Date: Tue, 13 Feb 2001 10:22:54 -0800 (PST)
From: Thomas Cheatham
Subject: Re: Ionic solutions


Sorry for the long followup on this topic, but it hits a big interest of
mine and may provide useful information for the whole reflector, and note
that these are just my *opinions* on the subject...

On Mon, 12 Feb 2001, Marios Philippopoulos wrote:

> After thinking this over, I'm still a bit skeptical in cases where
> ions are used only for the purpose of protein system
> electroneutrality, ie. low ionic strength situations. In the case of a
> protein placed in the environment of a few (say 4) ions, would these
> ions prefer to be surrounded by solvent or be attached to the protein?
> I'm inclined towards the former, but in any case, this cannot be quite
> the same situation as that of DNA solutions and the high ionic
> strength environments they entail.

Rebecca Wade's group has done some study of this issue and suggests that
"net-neutralizing" ions alone may lead to distortion of a (metastable)
protein structure (WW domain) in short simulations, distortion that is
not seen when higher salt concentrations are used.

See: Ibragimova GT and Wade RC "Importance of explicit salt ions for
protein stability in molecular dynamics simulation"
Biophys. J. 74, 2906-2911 (1998)

In trying to reproduce this (unpublished), it appears that with longer
initial equilibration of the water/salt (to reduce bias of the initial
placement), stable simulations of this WW domain can be obtained (with
~500 ps of water/salt equilibration) using the same protocol.

Another paper discussing this issue is:

Marti-Renom MA et al, "Effects of counterions and volume on the
simulated dynamics of solvated proteins. Applications to the activation
domain of procarboxypeptidase B" Prot. Eng. 11, 881-890 (1998).


In DNA simulation, early one we and other groups investigated the effects
of counterion placement and concentration on nucleic acid structure.
With DNA, it appears that the initial placement does not have such a
critical effect the structure or dynamics (with the exception that bound
ions can have long lifetimes and therefore any discussion of a specific
role of bound ion should be tempered by a discussion of the initial ion
placement, i.e. did the ion diffuse into this "strong binding pocket" on
its own or was it initially placed there?).

An *incomplete* list of references discussing the effects of ion/solvent
concentration and equilibration are:

Norberto de Souza O and Ornstein RL "Effect of periodic box size on
aqueous molecular dynamics simulation of a DNA dodecamer with
particle-mesh Ewald method" Biophys. J. 72, 2395-2397 (1997)

[shows that various box sizes from 5A to 15A of water surrounding the
DNA doesn't effect structure much; however when using less water be
aware of potential artifacts from periodicity; see work by
Hunenberger/McCammon]

Norberto de Souza O and Ornstein RL "Effect of warmup protocol and
sampling time on convergence of molecular dynamics simulations of a DNA
dodecamer using AMBER 4.1 and Particle-Mesh Ewald
method" J. Biomol. Struct. Dyn. 14, 607-611 (1997).

[suggests that the precise details of equilibration at not so
critical]

Cheatham TE III and Kollman PA "Molecular dynamics simulations of
nucleic acids in solution: How sensitive are teh results to small
perturbations in the force field and environment?" In Structure,
Motion, Interafction and Expression of Biological Macromolecules,
Proceedings of the 10th conversation, Albany, p 99-116 (1998).

[1 ns simulations in no salt, net-neutralizing and 1M all give similar
results. Also discusses the effect of changing box size,
parameters, etc. Currently we have studies reinvestigating this
issue of salt and box size in longer simulations...]

Young MA, Ravishanker G and Beveridge DL "A 5-nanosecond molecular
dynamics trajectory of B-DNA: Analysis of structure, motions and
solvation" Biophys. J. 73, 2313-2336 (1997).

[simulations are run with various ion placement schemes ranging from
MC to electrostatic potential. See also Young/Beveridges work in
JACS on specific ion association in the minor groove that predates
some of the more recent experimental work on this topic]


As multiple people have mentioned, equilibration is the key. I have noted
in recent studies of mine that, for example in producing ion-induced
bending results seen in phased A-tract DNA, that it may take upwards of
5ns before specific association of ions and DNA bending is seen. In
simulations of proteins I've done, ~500 ps of salt/water equilibration
was necessary to remove initial bias. Similarly, in high salt simulations
of DNA (> 4M), long equilibration times are necessary. As the amount of
salt increases or as the charge or lifetimes of bound ions increase,
longer equilibration is clearly necessary. At the same time, DNA itself,
with current force fields and proper treatment of the long range
electrostatics is remarkably stable and appears not particularly sensitive
to the initial ion placement (and even NO salt! or very low salt) such
that this issue is not critical for many applications (unless you are
trying to analyze the effect of specific ion association).

I have used multiple procedures for placing ions, ranging from the LEaP
ESP to hand randomization by picking random waters and editing the PDB to
turn a water into an ion, to automated methods in CHARMM involving random
placement (not overlapping solute) to cruder ESP methods (which use short
cutoffs on the interactions and are therefore more sloppy in placement)
and in each case, what is most important is that I am not biasing the
initial simulation and that long enough equilibration is performed rather
than the specific details of the procedure. The potential for bias is of
critical importance in current simulations that attempt to understand the
role of specific ion interaction on DNA structure and dynamics...


Thomas E. Cheatham, III