Catalyst selective and regiodivergent O- to C- or N-carboxyl transfer of pyrazolyl carbonates: synthetic and computational studies

The regiodivergent O- to C- or N-carboxyl transfer of pyrazolyl carbonates is described, with DMAP giving preferential N-carboxylation and triazolinylidenes promoting selective C-carboxylation (both with up to >99 : 1 regioselectivity). An enantioselective O- to C-carboxyl variant using NHC catalysis is demonstrated (up to 92% ee), while mechanistic and DFT studies outline the pathways operative in this system and provide insight into the reasons for the observed selectivity.


Introduction and background
The organocatalysed rearrangement of oxazolyl carbonates to the corresponding 4-or 2-carboxyazlactones was rst described by Steglich and Höe over forty years ago. 1 This reaction process has since attracted considerable attention thanks to its potential to generate synthetically useful a,a-disubstituted a-amino acid derivatives and is oen regarded as a benchmark for the evaluation of Lewis base-catalysed reaction processes. Initially promoted by the achiral Lewis bases DMAP or PPY, the groups of Fu, 2 Vedejs, 3 Richards 4 and Gotor 5 have shown that chiral DMAP or PPY derivatives can induce high enantiocontrol in this reaction process. 6 Alternative chiral Lewis base catalysts that have been applied to this enantioselective rearrangement include chiral phosphines by Vedejs, 3 asymmetric imidazoles by Zhang,7 and a variety of isothioureas by Gröger (acyl transfer), 8 ourselves 9 and Okamoto. 10 A dual-catalytic approach that utilises DMAP and a chiral thiourea has been demonstrated by Seidel, 11 while an ammonium betaine catalyst for this process has been utilised by Ooi. 12 Within this area we have previously shown that N-heterocyclic carbenes (NHCs) are versatile catalysts for the racemic Steglich rearrangement of oxazolyl carbonates. 13 Achiral triazolinylidenes promote this rearrangement process with low catalyst loadings and offer access to sterically hindered products, although chiral NHCs resulted in only modest enantiocontrol. 14 The potential of this strategy to access stereodened products with a quaternary stereogenic centre has seen this process extended to incorporate the asymmetric rearrangement of furanyl, indolyl and benzofuranyl carbonates, 15 alongside applications in total synthesis. 16 Notably, Vedejs et al. have investigated the regio-and enantioselective O-to C-carboxyl transfer of 5-arylfuranyl carbonates using TADMAP 1, 17 with the regioselectivity dependent upon the electronic nature of the 5aryl substituent (Fig. 1). Electron-donating aryl substituents favour a-functionalisation (a : g up to 92 : 8), while an electronwithdrawing substituent favours g-functionalisation (a : g up to 20 : 80). Building upon these precedents, we considered alternative molecular scaffolds upon which to investigate catalyst selective regio-and enantioselective O-to C-carboxyl transfer processes. 18 While originally exploited in the dyeing and photographic industries, pyrazolinones and their derivatives have displayed a wide range of medicinal and pharmacological activities such as analgesic 19 and antipyretic properties, 20 anti-inammatory, 21 anti-tumor, 22 anti-microbial, 23 anti-retroviral 24 as well as anti-ischemic effects 25 and neuroprotective properties. 26 These diverse applications have encouraged recent interest in novel synthetic methods to access enantioenriched pyrazolinones 27,28 and served as inspiration for our studies concerning the regio-and enantioselective O-to C-carboxyl transfer of pyrazolyl carbonates. Notably, triazolinylidene NHCs promote the rearrangement to generate C(4)-a,a-disubstituted pyrazolinones with high regioselectivity (up to >99 : 1 C : N) and in up to 92% ee, while catalytic DMAP gives N(1)-carboxyl pyrazolinones with high regioselectivity (up to 1 : 99 C : N). A mechanistic rationale for this observed selectivity is provided by computational studies on a representative model system.
Initial studies screened a range of Lewis base catalysts for their ability to promote the regioselective O-to C-or N-carboxyl transfer of a model N(2)-phenyl substituted pyrazolyl carbonate 2 that was readily prepared from commercially available materials (Table 1). In all cases, generation of the parent pyrazolinone as a side-product amounted to typically $5% of the crude reaction product mixture, so only a ratio of C-to Nregioisomeric products is given unless stated. 29 Treatment of 2 with an NHC catalyst (generated in situ by deprotonation of the triazolium salt 5 with KHMDS) in THF gave the C-regioisomer 3 with high selectivity (>99 : 1 C : N), isolated in 44% yield (entry 1). Further optimisation showed that this NHC-promoted transformation could be performed using lower catalyst loadings in toluene (entries 2-4) while still giving 3 with excellent regioselectivity (>99 : 1 C : N). Remarkably, the use of DMAP in CH 2 Cl 2 favoured N-carboxylation with high regioselectivity (7 : 93 C : N), giving 4 in 56% yield (entry 5). The regiochemistry of this carboxyl transfer was conrmed by X-ray crystal structure analysis of N-carboxylate 4. 30,31 Rearrangement with isothiourea DHPB also favoured the N-carboxyl regioisomer but with lower reactivity compared to DMAP (entry 6). Further investigation of the DMAP-promoted reaction showed that THF and toluene proved poor solvents for this process, giving only $25% conversion to product with modest C : N ratios (entries 7 and 8).
These results indicate that catalyst promoted regiodivergent selectivity is observed in this process under either NHC or DMAP catalysis.

