You are on page 1of 44

ANNUAL

REVIEWS Further
Annu. Rev. Phys. Chern. 1995. 46: 657-700 Quick links to online content
Copyright © 1995 by Annual Reviews Inc. All rights reserved

SALT-NUCLEIC ACID
INTERACTIONS

Charles F. Anderson' and M. Thomas Record, Jr. ',2


Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

Departments of 'Chemistry and 2Biochemistry, University of Wisconsin­


Madison, Madison, Wisconsin 53706
by University of Windsor on 07/16/13. For personal use only.

KEY WORDS: DNA, oligonucleotides, coulombic interactions, salt poly­


electrolyte effects, end effects

ABSTRACT

Coulombic interactions of salt ions with polymeric and oligomeric nucleic


acids in solution have large and distinctive effects on ion distributions, on
thermodynamic coefficients, and hence on equilibrium processes involving
nucleic acids, such as their conformational transitions and binding inter­
actions. In experimental or theoretical studies where an oligonucleotide
is taken to represent the corresponding polynucleotide, the impact of
coulombic end effects on molecular and thermodynamic properties must
be taken into account. Observable consequences of coulombic interactions
in nucleic acid solutions have been calculated by using models with varying
degrees of detail and methods formulated at varying levels of rigor. From
comparisons of experimental results with predictions of the prevalent
theoretical approaches, this review concludes that the more rigorous
methods have proved capable of accounting for thermodynamic (and some
molecular) consequences of coulombic interactions with a minimal number
of preaveraged parameters that represent the most important structural
features of the nucleic acid solution.

INTRODUCTION
Perspectives
The topic of salt-nucleic acid interactions may appear too specialized to
merit coverage in this series of Annual Reviews. However, nucleic acids
in solution are particularly favorable model systems for experimental (1,
657
0066-426X/95/1101-0657$05.00
658 ANDERSON & RECORD

2) and theoretical (3-5) studies of polyelectrolyte and oligo electrolyte


behavior, while RNA and DNA are the most prevalent and important
highly charged biopolyelectrolytes, which function in and also define the
concentrated polyelectrolyte environment of a living cell (6). The physical
properties of aqueous solutions of polymeric nucleic acids have been
extensively studied because of their intrinsic biological significance and
because of their favorable characteristics from the standpoint of applying
polyelectrolyte theories. This review examines experimental and theor­
etical progress that has been made toward understanding the physical
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

basis for effects due to salt ion interactions with polymeric and oligomeric
nucleic acids. These effects differ dramatically from those due to inter­
actions of salt ions with other salt ions, with mononucleotides, with most
proteins, or with uncharged polymers.
by University of Windsor on 07/16/13. For personal use only.

Stabilities of nucleic acid helices and of complexes of nucleic acids with


charged ligands, including proteins, are profoundly affected by changes in
salt concentration (cf 1-5, 7-1 0). The effects of salt ions on these processes
exhibit a different functional form and at low salt concentration are far
more pronounced than would be expected from (improper) application of
theory derived for nonpolymeric ionic solutions. Virtually any equilibrium
extent of helix formation or ligand-binding of nucleic acids can be achieved
simply by changing the salt concentration. Because ion concentrations in
living cells can change markedly in response to changes in growth
conditions, these effects may be of substantial biological relevance (6). In
dilute macromolecular solutions in vitro, the readily detectable effects of
salt concentration on processes involving nucleic acids can be analyzed
quantitatively by applying rigorous methods based on classical statistical
thermodynamics to suitably detailed models of the nucleic acid and other
solute and solvent components in the solution (11).
The combination of a large number, 121, of like charges (phosphate)
and a high average axial charge density is the ultimate origin of the
distinctive effects on molecular and thermodynamic properties that arise
from the interactions of a polymeric nucleic acid with salt ions. Regardless
of the detailed structural features of a rodlike polyion, its axial charge
density can be characterized analytically by the dimensionless parameter:

1.

Here b is the average axial distance between charges projected onto the
axis of the polyion, e is the electron charge, B the dielectric constant of
pure solvent, and kT has the usual meaning. (In water at 25°C, � � =

7. 14 b-\ if b is expressed in angstroms.) The role of � as the only structural


parameter needed to characterize the molecular, thermodynamic, and
SALT-NUCLEIC ACID INTERACTIONS 659
transport properties of solutions containing cylindrical polyions was
widely promulgated as the cornerstone of counterion condensation (cq
theory ( 12, 1 3). Contemporaneous studies based on the cylindrical Pois­
son-Boltzmann (PB) equation clearly recognized the significance of � as a
mathematical determinant of alternative functional forms of the PB poten­
tial and of some thermodynamic properties calculated therefrom ( 1 4, 15).
Various theories predict, and experiments have generally confirmed, the
following distinctive characteristics of solutions of rodlike polyions having
IZI » I and� � I , in which the concentration of excess salt is not too high
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

([NaCl] ;$ 0.5 M, for example).


I . The high local concentration of counterions close to the polyion surface
is a strong function of �, but it is relatively insensitive to changes in
by University of Windsor on 07/16/13. For personal use only.

salt concentration, and it persists with increasing dilution, in marked


contrast to the situation in ordinary salt solutions (4).
2. The contribution to the thermodynamic activity of the polyion due
to its interactions with salt ions has an approximately logarithmic
dependence on salt concentration and also a significant dependence on
�. Therefore, any process that alters � locally or globally (such as a
conformational change or ligand binding) can be driven in either direc­
tion by relatively small changes in the concentration of excess added
saIt (7, 1 6-1 8).
Neither small oligoions nor proteins, which generally are poly­
ampholytes, have a large number of like charges whose density is high in
any one dimension. Therefore, the typical molecular and thermodynamic
effects attributable to ion-ion or to protein-protein interactions are neither
so large in magnitude nor so persistent with reductions in salt con­
centration as those characteristic of solutions containing a polymeric
nucleic acid (7, 1 9).
The onset of the distinctive molecular and thermodynamic charac­
teristics of polynucleotides is exhibited by a homologous series of oligo­
nucleotides, all having the same�, so their lengths are proportional to I ZI.
Along the surface of any charged rodlike molecule of finite length, the
magnitude of the electric field diminishes steadily in the axial direction as
either end is approached. If I Z I is small enough, the nonuniform axial
profile of the electric field at the surface of the oligomer produces observ­
able coulombic end effects on, for example, the average number of counter­
ions close to an oligomer and on its thermodynamic activity. The sig­
nificant magnitude of these effects and their distinctive dependence on 1 Z I
has been deduced from comparative (oligomer vs polymer) experimental
studies of ion distributions (2, 20), conformational transitions (2 1 ), and
oligocation binding (W Zhang, unpublished data). Such information pro-
660 ANDERSON & RECORD

vides both a new dimension for testing the accuracy of theoretical cal­
culations (23-25) and a practical basis for determining how conclusions
drawn from in vitro studies of oligomeric nucleic acids in solution can be
applied to polymeric forms in vivo. The steadily growing use of oligo­
nucleotides as experimental and/or theoretical models for polynucleotides
(and the use of oligocations as model ligands that bind to oligomeric or
polymeric nucleic acids) motivates the need for comprehensive theoretical
and experimental studies of effects arising from the interactions of salt
ions with rodlike oligomers.
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

Scope
by University of Windsor on 07/16/13. For personal use only.

Most of the results reviewed here were obtained for nucleic acid solutions
containing salt ions of relatively low charge and small size, under con­
ditions such that their interactions with nucleic acid phosphates can be
described by classical electrostatics. Because these charges may interact
with dipolar or higher distributions that have no net charge, the more
explicit term coulombic is used in this article to refer to pair-wise inter­
actions between discrete (sometimes modeled as partial) charges in nucleic
acid solutions. (If the dielectric constant is not modeled as spatially
uniform, the interaction potential between two charges does not vary with
the reciprocal of their separation, and in this sense is not coulombic.)
Coulombic interactions have significant effects on a wide variety of proper­
ties, ranging from molecular, such as the radial distributions of salt ions
surrounding a nucleic acid, to macroscopic, including the equilibrium
thermodynamic and transport coefficients that characterize the entire solu­
tion. The equilibria and kinetics of processes involving nucleic acids, such
as their conformational transitions and binding interactions with proteins
and other ligands, also are strongly affected by their interactions with salt
ions.
Space does not permit adequate discussion of all recent research that is
connected in some way with salt-nucleic acid interactions but differs from
the work reviewed in this article in one or more of the following respects:
problems addressed, variables calculated, statistical mechanical approxi­
mations, and/or model assumptions incorporated. On the purely theor­
etical front, a number of reviews have been focused on applications of
different methods to the same model and/or of the same method to different
models (3, 4). In this article, which must be selective because of its length,
we emphasize those theoretical studies where actual contact with observ­
able properties has been made or where the results pertain directly to the
objective of analyzing and interpreting data.
SALT-NUCLEIC ACID INTERACTIONS 661

COULOMBIC INTERACTIONS IN NUCLEIC ACID


SOLUTIONS
Effects on Ion Distributions
In the immediate vicinity of a highly charged rodlike polyanion, such as
polymeric DNA, coulombic interactions produce a high local con­
centration of mobile counterions (e.g. Na+) and concomitantly a very low
local concentration of coions (e.g. Cl-). These pronounced non­
uniformities in the ion distributions persist even at high dilutions of salt
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

and polyion and (for the counterion) even in the total absence of added salt.
To describe the substantial enhancement in the local density of counterions
surrounding any type of highly charged polyion, the term condensation
has gained wide currency, even when no reference to the predictions of
by University of Windsor on 07/16/13. For personal use only.

CC theory is (or should be) intended. Counterion accumulation and coion


exclusion are preferable, model-independent terms to describe the pro­
found effects on the distributions of mobile ions that are due primarily to
their coulombic interactions with highly charged rodlike polyions.
The most complete and generally useful description of counterion
accumulation and coion exclusion due to any type(s) of interactions
between an ion (i) and a nucleic acid (NA) can be given in terms of the
radial distribution function gNA;- Knowledge of gNAi suffices, at least in
principle, to calculate the contribution of salt ion-nucleic acid interactions
to any thermodynamic property of the system. Some other characteristics
of ions in nucleic acid solutions, such as their diffusion coefficients and
their NMR relaxation rates, also can be expressed theoretically in terms
of the gNAi' Theoretical relationships between gNAi and measurable proper­
ties of nucleic acid solutions have been discussed in detail (4). For univalent
salt ions and highly charged cylindrical polyions, the local features ofgNA+
and gNA- are determined in large part by � and are relatively insensitive
to changes in C3 , the concentration of excess added salt. If the mass action
formalism is used to describe coulombic accumulation of salt cations near
a nucleic acid, the resulting equilibrium constant (as conventionally defined
in terms of bulk concentrations) generally has a pronounced dependence
on the concentration of that cation (26).
An experimental determination of the compete functional form of gNAi
(as a Fourier transform) in principle can be extracted from small angle
scattering measurements, but in nucleic acid solutions this approach has
proved to be feasible so far only for the heavy cations TI+ and BaH (27,
28). Information about gNA; at or near the surface of a nucleic acid can be
obtained by monitoring some spectroscopic property of the salt ion or of
some moiety-either natural or synthetically introduced-on the nucleic
acid itself (29-3 1 ) . The NMR relaxation rate of a quadrupolar monatomic
662 ANDERSON & RECORD

ion such as Na + or K + is enhanced substantially when the ion is close


enough to the nucleic acid surface, and quadrupolar nuclei in more com­
plicated ionic species, such as 59CO in the trivalent hexammine ion, also
have proved useful as probes of interactions with DNA (32, 33).
Despite some definite limitations, both on experimental feasibility and
on the accessibility of definitive quantitative interpretations, NMR spec­
troscopy is at present the most widely applied (and the most generally
applicable) of the various methods that have been utilized to probe the
interactions of salt ions with nucleic acids. As a means of investigating
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

equilibrium and dynamic features of these interactions, the advantages


and drawbacks of NMR spectroscopy of quadrupolar cations have been
discussed in an Annual Review addressing work prior to 1 990 (4). An
by University of Windsor on 07/16/13. For personal use only.

especially comprehensive article (2) has assessed the current body of infor­
mation about the character of the association of small ions with nucleic
acids, as indicated by NMR measurements. In summary, univalent cations
exhibit no sign of localization or dehydration due to interactions with
double-stranded (ds) nucleic acids (oligo- or polyanionic), although sub­
stantial effects indicative of site binding have been inferred from NMR
measurements probing the interactions of sodium ions with quadruplex
DNA (34). Some multiply charged ions, such as Mg2+, affect NMR
measurements in a way that suggests a modest noncoulombic affinity for
ds DNA (35), but the accumulations of others, such as divalent hexa­
methonium (36), appear due only to coulombic accumulation.

Effects on Thermodynamic Coefficients


Even though dielectric shielding by water and electrostatic screening by
salt ions reduce the effective range of coulombic interactions in nucleic
acid solutions, effects of these interactions on gNA+ and gNA- extend
radially for a considerable distance from the surface of the nucleic acid
( 1 0- 1 00 A depending on salt concentration and valence) (4). Generally
'"

difficult to detect directly [except, in favorable cases, by scattering measure­


ments (27, 28)] the long range effects of coulombic interactions on the 9NAi
have significant impacts on various measurable thermodynamic
coefficients that can be represented as appropriately weighted integrals
over the gNAi" For example, the contributions of ion-nucleic acid inter­
actions to ')' ±, the mean ionic activity coefficient of the excess salt in a
nucleic acid solution, could in principle be evaluated by subjecting the 9NAi
to the operations (integrations, differentiations) prescribed by statistical
thermodynamics. However, in practice some theoretical approaches (such
as those based on the PB cell model or grand canonical Monte Carlo (MC)
SALT-NUCLEIC ACID INTERACTIONS 663
simulations) provide much more direct and hence more accurate paths to
the evaluation of coefficients such as y± (37, 38) or r3,2 (37, 39).
The preferential interaction coefficient r3,2 provides a model-inde­
pendent thermodynamic characterization of the interactions of a solute or
cosolvent (component 3) with a relatively larger solute (component 2)
present at sufficiently high dilution in the solvent (component 1). The
excess solute 3 may be a nonelectrolyte or an electrolyte; the dilute solute
2 may be oligomeric or polymeric, charged or uncharged. According to
the usual definition (40),
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

2.

where C3 is the molar concentration of component 3, present at chemical


potential 11-3; C2 is the molar concentration of component 2; and the
by University of Windsor on 07/16/13. For personal use only.

solvent chemical potential is }11' In favorable cases these coefficients can


be measured by means of dialysis equilibrium experiments (lS). When C2
is low enough, r3,2 is equal to the experimental dialysis coefficient, defined
as

. o . C3-C�
hm r 3,2
cr>o
= r 3,2 = hm
Cr"O
' 3.
C2

where C; is the solute molarity in the dialysis solution.


