Biology Reference
In-Depth Information
substantially reorganized, as reflected by genetic interac-
tion changes
[151,198,199]
.
By overlaying transcriptome patterns with a binary
interaction network, Han and colleagues discovered two
types of highly connected hub protein. On the one hand,
'party' hubs are strongly co-expressed with most of their
interacting partners; on the other hand, 'date' hubs connect
to different partners at different times or contexts
[200]
.
These interactome dynamics differences are reflected in the
structures of hub proteins from the two groups. Relative to
date hubs, where different interacting partners may utilize
the same surface of interaction at different times, party hubs
tend to contain less disordered regions and to display more
interaction interfaces at their surface, as would be expected
for proteins with many simultaneous interacting partners
[201
Evolutionary Dynamics of Protein
Protein
e
Interactome Networks
A central hypothesis of systems biology is that genoty-
pe
phenotype relationships are mediated through physical
and functional interactions between genes and gene prod-
ucts that form intricate molecular networks within cells.
Genotype
e
phenotype relationships are also governed by
natural selection. Hence, understanding the principles
driving the evolution of molecular networks would
contribute to a deeper understanding of Life. Is there a 'core
interactome' shared by every form of Life? Do the
constraints of natural selection enforce constraints on
interactome network structure? Do interactomes grow over
evolutionary time? Does interactome complexity scale with
organism complexity? Are interactomes more stable or
more variable than genomes? Given these fundamental
questions
[207]
, it is not surprising that the evolutionary
dynamics of protein
e
203]
. The observation that date hubs are more
strongly associated with breast cancer phenotypes was used
to develop a co-expression signature that strongly differ-
entiated breast cancer patients on the basis of disease
outcome
[204]
.
Despite these initial successes, interactome dynamics
modeling will need to move beyond computational anal-
yses. Cell-type-specific transcriptome data may provide
intuitive approximation of protein expression levels, but
such estimations are bound to be imprecise. Detection of
a transcript does not necessarily imply that the corre-
sponding protein is present and stable, and the absence of
a transcript does not necessarily imply absence of the
corresponding protein, as proteins can remain stable long
past transcript degradation and can transit from cell to cell.
Relative protein concentrations must also be considered
when modeling interactome dynamics. Protein concentra-
tions influence the affinity of proteins for one another due
to mass action, and the effect of cell crowding on pro-
tein
e
protein interactome networks have
been a focus of investigation ever since the first large-scale
protein
e
protein interactome maps appeared.
If protein
e
protein interactomes were evolutionarily
stable systems, interactions between orthologous protein
pairs from distinct species should be largely conserved.
However, the observed fraction of interactions correspond-
ing to such 'interolog' pairs is consistently low across several
species
[53,208
e
211]
. The incompleteness of available
interactome maps, and/or the difficulties of orthology
mapping, may explain these apparently low proportions of
conserved interactions. Still, as even these low proportions
would not be expected at random, they are consistent with
natural selection acting on the conservation of at least
a subset of interactions throughout evolutionary time. A
complementary interpretation would be that protein
e
protein interactomes are evolutionarily dynamic systems,
constantly changing under the action of natural selection.
Cross-species comparisons indicate that ~10
e
5
inter-
actions are lost or gained per protein pair per million years,
leaving aside the interactome remodeling that necessarily
follows gene death and gene birth events
[25,212,213]
. This
corresponds to approximately 10
3
interaction changes per
10
6
years in the evolution of the human lineage. Different
types of protein interaction are rewired at different rates.
Transient interactions appear more evolutionarily volatile
than the more
e
protein interactomes remains unexplored. The intra-
cellular environment also affects protein interactions.
Proteins can be restricted to particular organelles bound by
membranes, as are the nuclei or mitochondria, or localized
in less sharply delimited regions such as nucleoli. Cellular
localization data are available at genomic scale for several
organisms
[205,206]
, but information about the dynamic
movements of proteins across cellular compartments is
lacking.
These caveats limit the scope of computational
approaches in modeling interactome dynamics. Experi-
mentally measuring protein
e
lasting interactions
forming protein
protein interactomes at high
resolution both in space, across subcellular locations and
across cell and tissue types, and in time, for example
through the course of development, may still appear
a distant goal, but this goal deserves to be actively pursued.
Conversely, evidence that the expression of interacting
proteins is tightly regulated shows that co-expression
should not be used as a benchmark for protein interaction
reliability.
complexes
[213
peptide interactions
appear to change more rapidly than interactions between
long proteins
[212]
. Evolutionary variation is observed
even for protein complexes participating in the cell cycle.
These complexes are globally conserved across several
yeast species, but differ in their regulatory subunit
composition and timing of assembly
[216]
. Incidentally,
this dynamic rewiring of interactome networks during
evolution is bound to limit the reliability of predictions of
215]
, and protein
e
e
e
