ferences between drugs. For example, only half as many doubly charged
`
`
`
`quinacrine molecules can bind to the outside of DNA as can the singly
`
`
`
`charged ethidium bromide (31). Stereochemical differences between
`
`
`
`drugs may also be important as evidenced by acridine orange, which
`
`
`exhibits more outside binding to DNA than does eithidium bromide
`
`
`
`(18). UnfortunatelY,acquisition of quantitative binding information
`
`about this weaker mode is hampered by the presence of the stronger
`
`
`intercalative complex that can affect the spectroscopic data and its
`
`
`analysis (18). This problem has been circumvented in the case of
`
`
`
`2,7-di-t-butyl proflavine, which is too bulky to intercalate but can still
`
`
`
`bind externally (74). Two binding modes are observed: the predominant
`
`
`one requires approximately three base pairs per bound drug, of which
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 94
`
`
`
`NUCLEIC ACID-DRUG COMPLEXES 95
`
`two must be A· T, and a weaker mode with much less specificity. In
`
`
`
`
`contrast, unmodified proflavine has a preference for binding externally
`at G· C sequences (93).
`Information about the thermodynamics of outside binding so far
`
`
`
`comes from kinetic studies at low drug/DNA ratio (62, 107). For
`
`proflavine, the only drug examined in detail, the enthalpy (JiB) is -1.0
`
`kcal mol-I for poly A·poly U and -9.8 kcal mol-1 for calf thymus
`
`DNA, whereas the outside binding to the glycosylated T2 DNA has a
`
`in the entropy JiB of -3.4 kcal mol-I. Differences are also reflected
`
`terms, negative for calf thymus DNA and positive for T2 DNA and
`
`poly A· poly U. Hence this form of binding is very sensitive to the
`
`chemical characteristics of the nucleic acid (107).
`
`BINDING TO NONCOMPLEMENTARY NUCLEIC ACIDS Investigation of the
`
`DNA
`
`nature of the binding of intercalating drugs to double-stranded
`
`inevitably raises questions about the extent to which such drugs can
`
`bind to single strand or noncomplementary regions. Such binding may
`
`be of some relevance to the role of these drugs in stabilizing the
`
`
`loop-outs and mismatches believed to form a fundamental part of the
`
`
`mechanism of frameshift mutagenesis (116).
`The binding of ethidium bromide and proflavine to homopolymers
`
`
`
`(e.g. poly A, poly U) is cooperative in nature at saturating drug levels,
`
`
`suggesting a condensation of the drugs on the phosphate backbone so
`
`
`that they form continuous tracts of aggregated drugs (26, 127). More
`
`detailed analysis of spectrophotometric binding curves in the case of
`
`
`
`proflavine suggests a second kind of binding involving drug-nucleic acid
`
`
`base interaction (26). The anticooperative nature of this binding mode is
`
`
`
`reminiscent of intercalation into DNA and suggests some kind of partial
`
`
`intercalation of the drug between the bases. Supportive of this interpre
`tation is the increased strength of such binding in the presence of
`
`greater secondary structure in the polymers (26).
`
`
`More explicit evidence for partial intercalation comes from NMR
`studies of ethidium bromide with UpU and poly U (50) and 9-
`
`
`
`aminoacridine with several different deoxydinucleoside phosphates
`
`(137). In both cases there is evidence for a partial intercalative single
`
`
`strand structure with the positively charged ring nitrogens directed
`
`toward the intemucleotide phosphate.
