Electrochemical Iodine‐Mediated Oxidation of Enamino‐Esters to 2H‐Azirine‐2‐Carboxylates Supported by Design of Experiments

Abstract An electrochemical iodine‐mediated transformation of enamino‐esters for the synthesis of 2H‐azirine‐2‐carboxylates is presented. In addition, a thermic conversion of azirines to 4‐carboxy‐oxazoles in quantitative yield without purification was described. Both classes 2H‐azirines‐2‐carboxylates and the 4‐carboxy‐oxazoles are substructures in natural products and therefore are of considerable interest for synthetic and pharmaceutical chemists. The optimization was not performed in a conventional manner with a one‐factor‐at‐a‐time process but with a Design of Experiments (DoE) approach. Beside a broad substrate scope the reaction was also employed to a robustness screen, a sensitivity assessment, and complemented with mechanistic considerations from cyclic voltammetry experiments.

Azirines are rather unusual strained unsaturated three-membered rings with an itrogen atom where two types of isomers are possible:1 H-azirines,c ontaining ac arbon-carbond ouble bond that are not stable and isomerize to 2H-azirines containing an imine subunit. One special class of azirines are the 2Hazirine-2-carboxylates (marked in blue, Figure 1), which occur in naturalp roducts such as dysidazirine,a nd antazirine from the Dysidea fragilis sponge, [1] as well as azirinomycin from Streptomycesaureus. [2] The common synthesiso fa zirines can be achieved in several ways like the usage of azides to form nitrenes under thermal conditions, [3] addition of nitrenes to alkynes or addition of carbenes on nitriles, [4] ring contraction of isoxazoles, [5] elimination reactions from oxime derivatives, [6] or by strongo xidants like PIDA or PIFAf rom enamines. [7][8][9][10] Unfortunately,t he use of mild oxidants, such as iodine [11][12][13] for the oxidation of enamines is rare and in all cases over stoichiometrica mounts of oxidants are needed. Hazardous reagents like azides, strong bases, strongo xidants, or high temperatures as well as ad isadvantageousa tom economya re great drawbacks of these reactions; therefore, as imple conversionw ith am ild oxidanti nc atalytic amountsu nder mild conditions would overcome most of the drawbacks of the described synthetic methods (Scheme 1). In initiale xperimentsb ased on previousw ork, we discovered that under electrochemical conditions an enamino ester,s uch as 1a,w as convertedi nto the azirine 2a in moderate yield (19 %) when iodine was presenti nt he anodic cell compartment of ad ivided electrolysis underg alvanostatic conditions. Intrigued by this finding, we decided to investigate this electrochemical cyclization reaction towards such as trained unsaturated heterocycle in more detail.
Nowadays,e lectrochemical processes are favorable compared to chemical oxidants and reductants, because lessw aste is accumulated beside the usage of supporting electrolyte and hazardousr eagents can be avoided;i na ddition to economic reasons. [14] The main challengeo fe lectrochemical methods are the additionalp arameters (e.g. electrode material, cell design, supporting electrolyte), [15] which need to be optimized on top of the reactionc onditions of an ordinary organic reaction (e.g. temperature, concentration, reaction time). The additional parameters aggravate the optimization,t hat is why we decided to utilize the design of experiments (DoE)a pproach [16,17] where al arge number of parameters can be optimized at once very effectively and with highers tatistical significance by runninga small set of experiments. The aim of this optimization procedure is to perform as mall set of experiments, where all parameters are changed at the same time in as ystematic order.T he interaction of each parameter among themselves and the influence on the yield are described mathematically by using DoE. The outcomeo ft his analysisi ss tatistically proven and the influence of each parameter can be quantified ( Figure 2). Cross interactions, outliers due to experimental errors, and the optimal reaction conditions can easily be found. In earlier work we already demonstrated that DoE in combinationw ith electrochemicalt ransformations can be av aluable tool to improve the reaction performance significantly. [18,19] For the screening process,w ec hoose 1b (ethyl (E)-2-(amino(4-fluorophenyl)methylene)-3-oxobutanoate;c ompareS cheme 2) as the test substrate in order to determine yields by 19 FNMR spectroscopy directly from the electrolyte because the product 2b decomposes under the thermal conditions of