For an uncharged solute, r3,2 can be interpreted directly as thermo­
dynamic binding (or association) of component 3 to component 2 (41 , 42).
If the only type of interac tion between components 3 and 2 is site binding
that produces an average density of V3 solutes per polymer, then r3,2 v3• =

More generally r3,2 can be either positive or negative, indicating, respec­


tively, net accumulation or exclusion of uncharged solute (over solvent)
in the vicinity of the polymer relative to the bulk solution (43, 44).
When solute 3 is a salt, the individual (but not independent) preferential
interactions of each ion with the polymeric species (charged or uncharged)
of component 2 may be characterized by single-ion preferential interaction
coefficients (45). Thus the accumulation (exclusion) of counterions (coions)
reflected in gNA+ (gNA-) at the molecular level is correlated with macro­
scopic effects that can be interpreted, in the context of a Donnan equi­
librium, as thermodynamic accumulation (exclusion) of individual ions.
For the particular case of a solution consisting of a nucleic acid oligo- or
polyanion with univalent cations and anions, the coion (anion) preferential
interaction coefficient, r -,2' is specified by r3,2, and the cation preferential
interaction coefficient, r + ,2, is specified by /Z2/+r-,2 /Z2/+r3,2 (46).
=

The net preferential interaction of the low molecular weight cation and
anion with the nucleic acid polyanion is
664 ANDERSON & RECORD

f +,2+f -,2 = IZ21+2f 3,2' 4.

Equation 4 defines the net thermodynamic extent of ion association with


the nucleic acid polyanion, resulting from counterion accumulation (f +,2)
and coion exclusion (f-,2)'
For cylindrical polyions, the dependence of r3,2 on the polyion charge
Z2 can be eliminated by defining f3," =: rdlz21, which is independent of
Z2 but remains a strong function of the axial charge density. (For oligoions,
r 3," exhibits a significant dependence on 22, owing to the end effects
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

considered in the final sections of this review. There the symbol f121,u is
used to describe the preferential interactions of oligoions with a single
type of excess solute, for which the subscript is understood.) The net
thermodynamic extent of univalent ion association per structural charge
by University of Windsor on 07/16/13. For personal use only.

on a poly- or oligoion, conventionally designated by the symboll-i (47),


can be expressed as

I -t' = f +,u+f -,u = I +2f -,U = I +2f3,u' 5.

For polyions like nucleic acids in solutions where C3 � 0.01 M, all theories
predict extensive counterion accumulation (r +,u � 0.8) and minimal coion
exclusion ( - f ;S 0.2), so 1- i is dominated by the thermodynamic
- U
,

accumulation of counterions (45). Because of this substantial net accumu­


lation of counterions, nucleic acids act like weak electrolytes in various
processes of biological significance.
A more explicit representation of the coefficients r +,2 and r -,2, can be
obtained on the basis of a two-domain (local region, bulk region) model
(46):

6.

Here, B+,2, B_,2, and B1,2 are, respectively, moles of univalent cations,
anions, and water accumulated in the local region per mole of the polyion
species of component 2; m3 and ml are the molalities of electrolyte and
solvent. This model also yields an interpretation of the effects of individual
salt ions and of water on a broad class of equilibria, reviewed below.
By appropriate thermodynamic transformations, f3,2 can be related to
a derivative that expresses the interdependence of the thermodynamic
activities of solute components 2 and 3 (40, 43). Consequently, effects of
the concentrations of ligands, salts, or other solutes on any equilibrium
(process) involving any type of sufficiently dilute species can be analyzed
directly in terms of preferential interaction coefficients, on the basis of the
development (11) reviewed in the next section.
SALT-NUCLEIC ACID INTERACTIONS 665

GENERALIZED THERMODYNAMIC ANALYSIS OF


EFFECTS OF SALT ACTIVITY ON EQUILIBRIA
Formulation
Processes involving nucleic acids (e.g. conformational transitions, self­
association, binding of small or large ligands) typically involve large changes
in the amount of charged, polar, and nonpolar surface exposed to the
solution, as well as changes in the interactions of functional groups (for
example, groups interacting with water in the initial state interact with one
another in the final state). The thermodynamic effects of these changes on
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

a given process can be characterized experimentally by determining Kobs>


the stoichiometric equilibrium quotient of concentrations of products and
reactants (1). In a sufficiently dilute macromolecular solution, Kobs is inde­
by University of Windsor on 07/16/13. For personal use only.

pendent of macromolecular concentration(s) but is almost invariably a


strong function of salt concentration and depends as well on temperature,
pressure, and in some cases other solution variables (e.g. pH). Cor­
responding to Kobs is �G�bs == RTln Kob" a "standard" free energy
-

change for the process that depends not only on T and P (pressure) but
also on the solution variables that affect Kobs. Under typical experimental
conditions, �G�bs can be interpreted as the free energy change for con­
version of reactants to products when each is at unit concentration in its
respective (ideal dilute macromolecular solution) standard state where it
interacts only with solvent 1 and solute 3. In contrast to �G�b" the cor­
responding conventional standard free energy change �Go for the process
in an ideal dilute solution (defined to be a function only of T and P)
generally cannot be determined experimentally for processes involving
nucleic acids.
Whereas �G�bs can be evaluated with reasonable accuracy by various
experimental methods (1, 9, 48), it has not yet been calculated accurately
a priori by any theoretical approach, presumably because of the complexity
and variety of the type of interactions that may be operative in addition
to coulombic effects. Contributions from some types of noncoulombic
interactions to nucleic acid processes can be estimated from empirical
correlations involving changes in water-accessible surface areas (49). Vari­
ous kinds of interactions, which in general may involve solvent as well as
solutes, determine the magnitude of �G�bs for a process involving a nucleic
acid. However, the dependence of �G�bs on the concentration C3 (or mean
ionic activity, at) of excess salt can, under typical experimental conditions,
be attributed entirely to the interactions of salt ions with the participants
in the equilibrium. Specifically, under the conditions of interest here (typi­
cal of in vitro experiments), all of the reactants and products are dilute in
comparison to C3, and no salt ions are detectably bound to any of them.
666 ANDERSON & RECORD

Even though the salt ions (generally) are not stoichiometric participants
in the equilibrium of interest, variations in Q± typically have a large effect
on the activity coefficients YJ of the reactants and products and hence on
Kobs and AG�bs' At a given T and P, Kobs is inversely proportional to K)"
the corresponding stoichiometric quotient of the YJ. (For equilibria in
which all of the participating species are charged, YJ is a single ion activity
coefficient, but the mean ionic activity coefficient for the corresponding
electroneutraI component, 2J, can be constructed simply by introducing
the activity coefficient for the oppositely charged salt ion, raised to the
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

appropriate power.) When the concentrations of all species J are


sufficiently dilute, each of the In YJ depends only on a±, T, and P. In
general, the dependence of In YJ on Q± arises from coulombic interactions
that predominate in dilute solutions of nucleic acids, as long as the salt
by University of Windsor on 07/16/13. For personal use only.

concentration is not so high that certain effects due to short-range inter­


actions become significant.
Under the conditions specified above, the experimentally accessible
derivative SaKobs == (0 In Kobs/OIn a± ) r,p - (81nKy/ olna± ) r,p can be
=

transformed by relating each of its constituent terms of the type


(a In YJla In Q± h,p to the corresponding preferential interaction coefficient
(as defined in Equation 2). A detailed thermodynamic analysis (11) yields
the following rigorous expression for the effects on Kobs of changes in a ±:

SaKobs == (a In Kobsla In Q± h,p


7.

Here, IZJI is the (net) charge on J. (In general the numerical factor mul­
tiplying [3,2 is the same as the exponent relating Q± to a3' ) If XJ represents
any derivative of /1" then the stoichiometric combination
.1(XJ) == �J VJX" where vJ is the signed stoichiometric coefficient of J,
either reactant ( - ) or product ( +). Each of the r 3,2J in Equation 7 can
be evaluated on the basis of any of the theoretical methods and models
outlined in the following section. The detailed procedures whereby these
coefficients can be calculated for electrolyte-poly (or oligo) electrolyte
interactions by various theoretical methods have been explained (25, 37,
50-52) and critically examined (11).
Under the broad range of conditions where Equation 7 is applicable, it
offers the most direct and hence the most numerically accurate route to
the theoretical evaluation of SaKobs' The available alternative approaches
first require evaluation of an excess electrostatic free energy function,
effectively equivalent to In YJ, for each of the J by numerical integrations
either over a charging parameter (51) or Over spatial coordinates (52-54)
in cases where the PB apprOXimation is assumed. Then, to proceed to
SALT-NUCLEIC ACID INTERACTIONS 667
the evaluation of SaKob" each of the In Y1 must be differentiated again
numerically with respect to a±. If the object is to make contact with data,
this differentiation cannot be avoided, because neither the PB equation nor
any other theoretical approach that takes into account only interactions of
salt ions with each participant in the equilibrium of interest can yield values
for Kobs that could be expected to be comparable with those determined
experimentally.

Applicability and Interpretations


TYPES OF REACTIONS The published derivation of Equation 7 specifically
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

addresses the binding of an oligocation to a nucleic acid polyanion (11).


In fact Equation 7 is of sufficient generality to describe effects due to
changes in the activity of an excess solute of any type (nonelectrolyte or
by University of Windsor on 07/16/13. For personal use only.

electrolyte) on any equilibrium involving species of all possible charge


types, provided that these species all are sufficiently dilute so that Y 1
depends only on the activity of the excess solute. For example, the dena­
turation of proteins by the ionic solute guanidinium chloride and a wide
variety of other salt effects on processes involving biomolecules, which
have been extensively reviewed elsewhere (e.g. 7, 55, 56), can be analyzed
on the basis of Equation 7. The derivation that led to Equation 7 can be
generalized further, for example, to describe the effects of more than one
type of excess solute.
A less general form of Equation 7, which has been known for some time
(41 ), is applicable to systems where all participants in the equilibrium of
interest, as well as the solute 3, are nonelectrolytes. In this case, 1211 0
=

for all J and SaKobs = d(f3•21). For helix-coil transitions, self-aggregation,


or any other type of reaction in which reactants and products consist
exclusively of species having the same charge sign, 12jl "# 0, but ,11 2jl 0,
=

as in the case where all participants are nonelectrolytes. Hence,


SaKobs = d(2f3,21), if component 3 is a uniunivalent salt. However,
dl211 "# 0 for the binding of a charged ligand to a nucleic acid or for any
other type of eqUilibrium where oppositely charged species appear together
as reactants and/or products. The origin of the term ,112,1 in the general
thermodynamic derivation of Equation 7 was clarified recently (11), and
its existence has not always been recognized. The term is correctly included
in an earlier expression for SaKobs that was derived on the basis of some
more specialized model assumptions (reviewed below) (18).
Conformational changes of nucleic acids usually are characterized by
some thermodynamic variable such as Tm, the temperature at the transition
midpoint, which can be determined by calorimetry or by spectroscopic
methods. If the transition is modeled as occurring between two con­
formational states (native and denatured forms of a cooperatively dena-
668 ANDERSON & RECORD

turing unit), then from the Gibbs-Helmholtz equation and the derivation
leading to Equation 7, the effect of salt activity on the conformational
transition can be expressed by either of the following derivatives:

or

8.
Here ilH�bs is the enthalpy change due to the transition whose midpoint
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

temperature Tm is determined by a± . For polymeric nucleic acids, the ratio


ilr 3.211ilH�bs is independent of chain length. Because AH�bs typically
increases with increasing Tm for nucleic acid transitions, both Tm and
T;;; 1 are typically approximately linear functions of ln a±, and the slope of
by University of Windsor on 07/16/13. For personal use only.

a plot of Tm vs ln a± is to a good approximation proportional to Ar3,2J


(57).
SINGLE ION PREFERENTIAL INTERACTIONS AND THE TWO DOMAIN MODEL
Equation 7 relates the experimental quantity SaKobs to the net change in
the thermodynamic extent of ion association (Equation 4) resulting from
conversion of reactant species to products. For equilibria in which the
excess solute 3 is a dissociated simple salt (strong electrolyte), the effects
of the individual ions on KObs can be expressed in terms of single ion
preferential interaction coefficients (45). Thus, by introducing Equation 4
into Equation 7,
9.
Each of the single ion preferential interaction coefficients can be rep­
resented in terms of the two-domain model (46) by combining Equation 6
with Equation 9:
1 0.
This expression may be especially useful for the interpretation of effects due
to noncoulombic preferential interactions, such as those that determine the
Hofmeister series (7). The steps leading to Equation 1 0 demonstrate the
more general thermodynamic basis for an earlier derivation of the same
expression (7), based on generalized binding polynomials.
POLYELECTROLYTE EFFECTS ON SALT DEPENDENCES OF Ko", AND Tm Significant
but often unappreciated differences exist in both the functional forms and
the magnitudes, especially at low solute concentrations, of the effects of
nonelectrolyte and electrolyte solutes on processes involving uncharged vs
charged polymers. If all solutes are uncharged and a solute 3 is locally
excluded from or accumulated near, but not site-bound to, at least one of
SALT-NUCLEIC ACID INTERACTIONS 669
the participants in an equilibrium, then typically .1r3,2 ex:. C3, and In K"bs is
approximately linear in C3 • If solute 3 is a 1 : 1 electrolyte and the equi­
librium involves small ionic species rather than highly charged oligo- or
polyelectrolytes, then Debye-Huckel theory predicts that In Kobs is linear
in q/2 (for C3 � 10- 3 mol dm-3). Thus, the effect of C3 on Kobs vanishes
at low C3, both for nonelectrolytes and for simple electrolytes. In contrast
to each of these situations, effects of accumulation (exclusion) of electrolyte
ions on conformational and ligand binding equilibria of highly charged
cylindrical polyions such as DNA and RNA exhibit typically large and
approximately constant power law dependences of Kobs on C3 that persist
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

even at low (excess) salt concentrations. Interpreted in the context of


Equation 7, these power law dependences indicate that SaKobs and therefore
.1r 3,2J are relatively large in magnitude and relatively independent of salt
by University of Windsor on 07/16/13. For personal use only.

concentration, even at extreme dilution.


Characteristic differences also are observed in the effects of salt activity
on midpoint temperatures Tm of conformational transitions of globular
proteins and helical nucleic acids. The effect of salt concentration on order­
disorder transitions (denaturation) of nucleic acids persists even at very
low salt concentration (16); Tm typically is observed to vary linearly with
the logarithm of the salt concentration (57). Because both dTm/dln a± and
T�/.1Hobs are found experimentally to be independent of salt concen­
tration, so also is .1r 3,2J for their different conformational states, according
to Equation 8 (58). By contrast, the [salt] dependence of Tm for the dena­
turation of globular proteins typically disappears at low salt concentration,
because Tm varies approximately as a linear rather than a logarithmic
function of salt concentration (55, 56).
In summary, for both the ligand binding equilibria and the con­
formational transitions of nucleic acids, the functional form and mag­
nitude (especially at low salt concentration) of Ar3,2J differ significantly
from those of .1r3,2J for the corresponding processes involving proteins.
This difference justifies the use of the term polyelectrolyte effect (9) to
describe the characteristically large magnitudes and distinctive functional
forms of the effects of salt concentration on these processes involving
polyanionic nucleic acids.

