A Screening Platform to Identify and Tailor Biocompatible Small-Molecule Catalysts

Article abstract of DOI:10.1002/chem.201904808

Interfacing biocompatible, small-molecule catalysis with cellular metabolism promises a straightforward introduction of new function into organisms without the need for genetic manipulation. However, identifying and optimizing synthetic catalysts that per

Full text of DOI:10.1002/chem.201904808

A Journal of
Accepted Article
Title: A screening platform to identify and tailor biocompatible small-
molecule catalysts
Authors: Rudy Rubini, Ilya Ivanov, and Clemens Mayer
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201904808
Link to VoR: http://dx.doi.org/10.1002/chem.201904808
Supported by

10.1002/chem.201904808
Chemistry - A European Journal
COMMUNICATION
A screening platform to identify and tailor biocompatible small-
molecule catalysts
Rudy Rubini, Ilya Ivanov, and Clemens Mayer*
Abstract: Interfacing biocompatible, small-molecule catalysis with
cellular metabolism promises a straightforward introduction of new
function into organisms without the need for genetic manipulation.
However, identifying and optimizing synthetic catalysts that perform
new-to-nature transformations under conditions that support life is a
cumbersome task. To enable the rapid discovery and fine-tuning of
biocompatible catalysts, we describe a 96-well screening platform that
couples the activity of synthetic catalysts to yield non-canonical amino
acids from appropriate precursors with the subsequent incorporation
of these nonstandard building blocks into GFP. Critically, this strategy
does not only provide a common readout (=fluorescence) for different
reaction/catalyst combinations, but also informs on the organism’s
fitness, as stop codon suppression relies on all steps of the central
dogma of molecular biology. To showcase our approach, we have
applied it to the evaluation and optimization of transition-metal
catalyzed deprotection reactions.
organisms contain a myriad of compounds that can poison
exogenously-supplied catalysts or reagents[17]
.
Consequently, the discovery and optimization of
biocompatible catalysts and reactions remains challenging.
Typically, a set of potential catalysts is first evaluated for a model
transformation under “biologically-relevant conditions” (i.e. in
presence of water, air and/or thiols) and promising candidates are
subsequently tested in biological settings.[12,18–20] The initial
evaluation, however, takes neither catalyst/reagent toxicity nor
catalyst poisoning by the organism into account. More recently,
evaluating biocompatible transition metal complexes has also
been attempted directly in biological settings by making use of
surrogate substrates that either become fluorescent[13,21,22] or are
converted to luciferins[23] upon a successful transformation.
Unfortunately, the observable phenotypes in these screens do not
depend on a cellular process and therefore, do not account for an
organism’s fitness. To address these challenges and further
streamline discovery and optimization efforts for biocompatible
catalysts, we here describe a 96-well screening platform that
rapidly reports on both the activity of a catalyst and the fitness of
the organism.
Inspired by the use of genetic circuitry for the directed
evolution of enzymes,[24–26] we reasoned that a rapid evaluation
and optimization of biocompatible catalysts requires a direct link
between a catalyst’s activity and an observable phenotype that
can only arise in living organisms. While replacing enzymatic
transformations in metabolic pathways with synthetic ones is a
possibility to establish such a link,[7] this strategy is not general
and would require genetic knock-outs that are prone to false
positives/negatives. Instead, we envisioned to introduce a simple
pathway that (1) is dependent on metabolism while not affecting
viability itself, (2) can readily be employed for different reaction
types and (3) can function in organisms ranging from bacteria to
mammalian cell lines.