gas chromatographic analysis. Initiallyi nt he DoE optimization,w es tartedt oi nvestigate categorical parameters in ar ough extensive screening. It seems that the optimization of categorical parameters very often relied on accidental discoveries of trial ande rror experiments because they have an on-linear correlation.F or system-atic optimizationo fc ategorical parameters in an one-factor-ata-time process, it is necessary to cover the whole reaction space by testinga ll possible combinations, which would lead to extensive experimentale fforts. The selection of the categorical parameters and the limits of the numerical parameters were made especiallyw ith respectt om ild and less hazardous reactionc onditions avoiding large excesses of additives or electric current to be applied. The following categorical parameters were investigated and for the solvents (MeOH, DMF, MeCN), the iodide source (Bu 4 NI, NaI,K I, PhI), the base (pyridine, DBU, KOPiv, 2,6-lutidine), the supporting electrolyte (NBu 4 BF 4 ,L iClO 4 , NEt 4 OTs), the anode material (graphite,p latinum, glassy carbon), and the cathode material( graphite, platinum,g lassy carbon) the bold formatted parameters were identified to be optimal in terms of the yield. The screening of the categorical parameters required 65 experiments by using DoE insteado f1 296 experiments (4·4·3·3·3·3) for coverage of the complete reactionspace (see the Supporting Information).
With these best categorical parameters,w es tarted with a doptimal optimization designw hich has been extended for the coverage of crossi nteractions. We needed2 5experiments for the numerical parameters containing mediator loading (0.20-0.30 equiv. ,o ptimal:0 .30equiv.), 2,6-lutidine equivalents (1.00-3.00, optimal:2 .40), electrolyte concentration (0.08-0.24 mol L À1 ,o ptimal:0 .24 mol L À1 ), temperature (25-55 8C, optimal:2 5 8C), applied charge( 1.8-2.4 Fmol À1 ,o ptimal: 2.4 Fmol À1 )a nd charge density (6-10 mA, optimal:1 0mA). The predicted yield by using DoE is correlated with the actual yield ( Figure 2) showingavery good representation of the influence between the numerical parameters and the yield. Some of the numerical parameters are at the limits of the prior defined range. We didn't want to extend the range to keep the reactionc onditions mild, the experimental set up easy,a nd avoid large excess of additives. The resulting modeli dentified 10 relevant interactions (see The Supporting Information) where aq uadratic term of the base equivalents had the greatest significance besides the electrolyte concentration and a cross interaction between temperature and electric current was also relevant (all p-values < 0.01);t hus, resulting in ah igh R 2 value of 0.98. After the initial optimization by the DoE approach,t he assessment of systematic and/or random errors [20] was investigated to increase the reproducibility of the electrochemicalm ethodology,w hichh as hitherto only been performed in photochemical reactions. [21,22] Therefore, the reaction was tested with the sensitivity assessment and examined beside typical parameters like concentrationo rm oisture con-  Predicted yields are plottedv ersus measured yields (in total 25 experiments:21f or the optimization, 4f or replicationofd ifferent experiments). All reactions were carried out on a0 .5 mmol scale using substrate 2b.Y ields were determined by 19 FNMR spectroscopy using 2-nitro-fluorobenzenea sthe internal standard. tent, also parameters unique for electrochemical conversions, such as the distance between the electrodes and the surface area of the electrodes ( Figure 3). Fortunately,t he electrochemical reactionp roved to be very insensitive to systematico r random errors so that the reactions hould be reproducible also for other researchers even if small deviations of the reaction conditions are applied.
After completion of the optimization and sensitivity assessment, we examined the scope and limitations of the electrochemicalc yclization reaction towards azirines (Scheme 2). Unfortunately,t he isolated yields differed from those determined by 19 FNMR (67 %c ompared to 87 %). Therefore, we performed an intensivea nalysiso fa ll individual work-up steps and evaluated how much product was lost in every work-up step (see SI). Sincet he azirine moiety is very labile the highest isolated yield was obtained when extraction with aqueous saturated NH 4 Cl solutiona sw ell as flash chromatography on neutral silica gel was performed. However,i ndependentf rom the work-up procedure, loss of af raction of the product could not be avoided.