`The ability of intercalating drugs to recognize and bind to noncom
`
`
`plementary base pairs in the Watson-Crick sense has also been docu
`mented. For example, it has been observed that the binding of ethidium
`
`bromide to the triple-stranded poly A·poly I is stronger than that to
`poly I· poly C (127), and indeed, when such A· I mismatches are
`
`for the drug introduced into poly I· poly C they have a higher affinity
`
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 95
`
`
`
`96 BERMAN & YOUNG
`
`than the surrounding sites (36). Ethidium bromide binding has also been
`
`
`
`
`observed for the poly d (G·1) helix (28). Similarly, poly U forms an
`
`
`
`ordered hairpin structure at low temperature involving
`U· U pairing
`
`
`
`
`(138), and this can intercalate proflavine (26). That mismatches may be
`
`plausible binding sites is also suggested by NMR studies of the binding
`
`
`
`dinucleotides or diof intercalating drugs to non-self-complementary
`of higher than I: I (di
`
`
`nucleoside phosphates, where stoichiometries
`(54, 137).
`nucleotide: drug) are often suspected
`
`
`bases has been noted (57). A further response to noncomplementary
`
`
`If these bases are surrounded on either side by normal Watson-Crick
`
`base pairs, then intercalation of the drug can force them to loop out.
`
`
`This was shown by NMR studies on the oligonucleotide pairs GpUpG
`
`+CpC and GpUpC+ GpUpC, where intercalation of ethidium bromide
`
`between the G·C base pairs forced the uracil residues to loop out (57).
`
`
`
`
`The ability of intercalating drugs to stabilize regions of non-W atson
`
`
`Crick complementarity may help explain the anomalously strong bind
`
`
`in magnitude ing of drugs to heat-denatured DNA, which is comparable
`
`
`to the binding to native DNA (15, 59).
`
`Effect of Intercalation on DNA
`
`A number of physical changes in the DNA have been associated with
`
`
`
`
`binding of intercalating drugs as described below. In all these measure
`
`
`
`ments, however, one is limited by knowledge of the extent of intercala
`tive binding versus other binding modes. This is probably least uncer
`
`
`
`
`at higher salt outside binding tain for ethidium bromide, which has little
`
`
`
`
`it is often concentrations and low ratios of drug/DNA; consequently,
`
`used as the standard to which other intercalators are compared. Also,
`the values obtained are an average for all DNA sites and hence do not
`
`detect any sequence variations that might exist.
`
`INCREASE IN DNA LENGTH As a drug intercalates, it causes an extension
`
`
`
`This base pairs. of DNA, owing to its insertion between two adjacent
`
`
`
`was first shown from autoradiographic results with proflavine and T2
`
`
`by studies DNA (21) and has been subsequently confirmed in solution
`
`
`
`
`
`of the viscosity, sedimentation, and electric dichroism behavior of
`
`DNA on addition of the drug (22, 37). The theoretical length
`sonicated
`for B-DNA is 3.4 A per intercalated
`extension
`drug, and this has been
`
`
`
`
`approximately observed for proflavine (22) and 9-aminoacridine deriva
`tives (118) in moderate salt conditions,
`which reduce the extent of
`
`
`
`
`
`
`suggest a measurements electric dichroism outside binding. In contrast,
`
`
`range of possible helix extensions (2.0-3.7 A) per intercalated
`
`drug,
`depending on the drug tested (37); however, the low ionic strength
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 96
`
`
`
`NUCLEIC ACID-DRUG COMPLEXES 97
`
`medium used for these measurements raises questions about the extent
`
`
`
`
`of outside binding and the behavior of DNA under such conditions.
`
`Assaying for the length increase is therefore a good way to check
`
`
`
`whether both chromophores of a potential bis-intercalating drug do
`
`indeed intercalate, since if they do, the increase will be double that for
`the monomer (58, 118).
`
`OF CLOSED CIRCULAR DNA The finding that intercalating
`UNWINDING
`(9),
`
`drugs can unwind closed circular DNA is based upon sedimentation
`
`viscosity (97, 113, 132), electron microscopic (64), and gel electro
`
`
`
`phoretic (49) observations. Considerable attention has focused on the
`
`precise experimental determination of an "unwinding angle" for
`ethidium bromide: this is the degree of unwinding of the DNA duplex
`as each drug molecule is bound, and it is now confirmed by several
`
`experiments to be 260 (49, 64, 124). The unwinding angles for other
`
`
`potential intercalating drugs have been determined relative to this
`standard (e.g. 47, 119, 128, 130).