Scope and limitations
The scope and limitations of these catalyst selective carboxyl transfer processes was next investigated through variation within the carbonate functionality and pyrazolyl scaffold at N(2)-, C(4)-and C(5)- (Table 2). Under DMAP catalysis in CH 2 Cl 2 , variation of the carbonate group gave the N-carboxyl products preferentially (#16 : 84 C : N) that were isolated in good to moderate yield. 31 Although benzyl carbonate 6 showed poor conversion even aer extended reaction times, trichloroethyl and aryl N-carboxylate products containing both electron-withdrawing and electron-donating substituents were produced with good conversions. The effect of structural perturbation within the pyrazolyl skeleton was next investigated. With an N(2)-methyl substituent, C(5)-aryl substitution resulted in modest conversion but still preferential N-carboxylation to 10, while C(5)-methyl substitution gave preferential N-carboxyl product 11 with high selectivity and yield. Variation of the C(4)  substituent also led to preferential N-carboxylation (products Having demonstrated the generality of the DMAP-promoted Oto N-carboxyl transfer process, the NHC-catalysed O-to C-carboxyl transfer process was explored with toluene chosen as the reaction solvent (Table 3). Using achiral NHC precursor 5, variation of the carbonate group (entries 1-5), as well as N(2)-, C(4)-and C(5)-substituents (entries 6-12) gave preferentially the C-carboxyl isomer with high selectivity (up to >99 : 1 C : N, up to 84% yield). The regioselectivity of the NHC-catalysed reaction appears essentially independent of the nature of the carbonate group and C(5)-substituent, however it is particularly sensitive to the steric constraint at C(4), with a C(4)-iso-butyl group giving reduced C : N selectivity (76 : 24 C : N, 27, entry 12) relative to less hindered methyl substitution (97 : 3 C : N, 22, entry 7). A more modest reduction in regioselectivity was observed on changing the N(2)-substituent from phenyl to methyl (for example, compare products 24 (99 : 1 C : N, entry 9) and 26 (86 : 14 C : N, entry 11)).

Enantioselective NHC-catalysed selective Oto C-carboxyl transfer
With the achiral NHC derived from salt 5 established as a regioselective catalyst for the formation of C-carboxyl pyrazolinones, the expansion of this methodology to the synthesis of enantioenriched products using chiral NHCs was probed. 14 A screen of chiral triazolium NHC catalysts for the asymmetric rearrangement of model substrate 2 into C-carboxyl 3 identied N-pentauorophenyl precatalyst 28 as the optimal catalyst with regards to regio-and enantioselectivity (toluene was again preferred as solvent over THF as it gave superior product yields and enantioselectivity). 33 The full scope and generality of this asymmetric process was then investigated using NHC precatalyst 28 (Table 4). Aryl carbonates containing both electronwithdrawing and electron-donating substituents were tolerated with moderate levels of enantioselectivity (up to 68% ee, entries 1-3). By contrast, trichloroethyl carbonate showed good reactivity but poor enantioselectivity (entry 4). With a common N-(2)phenyl substituent, other C-(4)-alkyl substituents were tolerated with promising enantioselectivity (up to 69% ee, entries 5 and 6). With an N-(2)-methyl substituent, a mixture of N-and C-carboxyl products favouring the C-carboxyl products was observed, with good to excellent levels of enantioselectivity for the C-carboxyl product achieved with methyl, ethyl and iso-butyl C-(4)-substitution (88-92% ee, entries 7-10). The absolute conguration within C-regioisomer 29 (entry 8) was assigned by X-ray diffraction with all other congurations assigned by analogy. 30