The isolated yields of the azirines of type 2,o btained from the electrochemical reaction, are quite acceptable and range for ac onsiderable large number of products between 40-67 %. The scope of the reaction with respect to the substituent R 2 ranges from electron-deficient to electron-rich arene moieties which are both tolerated with similar results. The reaction is Scheme2.Substrate scope of the electrochemical iodine-mediatedo xidation of enaminest oazirines. 19 FNMR yields (2-nitro-fluorobenzene as internalstandard) and the applied currents are given in parenthesis. Someyieldsare given additionally as based on recovereds tarting material (borsm) yields. limited to aromatic substituents as R 2 which is in line with the previously reported chemical reaction conditions, where an ew hypervalent iodine (III/V) oxidant was used to prepare azirines from enaminoesters. [23] The position of the substituent on the arene ring in R 2 ;e ither in meta or para position (substrates 2f/ 2g and 2k/2l)s eems to have an egligible effect upon the yield of the reaction. However,asubstrate of type 1 with a substituent in ortho position wasn ot accessible. The scope of the substituent R 1 rangedf rom simple methyl b-keto esters (2a-2m)o ver cyclopropyl (2n)a nd phenyl( 2o)s ubstituted carbonyl compounds and malonates (2q-2s). As expected, electron-poor substrates gave better yields:a ccordingly,t he nitrile 2i,p roducts 2f/2g with nitro substituents and ester 2e could be isolated in good yields. We expected that the yield of product 2d and 2h would be lower,s incee lectrochemical direct iodination could occur, [24] but in both cases such side products were not observed. Further functional groups like halidesw ere investigated (fluoride, bromide, CF 3 )a nd the corresponding products wereisolated in moderate to good yields. Interestingly,t he substrate 1p (4-amino-4-phenylbut-3-en-2one) led to ad ifferent product and the analytical data revealed that the dimer (2p)w as formed in 28 %y ield. The reaction was not successful when an additional double bond in conjugation to the enamine was present (2t). Likewise, the reaction was not applicable when at risubstituted enamine was utilized (compounds 2u and 2w,R 1 = H). Unfortunately,c ompound 2v does not react under electrochemical conditions, whileaconventional method with iodinew as successful. [11] We reasoned that the oxidation potential of the enamine or the corresponding azirine must be lower than iodine, therefore we tried an ex-cell approach where we electrolyzed 1.20 equiv of sodium iodide under equal conditions without enamine for 2.4 Fmol À1 and then added the enamine to the solution without applying furtherc harge;s urprisingly,s tilln of ormation of the azirine was observed.
Since many substrates of type 1 are difficult to synthesize [25,26] and it is challengingt os creen al arge amounto ffunctional groups very effectively,w ea pplied ar obustness screen [27,28] upon the electrochemical transformation of enamino esters to azirines ( Table 1, see The Supporting Information  for more examples). The reactionp roceeds in good yields in most cases even thougha dditives are present. Some additives are stable under the conditions (entries 1, 6a nd 13), while primary amines (entry 7), sulfolane (entry 8), or amides (entry 12) were not tolerated. N-Methyl indole (entry 11)i nhibit the reaction, because iodination of the indoleo ccurs as shown via GC-MS analysis( similar outcome for N-benzyl pyrrole, see SI for complete table).S urprisingly alkenes and alkynes (entries4 and 9) were toleratedi nm oderate yields, althoughi odination of the unsaturated units could occur but was not observed. However,t he presence of an additional conjugated double bond as in substrate 2t was not tolerated under the electrochemicalr eaction conditions. Ar easonf or the relativelyh igh functional group tolerance could be that the reactive iodine speciesi sg enerated continuously in lowc oncentrationi ns itu, which could be the decisive advantage of electrochemical methods over conventionalsynthesis in this case.
