Automated Determination of Oxygen‐Dependent Enzyme Kinetics in a Tube‐in‐Tube Flow Reactor

Abstract Enzyme‐mediated oxidation is of particular interest to synthetic organic chemists. However, the implementation of such systems demands knowledge of enzyme kinetics. Conventionally collecting kinetic data for biocatalytic oxidations is fraught with difficulties such as low oxygen solubility in water and limited oxygen supply. Here, we present a novel method for the collection of such kinetic data using a pressurized tube‐in‐tube reactor, operated in the low‐dispersed flow regime to generate time‐series data, with minimal material consumption. Experimental development and validation of the instrument revealed not only the high degree of accuracy of the kinetic data obtained, but also the necessity of making measurements in this way to enable the accurate evaluation of high K MO enzyme systems. For the first time, this paves the way to integrate kinetic data into the protein engineering cycle.

Rolf H. Ringborg + , [a, b] AsbjørnT oftgaard Pedersen + , [a] and John M. Woodley* [a] Enzyme-mediated oxidationisofparticularinterest to synthetic organic chemists. However,t he implementation of such systems demands knowledgeo fe nzymek inetics. Conventionally collecting kinetic data for biocatalytic oxidations is fraught with difficulties such as low oxygen solubility in water and limited oxygens upply.H ere, we presentanovel method for the collection of such kinetic data using ap ressurized tube-in-tube reactor,o perated in the low-dispersed flow regime to generate time-series data, with minimal materialc onsumption. Experimental development and validation of the instrument revealed not only the high degree of accuracyo ft he kineticd ata obtained, but also the necessity of making measurements in this way to enable the accurate evaluationo fh igh K MO enzyme systems. For the first time, this paves the wayt oi ntegrate kinetic data into the protein engineering cycle.
Selectiveo xidationi so ne of the most important transformations in synthetic organicc hemistry. [1][2][3] Then ecessity of achieving highr eactiony ield in such transformationsm akes enzymes particularly interesting as potentialcatalysts,onaccount of their exquisite selectivity in comparison witht heir chemo-catalytic counterparts.H owever,f or process applicationi ti so ften difficultt or each ther equiredr eaction intensity( reactionr atea nd product concentration). In particular,i ssuess ucha sl ow enzymatic activity,p roduct/substrate inhibition, co-factor regenerationa nd unfavorablet hermodynamice quilibrian eedt ob e solved using biocatalyticr eactione ngineering.T hese problems arec ommonly investigatedb ys tudying thek inetic behavior of an enzyme under differentc onditions. Subsequently,u sing these data, the challenges in reaching the required productivity canb eaddressede itherb yp rotein engineeringo r, alternatively, process engineeringt oc ircumvent kinetic limitations. However, it would be muchm oree ffectivei fs olutionsa rose from ac om-bination of botha pproaches. Regardless of thea pproach taken, enzyme improvementn aturally starts in theh ands of the protein engineer who typically screensf or improved enzymes using singlep oint measurements (i.e.a tasingle substrate concentration) to go throughm anye nzymev ariants. [4] In this way, protein engineeringi sa blet od eliver improved enzymes,a lso catalyzing the conversiono fn on-natural substrates. [5] However, single point measurements can only reveala pparentk inetic constants,s ucha st he so-called specificity constant (V max /K M ), which canb em isleadinga st he basis fors electing theo ptimal enzyme. [6][7][8] At points in developmenta tw hich selection is madef romasmallerp oolo fp rotein variants, it would be highly desirable to comprehensively quantifyt he kinetics, to havea na dequate basis for deciding on the best enzymef or a given reaction,and reactor configuration.Likewise,itisnecessary to determine the activity of an enzyme of interesto vert he fullr ange of potentialo peratingc onditionst ob ea blet ot ruly assesst he possibilities for process implementation.O nt his premise,w es uggestt hatc omprehensivek inetic investigations should be integrated intot he improvement cycleo fa ne nzyme for application.Inthisway it would be possible to direct screening to focus on evolving improved enzymatic kinetic properties, which arei deal for process implementation.T or ealize such a scheme, it is necessary to developa na utomatedc haracterizations ystem. [9] Herein,w ep resent ones uchs ystemf ocusedo n collectingkinetic datafor oxygen-dependentenzymes.
On studying enzymek inetics, it is important to measurei nitial rates at substratec oncentrations well above, as well as below,t he true Michaelis constant(s),t od etermine kinetic parameters with sufficient accuracy.I nt he study of oxygen-dependent enzymes, such investigations are notoriously difficult as ar esult of the limited solubility of oxygen in water,a nd to some extent, of the concomitant limited supply rate of oxygen. The challenge of controlling the oxygen concentration leads in many cases to conducting experiments at as ingle oxygen concentration (usually that in water,i ne quilibrium with air,a t2 76 mm). Air saturationi sh oweveri nsufficient to achieve enzyme saturation for severali ndustrially interesting oxidases [10][11][12] and, in any case, it introduces uncertaintyi nto parameter estimations. Indeed,c onventional experiments can only reveal apparent Michaelis constantsw hich are confined to the tested parameter space and should therefore be compared with great care. Likewise, oxygen supply is often carried out by bubbling air through the reactions olution. However, in doing so, it is necessary to consider the stripping of any volatile substrate(s)a nd product(s), as well as potential enzyme deactivation at the gas-liquid interface. [13] The constraint on the limited dissolved oxygen concentration in water can be alleviated by pressurizing the reactor or by using enriched air (to increase KGaA. This is an openaccessarticleunder the termsoft he Creative Commons AttributionL icense, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. the partial pressure), whereas the interfacial effect can only be alleviated by introducing ap hysicalb arrier betweent he gas and the liquid.