MODELS OF NUCLEIC ACID SOLUTIONS


Types of Interactions
Under typical experimental conditions, the laws of classical electrostatics
generally are considered sufficient to describe the interactions of mobile
univalent salt ions with nucleic acids in solution. The more advanced
theoretical methods of calculating ion distributions (see below) have begun
670 ANDERSON & RECORD

to include explicit (model) water molecules as interacting particles under


realistic solution conditions. However, this level of detail is not yet feasible
for computations of experimentally accessible thermodynamic coefficients
that characterize nucleic acid solutions or salt effects on their equilibria
(3). If the solvent is modeled as a structureless continuum, the interactions
between a salt ion and a phosphate charge on the nucleic acid still may be
affected by the dielectric discontinuity at the solution interface with the
nucleic acid and/or by the spatial variation in dielectric shielding attribu­
table to the gradient in the local concentrations of salt ions surrounding the
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

nucleic acid (59). Different approaches to modeling in detail the dielectric


properties of the solution near a rodlike polyion have led to qualitatively
different conclusions about their effects on local ion distributions (60-62).
Concerns have been raised about the sensitivity of theoretical predictions
by University of Windsor on 07/16/13. For personal use only.

of gNAi to assignments of dielectric parameters that cannot be verified on


any independent basis (63, 64). Effects of different dielectric models on
the calculated magnitudes of thermodynamic coefficients under typical
experimental conditions are relatively minimal (64, 65).
Pair-wise solvent-averaged interactions between partial or integral charges
here are designated coulombic, whether or not the dielectric properties of
the medium are modeled as spatially uniform. Although these intercharge
interactions are primarily responsible for the distinctive characteristics due
to nonideaIity in nucleic acid solutions, even at relatively low concen­
trations, short-range repUlsions also can affect the magnitudes of measur­
able properties. Often these repulsions are described by the primitive
model, in which each particle is assumed to be rigid, with dimensions
assigned by considering various structural factors, such as hydration (for
the salt ions) or the sums of van der Waals radii of atoms comprising the
macromolecular surface (for the nucleic acid). Less commonly, some form
of soft-core repulsive potential, such as the two-parameter Lennard-Jones
form, is assumed (66, 67).
Besides short-range repulsions, salt ions and the charged groups on
nucleic acids may be involved in various other kinds of noncoulombic
interactions, whose consequences generally become more significant in
concentrated solutions. Noncoulombic electrostatic interactions are
invariably of shorter range than coulombic, and the effects of quantum
mechanical interactions usually are confined to neighboring solute and/or
solvent molecules in solution. Some detailed quantum mechanical descrip­
tions of the interactions of small ions with nucleic acids have been reported
(68). Interaction potentials capable of accurately describing site binding
of cations to nucleic acids have not yet been incorporated into rigorous
calculations (for example by MC simulations) of any of the thermo­
dynamic coefficients considered in this review. In contrast, the relatively
SALT-NUCLEIC ACID INTERACTIONS 671
simple character of classical intercharge interactions has permitted numer­
ous rigorous and/or detailed computations of the thermodynamic and
molecular properties of nucleic acid solutions and hence of coulombic
effects on biologically important processes in such systems.
Theoretical calculations that take into account only coulombic inter­
actions and short-range repulsions between charges require, in addition to
composition variables and temperature, parameters sufficient to describe
the dielectric properties of the solution and to specify the dimensions and
distributions of charges (integral or partial) on the interacting particles.
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

For systems as complex as aqueous nucleic acid solutions, the theoretical


development must be simplified in some respects because of limitations on
the current capacities of computers and because of incomplete information
about the structure(s) of a nucleic acid in solution, as well as uncertainties
by University of Windsor on 07/16/13. For personal use only.

about other relevant model parameters and/or about the proper functional
formes) of the interaction potentials. In practice the molecular model
always is idealized in some respects. Approximations also are introduced
into each of the analytic equations reviewed below, and may be introduced
into the thermodynamic manipulations whereby the observable of interest
can be related to a variable that is more directly accessible to accurate
theoretical calculations (11 ).

Classification of Model Assumptions


To describe the effects of coulombic interactions in an aqueous solution
containing salt and a nucleic acid, the simplest model that has been exten­
sively investigated by all the currently prevalent theoretical methods has
been characterized as standard (4). A more informative description is one­
dimensional ( l D) coulombic primitive. These terms specify, respectively,
the dimensionality of the charge distribution on the polyion and the
characters of both the long- and the short-range potentials that govern its
interactions with salt ions. Although the model polyion is cylindrical and
hence three dimensional (3D), its charge distribution is assumed to be
uniform, continuous, and strictly I D . A more compact designation of this
model is introduced here: (b/e/a). Each of these symbols represents the
only parameter used to characterize, respectively, the polyion charge dis­
tribution, the pair-wise additive solvent-averaged coulombic interactions
between (unit) charges, and the repulsion at contact between ion and
polyion, modeled as hard particles. The (b/e/a) model takes no explicit
account of interpolyion interactions. For a rodlike oligoion with 1 ZI uni­
valent charges, the simplest model that describes coulombic and hard
particle interactions is designated (b; IZI/8/a) in this review.
The three parameters required by the (bI8/a) model generally are evalu­
ated on the basis of the following considerations (for either a poly- or an
672 ANDERSON & RECORD

oligoion). If the actual 3D structure of the polyion in solution is assumed


to be the same as in the crystalline state and if the crystallographic coor­
dinates of the polyion charges are known, then the mean axial charge
separation b is calculated by projecting the actual coordinates of the
p olyion charges onto its cylindrical axis and averaging over their separ­
ations. The distance of closest approach, a, between the center of a spheri­
cal sa lt
ion and the axis of the cylindrical polyion also can be estimated
independently from structural information about the average diameters
of these species. In the absence of sufficient experimental information to
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

permit an independent estimate of b and/or a, these parameters can be


fitted by comparisons between exp erim en tal
measurements and the cor­
responding predictions (obtained, for example, from the PB equation).
According to the (b/e/a) model, the single spatially uniform and com­
by University of Windsor on 07/16/13. For personal use only.

position-invariant parameter e characterizes the dielectric properties of the


entire system within which coulombic interactions between salt ions and
polyions are operative. In calculations applying either CC theory or the
PB equation to the (b/e/a) model, e almost always is set equal to the value
pertaining to pure solvent at the appropriate temperature.
Among the refinements on the ID coulombic-primitive model for rod­
the following have been investigated most extensively.
like nucleic acids,
Explicit 3D coordinates {rp} have been assigned to each interacting atom
on the nucleic acid. Some or all of these charges have been assigned
fractional values {qp}. The dielectric boundary at the nucleic acid-water
interface has been parameterized by introducing an additional dielectric
constant, e2 (59). Dielectric saturation of the surrounding solution has
been introduced by representing e as a function of r [usually continuous
(52), sometimes a series of step functions (64)]. The primitive model for
ion nucleic acid short-range repUlsions has been refined by using structural
information to assign distances {sp} of closest approach between all charges
on the nucleic acid and any point on its solvent-accessible surface (59).
Alternatively, short-range interactions have been characterized by assign­
ing Lennard-Jones-type parameters {O"k> ek} to some or all of the salt ions
and/or charges on the nucleic acid. As an example of the notation used in
this review to distinguish various models, an all-atom model incorporating
partial charges, dielectric discontinuity, dielectric saturation, and Lennard­
Jones short-range potentials would be designated ({rp}/{'lp}; e(r);
GZ/{O"bed)·
Choice of the Appropriate Model
Some studies have stressed the importance of including the greatest avail­
able amount of detail in modeling the charge distribution and solvent­
accessible surface of a nucleic acid in solution for calculations of both
SALT-NUCLEIC ACID INTERACTIONS 673
thermodynamic and molecular properties. Such effects of structure may
be viewed as preaveraged by the (b/e/a) model or models that allow for
the penetration of salt ions into one or more cylindrical shells of polyionic
charge (69). Even the most detailed structural models neglect the possible
consequences of conformational fluctuations and local diffusional motions
that may modulate coulombic interactions in solution. Some progress
has been made toward the difficult problem of assessing how molecular
flexibility may affect calculations on solutions containing ch ain like oligo­
or polyions, but so far only the simpler models have been considered (70,
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

7 1 ).
From a practical standpoint decisions about the number and kind of
model parameters that are needed to characterize the system of interest
should be dictated by the physical character of the property to be calculated
by University of Windsor on 07/16/13. For personal use only.

and by the level of resolution with which it can be determined experi­


mentally. Model assumptions that suffice to obtain accurate theoretical
predictions of some properties cannot be expected to be universally appli­
cable. Reliable predictions of certain local properties may require very
accurate knowledge of local structural details and of interaction potentials
involving solvent water. For example, the NMR relaxation rates of
quadrupolar cations near the surface of DNA are critically determined by
short-range interactions and may also be affected significantly by local
fluctuations in the structure of the nucleic acid (72). These complications
so far have thwarted reliable ab initio calculations of observable NMR
relaxation rates. The unavailability of an accurate means of calculating
(or directly measuring) the quadrupolar relaxation rates of nuclei close to
DNA does not prevent the extraction of some useful information from
NMR studies of nucleic acid solutions (2, 20). Though incapable of yield­
ing accurate predictions of certain microscopic variables, the (b/e/a) model
provides a sufficient basis, at least under typical experimental conditions,
for the calculation of thermodynamic coefficients, such as r3•2 for double
helical DNA, as well as for the experimentally observable values of SaKobs
and d Tm/d ln a± that characterize processes involving nucleic acids. This
conclusion is substantiated by studies that have been reviewed elsewhere
(3-5) or subsequently in this article.

THEORETICAL METHODS APPLIED TO NUCLEIC


ACID SOLUTIONS
Analytic Equations
Polyelectrolyte theories and their applications to DNA have not been the
subject of a review in the Annual Review of Physical Chemistry since 1 982
(47). At that time theoretical methods applied to nucleic acid solutions
674 ANDERSON & RECORD

were almost entirely of the analytic type, based either on the formulas of
CC theory in one of the versions developed by Manning (13, 73, 74) or on
the classical cylindrical Poisson-Boltzmann (PB) equation (51). Each of
these analytic approaches incorporates various kinds of approximations,
and they adopt very different strategies to accomplish the Boltzmann
equilibrium averaging required for calculations of macroscopic (such as
rd and/or microscopic (such as gNAi) equilibrium properties affected
by salt-nucleic acid interactions. Continual improvements in computer
hardware and software have enabled applications to nucleic acid solutions
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

of analytic methods more rigorous than either CC theory or the PB


equation. The hypernetted chain (HNC) equation, based on an approxi­
mation that is theoretically more accurate than that built into the PB
equation, also is significantly more difficult to solve, especially for systems
by University of Windsor on 07/16/13. For personal use only.

of cylindrical rather than spherical symmetry. Numerically accurate cal­


culations of gNAi have been obtained by treating both ion polyion and ion­
-

ion interactions in the HNC approximation (66, 67, 75), but this approach
has not yet been incorporated in calculations of thermodynamic
coefficients for nucleic acid solutions. In part because of the relative sim­
plicity with which calculations based on CC theory or the PB equation
can be performed, these methods continue to be developed and applied
extensively to calculate the molecular and thermodynamic properties of
nucleic acid solutions.
Manning's first (13) thermodynamic version of CC theory was for­
mulated entirely in terms of free energies. Its principal results are remark­
ably simple analytic limiting laws for contributions made by the inter­
actions of a rodlike polyion with salt ions to thermodynamic coefficients,
such as y± or r3,2 (Equation 2). The entire molecular input into the CC
limiting laws is contained in ¢, the nondimensional axial charge density
defined in Equation 1 . For highly charged polyions (Izl� > 1) in solutions
where the only type of counterion is Izi valent, the forms of the limiting
laws for the thermodynamic coefficients are consistent with the inference
that [ l - (lzl�)-l] counterions are condensed on the polyion, but this
physical interpretation at the molecular level is not required.
Determining where the condensed counterions are situated with respect
to a highly charged cylindrical polyion was the primary objective of molec­
ular CC theory (74), which minimizes a free energy expression with respect
to the extent of counterion condensation to obtain a formula for Vee,
the volume enclosing the condensed counterions. Although � is the sole
molecular input into Vee, the radius of the polyion also must be specified
to evaluate the thickness of the condensation layer, which is needed for
comparisons with the ion distributions predicted by other theoretical
methods (76). The average number(s) of condensed counterions within a
SALT-NUCLEIC ACID INTERACTlONS 675
thin annular volume that is (virtually) constant over an extended range of
salt concentrations can be deduced, but detailed functional forms of 9NA+
and gNA- are neither predicted nor assumed by molecular (or any other
version of) CC theory.
Of the various generations of CC theory, only the first ( 1 3) thermo­
dynamic version is completely analytic in the sense that it provides
algebraic expressions than can be applied to analyze measurements with­
out any numerical computations. Molecular CC theory also provides some
explicit analytic formulas, but certain applications, for example describing
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

the competitive condensation of counterions having different charges,


require simultaneous numerical solutions of two transcendental equations.
To predict the average extent of counterion accumulation surrounding a
finite line charge model for an oligonucleotide, a free energy expression
by University of Windsor on 07/16/13. For personal use only.

adapted from molecular CC theory has been minimized numerically (77,


78). In this and other numerical investigations based on the principles of
molecular CC theory, the consequences of introducing 3D polyion charge
distributions or other refinements on the (b/e/a) model have been exam­
ined, as have various structural features of nucleic acids affected by their
interactions with salt ions (79, 80).
In salt-free solutions of cylindrical polyions where the counterions are
all of the same charge type, the spatial dependence of the cylindrical
symmetric PB potential can be expressed in closed form, but evaluation
of the integration constants requires numerical solution of a transcendental
equation (14). If counterions of more than one charge type and/or coions
are present in solution surrounding a rodlike polyion, then no closed-form
solutions of the PB equation are known, even for the simplest possible
(b/e/a) model. For this model the characteristics of potentials obtained by
solving the cylindrical PB equation and their physical implications have
been reviewed in detail (4). Solving the two-dimensional (2D) PB equation
in the form appropriate for solutions containing rodlike oligoions also
requires more advanced numerical methods, because the resulting PB
potential depends on more than one independent variable even for the
simplest (b; I ZI/e/a) model (8 1 and references therein).
Three-dimensional forms of the PB equation have been solved by elab­
orate numerical computations, most often to predict ion distributions or
averaged features thereof for models that are in various respects more
detailed (hence supposedly more realistic) than (b/e/a). Among the cur­
rently feasible refinements of this model that have been investigated by
methods subject to the PB approximation, most attention has been given
to all-atom descriptions of the polyion, which amalgamate the available
(not always complete) structural information, usually inferred from solid­
state X-ray crystal structures, with a theoretically based assignment of
676 ANDERSON & RECORD

partial charges for some or all of the atoms on the nucleic acid (82-84).
The resulting sets of structural ({rp}, {sp}) and energetic ({qp}, e, e2 e(r))
parameters are input into a program that utilizes finite differences (or finite
elements) to solve a 3D PB equation in either the nonlinear form (85-93)
or (less commonly) the linearized (Debye-Hiickel) form (94). The output
from such calculations is used to predict, angstrom by angstrom, the
discretized contour map of the electric potential or, equivalently, the
spatial (angular and axial, as well as radial) distributions of small ion
charge densities in the vicinity of a nucleic acid surface. Although this
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

level of detail exceeds that accessible to any known experimental probe,


angular and axial averages of gNAi may be determined in favorable cases
by X-ray scattering measurements (27, 28), and some still more highly
averaged features of gNAi near the surface of a nucleic acid may be moni­
by University of Windsor on 07/16/13. For personal use only.

tored by various types of spectroscopy (2, 4). Solutions of the 3D PB


equation for detailed models also have been used to calculate the experi­
mentally accessible thermodynamic quantity SaKobs (Equation 7) (65).

