Preprogramming Complex Hydrogel Responses using Enzymatic Reaction Networks

Article abstract of DOI:10.1002/anie.201610875

The creation of adaptive matter is heavily inspired by biological systems. However, it remains challenging to design complex material responses that are governed by reaction networks, which lie at the heart of cellular complexity. The main reason for this slow progress is the lack of a general strategy to integrate reaction networks with materials. Herein we use a systematic approach to preprogram the response of a hydrogel to a trigger, in this case the enzyme trypsin, which activates a reaction network embedded within the hydrogel. A full characterization of all the kinetic rate constants in the system enabled the construction of a computational model, which predicted different hydrogel responses depending on the input concentration of the trigger. The results of the simulation are in good agreement with experimental findings. Our methodology can be used to design new, adaptive materials of which the properties are governed by reaction networks of arbitrary complexity.

Full text of DOI:10.1002/anie.201610875

Angewandte
Communications
Chemie
International Edition: DOI: 10.1002/anie.201610875
German Edition: DOI: 10.1002/ange.201610875
Life-Like Systems
Preprogramming Complex Hydrogel Responses using Enzymatic
Reaction Networks
Sjoerd G. J. Postma, Ilia N. Vialshin, Casper Y. Gerritsen, Min Bao, and Wilhelm T. S. Huck*
Abstract: The creation of adaptive matter is heavily inspired
by biological systems. However, it remains challenging to
design complex material responses that are governed by
reaction networks, which lie at the heart of cellular complexity.
The main reason for this slow progress is the lack of a general
strategy to integrate reaction networks with materials. Herein
we use a systematic approach to preprogram the response of
a hydrogel to a trigger, in this case the enzyme trypsin, which
activates a reaction network embedded within the hydrogel. A
full characterization of all the kinetic rate constants in the
system enabled the construction of a computational model,
which predicted different hydrogel responses depending on the
input concentration of the trigger. The results of the simulation
are in good agreement with experimental findings. Our
methodology can be used to design new, adaptive materials
of which the properties are governed by reaction networks of
arbitrary complexity.
contains two orthogonal types of crosslinks that can be
degraded or formed, respectively, through enzymatic activity
(Figure 1A). When a trigger, the endopeptidase trypsin (Tr),
is applied to the gel, degradation of the initial crosslinks
1 (C1) proceeds rapidly. Simultaneously, a crosslink precursor
(a copolymerized thioester) is slowly cleaved, creating thiols
that very quickly react with an added linker to form new
crosslinks 2 (C2). Thus, a gel–liquid–gel transition takes place
and a new gel is formed with potentially different properties
such as shape and stiffness as compared to the initial gel.
Next, we introduced an enzymatic reaction network into
the hydrogel of which the components all have their own
specific function. When all components are present in the
right concentrations, the network is able to sense the input
concentration of Tr, and determine the corresponding hydro-
gel response.
Importantly, the kinetics in the network are fully charac-
terized, and the programmed response of the network is
predicted by a computational model that is in good agreement
with our experimental results. In this way, we provide
a systematic approach for integrating reaction networks
within adaptive materials.
L
iving systems are adaptive and use enzymatic reaction
networks to detect changes in their environment, process
input information, and determine an appropriate response.[1]
Materials science has recently taken a keen interest in the
adaptivity of living systems,[2] and has created new materials
with life-like properties such as self-healing,[3] camouflaging,[4]
and control over surface characteristics.[5] Impressive exam-
ples include the incorporation of the oscillating Belousov–
Zhabotinsky reaction into a self-walking gel,[6] and the work
of Aizenberg and co-workers,[7] who used chemo–mechanico–
chemical feedback loops to produce a homeostatic material.
Others have pioneered control over hydrogel lifetimes with
preprogrammed feedback loops using organic[8] or enzy-
matic[9] reactions.