A process that matches these criteria is the site-specific
incorporation of non-canonical amino acids (ncAAs) into proteins
of interest through the suppression of a stop codon by the action
of an orthogonal translation system (OTS).[27–29] In order
to repurpose such OTSs for the evaluation and fine-tuning of
biocompatible catalysts, we surmised that the activity of small-
molecule catalysts to yield ncAAs from appropriate precursors
could be coupled with the subsequent incorporation of these
artificial building blocks into green fluorescent protein (GFP)
variants (Fig.1A). Based on these considerations, we constructed
a screening platform that comprises three main components: (1)
the exogenously supplied catalyst and ncAA precursor (=input),
(2) an OTS specific for the chemically synthesized ncAA
(=sensor), and (3) a GFP variant featuring an in-frame stop codon
(=reporter). Critically, only upon suppression of the in-frame stop
codon (UAG) full length GFP is produced. Thus, the fluorescence
signal detected should relate to ncAA production levels and, as a
result, report on catalyst proficiency. Moreover, ncAA
incorporation relies on all the steps of the central dogma of
molecular biology, and therefore should also report on the fitness
of the organism (Escherichia coli in this study).
Synthetic chemists and metabolic engineers pursue contrasting
approaches to make molecules.[1] While the former skillfully
employ synthetic catalysts and reagents to build up complex
molecules, the latter harness the reactivity of biocatalysts in living
organisms to produce compounds from fermentation.[2,3] Although
these approaches have been traditionally considered to be
incompatible, small-molecule catalysts that can interface with
cellular metabolism have the potential to expand biological
function without the need for genetic manipulation.[4–6] For
example, such biocompatible catalysts could be part of cellular
factories, in which they perform new-to-nature transformations to
diversify molecules produced by an organism.[7–10] Thus, such a
concerted effort of synthetic chemistry and metabolic engineering
could pave the way toward the direct synthesis of value-added
compounds in cellular settings. Additionally, biocompatible
catalysis holds promise for biomedical applications, such as
targeted drug release/synthesis,[11–15] the disruption of cell-cell
communication or rescuing dysfunctional enzymes involved in
human diseases.[16]
In order to enable such developments, biocompatible
catalysts have to perform a difficult balancing act and function
under conditions that both support life and allow for abiological
transformations to proceed. This task is complicated by the fact
that synthetic chemists routinely perform reactions in organic
solvents at temperature and pH regimes that are incompatible
with living organisms. Moreover, metabolite concentrations are
typically low (<1 mM), when compared to substrate
concentrations typically employed in organic synthesis.[5,6,16]
Conversely, the complex intra and extracellular environments of
[a]
R. Rubini, I. Ivanov, Dr. C. Mayer
Stratingh Institute for Chemistry
University of Groningen
Nijenborgh 4, 9474 AG Groningen (The Netherlands)
E-mail: c.mayer@rug.nl
Supporting information for this article is given via a link at the end of
the document.
This article is protected by copyright. All rights reserved.

10.1002/chem.201904808
Chemistry - A European Journal
COMMUNICATION
To evaluate the feasibility of the proposed screening
platform, we identified the uncaging of allyloxycarbonyl (alloc)-
protected amines as a model reaction from the collection of
bioorthogonal/biocompatible transformations reported in the
literature.[6,30] Specifically, this transformation is often employed
for the unmasking of prodrugs in vivo[13,31,32] and can readily be
adapted to our proposed screening platform by alloc-protection of
the amine functionality of a ncAA (Fig. 1B). A number of different
catalysts have been shown to catalyze this transformation with
varying efficiencies. Here, we selected a set of 12 catalysts (Fig.
1C): four commercially-available ones for which low activities
were reported (Cat1-4)[33–36] and a total of eight ruthenium-based
half-sandwich complexes featuring either
a
quinoline-2-
carboxylate (Ru1-4)[18] or a 8-hydroxyquinolinate (Ru5-8)[19] as
bidentate ligand, with some of them showing improved activities
when compared to Cat1-4.
Before evaluating these catalysts in presence of live E. coli,
we aimed to verify that suppression of an in-frame stop codon in
GFP by an OTS is both a reliable and quantifiable readout. For
this, E. coli was transformed with two plasmids that encode (1) an
OTS based on the promiscuous aminoacyl-tRNA synthetase,
pCNF-RS,[37] and (2) a GFP variant featuring a UAG stop codon
(either Y151* or Y182*, see Supplementary Information for
details). Next, we monitored GFP fluorescence in 96 well-plates
after induction of gene expression in LB media containing different
concentrations of p-chlorophenylalanine (p-ClF) or O-
methyltyrosine (OMeY) over a period of 10 hours. As anticipated,
GFP production was dependent on the concentration of the ncAA
(Fig. 2A and Fig S1). Plotting the relative fluorescence increase
after 200 minutes for the different concentrations yielded a linear
correlation, independent of which ncAA and GFP variant was
used (Fig. 2B and Fig. S2). In contrast, addition of alloc-protected
ncAAs did not result in a significant increase in GFP fluorescence,
ensuring a good signal-to-noise ratio over 2 orders of magnitude
(10 – 1000 µM).
