(563e) Computational Modeling of Synthetic Gene Circuits to Improve Stem Cell Differentiation

Induced pluripotent stem
cells (iPSCs) have emerged as a vehicle for manipulating and studying cellular processes
in ways that cannot be as readily achieved in a whole organism. For instance, in
the brain, various types of neurons are required for memory, and their loss
occurs with neurodegenerative disorders; iPSCs provide a way to study these
processes in a cell culture setting. In the lab, iPSCs are typically
differentiated by nucleofection, in which DNA is transferred into cells carrying
instructions to constantly produce a large amount of transcription factors (TFs).
In the case of inducing differentiation to basal forebrain cholinergic neurons
(BFCNs), the TFs are Lhx8 and Gbx1. However, this overexpression approach has
low yield and accuracy, which suggests that a more specific TF profile is required
for effective differentiation to the desired cell type. In vitro
experiments on neural progenitor cells have shown that providing bone
morphogenetic protein 9 (BMP-9), a more natural stimulus, yields a specific temporal
profile: Lhx8 peaks at 6 hours post-exposure and declines to baseline by 12
hours, and Gbx1 peaks at 48 hours and declines by 72 hours (Figure 1). These
sequential pulse dynamics significantly improve differentiation. The goal of my project was to
formulate computational models of zinc-finger transcription factor (ZF-TF) gene
circuits using ordinary differential equations (ODEs) and analyze their ability
to produce sequential pulse behavior.

Currently, there is a
limitation in the tools available for controlling cell behavior. To overcome
this challenge, the Leonard lab at Northwestern University has developed a
toolkit of synthetic molecular components called ZF-TFs that can be used to
create gene circuits comprised of interacting components (e.g., DNA,
RNA, and protein molecules). For my project, I integrated a recent experimental
and computational characterization of ZF-TFs to formulate dynamical models of new
ZF-TF gene circuits (Figure 2). The circuit components (i.e.,
ZF-TFs and promoter elements) have regulatory relationships that control each
other's levels, which I defined using ODEs. I then simulated the resulting TF
dynamics in Matlab to elucidate the regulatory relationships between the
molecular components in a circuit and quantify the circuit's ability to produce
sequential pulse behavior. In a circuit, there are parameters associated with
each ZF-TF that correspond to controllable physical properties of the molecular
components, and they affect the ZF-TF's expression and the extent of its
activating or inhibitory function. These parameters include the promoter
architecture, ZF-TFs' translational efficiency, plasmid dose, and ZF-TFs'
activating or inhibitory strength. Thus, there are a combinatorially immense
number of possible design choices in building a gene circuit. It is difficult
to intuit which choices have bigger impacts on the resulting dynamics and which
are better than others.

To alleviate this
challenge, I explored a higher-dimensional space of design choices in an
unbiased manner to perform an automated sweep of parameter values. Through this
analysis, I identified the parameters most important in determining TF dynamics
(Figure 3); I narrowed the hypothesis space by determining the optimal
parameter values for how to construct the components in the wet lab and how to
link them together in a circuit to achieve sequential pulse dynamics, as these
parameters ultimately do correspond to physical properties of the molecular
components over which we have control. Experimental collaborators and I took an
iterative approach, alternating between computation and experiments to execute
model-guided design of synthetic gene circuits (Figure 4), with the
objective to maximize fold induction (FI, increase in expression when a pulse
initiates) and fold reduction (FR, decrease in expression after a pulse has
peaked) in the TF expression to improve iPSC differentiation to BFCNs.  The performance (i.e., FI
and FR) of the gene circuits will be evaluated in HEK293 cells, in which the
ZF-TFs were originally characterized, prior to testing circuits in BFCNs. This
rigorous quantitative approach for predicting circuit dynamics over
wide-ranging conditions will enable tight control of gene expression in ways
that can also be applied to future biomedical applications of ZF-TFs.