We synthesized PAAm gels with two orthogonal cross-
linkers (Figure 1A): C1 is susceptible to cleavage by Tr, which
will be used as a trigger. Tr also triggers the cleavage of
thioester 3, revealing the cryptic thiol groups, which can react
with the poly(ethylene glycol)-bis-maleimide crosslinker 4
(MW= 2000 gmolÀ1) that is present in the gel, forming C2. Tr
rapidly cleaves amide bonds at the C-terminal end of
positively charged amino acids, and therefore we synthesized
C1 with an arginine–serine moiety in the middle of the
molecule (Figure 1A; see section S2 of the Supporting
Information (SI) for complete details on molecular structures
and synthesis of all molecules used). We measured a value for
kcat/KM (a measure of catalytic efficiency) of @ 24600 mmÀ1 hÀ1
for Tr cleaving C1, confirming that Tr will rapidly degrade gels
containing C1 (details of all kinetic studies are in section S3 of
the SI).
However, progress towards “life-like” materials has been
slow as we lack a general framework for constructing
materials with autonomous behavior and preprogrammed
responses to external stimuli. Designing materials with
complex responses requires the incorporation of chemical
reaction networks, where the kinetics within the system are
suitably balanced.[10] Here, we present a systematic approach
to program the complex response of hydrogels, inspired by
our previous work on enzymatic reaction networks.[11,12] First,
we developed a polyacrylamide (PAAm)-based hydrogel that
For the cryptic crosslink precursor, we studied a number
of amino acids that could serve as Tr-cleavable protecting
groups of thiol side groups. Interestingly, due to the relatively
high reactivity of thioesters, we found that the reactivity of Tr
towards lysine thioesters was as high as the arginine amide
bonds in C1 (kcat/KM > 29400 mmÀ1 hÀ1). Therefore, the less
[*] S. G. J. Postma, I. N. Vialshin, C. Y. Gerritsen, M. Bao,
Prof. W. T. S. Huck
reactive leucine thioester 3 was prepared (Tr hydrolysis kcat
/
KM = 575 mmÀ1 hÀ1). Importantly, thioester 3 is also quickly
cleaved by another enzyme, chymotrypsin (Cr; kcat/KM =
34400 mmÀ1 hÀ1), enabling more complex responses when
using both enzymes in an enzymatic reaction network as we
will show below.
Radboud University, Institute for Molecules and Materials
Heyendaalseweg 135, 6525 AJ Nijmegen (The Netherlands)
E-mail: w.huck@science.ru.nl
Supporting information for this article can be found under:
http://dx.doi.org/10.1002/anie.201610875.
Angew. Chem. Int. Ed. 2017, 56, 1 – 6
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1
These are not the final page numbers!

Angewandte
Communications
Chemie
Figure 1. Gel–liquid–gel transitions. a) The initial polyacrylamide hydrogel contains crosslinks 1 (C1) and a copolymerized thioester 3. C1 is
quickly degraded by trypsin (Tr), and thioester 3 is cleaved slowly by Tr or quickly by chymotrypsin (Cr) to produce thiols which subsequently form
new crosslinks 2 (C2) after reacting with a linker (poly(ethylene glycol)-bis-maleimide). Zigzag lines indicate the sites of enzymatic cleavage. b) A
cone-shaped gel (200 mL volume) containing C1 and thioester 3 is soaked with linker 4, after which Tr is applied ([Tr]0 =5 mm). A gel–liquid–gel
transition occurs during which the shape of the gel changes, and the magnetic stirring bar (that plays no further role in this process) is
encapsulated. c) Gels 1–3 contain different amounts of C1 and thioester 3, and consequently have different initial stiffness, but similar mechanical
properties after treatment with Tr. Error bars indicate the range of Young’s modulus in duplicate experiments (the error bar of gel 1 before
application of Tr is very small). Measurements were performed by atomic force microscopy (see SI for details). All experiments are performed in
100 mm Tris-HCl buffer pH 7.5 with 20 mm CaCl2.
During C2 formation, the enzymatic cleavage of thioester
should be the rate-limiting step to ensure control over the
kinetics of gel formation by changing enzyme concentration.
