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

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 (quantifiable readout). 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.

Abstract: Interfacing biocompatible, small-molecule catalysis with cellular metabolism promises as traightforward introduction of new functioni nto organisms withoutt he need for genetic manipulation. However, identifying and optimizing synthetic catalysts that perform new-to-nature transformations under conditions that supportl ife is a cumbersomet ask. To enable the rapid discovery and finetuning of biocompatiblec atalysts, we describe a9 6-well screening platform that couples the activity of synthetic catalysts to yield non-canonical amino acids from appropriate precursors with the subsequent incorporationo f these nonstandard building blocks into GFP (quantifiable readout). Critically,t his strategy does not only provide a common readout (fluorescence) for different reaction/catalyst combinations,b ut also informs on the organism's fitness, as stop codon suppression relies on all steps of the centrald ogma of molecular biology.T os howcase our approach,w eh ave appliedi tt ot he evaluation and optimization of transition-metal-catalyzed deprotection reactions.
Synthetic chemists and metabolic engineers pursuec ontrasting approaches to make molecules. [1] While the former skillfully employ synthetic catalystsa nd reagents to build up complex molecules, the latter harness the reactivity of biocatalysts in living organisms to produce compoundsf rom fermentation. [2,3] Although these approaches have been traditionally considered to be incompatible, small-molecule catalysts that can interface with cellular metabolism have the potentialt oe xpand biological function without the need for genetic manipulation. [4][5][6] For example,s uch biocompatible catalysts could be parto fc ellular factories,i nw hich they perform new-to-nature transformations to diversify molecules producedby an organism. [7][8][9][10] Thus, such ac oncertede ffort of synthetic chemistry and metabolic engineering could pave the way toward the direct synthesis of value-added compounds in cellular settings.A dditionally,b iocompatible catalysis holds promise for biomedical applications, such as targeted drug release/synthesis, [11][12][13][14][15] the disruption of cell-cellc ommunication or rescuing dysfunctional enzymes involvedi nh uman diseases. [16] To enables uch developments, biocompatible catalysts have to perform ad ifficultb alancing act and functionu nder conditions that both support life and allow an abiological transformations to proceed.T his task is complicatedb yt he fact that synthetic chemists routinelyp erform reactions in organic solvents at temperaturea nd pH regimes that are incompatible with living organisms. Moreover,m etabolite concentrationsa re typicallyl ow (< 1mm), compared to the standard substrate concentrationse mployed in organic synthesis. [5,6,16] Conversely, the complex intra and extracellular environments of organisms contain am yriado fc ompounds that can poison exogenously supplied catalysts or reagents. [17] Consequently, the discovery and optimization of biocompatible catalysts and reactions remain challenging. Typically,aset of potential catalysts is first evaluatedf or am odel transformation under "biologically relevant conditions"( i.e.,i np resence of water,air and/or thiols) andpromising candidates are subsequently tested in biological settings. [12,[18][19][20] The initial evaluation, however, takes neither catalyst/reagentt oxicity nor catalyst poisoning by the organism into account. More recently, evaluating biocompatible transition-metal complexesh as 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 as uccessful transformation. Unfortunately,t he observable phenotypesi nt hese screens do not depend on ac ellular process and therefore, do not account for an organism's fitness. To address these challengesa nd furthers treamline discoverya nd optimization efforts for biocompatible catalysts, herein, we describea96-well screening platform that rapidly reports on both the activity of ac atalyst andt he fitnessoft he organism.
Inspired by the use of geneticc ircuitry for the directed evolution of enzymes, [24][25][26] we reasoned that ar apid evaluation and optimization of biocompatible catalystsr equires ad irect link between ac atalyst's activity and an observable phenotype that can only arise in living organisms. Althoughr eplacing enzymatict ransformationsi nm etabolicp athways with synthetic ones is ap ossibility to establish such al ink, [7] this strategy is not general and would require genetick nock-outst hat are pronet of alse positives/negatives. Instead, we envisioned to introduce as imple pathway that 1) is dependent on metabo-lism while not affecting viability itself;2 )can readily be employed for different reaction types;a nd 3) can function in organismsranging from bacteria to mammalian cell lines.
