Biology Reference
In-Depth Information
Simulation and Prediction of Reprogramming
It was recently reported that overexpression of
Pdx1
1
,
Ngn3
1
, and
MafA
1
could reprogram
exocrine pancreas cells to the endocrine insulin-producing
cells
47
a clinically desired
transition because of its potential in cell therapy of diabetes. To model such
reprogramming, we first describe the virus-mediated ectopic gene overexpression with
temporal additional gain terms in the corresponding equations in our GRN model
(
Eq. 5.21
).
Figure 5.4E
presents the gene expression time profiles and trajectories in the
relevant phase planes during cell reprogramming using overexpression of the genes
Pdx1
1
,
Ngn3
1
, and
MafA
1
. An exocrine cell starts with high expression of
Ptf1a
. In the model,
reprogramming is implemented by the extra production terms for
Pdx1
,
Ngn3
, and
MafA
during a certain time window. We see that the cell switches its expression pattern from
that with a high-
Ptf1a
to one with a high-
Ngn3
which subsequently triggers the cell to go
through the
Pax4
-
Arx
branch point to finally reach the steady-state of the
β
β
cell. It should be
noted that some
α
cells are also produced in this process because of the stochasticity, and
'
'
such outliers that
go off in the wrong direction
are often observed in reprogramming
experiments.
Our model also predicts that
Ngn3
s role in reprogramming can be enhanced by the
inhibition of
Ptf1a
directly. This is important because an inhibitory perturbation of a
network node is technically much easier (e.g., via RNAi technology, small molecules) than
an activating perturbation. Since
Ptf1a
and
Ngn3
are cross-inhibitory, when
Ptf1a
expression
level is suppressed,
Ngn3
expression will increase. Our simulations also show that
combining the perturbation of the two network nodes,
Ptf1
(inhibition) and
Ngn3
(activation) can synergistically enhance efficiency of
'
β
cell reprogramming.
In conclusion, this example demonstrates how a gene regulatory network model, built
with qualitatively reported interaction schemes from the literature, which mostly represent
incomplete
97
rather than GRNs, govern cell-type diversification and
differentiation. Our results show that with a minimum of knowledge of the constraints
imposed by the gene network topology, pancreas cell differentiation can be explained as the
transitions among different cell attractors.
'
causal networks
'
CONCLUSION AND OUTLOOK
We have demonstrated the principles of manipulating the phenotype of cells
by actuating
switches between entire cell phenotypes and how such engineering can be informed by the
underlying gene regulatory network that governs normal development of these cell types.
The perturbations force cells to achieve a physiological, predestined phenotype, but going
there through a
'
'
during normal development. It is in this new rationale
of predictive model-based whole phenotype manipulation across uncharted terrain in gene
expression state space that warrants the placement of such manipulative approaches into
the domain of
road not taken
'
synthetic biology.
'
'
Controlling transitions between high-dimensional attractors in rugged
epigenetic
landscapes
is an elementary capacity that will help the design and engineering of more
complex biological systems.
'
The state transitions between attractors on a landscape is not just a helpful metaphor
but the direct mathematical manifestation of network dynamics that involve deterministic
constraints imposed by regulatory interactions encoded by the genome and modulated
and driven by external signals and gene expression noise. At the moment little specific
information about these genomic interactions is known, so that educated guessing is
a substantial ingredient in the modeling. But it turns out that, as modelers of biological
systems have long realized, a profound property of robust complex systems is that the
qualitative constraints of interactions, independently of quantitative details, capture much