Therefore, the thiols formed after cleavage of the thioester
should react as rapidly as possible with a crosslinker. Rate
studies showed that maleimides react almost instantaneously
with thiols in a Michael addition (section S3.3 of the SI). For
that reason, the maleimide-containing linker 4 was chosen as
the crosslinker to form C2 in all further studies.
Next, a gel was prepared that contained C1 and co-
polymerized thioester 3 (see Table S2 of the SI for the
compositions of all gels made), and the hydrogel was soaked
in a solution of linker 4. When Tr was applied to the hydrogel
([Tr]0 = 5 mm), the gel turned into a liquid after 3.5 hours
(Figure 1B). The liquid formed a gel again, but with a differ-
ent shape, after another 16.5 hours, which confirms that
crosslink degradation and formation take place at convenient
timescales.
initial gel was larger (gel 1), equal (gel 2) or smaller (gel 3)
than the amount of C2 in the final gel (see Table S2 of the SI
for gel compositions). The stiffness of the gels before and
after triggering the rearrangement of the crosslinks was
measured by atomic force microscopy coupled to a confocal
microscope (see section S4 of the SI). Figure 1C shows the
development of Youngꢀs moduli upon addition of Tr and
waiting overnight. Initial moduli of 37.2 kPa, 19.5 kPa, and
5.1 kPa were obtained for gels 1–3, respectively, which
changed to 22.4 kPa for both gels 1 and 2, and 17.0 kPa for
gel 3 after the crosslink rearrangement. Gel 1 had become
softer, gel 2 had retained its stiffness, and gel 3 had turned into
a stiffer gel. Thus, it was shown that three mechanically
different hydrogels can obtain a similar stiffness upon treat-
ment with Tr.
Next, we program the activity of Tr by incorporating the
enzyme in a reaction network. Figure 2A shows how Tr can
be formed autocatalytically from its own precursor trypsino-
gen (Tg), creating a positive feedback loop. This amplification
of [Tr] is expected to speed up both the degradation of C1 and
the activation of C2. Tr also rapidly converts chymotrypsi-
The gel–liquid–gel transition transforms the mechanical
properties of the gel, as the number of crosslinks C1 and C2
are independent of each other. We prepared three different
gels soaked with linker 4 in which the amount of C1 in the
nogen (Cg) into its active form chymotrypsin (Cr) with kcat
/
2
www.angewandte.org
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2017, 56, 1 – 6
These are not the final page numbers!

Angewandte
Communications
Chemie
Figure 2. Introducing an enzymatic reaction network in the hydrogel. a) Schematic representation of the network. Trypsin (Tr) catalyzes its own
formation by activating trypsinogen (Tg), creating a positive feedback loop. Tr also activates chymotrypsinogen (Cg) to form chymotrypsin (Cr).
Soybean trypsin inhibitor (STI) inhibits Tr and Cr strongly and weakly, respectively (indicated by the thickness of the inhibition arrows). b) The
degradation time of different gels in vial tests is dependent on the concentration of Tr ([Tr]0 =0.25–20 mm. For the cyan curve: [Tg]0 =80 mm,
[STI]=1.9 mm). In the cyan region, Tr is fully inhibited by STI. Error bars are standard deviations obtained from triplicate experiments. c) Influence
of Tg on the rate of gel reformation ([Tg]0 =80 mm, [STI]=1.9 mm, if present) in vial tests. The asterisk in the left column indicates that a gel was
formed again after 80 hours, which is off the scale. d) Influence of [Cg] on the rate of C2 formation in vial tests (in all gels: [Tg]0 =80 mm,
[STI]=1.9 mm, and [Tr]0 =3 mm). For the gel/liquid regions (orange and cyan stripes), the contents of the vial were difficult to classify as either
being a soft gel or a viscous liquid. All experiments are performed in 100 mm Tris-HCl buffer pH 7.5 with 20 mm CaCl2.
