Dynamic Covalent Organocatalysts Discovered from Catalytic Systems through Rapid Deconvolution Screening

The first example of a bifunctional organocatalyst assembled through dynamic covalent chemistry (DCC) is described. The catalyst is based on reversible imine chemistry and can catalyze the Morita–Baylis–Hillman (MBH) reaction of enones with aldehydes or N-tosyl imines. Furthermore, these dynamic catalysts were shown to be optimizable through a systemic screening approach, in which large mixtures of catalyst structures were generated, and the optimal catalyst could be directly identified by using dynamic deconvolution. This strategy allowed one-pot synthesis and in situ evaluation of several potential catalysts without the need to separate, characterize, and purify each individual structure. The systems were furthermore shown to catalyze and re-equilibrate their own formation through a previously unknown thiourea-catalyzed transimination process.


Introduction
Synthesis of new catalysts is critical for modern synthetic chemistry,b ut catalyst discovery is commonly based on timeconsuming and frustratingt rial-and-error protocols. To address this issue, many combinatorial approaches to accelerate the process have been developed. [1, 2} However, combinatorial catalysis has been hamperedb yl imiteda ccess to structurally diverse systems, in particularw ith bifunctional scaffolds.N ontrivial synthetic operations are commonly required for their assembly,w hich renders the systems unsuitablef or automated high-throughput synthesis. Furthermore, as ignificant drawback of most combinatorial catalytic protocols is the requirement for all candidates to be purified, characterized, and evaluated individually,r egardlesso ft heir activity.T herefore, collective catalyst screening is highly desirable, although only af ew pioneering reports have been described. [3] In recent years,s ubstantial effort has been invested into the design of modular and responsive catalysts, in which the activity can be controlled through secondaryi nputs. In particular, highly successfuls elf-assembled supramolecular catalysts with tunable activity have been developed for transition-metal cat-alysis [4] and organocatalysis, [5] providing quick and facile routes to bifunctional catalyst scaffolds. Elegant studies by the Reek [6] and Breit groups, [7] have also shown the potential for simplified screening of such systems by deconvolution methods. However,s upramolecular assemblies lack the robustness of covalent linkages.D ynamic covalent chemistry (DCC) uses reversiblec ovalentb onds to mimic the adaptive nature of supramolecular systems,w hile retaining the advantages of well-defined,s table covalentc ompounds. [8] For example, DCC has been successfully used for ligand/receptor identification, [9] molecular-interaction analysis, [10] kinetic processes, [11] biopolymers, [12] and chemical reactionnetworks. [13] Due to the high interesti nd eveloping tunable catalysts and catalytics ystems, we became interested in the possibility of creating such a" dynamic" catalyst and investigating itsproperties. There are furthermore no knownb ifunctionalc atalysts, in which the two functional parts are connected by ar eversible covalentb ond. The application of DCCf or catalyst discovery has otherwiseb een al ong-standingg oal. [14] Early examples relied on adaptive hosts ystems that re-equilibrate in the presence of at ransition-state analogue( TSA), leading to amplification of the host that in theory best stabilizes the transition state. [15] However,t his leads to an eed for design and synthesis of the TSA, andt he screening process mayr esult in ah ostt hat only binds the TSA withoutp ossessing anya ctual catalytic activity.
Because dynamic covalent chemistry is equipped with ad eveloped frameworkf or analysis of large mixtures, we imagined ap ossibility to directly find an optimal dynamic catalystf or ag iven reactionf rom al arge adaptive system.H erein, we have developed am ethod for the dynamic combinatorial synthesis of systems of bifunctionalc atalysts, followed by in situ identification of the optimalc atalyst.T he methodology was applied to the challenging Morita-Baylis-Hillman (MBH) reaction, and as electiveb ifunctional catalyst with interesting properties was discovered. [16] This method circumventsp revious issues with DCC and catalysis by directly screening towards the actual chemicaltransformation in ak inetic manner.

Results and Discussion
In bifunctional catalysis, two functional groups capable of activating substrates are mounted on one scaffold. [17] It was hypothesized that if such as caffold incorporated ar eversible bond as shown in Figure 1a,adynamic combinatorial system of potentialb ifunctional catalysts could be generated. By allowing the system to reach equilibrium,apredictable product distribution dictated only by the relative thermodynamic stability of the catalysts would be obtained. Thus, dynamic deconvolution with selective component removal can be used to evaluate the effect of each component (Figure 1b). [18] Note that the thermodynamic nature of the key bond connection is essential for the accuracy of the deconvolution approach. Performing the same deconvolutiono nm ixtures,i nw hich the bifunctional catalysthas been constructed under kinetic control is not afeasible methodology.S uch systems are highlyv ulnerable to kinetic traps, resulting in ar isk of active catalysts being unexpressed in the mixture. For ad ynamic system,a sl ong as the buildingb locks utilized for constructing the catalysts are relatively uniform in terms of the dynamic covalent functional group, all possible linear combinations should be expressed in the system in predictable ratios.