For mechanisticc onsiderations cyclic voltammetry experiments were performed. Yu andC hang [11] postulated an ucleophilic attack from the enamine on iodine (I 2 )t ot ransfer I + onto the substrate and liberation of iodide anions (I À ). Our initial hypothesis wast hat iodide was oxidized electrochemically to molecular iodine and then undergoes the desired reaction. However,d irect oxidation of iodide occurs only in aqueous media and it is described that oxidation of iodide in organic solvents can be more complex. [29] To verify that I 2 was generated electrochemically under these reaction conditions, we performed the reactioni naflask under the same reaction conditions except replacing NaI by 1.2 equivalents of I 2 and without applying electrical current to the reaction;t he product was formed only in trace amounts.E ven if the amount of I 2 was increasedu pt o3 .6 equivalents, only 25 %o fp roduct 2a was formed and6 0% of the starting materialr emaining unchanged. This result and the absence of any reaction with enamine nitrile 1v lead to the hypothesis that another oxidized iodine species might be involved. It is likely that iodide anions were oxidized electrochemically to form triiodide( I 3 À )i nt he first step (E p = 0.75 Vv s. Ag/AgCl) and another oxidation to molecular iodine occurredi nasecond step (E p = 1.05 Vv s. Ag/ AgCl;Scheme 3). [30][31][32] To verify the formationo fI + we replacedt he enamine carboxylate with 4-tolyl trimethylsilane under equal conditions. [19] The appearance of 4-iodotoluene in GC-MSa nalysis indicates the formation of I + ions in the electrochemical reaction which underwent an ipso-substitution reaction by iododesilylation. We assumed that under electrochemical conditions another oxidation of I 2 to I + or presumably al utidine stabilized adduct (A)t akes place. In the absence of 2,6-lutidine the reaction of startingm aterial 1a to 2a does not take place which could indicate that acomplex of an iodine species and lutidine (or alu-  Figure 4, red curve) but under galvanostaticconditions NaI will be oxidized first. The cyclic voltammetry also showed that neither 2,6-lutidine nor the azirine 2a or the supporting electrolyte were oxidized. Analog complexes of pyridinel igands with I + have been described and characterized by 13 CNMR. [33] The enamine carboxylate 1 then reacts with the formal I + -reagent A to form intermediate B which is stabilized by ah ydrogen bond with the adjacent carbonyl group. The stabilization of the intermediates B and C could rationalize why enamino keto esters react more readily than enamino diesters and enamino nitriles, while substrates such as 2u and 2w do not react. After deprotonation by 2,6-lutidine towards C,t he ring closure occurs to form intermediate D under liberation of iodide anionsa nd another de-protonation generates the desired product 2 with at otal number of 2electrons consumed in this oxidation.
When we started the investigation of the electrochemical cyclization reaction of enamino esters, also another isomer with the identical molecular mass was detected by GC-MS analysisw hile such an isomer was not detectable by 19 FNMR analysis. Accordingly, at hermalr earrangement in the GC oven led to the formation of the isomer.T herefore, thermalt reatment of the azirines of type 2 in asealed tube at elevated temperatures led to the isomer detectedb yG C-MS analysis. The product 3 could be isolated in all investigatede xamples (Scheme 4) in quantitative yield without anyt edious work-up or purification step;o nly removal of the solvent-like similar literaturek nownr earrangementso fa zirinest o4 -carboxy-oxazoles. [34,35] 4-Carboxy-oxazoles are as pecial class of oxazoles, which appear in natural products like virginiamycinM 1 or bistratamide C. [36] Notably,s tartingm aterial 2d with an electronpoor aryl substituent reacted slow compared to the other azirines.
In conclusion, we realized an ew electrochemical iodinemediated synthesis of 2H-azirine-2-carboxylates and their follow up reactiont o4 -carboxy-oxazoles, whicha re both interesting substructures of natural products. To reduce the number of experiments and at the same time increases tatistical significance of the optimization, we performed the reaction optimization with the Design of Experiments approach. We demonstrated the applicabilityo ft his methodb yabroad substrate scope and aw ide robustness screen;a bove all,w eperformed as ensitivity assessment in an electrochemical reaction. Also, cyclic voltammetrye xperiments were conducted to get mechanistic insights of the reaction. Further investigations,t argeting the electrochemical synthesis of the above-mentioned natural products are under current investigation.  Scheme4.Thermal rearrangement of azirines to 4-carboxy-oxazoles.Incase of substrate 2d 120 hr eaction time.