Fig. 1: A: Schematic representation of the envisioned screening platform. An
appropriate precursor (black sphere) is converted by a synthetic catalyst to a
ncAA (red sphere). Note, this transformation can occur outside or inside E. coli
cells. The synthesized ncAA will then be loaded onto an orthogonal tRNA by
an engineered aminoacyl-tRNA synthetase and incorporated by the ribosome
into a GFP variant that features an in-frame stop-codon. B: The uncaging of
alloc-protected ncAAs by palladium or ruthenium catalysts is used as a model
reaction in this study. C: Structures of complexes used in this study.
Fig. 2: A: Fluorescence (λex = 485 nm, λem = 528 nm) measured over time at different concentrations of p-ClF (concentrations refer to the racemic mixture of D/L-p-
ClF) with sfGFP_Y151*. B: Relative increase in fluorescence after 200 minutes for varying concentrations of p-ClF and alloc-p-ClF (concentrations refer to the
respective racemic mixtures). C: Legend depicting representative values for yields and turnover numbers (TONs) for deprotection with transition-metal complexes.
Bright red depicts high yields while bright blue depicts high TONs. D-F: Yields and TONs for the 12 complexes studied at decreasing concentrations. Colors represent
the average of at least three independent measurements (see Table S1 for averages and standard deviations).
This article is protected by copyright. All rights reserved.

10.1002/chem.201904808
Chemistry - A European Journal
COMMUNICATION
To evaluate the proficiencies of the selected transition-metal
complexes to catalyze the deprotection of alloc-p-ClF, we added
decreasing concentrations of each catalyst to the ncAA precursor
(1 mM as a racemate) at the time of induction. The 96-well setup
of the screening platform enabled the evaluation of 88 different
combinations (+8 samples with varying concentrations of p-ClF
for the calibration) in less than 4 hours. We used the relative
increase in GFP fluorescence after 200 minutes to estimate the
yields and turnover numbers (TONs) of a catalyst at a given
concentration (Fig. 2C). Consistent with their low levels of activity
in previous reports,[33,36] the commercial catalysts (Cat1-4, Fig.
2D) displayed at best moderate yields (~35% in presence of 50
µM Cat2) and low TONs (~18 for 6.25 µM Cat4, Table S1).
Conversely, the more efficient catalysts Ru1-8 gave rise to higher
conversions and TONs (Figs. 2E-F). For example, quantitative
conversion (>95%) was observed at high concentrations (≥100
µM) for the quinoline-2-carboxylate bearing complexes, Ru1-4,
with Ru3 also giving rise to >60 turnovers (Fig. 2E and Table S1).
While Cp-containing complexes Ru1 and Ru3 outperformed the
corresponding Cp* derivatives, this difference was more
pronounced for the related 8-hydroxyquinolinate-ligated
complexes, Ru5-8, for which only Cp complexes Ru5 and Ru7
displayed good activities (Fig. 2F). Another distinct feature of
these two complexes was a lack of activity at concentrations >25
µM. Consistent with lower OD600 values for high catalyst loadings
after the reaction, this observation presumably reflects toxicity of
Ru5 and Ru7 at high concentrations rather than a lack of activity.
Consequently, the highest yields (>65%) for Ru5 and Ru7 were
observed at a concentration of 12.5 µM. Notably, at low
concentrations both catalysts were able to perform ≈200
turnovers in LB media and in presence of live E. coli cells (Table
S1).