Simulations
Computational advances have enabled predictive calculations for the
(b/e/a) model, and some putative improvements thereupon, by methods
that are more rigorous than any of the analytic equations. Molecular
dynamics (MD) and stochastic dynamics (SD) have been used to calculate
ion distributions in solutions containing nucleic acids, either as oligo­
nucleotides or as segments intended to represent polyanionic DNA (95-
98). For all-atom models of oligonucleotide duplexes and triplexes (99,
1 00) in systems containing explicit water molecules, MD simulations have
provided detailed predictions about specific conformational features of
the oligomer and/or the distributions of solvent molecules and counterions
in the vicinity of particular oligomer structural groups. The strengths and
(current) limitations of MD and SD simulations, and results recently
obtained by these methods, have been amply covered in recent articles
(3, 95, 1 0 1 , 1 02). The thermodynamic coefficients that reflect coulombic
interactions in nucleic acid solutions have not yet been evaluated on the
basis of MD simulations.
For a specified set of ensemble constraints (canonical or grand canoni­
cal), MC simulations can be used to mimic the equilibration of various
molecular and thermodynamic properties of a model system of interacting
particles by sampling their configurations within a representative cell ( 103,
1 04). Implementations of MC methods to calculate ion distributions and
statistical thermodynamic variables have been described in detail (37, 75,
76). Calculations of gNAi by means of both canonical MC (CMC) and
grand canonical Me (GCMC) simulations have been extensively reviewed
SAL T-NUCLEIC ACID INTERACTIONS 677
(4). GCMC simulations have also been used to evaluate thermodynamic
coefficients y± (37, 38) and r3,2 (37, 39) for polymeric nucleic acids and for
oligonucleotides in both straight (23-25) and branched forms ( l OS).
Results of MC simulations are, in principle, subject only to statistical
errors that can be reduced to an arbitrary magnitude by expending enough
computer time. In practice the long-range character of coulombic inter­
actions and disparities in the sizes of the interacting particles may pose
special problems, including subtle systematic errors due to inadequate
equilibration times and/or truncation artifacts due to the necessarily finite
size of the MC cell and number of particles therein.
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

Comparative Theoretical Studies


Although improvements in the speed and power of computers continue to
by University of Windsor on 07/16/13. For personal use only.

expand the complexity of the systems and properties that can be inves­
tigated by MC simulations, for any set of model assumptions, much less
computer time and memory are required to solve the PB or even the HNC
integral equation, Thus, establishing the theoretical ranges of applicability
of these analytic alternatives by comparisons with more rigorous simu­
lations is ofpractical importance. Studies comparing the relative accuracies
of different theoretical treatments of nucleic acid solutions have been
extensively reviewed (e.g, 3, 4). Here only some recent and/or representa­
tive theoretical comparisons are considered. Measurements still provide
the only infallible tests of the ability of a given model and method to
calculate accurately a given property.
That CC theory, in one form or another, can account for a wide diversity
of observable phenomena with a bare minimum of molecular input appar­
ently indicates that the approximations implicit and explicit in this method
and its underlying model are in a remarkably stable state of balance under
a broad range of conditions. Theoretical studies [reviewed elsewhere (3-
5)] that have applied more rigorous methods to the (b/e/a) model have
demonstrated conclusively that it is not treated accurately, from the stand­
point of statistical thermodynamics, by either of the earliest versions of
CC theory (13, 74), except in the limit of high dilution. An application of
molecular CC theory to models more elaborate than (b/e/a) showed that
by incorporating both a helical charge distribution for (polymeric) ds
DNA and a dielectric saturation function, the predictions of the original
molecular CC theory for the magnitude and salt invariance of the extent
of counterion condensation are almost quantitatively retrieved (77). An
earlier CMC study also reported that a compensation between structural
detail in the model nucleic acid and dielectric saturation in the surrounding
salt solution produces a virtually constant extent of counterion accumu­
lation surrounding the nucleic acid surface, which does not differ sig-
678 ANDERSON & RECORD

nificantly from that predicted at infinite delusion by CC theory (52).


The recently reported close agreement between the predicted extent of
counterion condensation and the results of MC simulations (93) was
achieved by relaxing one of the principal features of molecular CC theory:
the virtual invariance of Vee to changes in salt concentration.
Theoretical efforts continue to be directed at finding a more fundamental
explanation for the widespread success of CC theory and at clarifying its
relationship to more rigorous statistical mechanical treatments of the
same or more detailed models ( 1 06, 1 07). Apparent extensions of the
applicability of CC theory to ordinary experimental conditions may be
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

attributable in some cases not to deep-seated compensations between


theoretical approximations, but rather to incidental compensations
between certain factors comprising observable properties. The following
by University of Windsor on 07/16/13. For personal use only.

section offers alternative explanations of the constancy of some spec­


troscopic and thermodynamic observables that has been taken as sup­
porting the extended validity of CC theory.

POLYELECTROLYTE EFFECTS IN NUCLEIC ACID


SOLUTIONS
Properties
EFFECTS ON ION DISTRIBUTIONS Theoretical and experimental studies of
the radial distributions of salt ions around polymeric nucleic acids have
been extensively reviewed (3, 4, 1 02). Coulombic interactions also cause
significant nonuniformity in the axial distribution of cations adjacent to
the surface of oligonucleotides, which are reviewed in the concluding
sections of this article. For both uni- and divalent counterions, in the
presence and absence of co ions, gNAi around a 50-nm (persistence length)
fragment of ds DNA, as monitored by X-ray scattering intensities, can be
fitted using the cylindrical PB equation for the (b/e/a) model (27). Although
this agreement was found both in the presence and absence of added salt,
the cylindrical PB equation, like Me simulations on the (b/e/a) model,
predicts a greater dependence on salt concentration of the local counterion
distribution near a nucleic acid than has been indicated experimentally
(75, 76).
The apparent insensitivity to changes in salt concentration of the extent
of counterion accumulation near the surface of a nucleic acid has been
inferred from various kinds of experiments ( 1 08), including in particular
NMR studies that monitor the change in the enhancement of the quadru­
polar relaxation rates of cations as a function of their concentration in
solutions of native B-DNA (2, 26, 1 09). A rigorous statistical mechanical
analysis of the PB equation pertaining to a cylindrical polyion in a salt
SALT-NUCLEIC ACID INTERACTIONS 679
solution indicates that the basic PB approximation could hold into the
ordinary experimental range of salt concentrations (C3 0 . 1 M) ( 1 1 0).
'"

Thus, the failure of the cylindrical PB equation to predict that the counter­
ion distribution near DNA is independent of C3 could be due to some
shortcoming(s) in the (b/e/a) model (which also forms the basis of molec­
ular CC theory).
Alternatively, the variation in the extent of counterion accumulation
predicted both by MC simulations and the PB equation for the (b/e/a)
model could be real but masked by compensating effects on the experi­
mental variable, at least in the case of NMR relaxation rates. Progress
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

toward more detailed theoretical interpretations of the NMR relaxation


rates of quadrupolar cations has been made by means of stochastic
dynamics simulations (72, 1 1 1-1 1 4). Some of these calculations have indi­
by University of Windsor on 07/16/13. For personal use only.

cated the possibility that the simple form of the concentration dependence
of the relaxation rates of univalent quadrupolar cations could result from
a compensation between increases in the local accumulation of counterions
and decreases in their local relaxation rate(s) as their total concentration
in the solution is increased.

EFFECTS ON THERMODYNAMIC COEFFICIENTS According to molecular CC


theory, both the number and physical location of the condensed counter­
ions are insensitive to changes in the total concentration of excess salt,
over a range extending from infinite dilution at least to 0. 1 mol dm- 3 (for
ds DNA) (74). This prediction could be viewed as an explanation of the
finding that the contribution of ion-polyion interactions to thermodynamic
properties of solutions of highly charged cylindrical polyions appears to
conform to CC theory even at ordinary experimental concentrations. A
diverse body of evidence for this extended applicability was collected in
a previous volume of the Annual Review of Physical Chemistry ( 1 1 5).
Comparisons of CC limiting laws with experiment generally depend on
introducing additional terms to allow for effects not included in the free
energy expression postulated by CC theory. For example, the limiting law
formula for l' ± must be supplemented by a term attributable to interactions
among salt ions in the absence of polyions. At moderate salt concentrations
the limiting law expression for r 3,u yields the experimental value for B­
DNA only after addition of a term that is directly proportional to the
salt concentration and to the volume excluded between ion and polyion
monomer ( 1 1 6).
The limiting law expressions deduced from thermodynamic CC theory
later were obtained without invoking any form of condensation hypothesis
by incorporating the distinctive limiting forms of the PB potential at
the surface of a highly charged (� > 1 ) cylindrical polyion into analytic
680 ANDERSON & RECORD

expressions for thermodynamic coefficients that had been derived from


the PB cell model (50, 1 1 7). The resulting analytic expressions do not
exhibit the apparent salt independence over a broad range of salt con­
centrations that has been inferred as a general property of the cor­
responding CC limiting laws. Nevertheless, the (relatively large) salt depen­
dence of r 3,2 that has been measured by membrane dialysis on solutions
of double-helical DNA can be fitted by applying the cylindrical PB equa­
tion to the (b/e/a) model with physically reasonable values of b and a (58,
1 1 8). Via the boundary condition imposed at the polyion surface, the
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

PB equation takes into account effects of excluded volume that are not
incorporated explicitly into either thermodynamic or molecular CC theory.
At sufficiently high salt concentrations, the excluded volume contribution
is predicted by the PB equation to dominate the salt dependence of r3,2'
by University of Windsor on 07/16/13. For personal use only.

Effects on Processes
Experimental determinations of SaKobs (Equation 7) or dTm/dln a± (Equa­
tion 8) provide valuable benchmarks with which computational or theor­
etical predictions of the thermodynamic consequences of coulombic inter­
actions can be compared or (more commonly) calibrated. In most
applications, not all of the parameters needed for a theoretical estimate of
SaKobs can be determined a priori. For example, for any calculation of the
effect of salt activity (a ± ) on helix formation or denaturation of DNA,
a detailed model cannot be constructed a priori because the structural
parameters of the denaturated state are not well known. The success of a
theoretical analysis of effects of a ± on these processes may be gauged by
comparison of fitted values of the structural parameters for the denatured
state with independent estimates (or intuition) (5, 58). The inherent com­
plexity of an aqueous nucleic acid solution is such that uncertainties remain
at present not only about the most appropriate values for some of the
molecular model parameters, but about the number and kind of these
parameters that are necessary to characterize the system. As a basis for
deciding about the level of detail needed in the theoretical model,
additional information about structures and interaction potentials in solu­
tion (especially if interactions with water are included) will be needed
before completely a priori predictions can be obtained for SaKobs or
dTm/dln a ± .

CONFORMATIONAL TRANSITIONS Effects o f salt concentration (C3 � 1 mol


3
dm - ) on Tm for nucleic acid order-disorder transitions in aqueous solution
appear to be primarily coulombic in character and thus to represent an
entropic contribution to helix stability (5, 1 19, 1 20). If C3 � 1 mol dm- 3 ,
interactions other than those that can be described by simple models based
SALT-NUCLEIC ACID INTERACTIONS 68 1
on coulombic interactions may well become significant (5, 7). Limiting law
expressions from thermodynamic CC theory were used originally in the
analysis ( 1 2 1 , 1 2 I a) of the experimentally observed linear dependence of
Tm on a± (excess uniunivalent salt) for the denaturation of a polymeric
nucleic acid double helix. This approach ( 1 6, 1 7) was generalized sub­
sequently to other order-disorder transitions of polymeric DNA:

11.
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

Here �(20 - I refers to the stoichiometrically weighted difference in the


axial charge density parameters (Equation I) pertaining to the final and
initial states of the transition. Equation 1 1 in effect introduces the limiting
law n,� = - (4�) - 1 ( 1 3, 50) into the general thermodynamic expression
by University of Windsor on 07/16/13. For personal use only.

(Equation 8). The utility of Equation 1 1 , based on the minimally par­


ameterized (b, e) model, is indicated by its ability to account ( 1 6, 1 7, 1 2 1 ,
1 2 I a) for the following classic observations. For transitions of DNA or a
polynucleotide from a helix to a completely denatured single-stranded (ss)
state, dTm/dln a ± increases as the number of strands in the helix increases.
For DNA, dTrn/dln a± decreases monotonically with increasing pH,
becoming negative above pH I I . For disproportionation of the
poly rA . poly rV duplex into a triplex [poly rA ' (poly rV)2] and a single
strand (poly rA), dTm/dln a + is negative. In the isothermal, salt-induced
B-Z transition of polynucleotides, originally investigated at salt con­
centrations C3 ;<; 1 mol dm-\ the lower charge density Z form is favored
by an increase in C3• Although this behavior appears contrary to Equation
11, a reconciliation of this system with other DNA conformational tran­
sitions has been obtained by the observation that the Z form also is favored
at sufficiently low C3 (cf 5 for a review). Salt effects on both Z-B and B-Z
transitions have been analyzed using the cylindrical PB equation and the
(b/e/a) model (5).
Values of dTm/dln a± for the denaturation of DNA have been analyzed
using the PB equation (and some GCMC simulations) (5, 58). The salt
dependence of the coulombic component of the free energy difference
between native and denatured states of polymeric B-DNA was shown to
be consistent with calculations based on the cylindrical PB equation for
the (b/e/a) model (5). Subsequently solutions of this equation were used
to evaluate �r 3,2J (in Equation 8) and thereby to fit experimental deter­
minations of dTrn/dln a± for the denaturation of B-DNA and for various
order-disorder transitions of complexes of poly rA and poly rV (58).
The observed constancy and magnitude of dTm/d ln a± for all of these
transitions are accounted for by PB calculations on the (b/e/a) model
with the conventionally accepted structural parameters for the native
682 ANDERSON & RECORD

conformation and with physically reasonable fitted values of the average


structural parameters for the denatured form of DNA in solution. These
calculations demonstrate that the semiquantitative successes of earlier
limiting law-based analyses ( 1 6, 1 7, 1 2 1 ) could result from compensations
between large dependences on C3 of the r3 . 2J pertaining to the native and
denatured forms.