`As with any solution measurement, it is desirable to know how much
`
`
`
`
`of the drug is truly intercalated. Since nonintercalative binding modes
`
`are more likely to be salt sensitive, observation of a salt dependence of
`the unwinding angle suggests the presence of such binding modes at
`
`
`
`
`lower salt concentrations; e.g. chloroquine (47). Differences in limiting
`(high salt) values of the unwinding angle more probably reflect funda
`
`
`
`mental differences in intercalation stereochemistry (47). For example,
`
`removal or chemical modification of the amino groups of ethidium
`
`bromide sharply reduces its unwinding angle, suggesting a role for these
`
`
`groups in orienting the intercalated ethidium cation (119).
`The ability of DNA ligands to unwind closed circular DNA mole
`
`cules can be considered as strong evidence for some form of intercala
`
`tive binding (128) and consequently is a good way to assay for intercala
`
`
`
`tion of both chromophores of a potential bisintercalator (58, 118). Other
`DNA binding drugs without the planar chromophores requited
`'
`for
`
`
`intercalation do not cause significant unwinding, with the exception
`of
`
`
`steroidal diamines which may have some partial intercalative ability
`(128, 131).
`
`INCREASE IN DNA STABILITY Intercalation usually leads to an increase
`
`
`
`
`in the Tm(midpoint of the thermal transition profile) of DNA, represent
`
`
`
`
`ing a stabilization of duplex structure relative to the single strands. This
`
`has been documented not only at the polynucleotide level (84, 85, 129)
`
`
`
`but also at the synthetic oligonucleotide (85, 86) and dinucleotide (87,
`137) levels.
`
`Annu. Rev. Biophys. Bioeng. 1981.10:87-114. Downloaded from www.annualreviews.org
`
` Access provided by 173.227.19.10 on 08/13/18. For personal use only.
`
`Page 97
`
`
`
`98 BERMAN & YOUNG
`
`It is clear that sequence differences exist in the relative thermal
`
`
`
`
`stability supplied by the intercalating drug. For example, ethidium
`
`
`bromide stabilizes poly A· poly U significantly more than poly I· poly C
`
`
`(129) and 9�aminoacridine stabilizes dGpG + dCpC more than
`
`dApG·dCpT (137), whereas daunomycin is found to stabilize self�
`
`
`
`
`complementary alternating (purine�pyrimidine)n polynucleotides more
`
`than complementary nonalternating (purine-purine)n . (pyrimidine
`
`
`pyrimidine)n ones (89). Also, the nature of the drug can determine the
`
`
`extent of duplex stabilization as shown by the greater stabilization of
`
`poly (dAdT) by ethidium bromide compared to proflavine (84, 85).
`
`
`
`DNA stabilization is manifested at the site of intercalation by a
`motional coupling of the drug and its adjacent base pairs (101) such that
`
`the relative internal motions of these base pairs is slowed (38). Ethidium
`and propidium are much more strongly coupled than the acridines
`
`
`
`(101), consistent with their relative effectiveness at duplex stabilization
`(84, 85).
`
`Intercalation Site Stereochemistry
`
`
`
`In contrast to the information available on gross changes in DNA
`
`
`
`
`structure, solution techniques are more limited in their ability to discern
`
`
`the conformational features at the intercalative site. These are im
`
`
`
`portant, however, for confirming in solution some of the general char
`
`
`
`acteristics learned from crystal structures and extending their range to
`
`other drugs and sequences not yet examined in such detail.
`
`
`
`Some general information about drug orientation relative to the DNA
`
`
`helical axis is available from flow dichroism and electric dichroism
`
`
`studies. The former established that the drugs are aligned roughly
`
`perpendicular to the DNA helical axis, thus lending early support for
`
`
`
`
`
`the intercalation hypothesis (60). Electric dichroism measurements sug
`
`gest, however, that the drugs are not exactly perpendicular but slightly
`
`
`t