Ap rocess that matches these criteria is the site-specific incorporation of non-canonical amino acids (ncAAs) into proteins of interest through the suppression of as top codon by the action of an orthogonal translation system (OTS). [27][28][29] To repurpose such OTSs for the evaluation and fine-tuning of biocompatible catalysts, we surmised that the activityo fs mall-molecule catalysts to give ncAAs from appropriate precursors could be coupledw ith the subsequent incorporation of these artificial buildingb locks into green fluorescent protein (GFP) variants ( Figure 1A). Based on these considerations, we constructed as creening platform that comprises three main components:1 )the exogenouslys upplied catalysta nd ncAA precursor (input);2 )anO TS specific for the chemically synthesized ncAA (sensor);a nd 3) aG FP variant featuring an in-frame stop codon (reporter). Critically,o nly upon suppression of the inframe stop codon (UAG) full length GFP is produced. Thus, the fluorescences ignal detected should relate to ncAA production levels and, as ar esult, report catalystp roficiency.M oreover, ncAA incorporation relies on all the steps of the centrald ogma of molecular biology,a nd therefore should also report on the fitnesso ft he organism (Escherichia coli in this study).
To evaluatet he feasibility of the proposed screening platform, we identified the uncaging of allyloxycarbonyl( alloc)protected amines as am odel reactionf rom 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 an cAA ( Figure 1B). A number of differentc atalysts have been shown to catalyze this transformation with varying efficiencies. Herein, we selected a set of 12 catalysts ( Figure 1C): four commercially available ones for which low activities were reported (Cat1-4) [33][34][35][36] and a total of eight ruthenium-based half-sandwichc omplexes featuring either aq uinoline-2-carboxylate (Ru1-4) [18] or a8 -hydroxyquinolinate (Ru5-8) [19] as bidentate ligand, with some of them showingi mproved 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 ar eliable and quantifiable readout. For this, E. coli was transformed with two plasmids that encode1 )anO TS based on the promiscuous aminoacyl-tRNA synthetase, pCNF-RS; [37] and 2) aG FP variant featuring aU AG stop codon (eitherY 151* or Y182*, see the Supporting Informationf or details). Next, we monitored GFP fluorescencei n9 6 well-plates after induction of gene expression in LB media containingd ifferent concentrations of p-chlorophenylalanine (p-ClF) or O-methyltyrosine (OMeY) over ap eriod of ten hours. As was anticipated, GFP production was dependent on the concentration of the ncAA (Figure 2A andF igure S1 in the Supporting Information). Plotting the relative fluorescenceincrease after 200 minutes for the differentc oncentrations yielded a linear correlation, independent of whichn cAA and GFP variant was used ( Figure 2B and Figure S2 in the Supporting Information). In contrast, addition of alloc-protected ncAAsd id not result in as ignificant increasei nG FP fluorescence, ensuringa good signal-to-noise ratio over two orderso fm agnitude (10-1000 mm).
To evaluate the proficiencies of the selected transition-metal complexes to catalyzet he deprotectiono fa lloc-p-ClF,w e added decreasing concentrationso fe ach catalyst to the ncAA precursor (1 mm as ar acemate) at the time of induction. The 96-well setup of the screening platform enabled the evaluation of 88 differentcombinations (+ 8samples with varying concentrations of p-ClF for the calibration)i nl ess than four hours. We used the relative increase in GFP fluorescencea fter 200 minutes to estimate the yields and turnover numbers (TONs) of a catalysta tagiven concentration ( Figure 2C). Consistent with their low levels of activity in previous reports, [33,36] the commercial catalysts (Cat1-4, Figure 2D)d isplayed at best moderate yields ( % 35 %i np resence of 50 mm Cat2) and low TONs ( % 18 for 6.25 mm Cat4, Ta ble S1 in the SupportingI nformation). Conversely, the more efficient catalysts Ru1-8 gave rise to higher conversionsa nd TONs (Figures 2E-F). For example, quantita- tive conversion (> 95 %) was observed at high concentrations (! 100 mm)f or the quinoline-2-carboxylate bearingc omplexes, Ru1-4, with Ru3 also giving rise to > 60 turnovers ( Figure 2E and Ta ble S1 in the Supporting Information). Although Cp-containing complexes Ru1 and Ru3 outperformed the corresponding Cp* derivatives, this difference was more pronouncedf or the related 8-hydroxyquinolinate-ligated complexes, Ru5-8,f or which only Cp complexes Ru5 and Ru7 displayed good activities ( Figure 2F). Another distinct feature of these two complexes wasalack of activity at concentrations > 25 mm.C onsistent with lower OD 600 values for high catalyst loadings after the reaction, this observation presumably reflectst oxicity of Ru5 and Ru7 at high concentrations rather than alack of activity.C onsequently, the highest yields (> 65 %) forR u5 and Ru7 were observed at ac oncentrationo f1 2.5 mm.N otably,a tl ow concentrations both catalysts werea ble to performa pproximately 200 turnovers in LB media and in presence of live E. coli cells (Table S1 in the Supporting Information).