To confirm that the fluorescence signal results from the site-
specific incorporation of the in-situ synthesized p-ClF, we purified
GFP variants after performing the deprotection reaction of alloc-
p-ClF with Ru3 (50 µM) and Ru7 (12.5 µM) in 100 mL E. coli
cultures (see Supplementary Information for details). Only in
presence of either transition metal catalyst, the addition of the
ncAA precursor resulted in production of full-length GFP (as
judged by SDS-PAGE), with UPLC/MS analysis confirming the
successful incorporation of p-ClF (Fig. S3).
To independently validate the observed yields/TONs and
apparent toxicities for some catalysts, we recovered samples
from the screen and (1) quantified the concentration of p-ClF by
HPLC and (2) determined the number of culturable cells on solid
media after overnight growth (see Supporting Information for
details). As neither of these methods lend themselves to the same
level of parallelization as the screening platform, the independent
validation was restricted to four complexes, Ru3-5 and Ru7.
When comparing yields determined by HPLC with those obtained
from relative GFP fluorescence, we observed a good correlation
for Ru3 and Ru4 for all concentrations, while Ru5 and Ru7 only
showed comparable yields at lower concentrations (Figs. 3A-D).
As expected, quantitative conversions measured by HPLC at high
concentrations for these two catalysts contrasted those obtained
by GFP fluorescence, a readout that takes biocompatibility into
account. Further evidence that high concentrations of Ru5 and
Ru7 are indeed not biocompatible derives from the fact that we
did not observe growth of E. coli on solid media following the
reaction (Fig. 3E and Fig. S4). While we still observe a significant
Fig. 3: A-D: Comparison of yields determined by GFP fluorescence (y-scale) and by HPLC quantification (x-scale) for varying concentrations of Ru3 (A), Ru4 (B),
Ru5 (C), and Ru7 (D). The dotted line indicates an ideal correlation between both quantification methods. Data points and error bars are the average yields and
standard deviations of at least 3 independent experiments (see also Table S2). E: Bar graph showing the number of culturable cells (in colony forming units (CFU)
per mL sample) after deprotection of alloc-p-ClF with varying concentrations of Ru3-5 and Ru7. Blue stars indicate that no colonies were obtained after overnight
incubation. F: Effect on the catalytic performance of varying concentrations of Ru3 in presence of different co-solvents (all 2.5 % (v/v)). Colors represent the average
yields (red) and TONs (blue) of three independent measurements (see Table S3 for values and standard deviations). G: Catalytic performance of Ru3 at decreasing
concentrations after incubation with growing E. coli cells for up to 3 hours prior to addition of the ncAA precursor. Colors represent the average of three independent
measurements (see Table S4 for values and standard deviations). Color code same as in Fig. 3F.
This article is protected by copyright. All rights reserved.

10.1002/chem.201904808
Chemistry - A European Journal
COMMUNICATION
decrease of culturable cells for Ru5 or Ru7 at concentrations we
observed the highest yields (12.5 µM), these conditions seem to
offer the best compromise between performance and
biocompatibility. Notably, neither Ru3 nor Ru4 display any
significant toxicity in the tested conditions (Fig. 3E and Fig. S4).
Based on its negligible toxicity and good performance, we
selected Ru3 to demonstrate that our screening platform could
guide the fine-tuning of reaction conditions in the future.
Anticipating that a biocompatible catalyst needs to perform under
varying conditions, we first studied the effect of different co-
solvents on catalyst performance. When comparing yields and
TONs in presence of either acetone, dioxane, ethanol, or DMSO
(all 2.5% (v/v)), Ru3 displayed comparable activities in all four co-
solvents at high concentrations, while DMSO was the preferred
solvent for low catalyst loadings (Fig. 3F and Table S3). Lastly, a
biocompatible catalyst should also function in concert with cells
and retain its activity over an extended period of time. To
determine the extent Ru3 undergoes deactivation in presence of
growing E. coli cultures, we added the catalyst up to 3 hours prior
to addition of the substrate. Notably, Ru3 proved durable and
retained >50% of its initial activity over the 3 hours period (Fig.
3G and Table S4). Combined, these results augur well that
biocompatible catalysts, such as Ru3, can perform in concert with
cells under varying conditions and over extended periods of time,
and thereby, will find future applications in constructing cellular
factories that produce high-value compounds on demand.