BINDING EQUILIBRIA Practically any value of Kobs for binding of an oligo­


cation (LZL+) to a nucleic acid polyanion can be obtained by varying the
salt concentration. The large negative power dependences of Kobs on salt
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

concentration observed experimentally for oligocation-nucleic acid inter­


actions were proposed to result principally from the polyelectrolyte charac­
ter of the nucleic acid charge distribution ( 1 8). A thermodynamic analysis
by University of Windsor on 07/16/13. For personal use only.

based on limiting law (cq theory was developed to derive the approximate
result

12.
where ZL is the number of charges on the oligocation, and � is defined in
Equation I. Equation 1 2 is a special case of the general equation (Equation
7) for SaKobs. The quantity 1 - (2�)- 1 is identified as the thermodynamic
extent of counterion association (per charge) with the uncomplexed nucleic
acid (cf Equation 5). Hence binding of an oligocation LZL to a polyanionic
nucleic acid at low salt concentration can be considered a thermodynamic
cation exchange interaction.
According to Equation 1 2, the proportionality of In Kobs to ln a± is salt
independent, at most a weak function of T (because e is approximately
inversely proportional to T), and directly proportional to ZL' These pre­
dictions are consistent with experimental data acquired in the ordinary
range of experimental concentrations for the binding to nucleic acids of a
considerable variety of oligocationic ligands, including oligopeptides (e.g.
1, 1 22-125), intercalating dyes (e.g. 126, 1 27), polyamines (e.g. 128), and
other oligocations (e.g. 1 29-1 32). These studies were performed for ss and
ds DNA and/or RNA at low to moderate concentrations of univalent salt
(typically 0.01 � C3 � 0 . 2 mol dm - 3), In some particularly comprehensive
quantitative studies, Mascotti & Lohman (1 23-125) investigated the bind­
ing ofa series ofoligolysines (ZL + 2 to 1 0) and oligoarginines (ZL + 2
= =

to 6), containing a tryptophan group as a fluorescent probe, to polymeric


ss polynucleotides [poly (dT, dU, rU, rA, rC, rI)] and ds pUC8 plasmid
DNA in dilute aqueous salt solutions. The observed large dependences of
Kobs on C3 exhibit (in representative cases) little variation with temperature
and thus appear to be entropic in character. For the binding of oligo­
peptides to ss DNA, SKobs == dIn Kobs/dln C3 does not depend on C3 and
SALT-NUCLEIC ACID INTERACTIONS 683
varies with ZL from - 1 .4 ± O.2 (ZL = + 2) to - S.S ± OJ (ZL = + 1 0) in ,

conformity with Equation 1 2. These thorough experimental investigations


of oligopeptides (LzL) with polymeric ss and ds nucleic acids constitute a
valuable data base for further tests of alternative theoretical calculations.
Of course the applicability of Equation 1 2 to describe the binding of
oligocationic ligands to nucleic acids does not necessarily confirm any
of the assumptions with which it was derived. As in the case of the
conformational transitions discussed above, the individual r 3 .2J specified
in Equation 7 for each of the participants in the binding equilibrium could
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

vary with salt concentration in ways that compensate to produce a saIt­


invariant value of SaKobs' Calculations of f 3,21 for an oligocationic ligand
LZL, for oligonucleotides of varying lengths, and for their central complexes
with the ligand all were obtained from GCMC simulations for essentially
by University of Windsor on 07/16/13. For personal use only.

the same model system as was assumed in deriving Equation 1 2 (25).


These results demonstrate that the decreases in I f3,2L I with increasing salt
concentration may be offset by increases in the remaining coefficients
comprising �r 3 . 2" so SaKobs is relatively insensitive to changes in salt
concentration,
The adequacy of the (ble/a) model as a basis for predicting experimental
values of SaKobs for the binding of simple ligands to nucleic acids was
demonstrated in a recent study (65) that compared results obtained by
solving the cylindrical PB equation for this model with those obtained by
solving a 3D PB equation for the ({rp}/{qp}; e; e2/{sp}) model. For the
binding of each of the divalent cationic antibiotics DAPI and netropsin to
the minor groove in the central regions of 1 2-base pair (bp) oligo­
nucleotides, the 3D PB calculati on s predict that - SaKobs = 2 1 , a value
.

identical to the corresponding predictions of the cylindrical PB equation,


which were, however, calculated for the binding of these ligands to a
polyion, The experimental values, each also obtained for binding to poly­
anionic DNA, are 2.3 ( 1 30) for DAPI and ", 1 .55 for netropsin ( 1 3 1 , 1 32),
The corresponding limiting law prediction obtained from Equation 12 for
the binding of a divalent cationic ligand to polyanionic DNA is
- SaK;;' 1.8.
=

To examine the possibility that a 1 2-bp model oligonucleotide was not


long enough to represent a polyion, SaKobs also was evaluated (65) by
solving the 3D PB equation for an all-atom representation of the binding
of DAPI to a 57-bp model oligonucleotide. (The coulombic end effects
that differentiate oligonucleotides from polynucleotides are reviewed
below.) No difference in SaKobs was found for the binding of DAPI to the
longer fragment. This finding was in accord with an inference based on
calculations using the 2D PB equation for rodlike oligoions, described by
the (b; / ZI/e, ezla) model, that coulombic end effects scale as (2K) - 1 , where
684 ANDERSON & RECORD

K is defined as the usual Debye-Huckel screening length ( 1 33). These


calculations were performed only for a relatively weakly charged oligoion
(¢ I) over a limited range of salt concentrations: 0.0025 < C3 , 0.025 mol
=

3
dm- • This scaling formula subsequently was found to be consistent with
the axial range of the coulombic end effect indicated by GCMC simulations
(23) for the (b/e/a) model of an oligonucleotide at the single salt con­
centration (0.001 5 mol dm - 3) . However, subsequent work has shown this
consistency to be coincidental (81).
A more thorough 2D PB study of oligonucleotides (¢ = 4.2) described
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

by the (b; IZI/8, 82/a) model demonstrates that over the concentration
range 0.001 � C3 � 0. 1 00 mol dm- 3 coulombic end effects do not scale as
(2K)- 1 but rather are virtually independent of C3 (8 1 ). These 2D PB
calculations predict that the central region of a 57-bp model oligo­
by University of Windsor on 07/16/13. For personal use only.

nucleotide is virtually polyanionic, both over the range of C3 investigated


experimentally (0.08--0.3 mol dm- 3) and at the concentrations for which
the 3D PB equation was solved, but they also predict that at C3 0.01 =

mol dm- 3 the reduced surface potential YPB(a) at the center of a I O-bp
oligonucleotide is lower by at least 0.3 kT units than at the center of a
1 00-bp oligonucleotide and, furthermore, that the salt dependence oflPB(a)
differs discernibly for these different lengths. This difference implies the
existence of an oligoion length effect on SaKob., though its predicted mag­
nitude cannot be quantified without knowledge of the excess free energy
pertaining to each participant in the binding equilibrium.
Essentially the same method (3D PB equation) and model ({rp}/{qp}; e;
82/{Sp}) as were used to analyze the binding of DAPI to 1 2- and 57-bp
oligonucleotides had been applied earlier to analyze the binding of the
(monomeric) amino terminal fragment of ). repressor to 9- and 45-bp
oligonucleotides over the range 0.025 � C3 � 0 . 1 00 mol dm- 3 ( 1 34). This
earlier study reported that the magnitUde of SaKobs for the binding of this
protein to the longer oligonucleotide was 30% greater, in qualitative
'"

accord with the 2D PB calculation cited above as indicating a significant


coulombic end effect on YPB(a) (81). Different ways of allowing for the
long range of coulombic interactions in 3D PB calculations based on all­
atom models apparently lead to qualitatively different conclusions about
the importance of coulombic end effects on SaKobs for the binding inter­
actions of nucleic acids. Alternatively, coulombic interactions with more
distant phosphate charges on a polymeric nucleic acid may be significantly
more important for a protein than for a smaller ligand such as DAPL As
reviewed in the following section, numerous theoretical calculations on
polymeric nucleic acids have demonstrated the necessity of taking proper
account of the long-range effects of coulombic interactions on ion dis­
tributions near highly charged polyions.
SALT-NUCLEIC ACID INTERACTIONS 685
To calculate SaKobs for the binding of small oligocationic ligands to
nucleic acids, no study has yet demonstrated the necessity of modeling the
participants in the binding equilibrium with a high degree of structural
detail or of including any other model refinements (such as explicit water
molecules) that might also be significant. Currently the same statement
holds for the binding of proteins to specific sites on polymeric nucleic
acids, even though for this type of binding some relatively minor variations
in the primary structure of nucleic acids are accompanied by significant
changes in the magnitude of SaKobs ' For example, substituting a different
base in the recognition sequence of lac-repressor causes a twofold change
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

in SaKobs (1 35) for its specific binding interactions with polyanionic


B-DNA. This effect was attributed to conformational adaptability of the
protein in response to the change in DNA sequence, rather than merely a
by University of Windsor on 07/16/13. For personal use only.

change in the free energy due to coulombic interactions with the altered
sequence. Significant effects of base sequence on SaKobs also have been
reported for other protein-DNA interactions, including the Ad repressor
( 1 36, 1 37) and EeoRI endonuclease ( 138). Relatively minor ( '" 20%)
changes in the magnitude of SaKobs have been predicted by calculations
using the 3D PB equation for all-atom models of the complexes of AcI
repressor to different OLI operator sequences (Table 3 of Reference 1 39),
but these predictions have yet to be verified experimentally. Conversely,
moderate effects ( ± 40%) on SaKobs for the binding of this protein to
different OR operator sequences have been observed (1 36), but the lack of
detailed structural information for these complexes has inhibited any
corresponding theoretical calculations of SaKobs '
For comparisons with experimental data, the effects of salt con­
centration on the specific binding of AcI repressor to its operator sites on
DNA have been calculated in three independent theoretical studies ( 1 34,
1 39, 1 40). The earliest (140) used canonical MC simulations in conjunction
with a perturbation method to calculate the contribution of coulombic
interactions with salt ions to the free energy of association of a dimer of
the amino-terminal fragment of the AcI repressor with a 1 7-bp oligo­
nucleotide. A ({rp}/e2, e(r)/{oA) model was assumed for each of the par­
ticipants in this binding equilibrium, and an effort was made to include
additional coulombic interactions with images of the oligonucleotide,
because the experimental value of SaKobs was obtained for binding to
polymeric DNA. Added NaCI concentrations over the range 0.025 to 0. 1
mol dm -3 were investigated, but no Mg2 + [a competitive ligand known (7,
8) to affect both Kobs and SaKobs in such systems] was included in the
simulation, which thus in this major respect did not match experimental
conditions. Despite the rigor of the method and the detail built into the
model, the resulting CMC calculation of I SaKobs l significantly under-
686 ANDERSON & RECORD

estimates (by at least a factor of six) the most recently reported experi­
mental values, in the range 3-7 (Table 3 of Reference 1 36). In a subsequent
theoretical study ( 1 34) covering the same range of salt concentrations
(again, without including Mg2+), the 3D PB equation was solved for two
different detailed models of the nucleic acid charge distribution to evaluate
SaKobs for the binding of the amino-terminal domain of the AcI repressor,
taken to be complexed as a monomer with either a 9-bp or 45-bp (model)
oligonucleotide. After the predicted value of SaKobs for the latter was
doubled (to account for the fact that the protein binding site actually
interacts as a dimer with the nucleic acid), it was considered in reasonable
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

agreement with experiment.


A third round of theoretical calculations on the AcI repressor-operator
interaction arrived at a result for SaKobs ( - 4.4) within the range of experi­
by University of Windsor on 07/16/13. For personal use only.

mental results ( 1 39). Again excess coulombic free energy changes were
evaluated (by a formally different route) from solutions of a 3D PB equa­
tion for an all-atom model, but a more recent set of crystallographic
coordinates were input into the model ( 1 4 1 ). When this theoretical cal­
culation was corrected (semiempirically) for the presence of Mg2+ in t he
experimental system, agreement between the calculated and theoretical
values of SaKobs was degraded (by '" 20%). Also (apparently) not included
in the theoretical calculation were long-range coulombic interactions with
phosphate charges on polymeric DNA (for which the experiments were
performed) that are located beyond the segment of the model oligo­
nucleotide for which the PB calculations were performed.
Efforts toward the important goal of calculating experimental values of
SaKobs for the binding of proteins to nucleic acids indicate that constructing
a detailed model for the system using known structural information (often
incomplete under actual experimental conditions), can pose a considerable
challenge. The level of detail in the model that is actually needed to yield
agreement with experiment may vary significantly with the identities of
the species involved in the binding interaction.

SIGNIFICANCE OF COULOMBIC END EFFECTS


Defining Characteristics
Differences in various average molecular and thermodynamic properties
(normalized per charge on the polymer or oligomer) that are affected by
coulombic interactions generally are observed when a sufficiently short
oligonucleotide is substituted for the corresponding polynucleotide in an
aqueous salt solution. These differences, which become progressively larger
with decreasing length of the oligonucleotide, are due to coulombic end
effects. As defined above, coulombic has the connotation of interactions
SALT-NUCLEIC ACID INTERACTIONS 687
between discrete charges, regardless of whether the dielectric is modeled
as spatially invariant. The modifier coulombic also is used to differentiate
the end effects of interest here from others, such as partial unwinding
(fraying) or enhanced binding affinity for charged ligands, that may arise
from some special structural and/or chemical feature at or near the end of
an oligonucleotide. Coulombic end effects are manifested not only at the
ends of an oligoion; more generally they must be understood as resulting
from the existence of these ends.
From the standpoint of statistical thermodynamics, coulombic end
effects can be described in terms of the mean electrostatic potential, <'1'),
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

and/or the potentials of mean force <<PNAi) for the interactions of each
type of salt ion i with the nucleic acid. These potentials determine, respec­
tively, all of the thermodynamic measures of nonideality, and the radial
by University of Windsor on 07/16/13. For personal use only.

distribution functions gNAi- A discussion of relationships between these


potentials and the properties of polyelectrolyte solutions can be found, for
example, in Reference 4. At points in solution sufficiently close to the
central region of a finite rodlike polyion, <'I') and the (<PNA;) exhibit no
systematic variation in a direction parallel to its axis. Some fluctuations
would appear close to the surface of any real polyion, because its charge
distribution does not have perfectly cylindrical symmetry. As either end
of the polyion is approached, the magnitudes of these potentials begin to
decline in the axial direction. Although the resulting coulombic end effects
cannot be discerned for a polyion, axial nonuniformities in ('I') and in
< <PNAi> for sufficiently short rodlike oligoions produce significant detect­
able effects on thermodynamic variables (2 1 ) and on ion distributions (2,
20).
Comparison of the 23Na NMR relaxation rates (Robs) in solutions con­
taining nucleic acids of different lengths shows that under salt free con­
ditions (where in both cases the nuclei .near the nucleic acid surface make
maximal contributions to Robs) , Robs of a 160-bp mononucleosomal DNA
fragment is 1 .8 times larger than that of a 20-bp synthetic oligonucleotide
(20). This factor actually could underestimate the magnitude of the cou­
lombic end effect on ion accumulation, because near either end of a nucleic
acid a component of the electric field gradient in the axial direction may
cause the local relaxation rate of a quadrupolar ionic nucleus to be some­
what greater than when the nucleus is near the central region of the nucleic
acid.
How short the oligoion must be before end effects are exhibited may
vary somewhat with the property of interest and with the method of
observation. The distance over which the termination of a rodlike array
of charges has any significant effect on <'I') or < <PNAi) depends on the
effectiveness of electrostatic screening by the saIt ions, and hence could be
688 ANDERSON & RECORD

expected to vary with the salt concentration (C3) and the axial charge
density e. Of variables that determine whether coulombic end effects are
detectable, the most basic is the length of the rod like molecule: I Z l b.
Regardless of the detailed form of the oligoion charge distribution, its
length can be expressed in units of b, as defined in connection with Equa­
tion I , provided that any part of the structure extending beyond the
terminal charge on each end is neglected.
In experimental and theoretical studies relatively little attention has
been paid to 1 ZI as a determinant of the consequences of coulombic
interactions in nucleic acid solutions. The transition from characteristics
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

of simple electrolyte solutions (l ZI 1 ) to the distinctively different


'"

characteristics of polyelectrolyte solutions (I ZI » 1 ) is of substantial inter­


est, because it exemplifies the broad spectrum of physical effects due
by University of Windsor on 07/16/13. For personal use only.

to coulombic interactions. In vivo most nucleic acids are polyanionic.


