Biology Reference
In-Depth Information
modify existing protein
protein complexes in a rational fashion would thus endow the
synthetic biologist with the capability to change cellular behavior at will. Potential synthetic
biology applications include the rewiring of signal-transduction pathways to turn on a
reporter gene in response to an environmental stimulus, or the design of proteins that bind
to functional sites on (and thus inhibit) target proteins.
In this section we will describe methods that are commonly used for the design of
protein
protein interactions, and introduce examples of several studies done in this regard
so far. The goals in CPD of protein interactions can broadly be divided into two areas:
redesigning existing interactions towards higher affinity or changed specificity; and the
design of novel binding interactions. Depending on the specific problem, either the
sequence of both binding partners in the complex may be modified by the design algorithm
(
), or the algorithm is only allowed to design the sequence of one partner
while leaving the other partner constant (
two-sided design
'
'
one-sided design
).
'
'
Computational Redesign of ProteinProtein Interactions
There are usually two motivations to redesign an existing protein
protein interface: (1)
increasing the affinity to make the interface more stable; and (2) changing the specificity of
the interface, meaning to redesign the interface in such a way that the affinity of one pair of
desired binding partners is retained, while the affinity towards another, competing binding
partner is reduced. In both cases, a structural model of the to-be-redesigned complex,
preferably a crystal structure, needs to be available, since this serves as an input for the CPD
calculations. When the goal is to increase the affinity, the computational workflow used is
usually some iterative combination of rigid-body docking algorithms developed for virtual
protein docking
14
and general side-chain placement algorithms, while also taking into
account a set of empirical rules governing affinity. When the goal is to modify specificity,
these docking and side-chain placement algorithms need to be augmented by specialized
algorithms that take into account and penalize the competing states.
105
There are several representative examples of increasing affinity by computational design.
Roberts et al.
15
presented a study where a peptide inhibitor of a PDZ domain involved in
cystic fibrosis was redesigned for higher affinity towards its target. Starting from an NMR
structure of the natural ligand
PDZ domain complex, three mutations were introduced into
the sequence to obtain a hexameric inhibitor that had 170-fold increased activity compared
to the natural ligand. Lippow et al.
16
applied computational design to the problem of
antibody affinity maturation, which is a field of broad therapeutic significance. In their
work, the authors increased the affinity of two antibodies: one, a lysozyme-binding model
antibody, by 140-fold through mutation of four residues; the other, an epidermal growth
factor receptor binding therapeutic antibody, by 10-fold through mutation of three residues.
Computational design was also used by Haidar et al.
17
to introduce four mutations into a
solubilized T-cell receptor, increasing the affinity toward its cognate peptide MHC complex
by 100-fold. The redesigned receptor is potentially better suited than the wild-type for
diagnostics applications.
There has also been significant progress in the field of specificity redesign in the last several
years. Perhaps the most challenging aspect of this problem from the standpoint of
computational design is the need to incorporate
into the calculation,
meaning to consider the unwanted, competing stages and design a sequence that disfavors
these. Another side effect of this requirement is that the designed sequences might not have
the highest possible affinity against the target of interest, since the identities of the interface
positions are not just determined by how well they interact with the target state, but also
how well they discriminate against the unwanted states. In virtually all CPD algorithms, the
designed sequence is a product of the side-chain placement algorithm, but most of these
algorithms, i.e. Monte Carlo schemes and FASTER (see above) have been developed to
negative design
'
'