KM = 12500 mmÀ1 hÀ1. Cr does not cleave C1 (Figure S21 of
the SI), but rapidly cleaves thioester 3, activating the
formation of C2. Finally, the enzymatic network is deactivated
by the presence of soybean trypsin inhibitor (STI) protein,
a strong, reversible inhibitor of Tr (KD 1 nm)[12a] that sets
a threshold for the Tr input that will activate the network. We
found that STI also reversibly inhibits Cr (KD = 634 nm,
Figure S14), but with a dissociation constant at least two
orders of magnitude larger than for Tr.
Then, we gradually increased the complexity of the
preprogrammed system to integrate the reaction network
within the hydrogels in a step-by-step approach. First, we
tested the degradation rate of a gel only containing C1
crosslinks using various [Tr]0 (0.25–20 mm, orange diamonds in
Figure 2B). We found a nonlinear correlation between [Tr]0
and the time required for a gel–liquid transition to occur, with
average gel degradation times ranging between 47 minutes
and 4.7 hours. Then, we prepared C1-only gels, but soaked
these in Tg (80 mm) and STI (1.9 mm), and again used the same
range of [Tr]0 to trigger degradation (cyan circles, Figure 2B).
Now, we observe a similar trend in gel degradation times as in
the previous experiment, but below a certain input threshold
(cyan region in Figure 2B), Tr is fully inhibited by STI and
the gel does not degrade, even when left for more than
a week.
In addition, when gels containing both C1 and thioester 3
were prepared without network components or linker 4 (so no
gel reformation could occur), we did not see any significant
difference from the previous experiments ([Tr]0 = 0.25–20 mm,
green triangles, Figure 2B). This observation confirms that Tr
Angew. Chem. Int. Ed. 2017, 56, 1 – 6
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3
These are not the final page numbers!

Angewandte
Communications
Chemie
cleaves C1 much faster than thioester 3, and when the two are
combined in one hydrogel, Tr will first degrade most of C1
crosslinks before significant activation of the cryptic crosslink
sites occurs.
In Figure 2C, the influence of Tg on the rate of gel
reformation is shown. In this case, gels containing both C1 and
thioester 3 were prepared, and soaked with linker 4 to enable
gel reformation. The gel reformation time decreased from
80 hours (indicated by an asterisk in Figure 2C) in the
absence of Tg and STI to 12 hours in the presence of both
proteins (80 and 1.9 mm, respectively) when [Tr]0 was 3 mm.
The influence of the positive feedback was less pronounced at
a higher [Tr]0 (8 mm) applied to the gel, but still the gel
reformation time decreased 2.5 times in the presence of Tg
and STI. Interestingly, we do not observe an influence of the
positive feedback on the gel degradation time, indicating that
again C1 cleavage is dominant over other processes catalyzed
by Tr at the early stages after application of Tr.
Finally, C1- and thioester 3-containing gels were soaked in
solutions with varying concentrations of Cg in the presence of
Tg (80 mm) and STI (1.9 mm), and in all cases [Tr]0 was 3 mm
(Figure 2D). The observed response of the gel was highly
dependent on Cg concentration. At 0.5 mm Cg, gel–liquid–gel
transitions were still observed, but at concentrations of 1.5
and 2.5 mm Cg the hydrogel went through a gel/liquid phase in
which it was difficult to classify the material as either a soft gel
or a viscous liquid (indicated by orange/cyan stripes in
Figure 2D). Increasing the concentration of Cg even further
led to gel–gel transitions without an intermediate liquid state,
because the rate of formation of C2 is greatly enhanced by the
production of Cr and ultimately outpaces the degradation of
C1.
In the previous experiments, we have seen three different
gel responses: 1) inhibition of Tr by STI, 2) gel–gel transitions
when the rate of C1 degradation is slower than of C2
formation, and 3) gel–liquid–gel transitions when C1 degra-
dation is faster than C2 formation. As our final demonstration
of the ability to preprogram the complex response of the
hydrogel materials, we coupled the different hydrogel
responses to the input Tr concentration. A computational
model in MATLAB was constructed using all the determined
rate constants in ordinary differential equations (see sec-
tion S5 of the SI), to explore the hydrogel response at
different initial concentrations of Cg and Tr.