Keywords: biocompatible catalysis • non-canonical amino acids
• catalyst screening • uncaging • metabolism
[1]
J. D. Keasling, A. Mendoza, P. S. Baran, Nature 2012, 492, 188–
189.
[2]
[3]
[4]
R. W. Hoffmann, Angew. Chemie Int. Ed. 2013, 52, 123–130.
J. D. Keasling, Science 2010, 330, 1355–1358.
S. Wallace, E. E. Schultz, E. P. Balskus, Curr. Opin. Chem. Biol.
2015, 25, 71–79.
[5]
J. G. Rebelein, T. R. Ward, Curr. Opin. Biotechnol. 2018, 53, 106–
114.
[6]
A. H. Ngo, S. Bose, L. H. Do, Chem. Eur. J. 2018, 24, 10584–
10594.
[7]
Y. Lee, A. Umeano, E. P. Balskus, Angew. Chemie Int. Ed. 2013,
52, 11800–11803.
[8]
G. Sirasani, L. Tong, E. P. Balskus, Angew. Chemie Int. Ed. 2014,
53, 7785–7788.
[9]
S. Wallace, E. P. Balskus, Angew. Chemie Int. Ed. 2016, 55, 6023–
6027.
[10]
[11]
[12]
D. W. Domaille, G. R. Hafenstine, M. A. Greer, A. P. Goodwin, J. N.
Cha, ACS Sustain. Chem. Eng. 2016, 4, 671–675.
R. M. Yusop, A. Unciti-Broceta, E. M. V. Johansson, R. M. Sánchez-
Martín, M. Bradley, Nat. Chem. 2011, 3, 239–243.
J. T. T. Weiss, J. C. C. Dawson, K. G. G. Macleod, W. Rybski, C.
Fraser, C. Torres-Sánchez, E. E. E. Patton, M. Bradley, N. O. O.
Carragher, A. Unciti-Broceta, Nat. Commun. 2014, 5, DOI
10.1038/ncomms4277.
[13]
[14]
A. M. Pérez-López, B. Rubio-Ruiz, V. Sebastián, L. Hamilton, C.
Adam, T. L. Bray, S. Irusta, P. M. Brennan, G. C. Lloyd-Jones, D.
Sieger, et al., Angew. Chemie Int. Ed. 2017, 56, 12548–12552.
J. P. C. P. C. Coverdale, I. Romero-Canelón, C. Sanchez-Cano, G.
J. J. Clarkson, A. Habtemariam, M. Wills, P. J. J. Sadler, Nat.
Chem. 2018, 10, 347–354.
[15]
[16]
[17]
M. Yang, J. Li, P. R. Chen, Chem. Soc. Rev. 2014, 43, 6511–6526.
S. Wallace, E. P. Balskus, Curr. Opin. Biotechnol. 2014, 30, 1–8.
Y. M. M. Wilson, M. Dürrenberger, E. S. S. Nogueira, T. R. R. Ward,
J. Am. Chem. Soc. 2014, 136, 8928–32.
In summary, our work introduces
a versatile and
operationally-simple screening platform to evaluate and optimize
biocompatible small-molecule catalysts. Specifically, the
incorporation of an in-situ synthesized ncAA into GFP through
stop-codon suppression yields a fluorescence readout that
accurately accounts for both the performance and biocompatibility
of a catalyst. Moreover, the screen can be performed in 96-well
format, enabling a rapid and straightforward assessment of
potential catalysts in parallel in small volumes. As long as the
activity of a non-enzymatic transformation can be linked to the
synthesis of a genetically-encodable ncAA,[27–29] the platform
should readily be applicable to evaluate biocompatible catalysts
for other types of transformations. Beyond discovery efforts, we
expect that our method will also allow for the fine-tuning of
reaction conditions in a workflow that is akin to method
development in organic synthesis. While we are only at the
beginning of seamlessly merging small-molecule catalysts with
cellular metabolism, biocompatible catalysts identified via this
screen could ultimately upgrade cellular metabolism and find use
in biotechnological or biomedical applications.
