As scientists, we occasionally need to step back and scrutinize
important issues within developing scientific fields. Molecular dynamics simulations of various G
protein-coupled receptors (GPCRs) have attempted to find the key conformations
of these receptors that describe the activation and modulation of their experimentally
observed responses. Although this quest has been pursued for over half a
century, most current articles on this subject stress that simulations have made
great progress, but that much more needs to be done to accurately model drug-receptor
interactions.
Although thousands of crystal structures have been deposited
in databanks, such as the Protein Data Bank (PDB), with consistently better
resolutions, the conditions that led to their successful crystallization are vastly
different from their in vivo environments. This naturally raises
questions about the relevance of these structures for our understanding of
their natural functioning states. These artificially produced crystal structures
are far different than the GPCR’s transmembrane embedded, pH-dependent and
REDOX (Reduction (RED) and Oxidation (OX)) sensitive states in vivo. After decades of work, we’re not much closer
to discovering GPCR’s natural modes of activation that necessarily include these
pH-dependent and REDOX sensitive states for these fascinating proteins.
If one critically examines the dogma that has crept into the
field of molecular simulations, then one must begin to question some of that
dogma. For instance, the dogma surrounding the essential disulfide bond
(linking together two cysteine amino acids by their thiol or sulfhydryl groups)
found in many GPCRs, suggests that those studying the molecular dynamics of
these proteins may have forgotten some of their biochemistry concerning the
stability of disulfide bonds in vivo. Many of these disulfide bonds are
easily broken and reformed under natural conditions. In addition, many GPCRs
such as rhodopsin contain an odd number of extracellular cysteines so that
there may always exist at least one free cysteine that remains reactive. A free
cysteine also accounts for the pH-dependence and REDOX properties
experimentally associated with many GPCRs.
There have been relatively few attempts to model these extracellular
cysteines in their free acid (SH) and base (S-) states. The complexities of
cysteine sulfhydryl chemistries in vivo add multiple layers of complexity
and confusion when ascertaining the functions of GPCRs. Even the seemly simple
treatment by Dithiothreitol (DTT) to liberate the two cysteine thiols, or sulfhydryls,
from a disulfide bond, may also block their subsequent reactions with other drugs
or ligands in binding or activation assays. In addition, once the sulfhydryl
groups are free, their pKa, which determines when the free thiol, or SH, group
becomes deprotonated, may vary over a very wide range depending on the polarization
of neighboring groups and the surrounding membrane charges. These free thiols,
or sulfhydryls, are also sensitive to oxidation under normal atmospheric
conditions, which greatly complicates the experimental study of receptors under
normal laboratory conditions.
Least we become too confident that our simulations are
completely accurate and predictive, many molecular dynamic simulations are done
with the hope of discovering unique conformational changes, or “mechanistic
hypotheses”, for receptor activation, but this approach may be digging a much
deeper hole than initially intended. The almost endless search for better, more
meaningful simulations with more powerful computers, stretches the horizon of
drug discovery into vastly more complex simulations with lipids, water,
counter-ions and other proteins necessary for receptor activation. These
endless cycles of refinement and simulation don’t provide us with a good model
for the necessary molecular switch between an off and on state for receptor
activation. We’re usually happy if we can see clear conformational differences between
the agonist and antagonist binding with their targeted receptor molecules, but
what about the partial agonists, allosteric modulators, inverse agonists, rapid
desensitization and tachyphylaxis? These present enormous challenges to our
present simulation methods. Many laboratory experiments are and were previously
done with chiral mixtures of enantiomers, whereas, the molecular dynamic
simulations usually use only one enantiomer that is considered the most active.
What are the implications comparing these mixtures used for laboratory experiments
versus molecular dynamic simulations using only one enantiomer?
Similar criticisms can be made about QSAR (Quantitative
Structure-Activity Relationship) studies of biological molecules, because they
assume that the observed structure and properties that are modeled will help to
predict the molecular behaviors. The idea that structure and its accompanying
properties can predict function is appealing, but fraught with difficulties. We’re
tempted by our ability to make molecular and protein structures with properties
such as charges. We examine these charge patterns to see what we can learn
about their reactive properties and then make inferences about how they react
with another molecule, but in the environment of a cell, these charges may be
fleeting or nonexistent if they are surrounded by a sea of catalytic molecules
such as enzymes that can add or remove groups such as a phosphate or a plethora
of many other functional groups.
We know the structures and properties of buildings and cars,
but their functions depend upon many other variables such as their surrounding
environments and the people who occupy those buildings and cars. Similarly, biological
proteins and molecules are embedded within a cellular milieu with many other
molecules that act as cofactors, energy providers, anchors, etc. These exist in
a range of environments that have varying charges, fields, lipophilicities, gradients,
etc. The whole is much greater than the sum of its parts when it comes to
understanding the functioning of biological molecules.
However, as hard-nosed scientists, we should ask the
toughest questions, which suggests that we should ask what molecular mechanisms
might function as a distinct on and off switch for the GPCRs? We know that
something like a chemical, net shift must drive receptors from their inactive
to active states. Current debates center around conformational selection versus
conformational induced fit. The presence of constitutively active receptors due
to receptor overexpression suggests that conformational selection may be the
preferred mechanism for receptor activation. Alas, this doesn’t tell us what
that active conformation is. Because the net binding energies of many drugs, or
ligands, are similar in magnitude to the background thermal noise at normal
temperatures, “mechanistic hypotheses” don’t describe a clean off and on switch.
The primary criticism being that there are no distinct boundaries between the
on and off states that determine the extent of movements of any residue, helix,
beta sheet, or loop that are necessary to select the active conformational
state.
A better model for the molecular on and off switch would be something
such as a distinct change from an acid to base state that would also be
accompanied by an electrostatic change within the receptor. This would also be
consistent with the experimental observations that many GPCRs show an increase
in their activities at higher pH levels. There’s an ongoing need to critically
analyze and incorporate much more of the available and reliable experimental
data into our molecular dynamic simulations.
With the enormous scientific talent out there, we can and will
explore more productive models for receptor activation. Having any model that provides
some possible insights into the functions of GPCRs, may be pleasing, but we must
not abandon many years of previous experimental observations that have been
repeatedly checked and verified.
We may not be making more timely progress because the
biochemists and enzymologists aren’t communicating enough with the
pharmacologists and biophysicists and those who perform theoretical
simulations. By combining our collective
expertise and maintaining a skeptical, but open mind, we will greatly enhance
our understanding of how our sensory and drug-targeted receptors function.
Richard G. Lanzara, MPH, Ph.D.