To confirm that the fluorescences ignal results from the sitespecific incorporation of the in situ synthesized p-ClF,w ep urified GFP variants after performing the deprotection reaction of alloc-p-ClF with Ru3 (50 mm)a nd Ru7 (12.5 mm)i n1 00 mL E. coli cultures( see the Supporting 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 was judged by SDS-PAGE), with UPLC/MS analysis confirming the successful incorporation of p-ClF ( Figure S3 in the Supporting Information).
To independently validate the observed yields/TONs and apparentt oxicities for some catalysts,w er ecovered samples from the screen and 1) quantifiedt he concentration of p-ClF by HPLC;a nd 2) determined the number of culturable cells on solid media after overnight growth (see the Supporting Information for details). Because neither of these methods lend themselves to the same level of parallelization as the screening platform,t he independentv alidation was restricted to four complexes,Ru3-5 and Ru7. When comparing yields determined by HPLC with those obtained from relative GFP fluorescence, we observed ag ood correlation for Ru3 and Ru4 for all concentrations,w hereas Ru5 and Ru7o nly showed comparable yields at lower concentrations ( Figures 3A-D). As was expected, quantitative conversions measured by HPLC at high concentrations for these two catalysts contrasted those obtained by GFP fluorescence, ar eadout that takes biocompatibility into account.F urther evidencet hat high concentrations of Ru5 and Ru7 are indeed not biocompatibled erives from the factt hat we did not observe growth of E. coli on solid media following the reaction( Figures 3E and S4 in the SupportingI nformation). Although we still observed as ignificant decreaseo fc ulturable cells for Ru5 or Ru7 at concentrationst hat give rise to the highest yields (12.5 mm), these conditions seem to offer the best compromise between performance and biocompatibility. Notably,n either Ru3 nor Ru4 display any significant toxicity in the tested conditions ( Figures 3E andS 4i nt he Supporting Information).
Based on its negligible toxicity and good performance, we selected Ru3 to demonstrate that our screening platform could guide the fine tuning of reactionc onditions in the future.Anticipating that ab iocompatible catalystn eeds to perform under varying conditions, we first studied the effect of differentc o-solvents on catalyst performance. When comparing yields and TONs in presence of either acetone, dioxane, ethanol,o rD MSO (all 2.5 %( v/v)), Ru3 displayed comparable activitiesi nall four co-solvents at high concentrations, whereas 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 three hours prior to addition of the substrate. Notably,R u3 proved to be durable and retained > 50 %o fi ts initial activity over the three hours period ( Figure 3G and Table S4 in the Supporting  Information). Combined, these resultsa ugur wellt hat biocompatible catalysts, such as Ru3, can perform in concert with cells under varying conditions ando ver extended periodso ft ime, and thereby,w ill find future applicationsi nc onstructing cellular factories that produce high-value compounds on demand.
In summary,o ur work introduces av ersatile and operationally simple screening platform to evaluatea nd optimize biocompatible small-molecule catalysts. Specifically,t he incorporation of an in situ synthesized ncAA into GFP through stop-codon suppression yields af luorescence readout that accurately accounts for both the performance and biocompatibility of ac atalyst. Moreover,t he screen can be performed in 96-well format, enabling ar apid ands traightforwarda ssessment of potential catalysts in parallel in small volumes. As long as the activity of an on-enzymatic transformation can be linked to the synthesis of ag enetically encodable ncAA, [27][28][29] the platform should readily be applicable to evaluateb iocompatible cata-lysts for other types of transformations. Beyondd iscoverye fforts, we expect that our methodw ill also allow the fine tuningo fr eaction conditions in aw orkflowt hat is akin to method development in organic synthesis. Although we are only at the beginning of seamlessly merging small-molecule catalysts with cellular metabolism, biocompatiblec atalysts identified through this screen could ultimately upgrade cellular metabolism and find use in biotechnological or biomedical applications.