To utilize this DCC methodology for discoveryo fd ynamic bifunctional catalysts, we required as ynthetically relevant model transformation. The MBH reaction was chosen, because organocatalysis has provent ob eh ighly successful for this transformation, and the importance of bifunctionality has been well investigated. [19] Furthermore, studies have found that optimal catalysta rchitectures were difficult to predict through rational design,w hich together with the often very long reaction times highlighted an eed for rapid catalyst screening methods. [19b, e] Traditionally,M BH reactions utilizing a,b-unsaturatedk etones as donors are also hard to control, with polymerization and side-reactions often diminishing the efficiency. Accurate catalyst predictions for such ar eactionw ould indicate that the dynamic screening methodology possessed ahigh level of generality.
Thus, ar acemic catalysts ystem that incorporated an ucleophilic Lewis base, an H-bond donor and ad ynamic imine bond connecting the two components was designeda ss hown in Scheme1a. Acids andw ater render the imine bond labile,b ut removal of either component leads to as tructurally robust linkage.T his "conditional reversibility" is essential, because ad ynamic catalyst should be able to equilibrate under one set of conditions and stay inert under another.A si llustrated in Scheme1b, the catalyst shoulda ctivate both the enone and the aldehyde, and preorganize the substrates for conversion towardsthe MBH adduct.
The initial strategy was to first form the imines,a nd then allow the dynamic system to reache quilibrium in situ using an equilibration catalyst. This approachw as tested for the model system shown in Scheme 2, using components A, B, 1,a nd 2 to form imines A1, A2, B1,a nd B2 quantitatively.H erein, only component B2 fulfills the criteria for bifunctionality,b ecause it possessesb oth an ucleophilic tertiary amine moiety and an Hbond donating thiourea group.
However,u pon attempted re-equilibration by addition of catalytic amounts of water and widely used transimination catalysts, such as benzoic acid or Sc(OTf) 3 ,i tw as noticed that the component distribution in the imine system did not change. Controle xperimentsc onfirmed that the system had in fact already reached equilibrium during condensation (see the Supporting Information). This result was surprising, because amines and aldehydes in the absence of acida re known to condensate irreversibly under kinetic control.  It is known that thiourea moieties form strong Hb onds with imines.I tw as thus hypothesized that the thiourea NÀHp rotons could act as general acid catalysts for the system and selfcatalyzet he system synthesis, in which it takes part. Further control experiments indicated that thioureasa re indeed able to induce equilibration of dynamic imine systems, as long as water and/or amines are still present in the mixture (see the Supporting Information). We also confirmed that transimination did not proceed at all in the absence of these species, which supportsahydrolysis/condensation mechanism for the re-equilibration.T his effectively led to dynamic systems that were "locked" at equilibrium under dry conditions, because the water necessary for re-equilibration was continuously removed during the condensation phase. Furthermore, it was also confirmed that thiourea structures were capable of catalyzing the exchange even in the absence of primary amines,i ndicating that aliphatic aminet ransimination catalysis was not the sole factor at play.T ot he best of our knowledge, this is the first report of H-bond-catalyzed transimination outside of biological systems. Thisf inding greatlys implified our method, because the re-equilibration step shown in Scheme 2c ould be entirely omitted.F urthermore, it added af urtherl ayer of complexity to this potentialc atalyst class, because these dynamic thioureaimine catalysts are, in as ense, ablet om odify and catalyze their own formation.
With equilibration conditions in hand, the system was next expanded to four aldehydes and four amines, as shown in Scheme3,t oi ncrease the chances of findinga na ctive catalyst. Aldehydes 2, 3,a nd 4 comprise nucleophilic sites in the ortho positiont ot he imine linker,w hereas amines B, C,a nd D incorporate H-bond donors. Cyclohexylamine A and benzaldehyde 1 were used as controls. Ad ynamic catalyst system composed of 16 different imines was formed analogously to the model reaction, and equilibrium was again attained duringt he condensation phase. Next, ethyl vinyl ketonea nd p-nitrobenzaldehyde were added directly to the system as shown in Figure 2a.T he MBH reaction proceeded readily,a nd 20-25 %y ield of the desired adduct 5 was obtained after 24 h, as indicated by NMR analysis.T hus, at least one of the 16 potential catalysts in the mixturep ossessed MBHa ctivity.