Figure 3. Phase diagram of [Cg] vs. [Tr]0 obtained by mathematical
modelling. Experiments are shown in squares (in all cases
[Tg]0 =80 mm and [STI]=1.9 mm). The blue square with an asterisk
denotes experiments conducted at [Tr]0 =7, 20, or 50 mm, all of which
show gel–liquid–gel transitions. Cg=chymotrypsinogen, Tr=trypsin.
In short, we have shown a stepwise design of hydrogel
materials with preprogammable responses to environmental
triggers (in this case [Tr]). The incorporation of a small
reaction network that is coupled to the crosslinking chemistry
in the hydrogels can be used to create materials that sense and
process a biochemical input. We believe that this approach
can be extended further to create adaptive matter based on
hydrogels containing more complex reaction networks that
can control the time profile of the chemical response within
the material. In addition, the methodology can be applied to
other networks based on DNA,[13] or small organic mole-
cules.[14]
Acknowledgements
We thank Simon Yang and Hugo Veldhuizen for their
contributions to the synthesis of thioester 3 and lysine
thioester, respectively. This work was supported by funding
from the Dutch Ministry of Education, Culture and Science
(Gravity programme 024.001.035), and the Radboud Nano-
medicine Alliance.
In Figure 3, the phase diagram generated by these
simulations is depicted, in which three regions corresponding
to the three different gel responses are observed. Importantly,
the phase diagram shows that the strength of Tr trigger can
lead to any of the three responses if the concentration of Cg
exceeds 0.5 mm. We verified these computational results in
experiments using identical gels ([Cg] = 1.5 mm, [Tg] = 80 mm,
[STI] = 1.9 mm), and we indeed obtained different responses
depending on the Tr concentration (squares in Figure 3,
pictures and procedures in section S4.1.7 of the SI). Our
model is in good agreement with the experimental results,
although the influence of STI is slightly overestimated in the
simulations. Additional modelling indicated that we indeed
need the full reaction network to obtain the complex gel
responses (Figure S26 and S27).
Conflict of interest
The authors declare no conflict of interest.
Keywords: enzymes · gels · life-like systems ·
programmable materials · reaction networks
[1] a) J. E. Purvis, G. Lahav, Cell 2006, 152, 945 – 956; b) B. N.
Kholodenko, Nat. Rev. Mol. Cell Biol. 2006, 7, 165 – 176; c) D.
Bray, Nature 1995, 376, 307 – 312.
4
www.angewandte.org
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2017, 56, 1 – 6
These are not the final page numbers!

Angewandte
Communications
Chemie
[2] a) M. A. Cohen Stuart et al., Nat. Mater. 2010, 9, 101 – 113;
8237 – 8241; c) T. Heuser, E. Weyandt, A. Walther, Angew.
b) B. A. Grzybowski, W. T. S. Huck, Nat. Nanotechnol. 2016, 11,
585 – 592.
Chem. Int. Ed. 2015, 54, 13258 – 13262; Angew. Chem. 2015, 127,
13456 – 13460; d) E. Jee, T. Bansagi, Jr., A. F. Taylor, J. A.
Pojman, Angew. Chem. Int. Ed. 2016, 55, 2127 – 2131; Angew.
Chem. 2016, 128, 2167 – 2171.
[3] a) S. R. White, N. R. Sottos, P. H. Geubelle, J. S. Moore, M. R.
Kessler, S. R. Sriram, E. N. Brown, S. Viswanathan, Nature 2000,
409, 794 – 797; b) P. Cordier, F. Tournilhac, C. Souliꢁ-Ziakovic, L.
Leibler, Nature 2008, 451, 977 – 980; c) T.-S. Wong, S. H. Kang,
S. K. Y. Tang, E. J. Smythe, B. D. Hatton, A. Grinthal, J.
Aizenberg, Nature 2012, 477, 443 – 447.
[4] S. A. Morin, R. F. Shepherd, S. W. Kwok, A. A. Stokes, A.
Nemiroski, G. M. Whitesides, Science 2012, 337, 828 – 832.
