|
Anti-viral
therapy for HIV is one of the most dramatic success
stories in modern medicine, but it is far from the cure
that researchers originally hoped for. Scientists keep
aiming for a brand-new approach that can either truly
cure the disease, or at least make treatment less costly
and complex.
Debananda
Das of the National Cancer Institute’s HIV & AIDS
Malignancy Branch, Experimental Retrovirology Section,
is one of those researchers. He’s been focusing on an
intriguing target called CCR5. This receptor lets HIV
into cells, where it begins its ruthless cycle of
replication, causing lifelong infection. CCR5 came into
the limelight when scientists realized that about 1% of
people have certain mutations to this receptor that
render them essentially immune to HIV infection. Even
when people with these mutations are exposed to the
virus, they often do not stay infected.
Inhibiting
CCR5 could thus be a neat way to bar HIV from entering a
patient’s cells in the first place. Das and his
colleagues, in the group of Hiroaki Mitsuya, had a
couple of candidate molecules for this in hand. But they
knew they’d have to make adjustments even to those if
they wanted a powerful and very specific inhibitor.
Unfortunately, CCR5 is a G-protein coupled receptor (GPCR).
Hundreds of this class of protein molecules exist, but a
crystal structure (with good resolution) has been solved
only for one-bovine rhodopsin. Structure-based design is
tough even when a detailed picture of the actual target
structure is available. When it isn’t, it’s a shot
in the dark.
Enter
predictive modeling. While they knew they’d have to
use modeling to get to the target’s structure, Das and
his colleagues wanted to maximize their chances of
success. “With traditional modeling, you end up with a
lot of false positives,” says Das. “The model tells
you something is likely to work, but it really
doesn’t.”
And
bovine rhodopsin is not the ideal starting point for
figuring out what CCR5 really looks like: The two
proteins have only 20% sequence homology. But, “If it
is a GPCR, you already know it has seven transmembrane
helices, so the basic fold is not the issue,” says Das.
“The question is the side chains, and what directions
they are going in.”
Piecing
Together a Protein
It
helped a lot that Das’ collaborator, Kenji Maeda, had
already generated a lot of site-directed mutagenesis
data: Maeda introduced mutations into CCR5, and then
studied how different inhibitors bound to it, and how
their binding changed from one mutation to the next (see
this issue’s cover photo). “That tells you which
amino acids are important,” Das says. “It does not
tell us which atom in the residue is interacting with
given atoms in the inhibitor.” That’s where the
modeling comes in.
To
improve their chances of success, the NIH group used
three inhibitors that were structurally quite different,
which gave them a wide range of data. They also combined
homology modeling with a novel Induced Fit solution from
Schrödinger, which combines its Prime and Glide
programs for predicting ligand-induced conformational
changes in receptors. (See “Structure Prediction’s
New Frontier,” p.
23, for more.) “The algorithm iteratively refines the
orientation of the side chains,” explains Das.
The
group was pleased with their results (J. Biol. Chem.
2006; 281:12688–12698). “It worked well,” he says.
“We used the site-directed mutagenesis data to help us
get rid of the false positives.” After building a
model of CCR5, the scientists predicted some mutations
that would change the binding affinity of inhibitors.
Most of their predictions proved correct when compared
with the site-directed mutagenesis results.
“Now
we think we have a good structural model,” says Das.
“It is almost like actually having an experimentally
refined structure. We are at the same point in our work
where we would be if someone had crystallized the
protein.”
Das
emphasizes that they still have a long way to go. But
because CCR5 is a relatively new target, this type of
progress is still very encouraging. “The site-directed
mutagenesis data are expensive and time consuming to
get,” he notes. “These are challenging
experiments.” But taking the time to do them, and
working out a better model of the protein structure,
should help speed their work from here.
|