However, the execution and interpretation of various kinds of experiments
in vitro are facilitated by the use of oligonucleotides having a definite
number and sequence of bases. Typically the smallest possible oligo­
nucleotides are used because the cost of synthesizing and purifying such
well-characterized samples increases rapidly with increasing I ZI . Com­
putational costs also dictate that the shortest possible segment be chosen
to represent the polynucleotide of interest when properties affected by its
interactions with salt ions are calculated by any of the more detailed
models and/or more rigorous methods. In such calculations coulombic
end effects generally have been recognized as significant, even though (as
noted above) they may be viewed as obstacles to arriving at predictions
for a polyanionic nucleic acid, particularly if it is modeled in great detail.
From a positive perspective, quantitative calculations of coulombic end
effects for model oligoions provide an opportunity to test the applicability
of various theoretical approaches, as compared with each other and, where
possible, with experimental data. To determine the conditions under which
coulombic end effects can be neglected or, when they cannot, to decide the
proper way of taking them into account requires a systematic examination
of the complete grid of physical variables that govern the approach to the
polyion limit. Toward this goal results that have been achieved thus far in
theoretical and experimental studies of oligonucleotides are reviewed in
the remainder of this article.

Axial Range
In the most rigorous available theoretical calculations on systems intended
as models of polyelectrolyte solutions, care has been taken to avoid,
or compensate for, coulombic end effects. Directly or indirectly, such
calculations provide a clear indication that the ion distributions sur-
SALT-NUCLEIC ACID INTERACTIONS 689
rounding even the central region of model oligoions with I ZI $ 60 differ
from those surrounding that of the corresponding model polyion, at least
when C $ 10- 2 mol dm- 3 • In the first study reporting numerical solutions
3
of a 3D form of the PB equation for a structurally detailed model of ds B­
DNA (82), coulombic end effects were examined with the objective of
determining how long the model segment must be to ensure that the
counterion charge density surrounding its middle section (extending axi­
ally over 34 A and radially over 60 A from the surface) is representative
of polymeric DNA. Two segment lengths (l ZI = 80 and 60) at two con­
centrations of uniunivalent salt (e = 1 0-5 mol dm- 3 and 1 0-2 mol dm- 3)
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

3
were investigated. At the lower salt concentration, the results indicate that
coulombic end effects on ion distributions near the middle of a DNA
molecule cannot be neglected, even for the longer segment. At C = 1 0 - 2
by University of Windsor on 07/16/13. For personal use only.

3
mol dm - 3 , the equivalence of the average counterion densities surrounding
the middle sections of both segments was deemed to indicate that the
surrounding ion distributions are characteristic of polymeric DNA. The
scatter in these PB calculations precludes any exact quantification of the
range of the coulombic end effect or of the decrease in the number of
counterions surrounding the oligonucleotide segment from the value
expected if the entire segment were in the interior of polyionic DNA. The
average charge density along the axis of both of the model segments
investigated does appear to be nonuniform over a distance extending at
least 1 . 7 nm and possibly as much as 3.4 nm from either end. For the
shorter (lZI = 60) segment, when C3 10- 2 mol dm- 3 , at least one third
=

and possibly as much as two thirds of the axial profile of the average
charge distribution deviates from the uniform value that is presumably
characteristic of polymeric B-DNA.
In the earliest (142) application of canonical Me simulations to inves­
tigate ion distributions surrounding polyanionic ds B-DNA in aqueous
solution (modeled as a dielectric continuum), constraints on computer
time dictated the simplest possible model for the nucleic acid charge
distribution (linear or surfacial helical) and the smallest possible number
of mobile ions. Accordingly, the simulations were conducted on solutions
containing either no or minimal added salt, and I ZI = 1 0 for the model
segment enclosed by the Me cell. To compensate for the small size of this
Me cell, configurational energies of the charges enclosed within it were
calculated by including interactions with identical configurations of charges
in a total of four image cells symmetrically situated along the nucleic acid
axis. Thus, interactions were in some sense included over a total segment
length corresponding to I Z I 50. In the next study to report Me simu­
=

lations of ion distributions surrounding polymeric B-DNA (75), sub­


stantial computer time was expended to obtain accurate results at excess
690 ANDERSON & RECORD

3 3
salt concentrations (C3 1 0 and 10- 2 mol dm- ). Even though model
= -

segment lengths consisting of at least I ZI 60 were investigated, a cylin­


=

drically symmetric external potential also was imposed (in conjunction


with minimum image cells) to account for the interactions of ions enclosed
by the MC cell both with charges on the external segments of the (indefi­
nitely) long model polyion and with the ion distributions surrounding
them. An external potential had been used analogously for the first time
in an MC study of ion distributions perpendicular to a charged (infinite)
plane ( 1 43). Although external potentials have not been universally
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

accepted (5) as a means of including interactions with charges outside the


largest computationally tractable MC cell, the necessity of allowing for
the long range of coulombic interactions in MC simulations of ion dis­
tributions surrounding highly charged rodlike polyions has been well
by University of Windsor on 07/16/13. For personal use only.

established. Canonical and grand canonical MC studies of the (b/e/a)


model for polyanionic DNA (as reviewed, for example, in 3, 4) now
routinely use external potentials, but for more complicated all-atom
models, implementation of this approach presents difficulties that though
not insuperable, may be avoided entirely if the model polyion segment is
long enough.
The significance of coulombic end effects inferred from the theoretical
calculations on polynucleotides cited above is illustrated explicitly by
GCMC simulations of the axial profiles of counterion distributions along
the surface of rodlike oligoions described by the (b; I ZlJe/a) model (23).
(This analysis also demonstrates that for a homologous series of oligo­
electrolytes that are long enough to include a polyelectrolytelike interior
region, both the average molecular and thermodynamic properties
approach their polymeric limits as linear functions of 1 2 1 - 1 .) Subsequent
CMC simulations of sodium ion distributions near d(AT)s ' d(AT)s in each
of three conformations (B, A, and "wrinkled D") were based on either the
all-atom ({r;}/{q;}, e/{s;}) or the (b; I ZI/e/a) model (144). The short length
of the oligonucleotide (1 21 1 8) permitted simulations at each of four salt
=

concentrations in the range 0.03-0.30 mol dm-3, all in excess over the
oligonucleotide monomer concentration [P] 2.2 X 10-2 mol dm- 3 • For
=

each of the conformations investigated, the results presented in Figures I ,


2, and 4 of Reference 144 demonstrate that even at relatively high salt
concentrations, and for a I SO-fold excess, end effects on the axial profile
of the cation distribution are clearly evident. With at most two exceptions,
the number of ions located in any of four concentric annular volumes
surrounding two terminal regions, whose axial extents together comprise
at least 40% of 1 21 b, differs significantly from the number of ions located
in the corresponding volume surrounding the central region. Moreover,
within the annular volume that corresponds most closely to the con-
SALT-NUCLEIC ACID INTERACTIONS 69 1
densation layer, the sensitivity to changes in total salt concentration of the
cations accumulated in the terminal region is distinctly greater than in the
central region.
Axial nonuniformity in the counterion charge distribution near the
surface of a rodlike oligo- or polyanion can arise not only from proximity
to the ends of the molecule (23), but also from the local reduction in axial
charge density of the oligo- or polyion that occurs upon binding of an
oligocationic ligand (with ZL < I ZI charges). Unless the ligand is so long
that it effectively splits the polyion into two noninteracting segments,
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

coulombic interruption effects, unlike end effects, do not have an ever­


widening axial range with increasing radial distance from the complexed
ligand. GCMC simulations have been used to predict the axial non­
uniformity in the surface counterion distribution caused by binding a
by University of Windsor on 07/16/13. For personal use only.

single rodlike ligand (I ZI = 8) to the center ofa B-DNA rodlike oligoanion


(24 < IZI < 250) (25). The complex was modeled by eliminating ZL
charges centered on the midpoint of a (b; I ZI/ela) oligoanion. These axial
profiles illustrate directly the cation-exchange character of binding of LZL.
Knowledge of the axial range over which the presence of a bound ligand
affects the local counterion distribution near a nucleic acid is useful as a
gauge of the maximum binding density at which ligands can be considered
noninteracting in analyses of their binding isotherms ( I I ). This con­
sideration is important, for example, in establishing the range of conditions
over which experimental values of Kobs can be analyzed with Equation 7.

STUDIES OF OLIGONUCLEOTIDE SOLUTIONS


Scope of Investigations
Recent reviews (3-5, 59, 145, 146) have covered applications of CC, PB,
HNC, or MC methods to calculate ion distributions and/or some thermo­
dynamic coefficient(s) for solutions containing rodlike polyions. Few anal­
ogous calculations have been performed for rodlike oligoions, and still
fewer specifically for oligonucleotides. Some recent papers (23, 25, 1 44)
have discussed pertinent previous work on oligoions with values of �
different from 4.2 (characteristic of ds B-oligonucleotides). Coulombic end
effects on the following properties of oligonucleotides in solution have been
studied: the mean electrostatic potential, <\}II Z I> (8 1 ); the local counterion
concentration near the oligonucleotide surface, CIHx, a) (23, 25, 144); the
axial average of this concentration CI"ir(a) (23, 77, 1 44); and the salt­
oligonucleotide preferential interaction coefficient, rrzr,u (23, 25) (Equation
2). These quantities have been calculated by applying GCMC (23-25),
CMC (144), PB (8 1 , 82), or CC (77, 1 47) to the following models for
oligonucleotides in solution: (b; I Zl/ela) (23-25, 77, 1 47), (b; I Zl/e; e2la)
692 ANDERSON & RECORD

(8 1 ), and ({rp}/{qp}, I:/{sp}) (82, 144). Some results computed for different
models by the same method have been compared (8 1 , 1 44). Although all
of the results reported in this section were obtained for rodlike oligo- or
polyions having the same value of e, in most cases the other relevant
variables (1 2\, C3) were different, and/or different measures of coulombic
end effects were reported. Consequently, few quantitative comparisons
can be made.
The theoretical calculations on oligonucleotide solutions reviewed here
do yield some clear-cut indications of the importance of coulombic end
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

effects. Of greater weight are the several cases where theoretical descrip­
tions that take into account coulombic end effects have proved capable of
3
fitting experimental observations. A comparative 2 Na NMR study (20)
of two oligonucleotides (\ Z\ = 38 and 320) demonstrated that the depen­
by University of Windsor on 07/16/13. For personal use only.

dence on \ Z\ of the extent of counterion accumulation is consistent with


GCMC simulations on the (b; \ 2 \/I:/a) model. Values of [121,. obtained
from GCMC simulations on the (b; \ 2\/e/a) model have been used in
quantitative analyses of experimental studies of the salt dependence of
conformational (24) and ligand binding equili bria involving oligo­
nucleotides (J. Bond, unpublished results).

End Effects on Cation Accumulation Near an Oligonucleotide


Although the M e studies of both polyanionic DNA and oligonucleotides
reviewed above clearly indicate the importance of coulombic end effects,
a qualitatively different conclusion could be drawn from two different
forms of CC theory (77, 1 47), formulated independently to describe the
association of univalent counterions with a finite rodlike oligoion. These
adaptations of CC theory-one (77) descending from the molecular
approach developed for cylindrical polyions (74) and the other ( I 47) incor­
porating a ligand binding formalism-yield predictions that differ by less
than 20% for Ot.IZI' the average extent of counterion condensation of
'"

univalent cations on an oligonucleotide having \ 2\ phosphate charges.


According to both of these versions of CC theory, coulombic end effects
are significant only at salt concentrations below the typical experimental
range (C3 ;:5 10-3 mol dm - 3), even for oligonucleotides shorter than the
segments needed to eliminate end effects in the more rigorous calcula­
tions of ion distributions around polyions. Much effort has been de­
voted to comparing the predictions of CC theory for cylindrical polyions
with those of alternative theoretical approaches and with experimental
measurements (cf 4). Accordingly, and in view of the widespread popu­
larity of CC theory among experimentalists, the following somewhat
detailed consideration of analogous comparisons for oligonucleotides is
warranted.
SALT-NUCLEIC ACID INTERACTIONS 693
For aqueous solutions containing ds B-oligonucleotides with / Z/ ranging
from 8 to 100 at a fixed phosphate concentration, eu 2.49 x 10-3 mol
=

dm - 3 , in the presence of a fixed concentration of added salt,


e3 1 .46 x 1 0- 3 mol dm- \ GCMC simulations of ion distributions were
=

calculated for the (b; / Z//8/a) model (23). Ion distributions near the oligo­
nucleotide were quantified in terms of CI"iI(a), the average over all axial
positions of the local counterion concentrations within an annular volume
extending 0 . 1 nm from a, the distance of closest approach of ions to the
surface of the (cylindrical model) oligonucleotide. (For shorter oligo­
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

nucleotides, whose length / Z / b is comparable to their diameter, 0.2 nm,


CI"iI(a) could be somewhat overestimated because the average taken in
evaluating this surface concentration neglects the ions in the capping
volumes extending outward into the solution beyond either end of the
by University of Windsor on 07/16/13. For personal use only.

model oligomer.) At a given / Z / and e3 , C,tia) is expected to differ from


81,IZI, because the radial thickness of the annular condensation layer
exceeds 0. 1 nm. However, the ratios RMC(Z) == Cit:,(a)!C,�,(a) and
RCc(Z) == 81,I ZI/()I,lool are more directly comparable as theoretical pre­
dictions of the deviation from the polyion limit of the cation accumulation
near an oligonucleotide as a function of /Z/ at fixed e3 • In fact, RMC(Z)
constitutes an upper bound on the number that would be computed by
integrating over the exact extent of the condensation layer. If C,Ha) and
Citola) had been evaluated from MC ion distributions by integrating over
a larger annular thickness (equivalent to that of the condensation layer),
the resulting ratio of these integrated ion densities would be even smaller
than that evaluated for the O. l -nm thickness, because at larger radial
distances from the surface of a rodlike oligoion variations in elHr) are
manifested over a longer distance along any line parallel to its axis, and
this effect becomes more accentuated at smaller values of / Z/. Moreover,
according to CC theory the counterion radial distribution within the
condensation layer is supposed to be uniform, whereas the trend predicted
by any alternative theoretical approach is for eI ZI(r) to decrease (not
necessarily monotonically) with increasing distance from the surface of
the oligoion. A final consideration bearing on the comparability of RMC(Z)
with RCC(Z) is that the GCMC simulations were performed (for the sake
of computational economy) on solutions where Cu 2.49 M and =

C3/Cu '" 0.6, whereas the CC theories explicitly consider only the case
of Cu --+ 0 and C3/Cu » 1 . However, the results of subsequent GCMC
simulations (20) predict that (at least in the concentration range of interest
here) CI�I(a) decreases with decreasing Cu at a given C3 and furthermore
that this decrease is more pronounced for smaller values of / Z/. Thus,
if RCC(Z) - RMC(Z) > 0, this difference actually underestimates the true
difference between the predictions of these different theoretical approaches
694 ANDERSON & RECORD

with respect to the importance of coulombic end effects on cation accumu­


lations near oligonucleotides.
From the results presented in Table 1 of Reference 23, RMC(lOO) 0.85
=

and RMC(lO) 0.06. (The latter value was obtained by linear interpolation
=

between points at I ZI = 8 and 12.) The corresponding ratios predicted at


C '" 1 0- 3 mol dm -3 by molecular CC oligo theory are (as shown in Figure
3
3 of Reference 77) Rcc(I OO) = 0.96 and RCC(1O) 0.5 1 . Similar, but
=

slightly higher ( � 1O%) predictions for these ratios can be obtained by


interpolating to C3 ", 1 0-3 mol dm-3 the results of ligand binding CC
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

theory presented in Table 3 of Reference 1 47. Thus, when C "" 10-3 mol
3
dm- 3 RMC(Z) is either significantly (at I ZI = 1 00) or substantially (at
1 ZI = 1 0) lower than RCC(Z). Because, for the reasons given above, RMC(Z)
is actually an upper bound on a ratio more exactly comparable with
by University of Windsor on 07/16/13. For personal use only.

