BioSystems 59 2001 27 – 34
A computational model of membrane lipid electronic properties in relation to neural signaling
Harry L. Price
a,
, Ron Wallace
b,1
a
Department of Chemistry, Uni6ersity of Central Florida, Orlando, FL
32816
-
2366
, USA
b
Department of Sociology and Anthropology, Uni6ersity of Central Florida, Orlando, FL
32816
-
1360
, USA Accepted 24 September 2000
Abstract
We present a computational model of a transiently-organized neural membrane molecular system with possible information-processing capacity. The model examines field-induced dipole and quadrupole moments and polarizabil-
ity in monomeric, dimeric, and trimeric ethenes. Polarization of the ethenes is strongly indicated. This result is interpreted as a significant electronic feature of a molecular computing system based on organization of membrane
lipids into a transient 10
− 4
s crystalline state due to lipid-protein hydrophobic mismatch at the membrane-ion- channel interface. Predictive implications of the model’s electronic features are briefly discussed. © 2001 Elsevier
Science Ireland Ltd. All rights reserved.
Keywords
:
Hydrophobic mismatch; Neural membrane; Lipids; Computational modeling; Molecular computing; Spectroscopy www.elsevier.comlocatebiosystems
1. Introduction
Polyunsaturated fatty acids PUFAs consist of a phosphorus head group joined to two hydrocar-
bon chains containing at least one ethene bond CC. These molecules are present in all biologi-
cal membranes including that of the neuron. The degree of unsaturation varies markedly but the
most common ratios are 18:1, 18:2, 18:3 and 20:4, where the first number designates the chain length
and the second indicates the number of ethenic bonds Gennis, 1989. Since nearly all the ethenes
are cis, there is a kink in the molecule, which prevents orderly packing of bilayer diacyls. This
feature, in combination with the lack of a func- tional explanation for lipid hetereogeneity \ 100
species in any membrane has suggested that biomembrane molecular structure is stochastic.
The Singer – Nicolson ‘fluid-mosaic model’ of the membrane Singer and Nicolson, 1972 construes
the structure as a homogeneous 2-D liquid in which the constituent molecules have considerable
lateral mobility. However, more recent studies suggest that biomembranes are liquid crystals in
which protein-associated lipid microdomains of 10
8
– 10
10
molecules with lateral lengths of 10 – 300 A
, and typical associative lifetimes of 10
− 4
s
Corresponding author. Fax: + 407-823-2252. E-mail addresses
:
hpricepegasus.cc.ucf.edu H.L. Price, rwallacepegasus.cc.ucf.edu R. Wallace.
1
Fax: + 407-823-3026. 0303-264701 - see front matter © 2001 Elsevier Science Ireland Ltd. All rights reserved.
PII: S0303-26470000137-4
play important regulatory roles in membrane per- meability to Na
+
and other ionic species Alfsen, 1989; Mouritsen and Jørgensen, 1992; Jørgensen
and Mouritsen, 1995; Marsh, 1995; Simons and Ikonen, 1997. This unordered-to-ordered phase
transition may also affect formation of exocytotic vesicles as well as voltage- and ligand-gated chan-
nel activity, all of which are important in nerve cell communication. Evidence from myocyte and
neural membranes suggests that these microdo- main functions may be modulated by PUFAs in a
dose-dependent fashion Vreugdenhil et al., 1996. Increased membrane omega-3 PUFA concentra-
tion in guinea pig cardiac myocytes reduces action potential duration while prolonging its duration
in rats. PUFA-concentration-dependent shifts to more hyperpolarized potentials have been demon-
strated in vitro for rat hippocampal CA1 neurons.
These molecular data appear consistent with neurobiological evidence for axonal and dendritic
conduction block documented in a wide variety of species leech, crayfish, rabbit, rat. In this well-
known but poorly understood phenomenon, a neural impulse, once initiated, is not inevitably
propagated in a cable-like manner to the presy- naptic terminal Krnjevic´ and Miledi, 1959; Par-
nas, 1972; Swadlow and Waxman, 1976. The evidence suggests that subneural molecular sys-
tems may ‘decide’ at several stages of impulse propagation, whether or not the impulse should
be ‘allowed’ to continue. Put somewhat differ- ently, the data suggest that the neuron is not a
single switch, as postulated originally in the Hodgkin – Huxley model and more recently in the
McCulloch – Pitts MP heuristic utilized in artifi- cial intelligence McCulloch and Pitts, 1943;
Hodgkin and Huxley, 1952 but may in fact be ‘a series of switches’ Scott, 1995. Based on these
findings, a number of possible mechanisms by which membrane electronic properties could mod-
ulate microdomain functions, and ultimately regu- late neuron electrical activity, have been examined
through simulation studies and analyses of artifi- cial membranes Kinnunen and Virtanen, 1986;
Wallace et al., 1998; Wallace and Price, 1999.
In this article, we present a computational model of molecular computing in the neural mem-
brane. The most significant feature of the model is a two-stage potential energy search of protein-as-
sociated membrane phospholipids. The first stage of the search is initiated by ligand or voltage
gating of the ion-channel protein, a perturbation which generates protein conformational change
from a-helix to random coil, in turn producing lipid-protein hydrophobic mismatch Mouritsen
and Bloom, 1993; Sperotto and Mouritsen, 1993; Mouritsen and Jørgensen, 1994; Lehtonen et al.,
1996; Lehtonen and Kinnunen, 1997; Mouritsen and Jørgensen, 1997; Mouritsen, 1998; Sabra and
Mouritsen, 1998; Killian, 1998. As a consequence of the mismatch, membrane lipids self-assemble
such that lipids with hydrophobic lengths most closely approximating that of the protein become
more abundant in the protein’s vicinity Fig. 1. The effect of the self-assembly in combination
with the lipid-condensing effect of cholesterol is the alignment of lipid ethenes in the plane of the
membrane bilayer Hyslop et al., 1990; Raffy and Teissie´, 1999. Subsequent permeant ion move-
ment through anion and cation specific channels generates a field orthogonal to ion movement,
which polarizes the aligned ethenes. Ethene reor- ganization due to dipole – dipole interactions con-
stitutes the second stage of the potential energy search. When the difference
E
c
− E
m
=d
MIN
where, E
m
is the potential energy of the lipid microdomain, E
c
is the potential energy of the ion-channel protein in the random coil conforma-
tion, and d
MIN
is the minimal energy difference
Fig. 1. Schematic illustration of membrane lipid selectivity resulting from lipid-protein hydrophobic mismatch. Shaded
area indicates protein hydrophobic region.
required to close the channel pore the system is at the threshold value. Following cessation of ion
movement and concomitant neutralization of membrane ethenes, the channel protein relaxes to
the a-helix conformation which restores hydro- phobic matching Leuchtag, 1994. In this man-
ner, microdomain dynamics regulate the duration of channel opening, a model consistent with in
vitro experimental manipulation of ion channel activity by variations in concentration of unsatu-
rated membrane lipids Vreugdenhil et al., 1996. We describe the above process in terms of
biomolecular computing Wallace and Price, 1999, i.e. as a mechanism by which the frequency
coding of classical neural networks is regulated by molecular minimum potential energy searches
Conrad, 1992. Finally, we propose experiments involving Raman and fluorescence resonance en-
ergy transfer FRET spectroscopy as a means of investigating the above features in a liposomal
system.
2. Computational modeling of membrane ethene system stability and polarization