Scheme3.Formation of dynamic 16-component imine system and "locking" by water removal. To minimizet he number of experiments requiredt oi dentify the activec omponents in the mixture, ad ynamic deconvolution scheme was devised, the resultso fw hich are shown in Figure 2b.E quimolara mountso ft he amine and aldehyde species were generallyr equired, because the formed imines were inert under MBH conditions even in the presence of thioureas. Hence, deconvolution could be efficiently accomplished through selective replacement of the evaluated component by an equivalent amount of ar eference compound (A for amines, 1 for aldehydes). Initialr ates were then measured to fully correlate systemic catalytic activity with changes in system compositionu pon component replacement. [20] Replacement of potentiallya ctivec omponents by inactive species would lead to retarded rates of the investigated reaction, compared with the complete system with all functionalities present (the reference bar in Figure 2b). Conversely, removal of ac omponent that is detrimental to catalytic activity should give enhanced initial rates.
As can be seen from Figure 2b,r eplacement of the dimethylamino-containing component 2 gave as light rate increase. A potentiale xplanation for this observation can be the systemic effects of bifunctionality in the catalyst system. Assuming one or more optimal combinations of nucleophile and H-bond donor,ascenario,i nw hich pairing of an inactive component with ap otentially active speciesw ould produce ab ifunctional catalystt hat exhibits low activity,c an be envisaged. If this pairing would be thermodynamically more preferred than pairing of two active components, then removal of the inactive component would lead to re-equilibration in favor of the more active catalyst combination and thus increased rates. This scenario may be well applicablet ot he case of component 2. However,r emoval of diphenylphosphine-containing aldehyde 3 led to complete loss of catalytic activity,i mplying that the highly nucleophilic phosphine was the only nucleophile in the system capable of catalyzing the reaction. In furthers upport of this observation, imidazole-based aldehyde 4 showed almost no rate change when replaced.
The results from the H-bond donor screens howedl ess pronounced differences. Removalo ft he weakerH -bondingt hiourea C provided the largests ystemic effect, with the product formation rate decreasing by almost 30 %. Replacement of the strongerH -bond donor B insteadl ed to ar ate increase, suggesting that B had deleterious effects on the catalysis.
To evaluate the accuracy of the deconvolution predictions, ap arallel screening test was subsequently performed. All linear combinationso ft he catalysts were synthesized in situ by direct condensation of the corresponding amine and aldehyde, and tested in singlee xperiments. Only the four reactions involvingt he iminesr esulting from aldehyde 3 showed any product formation after 24 h. These four catalysts were then synthesized and purified, giving bench-stable compounds that were subsequently tested in controlled single experiments. The resultsa re summarized in Figure 3a nd are in accordance with the dynamic deconvolution results. Compound C3 turned out to be the most activec atalyst, with a1 9% yield of the MBH product 5,c ompared to 15 %f or B3 and only 3% for A3 and D3.
The relatively high catalytic ability of B3 was initially surprising, because the system experiments actually predicted the compound to be detrimental to catalysis. However,subsequent experiments showed that B3 was highly unselective, with formation of large amountso fb yproducts. Furthermore, product 5 was shown to be unstable in the presenceo fB3,a nd decomposed over time. These effects are an example of why care has to be taken in the collective screening of catalyst mixtures, because simple determinationo ft he yield of 5 upon completed reactionw ould not lead to accurate predictionso ft he optimal catalysta ctivities. However, this study has showcased that kineticm easurements of initial rates is ap ossible way to measure systemic activities of catalyst mixtures.
Although C3 is by no meansastate-of-the-art catalysta ctivity-wise, these results provide compelling evidence that the deconvolution methodology has accurately predicted the most active catalyst from ad ynamic system. This protocols eems to be highly suited for detecting components crucial for activity, but it can also differentiate between less important functional groups that still contributet ot he catalysis in the system. The methodi ss imple and straightforward, and allows one-pot synthesis and subsequent screening of well-defined, covalently linked bifunctional organocatalysts without the need for separation, purification,and characterizationofeach individual molecule.T he small model system investigated in this study is easily amenable to expansion, and the deconvolution protocol would be expected to increase furtheri ne fficiency with larger systems. Furthermore, considering the range of dynamic covalent linkages developed in recent years, aw ide range of potential dynamic catalysts architectures couldb eenvisaged.