RCC(Z), the unequivocal conclusion emerges that RMC(Z) approaches


unity much more slowly as a function of Z than does RCC(Z), at least
when C3 '" 1 0- 3 mol dm-3.
At higher salt concentrations the approach of RCC(Z) to 1 is predicted
to be complete for even smaller values of I ZI: RCC(lO) 0.92 at C = 0. 1
=

3
mol dm-3• Although the I ZI dependence of C,Ha) has not yet been tho­
roughly investigated by GCMC simulations at higher salt concentrations,
preliminary results (J Bond, unpublished data) do not give any indication
of a significantly more rapid approach of RMC(Z) to unity at higher C •
3
This inference also is in accord with recent PB calculations (8 1) showing
that for oligonucleotides with 20 < I Z I < 1 00 the axial range of the cou­
lombic end effect on ('I') at the surface of the model oligoion varies by
� 20% when the C (uniunivalent salt) is increased from 10-3 mol dm- 3
3
to 1 0 - 1 mol dm -3. Both the molecular and ligand binding versions of CC
theory do concur qualitatively with the GCMC simulations insofar as
RCC(Z), like RMC(Z), is predicted to approach unity as a linear function
of I Z I - I . However, the agreement mentioned explicitly in Reference 77
between the predictions of the molecular CC theory for oligonucleotides
and those of the earlier GCMC simulations (23) is restricted to the func­
tional form, but not the magnitude, of the variation in the average amount
of counterion accumulation with I Z I - I at fixed C •
3
The approach of RCC(Z) to unity actually may be accelerated by some
assumptions built into each of the adaptations of CC theory for rodlike
oligoions. To implement minimization of the free energy expression
assumed for a finite line of charges, molecular CC theory posits that all
of the discrete model charges along the axis are reduced by counterion
condensation to the same fraction. Consequently, coulombic end effects
on the axial distribution of condensed counterions are neglected, as if the
oligoion were a segment in the middle of a polyion. For sufficiently small
SALT-NUCLEIC ACID INTERACTIONS 695
1 21, e\,\Z\ does differ from e\,)oo), but this difference could be spuriously
small because the energetic inequivalence of charges along the axis of the
oligoion has been preaveraged. Similarly, in the formulation of the ligand
binding version of CC theory each of the sites on the oligoion is taken to
be equivalent and independent ( 147). The introduction of this binding
polynomial formalism was considered necessary to avoid the divergence
as C3 ...... 0 of the condensation volume, VCC,lzl' pertaining to a finite line of
I Z I charges. This problem was circumvented in the alternative adaption
of "molecular" CC theory (77) by assuming that for any value of 1 21,
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

Vee,lzl = Vee,lool' the salt-invariant condensation volume pertaining to the


corresponding polyion. The presumed 1 21 independence of Vee also could
produce a tendency in 0[,121 toward premature attainment of the polyion
limit.
by University of Windsor on 07/16/13. For personal use only.

The salt invariances predicted by molecular CC theory for both Vee,I "' 1
and 81 ,l eo l are consistent with analyses of NMR studies on various quadru­
polar cations in solutions of polyanionic DNA (2). However, when ana­
lyzed with the same set of assumptions, exactly analogous 23Na NMR
measurements (20) on an oligonucleotide (I Z I 38) indicate that rNa (38),
=

the extent of counterion accumulation per oligonucleotide charge, as


sensed by the NMR measurement, increases by at least 30% when the
ratio of sodium ions to phosphate monomers ([Na]/[PJ) is increased from
1 to 50 (20). Over this same range of salt concentrations the ration
rNa(38)!rNa(320) varies from 0,55 to 0.75. These observations can be
accounted for by GCMC simulations on the (b; 1 21!e!a) model but not by
the results of molecular CC theory for oligoions shown in Figure 3 of
Reference 77. The latter comparison may be viewed as inappropriate,
because the CC theory is formulated subject to the conditions [Na]/[P] » 1
and [P] ...... 0, which are inaccessible to NMR (and many other kinds of
experimental measurements). However, this concern has not prevented
frequent citations of analogous NMR studies of rodlike polyions (which
necessarily are conducted at [Na]/[P] ;:S 1 0) as supporting the qualitative
predictions of molecular CC theory (formally derived under the constraint
[Na]/[P] » I) that 8[,1 001 and Vee,1 I are insensitive to changes in C (74).
eo
3
End Effects on Processes Involving Oligonucleotides
CONFORMATIONAL TRANSITIONS Short two-stranded (dimer) duplexes and
one-stranded (hairpin) duplexes having the same number of phosphates
I ZI exhibit strikingly different values of dTm/dln a± . Dimer and hairpin
duplexes also exhibit different dependences of dTm!dln a± on I ZI. The
classic experiments of Baldwin and coworkers (2 1) on hairpin duplexes
of dTA oligomers (N-mers) demonstrated conclusively that dTm!dln Q ±
decreases with decreasing IZI. Analysis of experimental data on the salt
696 ANDERSON & RECORD

concentration dependence (10- 3 to 1 0 - 1 mol dm-3) of Tm of hairpin


oligonucleotides ( 1 8 < I ZI < 44) indicates that l �rlzl,ul is significantly less
than I �r l oo l,ul, the value determined for the denaturation of the cor­
responding polynucleotide, and that I �rlzl,u l approaches I �rlool,u l as a
linear function of 1 II ZI (23, 1 48), These thermodynamic coefficients exhibit
a significant coulombic end effect. In contrast, for the denaturation of short
oligonucleotide two-stranded helices (I ZI � 22) ( 1 49, 1 50), dTmldln a± has
been reported to be similar to the polymer value. According to Equation
8, these results imply that �rI Z I,u is similar to Llr1 oo l,u and hence that little,
if any, end effect is apparent.
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

The concept of thermodynamic association, and some explicit


expressions obtained by introduction of an assumed form of the end effect
into the thermodynamic limiting law version of CC theory, were shown
by University of Windsor on 07/16/13. For personal use only.

( 148) to account for the I ZI dependence of dTmldln a± for hairpin oligo­


nucleotides and to predict a smaller I Z I dependence of dTmldln a ± for
dimer duplexes, which at that time had not been investigated, Although
this limiting law approach correctly predicts the functional form and the
trends in the I Z I dependence of dTm/d ln a ± , its quantitative predictions
have been superseded by more recent MC calculations for the same model
(23, 24), In contrast, the duplex but not the hairpin denaturation data can
be rationalized, at least qualitatively, by an adaptation of CC theory (77)
that predicts the virtual absence of end effects on the extent of counterion
condensation surrounding short oligonucleotides, A corresponding
absence of end effects can be inferred as the CC prediction for thermo­
dynamic coefficients such as rIZI,u ' However, GCMC simulations on the
same model (b; I ZI/E/a) as that assumed by CC theory have shown that
both the hairpin and the dimer data are consistent with a pronounced
coulombic end effect on rlZI,u for nucleic acid oligomers (24), These theor­
etical calculations are in quantitative agreement with the experimental
findings that for d(TA) hairpin oligomers Llrl ZI,u decreases as a linear
-

function of 1 Z I - 1 and that the salt dependence of hairpin denaturation is


far less than that of denaturation of two-stranded helices for the same
helical length, The two-stranded oligonucleotides for which denaturation
data are available appear to fall in a range of lengths where substantial
coulombic end effects are masked by the reduction in I ZI per molecule
upon strand separation,

LIGAND BINDING To test the predicted importance of the polyelectrolyte


character of DNA as the principal determinant of the large effect of
uniunivalent salt concentration C3 on its binding interactions with oligo­
cations and other ligands, effects of varying C3 on the binding of a tryp-
SALT-NUCLEIC ACID INTERACTIONS 697
tophan-containing oligo lysine octacation, KWK6 (designated as V + ), to
polymeric [poly(dT)] and to oligomeric [dT(pdT)lO] single-stranded DNA
have been studied (W Zhang, unpublished data). Strengths and salt con­
centration dependences of these interactions are characterized experi­
mentally by Kobs and SKobs == dIn Kobs/d ln C3. When C3 ;S OJ mol dm- 3 ,
both Kobs and I SKobs I are significantly larger for binding of L8 + to a site
on poly (dT) than to dT(pdT) l O ' Because of the large differences between
SKobs associated with L8+ binding to oligomeric ( - SKobs = 3.6 ± 0 . 1 ) and
polymeric ( - SKobs 6. 1 ± 0. 1 ) DNA, the difference between the corres­
=

ponding values of Kobs increases dramatically as C3 is decreased (approxi­


Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

mately 80-fold near 0. 1 M salt). PB calculations on the I D coulombic­


primitive models for the ligand, the free oligo- or polynucleotide, and
their complex (J Bond, unpublished results) agree with values of SKobs
by University of Windsor on 07/16/13. For personal use only.

determined experimentally for both lengths of DNA. Thus, observed and


calculated differences in SKobs for V + binding to dT(pdT) l O and poly(dT)
demonstrate that inclusion of the flanking DNA charges, and therefore
the polyelectrolyte character of the DNA phosphate charge distribution,
profoundly increases both Kobs and I SKobs I at salt concentrations in the
usual experimental range.

ACKNOWLEDGMENTS

Research in this laboratory in the area of this review and preparation of


this review have been supported by NIH Grant GM3435 1 and the UW
Graduate School.

Any Annual Review chapter, as well as any article cited in an Annual Review chapter,
may be purchased from the Annual Reviews Preprints and Reprints service.
1-800-347-8007; 415-259-5017; email: arpr@c1ass.org

Literature Cited

1 . Lohman TM, Mascotti DP. 1992. cord MT Jf. 1 99 1 . J. Mol. Bioi. 222:
Methods Enzymol. 2 1 2: 400--24 28 1-300
2. Braunlin WHo 1 995. Adv. Biophys. 7. Record MT Jr, Anderson CF, Lohman
Chem. 1 5 : 89-139 TM. 1978. Q. Rev. Biophys. I I: 1 02-
3. Jayaram B, Beveridge DL. 1 996. Annu. 78
Rev. Biophys. Biomol. Struct. 25: In 8. Record MT Jr, Spolar RS. 1 990. In
preparation Nonspecific DNA-Protein Interactions,
4. Anderson CF, Record MT Jf. 1 990. ed. A Revzin, pp. 33-69. Florida: CRC
Annu. Rev. Biophys. Biophys. Chem. 1 9: 9. Record MT Jr, Ha JH, Fisher M . 1 99 1 .
423-65 Methods Enzymol. 208: 291-343
5. Frank-Kamenetskii MD, Anshelevich 1 0. Lohman TM. 1986. CRC Crit. Rev.
VV, Lukashin AV. 1 987. Sou. Phys. Biochem. 1 9: 1 9 1-245
Usp. 30: 3 1 7-30 I I. Anderson CF, Record MT Jr. 1993. J.
6. Cayley S, Lewis BA, Guttman HJ, Re- Phys. Chem. 97: 7 1 1 6-26
698 ANDERSON & RECORD

1 2 . Oosawa F. 1 957. J. Polymer Sci. 2 3 : 38. Jayaram B, Beveridge DL. 1 9 9 1 . J.


421-30 Phys. Chern. 95: 2506--16
1 3 . M anning GS. 1969. J. Chem. Phys. 5 1 : 39. Vlachy V, Haymet ADJ. 1986. J. Chern.
924-33 Phys. 84: 5874-80
1 4. Fuoss RM, Katchalsky A, Lifson S. 40. Eisenberg H. 1976. Biological Macro­
1 95 1 . Proc. Natl. A cad. Sci. USA 37: molecules and Polyelectrolytes in Solu­
579-89 tion, pp. 29-47. Oxford: Clarendon
1 5. Gross LM, Strauss UP. 1966. In 4 1 . Wyman J Jf. 1964. Adv. Protein Chem.
Chemical Physics of Ionic Solutions, ed. 1 9 : 223-86
BE Conway, RG Barradas, pp. 3 6 1 -89. 42. Schellman JA. 1978. Biopolymers 1 7:
New York: Wiley 1 305-22
1 6. Record MT Jr. 1 975. Biopolymers 14: 43. Schellman JA. 1 990. Biophys. Chem.
2 1 37-58 37: 1 2 1-40
1 7. Record MT Jr, Woodbury CP, Loh­ 44. Timasheff SN. 1 992. Biochemistry 3 1 :
man TM. 1 976. Biopolymers I S : 893-
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

9857--64
915 45. Record MT Jr, Richey B. 1988. In ACS
1 8. Record MT Jr, Lohman TM, Sourcebook for Physical Chemistry
DeHaseth PL. 1976. J. Mol. Bioi. 1 07: Instructors, ed. T Lippincott, pp. 145-
1 45-58 59. Washington, DC: Am. Chem. Soc.
1 9. Warshel A. 1 99 1 . Annu. Rev. Biophys.
by University of Windsor on 07/16/13. For personal use only.