[5] a) X. Yao, Y. Hu, A. Grinthal, T.-S. Wong, L. Mahadevan, J.
Aizenberg, Nat. Mater. 2013, 12, 529 – 534; b) K.-C. Park, P. Kim,
A. Grinthal, N. He, D. Fox, J. C. Weaver, J. Aizenberg, Nature
2016, 531, 78 – 82.
[10] L. Heinen, A. Walther, Soft Matter 2015, 11, 7857 – 7866.
[11] a) S. N. Semenov, A. S. Y. Wong, R. M. van der Made, S. G. J.
Postma, J. Groen, H. W. H. van Roekel, T. F. A. de Greef,
W. T. S. Huck, Nat. Chem. 2015, 7, 160 – 165; b) A. S. Y. Wong,
S. G. J. Postma, I. N. Vialshin, S. N. Semenov, W. T. S. Huck, J.
Am. Chem. Soc. 2015, 137, 12415 – 12420.
[12] a) S. N. Semenov, A. J. Markvoort, W. B. L. Gevers, A. Piruska,
T. F. A. de Greef, W. T. S. Huck, Biophys. J. 2013, 105, 1057 –
1066; b) S. N. Semenov, A. J. Markvoort, T. F. A. de Greef,
W. T. S. Huck, Angew. Chem. Int. Ed. 2014, 53, 8066 – 8069;
Angew. Chem. 2014, 126, 8204 – 8207.
[13] a) J. Stricker, S. Cookson, M. R. Bennett, W. H. Mather, L. S.
Tsimring, J. Hasty, Nature 2008, 456, 516 – 519; b) A. Padirac, T.
Fujii, A. Estꢁvez-Torres, Y. Rondelez, J. Am. Chem. Soc. 2013,
135, 14586 – 14592.
[14] S. N. Semenov, L. J. Kraft, A. Ainla, M. Zhao, M. Baghbanzadeh,
V. E. Campbell, K. Kang, J. M. Fox, G. M. Whitesides, Nature
2016, 537, 656 – 660.
[6] S. Maeda, Y. Hara, T. Sakai, R. Yoshida, S. Hashimoto, Adv.
Mater. 2007, 19, 3480 – 3484.
[7] X. He, M. Aizenberg, O. Kuksenok, L. D. Zarzar, A. Shastri,
A. C. Balazs, J. Aizenberg, Nature 2012, 487, 214 – 218.
[8] a) J. Boekhoven, A. M. Brizard, K. N. K. Kowlgi, G. J. M. Koper,
R. Eelkema, J. H. van Esch, Angew. Chem. Int. Ed. 2010, 49,
4825 – 4828; Angew. Chem. 2010, 122, 4935 – 4938; b) J. Boek-
hoven, W. E. Hendriksen, G. J. M. Koper, R. Eelkema, J. H.
van Esch, Science 2015, 349, 1075 – 1079.
[9] a) S. Debnath, S. Roy, R. V. Ulijn, J. Am. Chem. Soc. 2013, 135,
16789 – 16792; b) C. G. Pappas, I. R. Sasselli, R. V. Ulijn, Angew.
Chem. Int. Ed. 2015, 54, 8119 – 8123; Angew. Chem. 2015, 127,
Manuscript received: November 7, 2016
Final Article published: && &&, &&&&
Angew. Chem. Int. Ed. 2017, 56, 1 – 6
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
5
These are not the final page numbers!

Angewandte
Communications
Chemie
Communications
Life-Like Systems
Question-and-answer game: A system-
atic approach is presented to integrate
reaction networks within responsive
hydrogels. Different gel responses
depending on the input concentration of
the network trigger were obtained after
understanding the kinetics in the system.
S. G. J. Postma, I. N. Vialshin,
C. Y. Gerritsen, M. Bao,
W. T. S. Huck*
&&& — &&&
Preprogramming Complex Hydrogel
Responses using Enzymatic Reaction
Networks
6
www.angewandte.org
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2017, 56, 1 – 6
These are not the final page numbers!