Recently,t he Te flon AF-2400 fluoropolymer, [14] which is characterized by high gas permeability,h as been used as am embrane in the latestd evelopment of the so-called Tube-in-Tube Reactor (TiTR) design [15] which has previously proven to be useful for the supply of gaseous substrates to liquid reaction media while retaining the chemical resistance of traditional fluoropolymers. [16,17] The TiTR is made of an inner Te flon AF-2400 tube encased within an outer PTFE tube with low oxygen permeability.Amixture of oxygen and nitrogen is supplied in the space between the two tubes, whereby the oxygen can be transferredt ot he liquid reactionm ixture in the inner tube through the membrane. We reasoned this would make the TiTR ideal for studying the kinetics of oxygen dependentb iocatalytic reactions, since the challenges of conventional systems can be avoidedb yc reatingabubble-free aeration system.T he small dimensions of the inner tube (I.D/O.D. 230/ 410 mm) maximize the surface-to-volumer atio, which combined with the high oxygen permeability of Te flon AF-2400, enablesv ery high oxygen supply rates. This TiTR allows operation at dissolved concentrations of oxygen very closet ot he equilibriumv alue between the gas phase and the reaction medium, despite al ow driving force (i.e. the reactor will operate at ad issolved oxygen concentration within 99 %o fs aturation). Additionally,b yp ressurizing both the inner and outer tube, the oxygen solubility in the reaction mixture can be increasedp roportionally.T he setup therefore allows control over oxygen as as ubstrate in oxygen-dependent enzyme reactions. Furthermore, the TiTR satisfies the requirement for negligible change in substrate concentration for measurement of initial rates, since oxygen can be supplied along the reactor as it is consumed. Based on this concept, as ystem suitable for kinetic characterization of oxygen dependente nzymes was developed by combining the TiTR with precise liquid and gas supply systems and connecting the outlet of the inner tube to aU V/Vis detector.B ym eans of as witch valve, samples werec arried from the injection loop into the detector, where the solution was subjected to flow injection analysis.
Although such ar eactor is very useful for conducting oxygen-dependent enzyme reactions (under pressure), we realized that af urther development was still necessary for the meaningful collection of kinetic data. Laboratory flow reactors typicallyo perate in the laminar flow regime with large axial dispersion, which requires steady-statee xperiments. Such experiments often consume more material over al ongert ime period and with al ower sampling frequency than those performed in equivalent batch apparatus. [18] Recently,areview of Ta ylor'sw ork regarding mixing and dispersion [19] hasl ed to the application of low dispersed flow in microreactors. [20] This is a unique regime of laminarf low that occurs only at am icrofluidic scale. [20] In this flow regime, the radial mixing from the center of the tube to the edges is governed solely by diffusion. At the microscale, the diffusion lengths are by definition very small and this will in turn give very short radial mixingt imes. Low dispersed flow will therefore flatten the well-known "tongue" profile of laminar flow,a nd solute concentrations will therebyo nly change along the length of the reactor.C onsequently,t he reactor can be described by plug-flow behavior, which was used in am ethod recently reportedb yM oore and Jensen. [21] In this method,a tl ow residence time, steady-state is obtained and the flow rate is subsequently ramped down. By following the conversion during the ramp, initial rate measurements( i.e. concentration-time profiles) are possible without the need to obtain multiple steady-states. Nevertheless, the re-portedM oore and Jensen method requires modification for biocatalysis. Low dispersed flow is very dependent on the diffusivity of the solutes,a nd the larges ize of enzymec atalysts translates into at wo order-of-magnitude lower diffusivity compared to smallm olecules (10 À11 cf. 10 À9 m 2 s À1 ). [22,23] Thea xial dispersion of enzymes will therefore be much more pronounced, indicating that enzymes are more dispersed along the length of the channel compared to the small molecule reactants and the resulting products.I tw as therefore necessary to make sure that the enzyme concentrationi nt he entire reactor volume remained constant. This was ensured by achieving steady-state with respect to the enzymec oncentration and thereafter keeping the enzymefeed concentration constant,independento ft he liquid flow rate. In this way,i tw as assumed that the degree of dispersion would be dependent on the diffusion coefficients of the substrate(s)a nd product(s) alone. The integrated combination of each of the aforementioned developments has led to the establishmento ft he currenti nstrument, which now gives an ovel and automated way of kinetically characterizingo xygen-dependent enzymes, see Figure 1. The specific details of the setup are described in the Supporting Information (SI).
