Page 121 - Read Online

Zhong et al. Chem Synth 2023;3:27 https://dx.doi.org/10.20517/cs.2023.15 Page 15 of 25

Recently, Dehne et al. reported that PEN-based reaction networks could generate periodic signals for the
[139]
dynamic control of colloidal self-organization, as shown in Figure 9 . To program the complex dynamic
behaviors on a mesoscopic scale, two different fluorescent-labeled colloidal particles with a size of 1 µm are
functionalized with two different DNA single strands. The dynamic aggregation of these particles can be
controlled using complementary linker strands. And a dNTP-driven predator-prey oscillating network is
designed to guide the periodic aggregation of colloidal particles. The system includes four different
functional modules. An autocatalytic reaction module involves a DNA polymerizing process and the
following sequence-specific nicking reactions that lead to the primer-activated production of itself for
amplifying its concentration. The primer strand can initiate two downstream reactions: the production of
the complementary linker strand for triggering the aggregation of colloidal particles and the negative-
feedback polymerizing reaction to block the primer strand by forming a stable DNA duplex. The
exonuclease degrades the primer, the blocked primer, and the complementary linker in their single-strand
state. With this design, the periodic concentration change of the complementary linker strand is realized to
drive the autonomous oscillatory colloidal structure formation and dissociation without the addition of
external stimuli.

Nucleic acids have been used to store and process information for a long time . DNA circuits that act as
[51]
[141]
the central processing unit (CPU) execute diverse analog computing operations . As artificial neural
networks (ANN) are revolutionizing electronic computing, DNA computing researchers turn their gaze to
mimic biological information processes [141-143] . Biomimetic networks were used to imitate massively parallel
and recurrent architectures, analog and asynchronous operation, and fault-tolerant and redundant
computations. However, biological information processes, such as genetic information encoding and
decoding and signaling dynamic processes, typically operate nonlinearly, allowing the biological system to
manipulate molecular information precisely . This is essential to transduce biological information to
[57]
[144]
[145]
[146]
complex natural behaviors, such as self-organization , cell differentiation , and migration . Recently,
Okumura et al. used PEN tools to build neuron networks with tunable weights and biases, including
[143]
positive weight, negative weight, weight adjustment, amplification, bias, and reporting units . Specifically,
the inputs are divided into positive and negative weights. The positive weight inputs will bind to the
converter templates and produce an output strand with DNA polymerase and nickase. The negative weight
operation produces the complementary strand, which deactivates the strand generated by the positive
weight process. At the same time, the weight adjustment template competes with the above two weight
templates to adjust the weights. When the weighted sum exceeds a given tunable threshold, the neuron is
transformed into an ON state by the autocatalytic exponential replication of the output signal. Throughout
this process, the exonuclease continuously degrades all single-stranded DNA to maintain nonlinear
dynamic behaviors. The neuron is powered by these far-from-equilibrium DNA weighting reactions for
discriminating tiny differences in the input concentration to generate an output signal. In addition, the
output signal can then regulate a downstream, connected neuron. Designed neurons are assembled in
multilayer architectures to classify nonlinearly separable regions and perform nonlinear decisions with
higher sensitivity than simple DNA networks. The modular design of the PEN-based neuron network
highlights the potential to expand and enhance the nucleic acid-based dynamic network to make it
“smarter”.

SIGNALING DYNAMICS IN CELL-LIKE ENVIRONMENTS
A cell is the smallest unit of life, in which tens of thousands of genes and countless biomolecules coordinate
to perform cellular functions through complex dynamic networks . Despite the progress of nucleic acid-
[147]
based dynamic networks operated in the test tube, it is still challenging to integrate nucleic acid-based
dynamic networks into cell-like containments to design and construct “artificial cells” for mimicking