Havings hown that the dynamic covalent chemistry enabled accelerated activity screening, we turned to investigating the behavior of the dynamic bifunctional catalyst C3 in more detail.W hen the MBH reaction was performed with 20 %l oading of C3,ayield of 87 %c ould be provided after an extended reaction time (240 h). In comparison, am aximum of only 27 % yield could be obtained using B3.Also, C3 could efficientlycatalyze an aza-MBH reactionw ith highly electrophilic phenyl Ntosyl imine 6 to give aza-MBH adduct 7 in av ery good 85 % yield over72h(Scheme 4). [21]  Furthermore,w ew erei nterested in investigating if the dynamic covalentb ond could be utilized to modulate the MBH activity.R unning the reaction with only amine C predictably only led to imine formation with p-nitrobenzaldehyde, but more surprisingly,u tilizing aldehyde 3 as the sole catalystl ed to almostn op roduct and quick decomposition (Table 1). When adding C and 3 together, the MBH reaction proceeded with very low selectivity and yield, with decomposition of the aldehydep resumably occurring over MBHa dduct formation. However,w hen C and 3 were pre-stirred with 4 MS overnight, C3 was formed in quantitative yield, and the corre-spondingMBH reactionp roceeded readily and selectively. Conversely, pre-stirring four equivalentso fH 2 Ow ith C3 followed by reagent addition again produced almostn op roduct formation, because the thiourea seemed to have catalyzed the partial hydrolysis of the imine back to the unfavorable aldehyde-amine pair.These resultsindicate that the dynamic bifunctional organocatalysts might be utilized as primitive switches, especially given the discovered self-modifying capabilities of this class of catalysts.
The inclusion of ad ynamic imine bond, as well as at ransimination catalyst, into the same structure also opens further interestingp ossibilities. For the catalyst screening, the dynamic system was "locked" during the entire catalytic event to maintain accuracy in reaction kinetics measurements. However,i ti s also straightforwardt o" unlock" the dynamic system anda llow living dynamic catalyst behavior,i nw hich the catalyst structure is continuously changingd uring the reaction. In theory,o rganocatalystsc apable of in situ error correction of their own moleculara rchitecture couldt hen be envisaged.

Conclusion
An ew class of dynamic bifunctional catalysts capable of catalyzing modificationso ft heir own constitution was developed, and it was showcased how this property allows one-pot synthesis and evaluation of large systemso fc atalysts. The methodology uncovered ar elativelye ffective catalystf or the Morita-Baylis-Hillman reaction, and catalyst effectiveness could be regulated through manipulations of the dynamic covalent bond. DCC is integral for the screening approach, because it enables adeconvolution strategy that rapidly identifies the system components that contribute mostt oc atalytic activity.T he dynamic imine linkage allows proofreading of the dynamic system, witht he reversibility ensuring au niform catalyst distribution. The methodology can be utilized for catalyst discovery,and the obtained dynamic bifunctional scaffolds exhibit the potential for use as adaptable organocatalysts. Furthermore, this also marks the first report of thiourea-catalyzed transimination. Further investigations on the screening methodology and the self-modifying ability of the dynamic catalysts are currently in progress.

Experimental Section
Experimental procedure for dynamic system generation Aldehydes and amines (0.075 mmol each) were dissolved in anhydrous THF (0.5 mL) in an Eppendorf vial, and the solution was transferred to ad ry reaction vial containing pre-activated 4 MS (300 mg) under N 2 .T he mixture was stirred at room temperature for 20 ha fter which time the equilibrated system was obtained. Te sts for thiourea system equilibration were performed (see the Supporting Information), showing that the systems were at equilibrium after condensation.
Kinetic analysis of Morita-Baylis-Hillman reactions with dynamic systems catalysis Ad ynamic system was generated according to the description above. Afterwards, p-nitrobenzaldehyde (18.1 mg, 0.12 mmol) in anhydrous THF (0.120 mL) was added under N 2 ,f ollowed by addition of ethyl vinyl ketone (23.9 mL, 20.8 mg, 0.24 mmol). The mixture was stirred at room temperature under N 2 .A na liquot of the reaction mixture (30.0 mL) was withdrawn and added to 0.550 mL  CDCl 3 in an NMR tube, with PhSiMe 3 (0.020 mL/mL CDCl 3 )a si nternal standard. NMR measurements were performed within 5min, although control experiments indicated that the aliquot composition was stable for several hours in anhydrous CDCl 3 .P roduct formation was monitored by integrating the characteristic peaks at d = 5.66 and 6.00 ppm and comparing to the internal standard.