To demonstrate the performance of the instrument, the well-known enzyme, glucose oxidase(GOx, E.C. 1.1.3.4), was selected. The GOx enzyme catalyzes the oxidation of glucoset o glucono-d-lactone, using molecular oxygen (which is itself reduced to hydrogen peroxide). Following the enzymatic reaction, glucono-d-lactone is spontaneously hydrolyzed to gluconic acid, which formation can be followed spectrophotometrically (see Supporting Information). The hydrogen peroxide formed is removed instantaneously by the addition of catalase, which enables its conversion into water and half the stoichiometric amount of oxygen. The removal of hydrogen peroxide forces the reaction to proceed in au nidirectionalm anner and also protects GOx from oxidation. GOx has been shown to follow ap ing-pong bi-bi reactionm echanism (Scheme 1) [24] for which ar ate expression can be derived (Equation (1)).
The flow manipulation method appliedt op roduce the equivalent batch data from the setup, requires an accurate determination of the reactor volume. Hence, initially,r esidence time distribution experiments were conducted to determine the volume of the reactor (155 AE 1.8 mL, see Supporting Information). Next, the results of the flow methodw erec ompared with steady-state operation, and it wass hown that the setup indeed produces time-series data even with the addition of a slow diffusing (bio)catalyst (see SI). Finally,t ov alidate the enzyme kinetics measured in the TiTR, equivalent experiments to those carried out in batch by Toftgaard Pedersen and coworkers [25] were conducted. In the batch experiments, the setup used an aerated stirred tank reactor with adjustable oxygen/nitrogen feed. The comparison revealed an excellent correlation between the two systems and the combinedr esults of the validation experimentsc onfirmed that the kinetics determined using the TiTR setupare reliable (Figure 2).
The fit of Equation (1) to thesed ata revealed ar elatively high Michaelis constant of 0.52 mm for oxygen (Table 1), which is also obtained from the unsaturated enzymek inetics observed at high glucosec oncentrationsa nd atmospheric pressure ( Figure 2). It is generally accepted, that to reliably quantify Michaelis constantsi ti sn ecessary to measure enzymek inetics in as ufficiently large range of substrate concentrations, comprising values that are 5-fold (as am inimum, and preferably 10-fold)h igher and lower than the true K M .I nt he TiTR setup, this was achieved by increasing the operating pressure of the setup to 6bar to increase the maximum dissolved oxygen concentration to 7.13 mm (using pure O 2 at 25 8C). Enzymes aturation was therebyo btained even at the highest concentration of glucose ( Figure 3), enabling am ore reliable prediction of all the kinetic parameters ( Table 1).
The TiTR setup wasf ully automateda nd computer controlled,t hereby enabling characterization of an oxygen-dependent enzyme within 24 hours with minimal manual labor. While the preparation of solutions is identicalf or both batch Figure 1. Experimental setup of the Tube-in-Tube Reactor.The threesyringe pumps on the left deliver al iquid solution to the inner membranet ube, illustrated by the orangeline. Twom assflow controllers are used to vary the gas composition in the range 5-100 %O 2 ,supplied to the outer tube.T he gas is wetted and heated beforee ntering the reactort oa void the stripping of waterf rom the inner tube. The gas was fed througha noutert ube, madeo fP TFE. A pressure regulator and am anometer were located at both ends of the two tubes to control the pressure,asw ell as to ensure an equalo rh igherp ressure on the liquid side of the membrane.
Scheme1.Cleveland representationo ft he glucose oxidaseping-pongb ibi mechanism. Ed enotes the oxidized free form of the enzyme whereas Fd enotes the reduced form of the free enzyme.  and TiTR, the batch setup requires four full days of labor.F urthermore, the smalld imensions of the system makei tp ossible to collect one initial rate measurement per 1.4 mL of reaction mixture,w hich is considerably less than the 150 mL required in the alternative sparged batch setup. In summary,w eh ave developed and validated an automated flow reactor system that rapidly and accurately determines the kinetics of oxygen-dependent enzymes. The tool allows perfect controlo ft he oxygen concentration in solution,w hich by pressurizing the system can enablev alues that are up to 25-fold higher than the values achievable by using merely air under atmospheric conditions. Operation in the lowd ispersed flow regime allowed the generation of time-series data with an enzymatic catalyst, despite its low diffusivity,a nd the resulting data were in good agreementw ithe xperiments conducted in ab atch system.T he system is capable of characterizing the kinetics of any enzyme within the oxidoreductase class (EC 1), for which reactions frequently result in changes to the UVspectra,t oe nablef acile quantification of conversion.T he application is however not limited to oxygen-dependente nzymes alone, but can in principle be used to study many other enzymesu sing gaseous substrates, such as hydrogenases (using H 2 ), [26] formated ehydrogenases (using CO 2 ) [27] or methane monooxygenases (using CH 4 ). [28] The tool presented here could introduce kinetic characterization of oxidoreductases into the catalyst development cycle, where biocatalytic reaction engineering can be used to guide both process and protein engineering. [9,29] The need to improvet his development cycle further is particularly important to facilitate the wider and more effective implementation of biocatalytic reactions, especially in the pharmaceutical industry. [30]