[18]
T. Völker, F. Dempwolff, P. L. L. Graumann, E. Meggers, Angew.
Chemie Int. Ed. 2014, 53, 10536–10540.
[19]
[20]
T. Völker, E. Meggers, ChemBioChem 2017, 18, 1083–1086.
B. Rubio-Ruiz, J. T. T. Weiss, A. Unciti-Broceta, J. Med. Chem.
2016, 59, 9974–9980.
[21]
[22]
[23]
[24]
M. Jeschek, R. Reuter, T. Heinisch, C. Trindler, J. Klehr, S. Panke,
T. R. Ward, Nature 2016, 537, 661–665.
J. Zhao, J. G. Rebelein, H. Mallin, C. Trindler, M. M. Pellizzoni, T. R.
Ward, J. Am. Chem. Soc. 2018, 140, 13171–13175.
H.-T. Hsu, B. M. Trantow, R. M. Waymouth, P. A. Wender,
Bioconjug. Chem. 2016, 27, 376–382.
S. L. Choi, E. Rha, S. J. Lee, H. Kim, K. Kwon, Y. S. Jeong, Y. H.
Rhee, J. J. Song, H. S. Kim, S. G. Lee, ACS Synth. Biol. 2014, 3,
163–171.
[25]
[26]
K. M. Esvelt, J. C. Carlson, D. R. Liu, Nature 2011, 472, 499–503.
Y. Okamoto, R. Kojima, F. Schwizer, E. Bartolami, T. Heinisch, S.
Matile, M. Fussenegger, T. R. Ward, Nat. Commun. 2018, 9, 1–7.
A. Dumas, L. Lercher, C. D. Spicer, B. G. Davis, Chem. Sci. 2015,
6, 50–69.
[27]
[28]
[29]
[30]
D. D. Young, P. G. Schultz, ACS Chem. Biol. 2018, 13, 854−870.
J. W. Chin, Nature 2017, 550, 53–60.
J. G. Rebelein, Y. Cotelle, B. Garabedian, T. R. Ward, ACS Catal.
2019, 9, 4173–4178.
[31]
[32]
J. Li, P. R. Chen, Nat. Chem. Biol. 2016, 12, 129–137.
C. Adam, A. M. M. Pérez-López, L. Hamilton, B. Rubio-Ruiz, T. L. L.
Bray, D. Sieger, P. M. M. Brennan, A. Unciti-Broceta, Chem. Eur. J.
2018, 24, 16783–16790.
[33]
[34]
[35]
[36]
[37]
J. Li, J. Yu, J. Zhao, J. Wang, S. Zheng, S. Lin, L. Chen, M. Yang,
S. Jia, X. Zhang, et al., Nat. Chem. 2014, 6, 352–61.
S. Tanaka, H. Saburi, M. Kitamura, Adv. Synth. Catal. 2006, 348,
375–378.
N. Li, R. K. V. Lim, S. Edwardraja, Q. Lin, J. Am. Chem. Soc. 2011,
133, 15316–15319.
C. Streu, E. Meggers, Angew. Chemie Int. Ed. 2006, 45, 5645–
5648.
D. D. Young, T. S. Young, M. Jahnz, I. Ahmad, G. Spraggon, P. G.
Schultz, Biochemistry 2011, 50, 1894–1900.
Acknowledgements
This work was supported the Netherlands Organisation for
Scientific Research (NWO, Veni grant 722.017.007).
This article is protected by copyright. All rights reserved.

10.1002/chem.201904808
Chemistry - A European Journal
COMMUNICATION
Entry for the Table of Contents
COMMUNICATION
Rudy Rubini, Ilya Ivanov, Dr. Clemens
Mayer*
Page No. – Page No.
A screening platform to identify and
tailor biocompatible small-molecule
catalysts.
A multi-well screen that rapidly reports on the catalytic activity of transition-metal
catalysts in presence of live cells is reported. The activity of a catalyst to yield a
non-canonical amino acid (ncAA) from an appropriate precursor is coupled to the
incorporation of the nonstandard building block into GFP (=quantifiable readout).
This article is protected by copyright. All rights reserved.