46. Record MT Jr, Anderson CF. 1 995.


Biophys. Chem. 20: 267-98 Biophys. J. 68: 786-94
20. Stein V, Bond JP, Capp MW, Ander­ 47. Anderson CF, Record MT Jr. 1 982.
son CF, Record MT Jr. 1 995 . Biophys. Annu. Rev. Phys. Chem. 33: 1 9 1 -222
J. 68: 1 063-72 48. Senear DF, Brenowitz M , Shea MA,
2 1 . Scheffler IE, Elson EL, Baldwin RL. Ackers GK. 1 986. Biochemistry 25:
1 970. J. Mol. Bioi. 48: 145-71 7344-54
22. Deleted in proof 49. Spolar RS, Record MT Jr. 1 994. Sci­
23. Olmsted MC, Anderson CF, Record ence 263 : 777-84
MT Jr. 1989. Proc. Natl. A cad. Sci. 50. Anderson CF, Record MT Jr. 1983. In
USA 86: 7766-70
Structure and Dynamics: Nucleic Acids
and Proteins, ed. E Clementi, R Sanna,
24. Olmsted MC, Anderson CF, Record
MT Jr. 1 99 1 . Biopo/ymers 3 1 : 1 593-604
pp. 301-/9. New York: Adenine
5 1 . Stigter D. 1 975. J. Colloid Interface Sci.
25. Olmsted MC, Bond JP, Anderson CF,
Record MT Jf. 1 995. Biophys. J. 68:
53: 296-306
634-47
52. Jayaram B, Swaminathan S, Beveridge
26. Anderson CF, Record MT Jr, Hart PA.
DL, Sharp K, Honig B. 1990. Macro­
1978. Biophys. Chem. 7: 3 0 1 - 1 6
molecules 23: 3 1 56--6 5
2 7 . Chang S L , Chen S-H, Rill R L , Lin JS.
53. Sharp KA, Honig B. 1 990. J. Phys.
1 990. J. Phys. Chem. 94: 8025-28
Chel11 . 94: 7684-92
28. Wu CF, Chen S-H, Shih LB, Lin JS.
54. Reiner ES, Radtke CJ. 1990. J. Chern.
1 988. Phys. Rev. Left. 6 1 : 645-48
Soc. Faraday Trans. 86: 390/-12
29. Wensel TG, Meares CF, Vlachy V,
Matthew JB. 1986. Proc. Nat!. Acad. 55. von Hippel PH, Schleich T. 1969. In
Biological Macromolecules, Vol. 2.
Sci. USA 83: 3267-7 1
30. Wemmer DE, Srivenugopal KS, Reid Structure and Stability of Biological
BR, Morris DR. 1 985. J. Mol. BioI. Macromolecules, ed. SN Timasheff, G
1 85: 457-59 Fasman, pp. 4 1 7-574. New York:
3 1 . Shin YK, Hubbell WL. 1992. Biophys. Dekker
J. 6 1 : 1 443-53 56. von Hippel PH, Schleich T. 1 969. A ce.
32. Braunlin WH, Anderson CF, Record Chem. Res. 2: 257--65
MT Jr. 1 987. Biochemistry 26: 7724-3 1 57. Krakauer H, Sturtevant JM. 1968.
33. XU Q, Jampani SRB, Brau1in WHo Biopolymers 6: 49 1 -5 1 2
1993. Biochemistry 32: 1 1 754-60 58. Bond J , Anderson CF, Record MT Jf.
34. XU Q-W, Deng H, BraunIin WHo 1993. 1 994. Biophys. J. 67: 825-36
Biochemistry 32: 1 3 1 30-37 59. Sharp KA, Honig B. 1 990. Annu. Rev.
35. Paulsen MD, Richey B, Anderson CF, Biophys. Biophys. Chem. 1 9 : 301-32
Record MT Jf. 1988. Biopolymers 1 7 : 60. Conrad J, Troll M, Zimm BH. 1988.
1 249--65 Biopolymers 27: 1 7 1 1-32
36. Padmanabhan S, Richey B, Anderson 6 1 . Troll M, Roitman D, Conrad J, Zimm
CF, Record MT Jr. 1 988. Biochemistry BH. 1 986. Macromolecules 19: 1 1 86-94
27: 4367-76 62. Jayaram B, Sharp KA, Honig B. 1 989.
37. Mills P, Anderson CF, Record MT Jr. Biopolymers 28: 975-93
1 986. J. Phys. Chem. 90: 6541-48 63. Lukashin AV, Beglov DB, Frank-
SALT-NUCLEIC ACID INTERACTIONS 699
Kamenetskii MD. 1 99 1 . J. Biomol. 92. Pack GR, Wong L, Lamm G. 1 989. Int.
Struct. Dyn. 9: 5 1 7-23 J. Quant. Chem. 1 6: 1 - 1 5
64. Pack GR, Garrett GA, Wong L, Lamm 93. Lamm G, Wong L , Pack G R . 1 994 .
G. 1 993. Biophys. J. 65: 1 3 63-70 Biopolyrners 34: 227-37
65. Misra VK, Sharp KA, Friedman RA, 94. You TJ, Harvey SC. 1 993. J. Comp.
Honig B . 1994. J. Mol. BioI. 238: 245- Chem. 1 4: 484-501
63 95. Davis ME, McCammon JA. 1 990.
66. Bacquet RJ, Rossky PI. 1984. J. Phys. Chem. Rev. 90: 509-21
Chem. 88: 2660--69 96. Ptaszek LM, Vijayakumar S, Ravi­
67. Bacquet RJ, Rossky PI. 1 988. J. Phys. shanker G, Beveridge DL. 1 994. Bio­
Chem. 92: 3604-- 1 2 polymers 34: 1 1 45-5 3
68. York O M , Darden T, Deerfield D , 97. Swaminathan S, Ravishanker G,
Pedersen LG. 1 992. Int. J. Quant. Beveridge DL. 1 99 1 . J. Am. Chern. Soc.
Chern. Quant. Bioi. Syrnp. 1 9 : 1 4 5-{)6 1 1 3: 5027-40
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

69. Demaret GP, Gueron M. 1 99 3 . 98. Fritsch VG, Ravishanker G, Beveridge


Biophys. J. 65: 1 700-- 1 3 DL, Westhof E. 1 993. Biopoiyrners 33:
70. Valleau IP. 1 989. J. Chem. Phys. 1 29: 1 537-52
1 63-75 99. Laugh ton CA, Neidle N. 1992. J. Mol.
71. Fixman M. 1 990. J. Chem. Phys. 92 : Bioi. 223: 5 1 9-29
by University of Windsor on 07/16/13. For personal use only.

6283-93 1 00. Mohan V, Smith PE, Pettitt BM. 1 993.


72. Reddy MR, Rossky PI, Murthy CS. J. Phys. Chem. 97: 1 2984--90
1 987. J. Phys. Chem. 9 1 : 4923-33 1 0 1 . Beveridge DL, Ravishanker G. 1 994.
73. Manning GS. 1 977. Biophys. Chem. 7: Curro Opin. Struct. Bioi. 4: 246-55
9 5- 1 02 1 02. van Gunsteren WF, Luque FJ, Timms
74. Manning GS. 1 978. Q. Rev. Biophys. D, Torda AE. 1 994. Annu. Rev.
1 1 : 1 79-246 Biophys. Biomol. Struct. 23: 847-{)3
75. Murthy CS, Bacquet RJ, Rossky PJ. 1 03. Allen MP, Tildesley DJ. 1987. Com­
1 985. J. Phys. Chem. 89: 70 1 - 1 0 puter Simulation of Liquids. Oxford:
76. Mills P , Anderson CF, Record M T Jr. Clarendon. 385 pp.
1 985. J. Phys. Chern. 89: 3984-94 1 04. Kalos MH, Whitlock PA. 1 988 . Monte
77. Fenley MO, Manning GS, Olson WK. Carlo Methods, Vol. I, Basics. New
1 990. Biopolymers 30: 1 1 9 1-203 York: Wiley. 1 96 pp.
78. Fenley MO, Manning GS, Olson WK. 1 05. Olmsted MC, Hagerman PJ. 1 994. J.
1 990. Biopolymers 30: 1205- 1 3 Mol. Bioi. 243: 91 9-29
79. Fenley MO, Manning G S , Olson WK. 1 06. Hao MH, Harvey SC. 1 992. Macro­
1 992. J. Phys. Chern. 96: 3963--69 molecules 25: 2200--8
80. Fenley MO, Olson WK, Tobias J, Man­ 1 07. Rajasakaran E, Jayaram B. 1 994.
ning GS. 1 994. Biophys. Chem. 50: 255- Biopolyrners 34: 443-4 5
71 1 08. Manning GS. 1 990. See Ref. 1 5 1 , pp.
81. Allison SA. 1994. J. Phys. Chem. 98: 191-205
1 2091-96 1 09. Padmanabhan S, Paulsen M, Anderson
82. Klein BK, Pack G. 1 9 83 . Biopolymers CF, Record MT Jr. 1 990. In Mono­
22: 233 1-52 valent Cations in Biological Systems, ed.
83. Gilson M, Sharp K, Honig B. 1 988. J. C Pasternak, pp. 3 2 1 -38. Florida: eRC
Compo Chem. 9: 327-35 1 1 0. Fixman M. 1 979. J. Chern. Phys. 70:
84. Honig B, Sharp K, Yang AS. 1993. J. 4995-5005
Phys. Chern. 97: 1 1 0 1 -9 1 1 1 . Chen SWW, Rossky Pl. 1 993. J. Phys.
85. Oberoi H, AlIewelI N. 1 993. Biophys. Chern. 97: 10803- 1 2
J. 65: 48-55 1 1 2. Chen SWW, Rossky P l . 1993. J. Phys.
86. Pack GR, Klein B. 1 984. Biopolymers Chent. 97: 6078-82
23: 2801-23 1 1 3 . Forester TR, McDonald IR. 1 99 1 .
87. Lamm G, Pack GR. 1 990. Proc. Nat/. Mol. Phys . 72: 643-{)0
Acad. Sci. USA 87: 9033-36 1 1 4. Guldbrand LE, Forester TR, Lynden­
88. Pack GR, Lamm G. 1 993 . Int. J. Quant. Bell RM. 1 989. Mol. Phys. 67: 473-93
Chem. Quant. Bioi. Symp. 20: 21 3-30 1 1 5. Manning GS. 1972. Annu. Rev. Phys.
89. Pack GR, Wong L, Prasad CV. 1986. Chern. 23: 1 1 7-40
Nucleic Acids Res. 14: 1479-93 1 1 6. Stigter D. 1978. J. Phys. Chem. 82:
90. Pack GR, Prasad CV, Salafsky JS, 1 603-6
Wong L. 1986. Biopolymers 25: 1 697- 1 1 7. Marcus RA. 1 955. J. Chern. Phys. 23:
715 1 057-{)8
91. Pack GR, Lamm G , Wong L , Clifton 1 1 8. Paulsen MD , Richey B, Anderson CF,
D. 1 990. See Ref. 1 5 1 , pp. 237-46 Record MT lr. 1 987. Chent. Phys. Lett.
700 ANDERSON & RECORD

1 39: 448-52. Erratum: Chem. Phys. 1 35. Mossing MC, Record MT Je. 1 985. J.
Lett. 1 43: 1 1 5 Mol. BioI. 1 86: 295-305
1 1 9. Filimonov VV, Privalov PL. 1 978. J. 1 36. Senear DF, Batey R. 1 99 1 . Bio­
Mol. Bioi. 1 22: 465-70 chemistry 30: 6677-88
1 20. Record MT Jr, Mazur SJ, Melancon P, 1 37 . Koblan KS, Ackers GK. 1 99 1 . Bio­
Roe J-H, Shaner SL, Unger L. 1 98 1 . chemistry 30: 7822-27
Annu. Rev. Diochem. 50: 997-1 024 1 38. Lesser DR, Kurpiewski MR, Jen­
1 2 1 . Manning GS. 1 972. Biopolymers 1 1 : Jacobsen L. 1 990. Science 250: 776-85
937-49 1 39. Misra VK, Hecht JL, Sharp KA, Fried­
1 2 1 a. Manning GS. 1 972. Biopolymers I I : man RA, Honig B. 1 994. J. Mol. Bioi.
95 1-55 238: 264-80
1 22. Latt SA, Sober HA. 1 967. Biochemistry 1 40. Jayaram B, DiCapua FM, Beveridge
6: 3293-306 DL. 1 99 1 . J. Am. Chem. Soc. 1 1 3:
1 23 . Mascotti DP, Lohman TM. 1 990. Proc. 52 1 1-1 5
1 4 1 . Beamer LJ, Pabo CO. 1 992. J. Mol.
Annu. Rev. Phys. Chem. 1995.46:657-700. Downloaded from www.annualreviews.org

Natl. Acad. Sci. USA 87: 3 1 42-46


1 24. Mascotti DP, Lohman TM. 1 992. Bio­ Bioi. 227: 1 77-96
chemistry 3 1 : 8932-46 142. LeBret M, Zimm BH. 1 984. Bio­
1 25. Mascotti DP, Lohman TM. 1 993. Bio­ polymers 23: 27 1 -85
chemistry 32: 10568-79 143. Torrie GM, Valleau JP. 1 980. J. Chem.
by University of Windsor on 07/16/13. For personal use only.

1 26. Hopkins HP, Wilson WD. 1 987. Bio­ Phys. 73: 5807- 1 6
polymers 26: 1 347-55 144. Mills PA, Rashid A, James T L . 1 992.
1 27. Laugaa P, Markovits J, Delbarre A, Le Biopolymers 32: 1 49 1 -501
Pecq J-B, Roques BP. 1 985. Bio­ 145. Beveridge DL, Di Capua FM. 1 989.
chemistry 24: 5567-75 Annu. Rev. Biophys. Biophys. Chem. 1 8:
1 28. Braunlin WH, Strick TJ, Record MT 431-92
Jr. 1982. Biopolymers 2 1 : 1 30 1-14 146. Jayaram B, Anaja N, Rajasakaran E,
129. Plum GE, Bloomfield VA. 1 988. Bio­ Arsra V, Das A, et al. 1 994. J. Sci.
polymers 27: 1 045-5 1 Indust. Res. 53: 88-105
1 30. Wilson WD, Tanious FA, Barton HJ, 147. Dewey TG. 1 990. Biopolymers 29:
J ones RL, Fox K, et al. 1 990. Bio­ 1 793-99
chemistry 29: 8452-6 1 148. Record MT Jr, Lohman TM. 1 978.
1 3 1 . Breslauer KJ, Remeta DP, Chou W -Y, Biopolymers 1 7: 1 59-66
Ferrante R, Curry J, et al. 1 987. Proc. 149. Braunlin WH, Bloomfield VA. 1 99 1 .
Natl. Acad. Sci. USA 84: 8922-26 Biochemistry 30: 754-58
1 32. Marky LA, Bres1auer KJ. 1 987. Proc. 1 50. Williams AP, Longfellow CE, Freier
Natl. Acad. Sci. USA 84: 3 1 42-46 SM, Kierzek R, Turner DH. 1 989. Bio­
1 33 . Katoh T, Ohtsuki T. 1 982. J. Polymer chemistry 28: 4283-9 1
Sci. 20: 2 1 67-75 1 5 1 . Beveridge DL, Lavery R, eds. 1 990.
\ 34. Zacharias M , Luty BA, Davis ME, Theoretical Biochemistry and Molecular
McCammon lA. 1992. Biophys. J. 63: Biophysics: A Comprehensive Survey.
1 280-85 New York: Adenine

You might also like