Mechanistic investigations
With this rearrangement reaction producing two regioisomeric products, the possibility of product interconversion due to the reversibility of the C-C and C-N bond-forming processes was investigated on model N(2)-Me substrates 11 and 22. First, N-carboxylate 11 (1 : 99 C : N) was resubmitted to both DMAP   However, while treatment of enantioenriched C-carboxylate 22 (87% ee) with either DMAP or chiral NHC 28 returned 22 exclusively (87% ee), treatment with the achiral NHC derived from 5 gave C-carboxyl 22 in racemic form. 34 Treatment of (AE)-Ccarboxylate 22 with chiral NHC 28 also returned (AE)-22 (Scheme 2).
These results, combined with a crossover experiment upon a mixture of N-carboxylate products, 35 indicate that O-to C-or Ncarboxyl transfer reactions with DMAP are irreversible in this model system, with N-carboxylation kinetically preferred; N-to C-carboxyl transfer is disfavoured with DMAP. With the achiral NHC derived from precatalyst 5, O-to C-or N-carboxyl transfer reactions are reversible, with the C-isomer thermodynamically preferred, while N-to C-carboxyl transfer is also favoured. However, with the chiral NHC, O-to C-carboxyl transfer is irreversible with high enantiocontrol, while N-to C-carboxyl transfer is also allowed with good enantiocontrol. To probe this latter hypothesis, the reaction conversion, C : N product ratio and ee of the O-to C-rearrangement of 30 to 22 using precatalyst 28 was monitored ( Table 5). 1 H NMR spectroscopic analysis of the reaction mixture showed increasing ratios of C : N products over time, further evidence of catalysed N-to C-carboxyl transfer. 30 The ee of the C-carboxylate product 22 was however essentially independent of the reaction time and C : N ratio, consistent with our previous observations.

Computational insight
Computations were next performed on a simplied model substrate to elucidate the mechanism and origins of the observed regioselectivity. We employed M06-2X 36 /6-31+G** 37 / PCM 38 //M06-2X/6-31G*/PCM in toluene for NHC catalysis and dichloromethane for DMAP, as implemented in Gaussian09. 39 Manual, exhaustive conformational searches were performed to ensure all relevant intermediates and transition state structures (TSs) were located. Intrinsic reaction coordinate (IRC) computations were performed on all TSs to verify reaction pathways. Cand N-carboxylations share the same general mechanism shown in Fig. 2. Initial O-carboxylate attack by catalyst (TS-II) and subsequent tetrahedral intermediate collapse (TS-IV) leads to common intermediates, enolate 30 and carboxylated catalyst (CO 2 Me-Cat). Regiodivergence occurs by recapture of carboxyl group by enolate 30 at either C(4)-or N(1)-(TS-VI). The dissociation of the catalyst from the resulting tetrahedral intermediate (TS-VIII) releases the nal products.

DMAP catalysis
The DMAP-mediated carboxyl transfer preferentially results in N-carboxylation. Shown in Fig. 3, initial O-carboxylate attack (TS-II) by DMAP affords tetrahedral intermediate III, collapse of which (TS-IV, 20.4 kcal mol À1 ) affords ion pair intermediates V (CO 2 Me-DMAP and enolate 30). The C-vs. N-regiocontrol is established when the substrate enolate 30 attacks the carboxylated DMAP either via C(4)-or N(1). Consistent with the experimental results, the N-carboxylation process is favoured computationally by $5 kcal mol À1 (DMAP-TS-VI-N, DG ‡ ¼ 21.9 kcal mol À1 vs. DMAP-TS-VI-C, DG ‡ ¼ 26.6 kcal mol À1 ). Interestingly, the N-carboxylation is stepwise addition of enolate and extrusion of catalyst, whereas the C-carboxylation process  (black, Fig. 3) proceeds via a concerted, asynchronous carboxylation. 40 This difference in concerted/stepwise behaviour reects the lack of electrostatic stabilizing effects in DMAP-TS-VI-C vs. DMAP-TS-VI-N. In DMAP-TS-VI-N, there is a substantial spatial overlap between the positively charged carboxylated DMAP and the attacking enolate 30, as the enolate oxygen and p bond is in closer proximity to the carboxylated DMAP. This is in contrast to DMAP-TS-VI-C, where there is a relatively poor spatial overlap, with only the enolate oxygen in proximity to the positively charged DMAP ring. The forward process leading to the C-carboxylation is favoured by 2.6 kcal mol À1 over the forward process for the N-carboxylation (Fig. 3), resulting in exclusive production of C-carboxylated product upon retreatment of N-carboxylated product with achiral NHC (as observed in Scheme 1).

Structural comparison of enolate p vs. s reactivity
The remarkable switch in regioselectivity observed between DMAP and NHC catalysis in this system is a result of the markedly different reactivity patterns of the intermediate carboxylated DMAP or NHC and their interaction with the pyrazoline enolate as illustrated in Fig. 3. This is most striking in DMAP-TS-VI-N, where favoured nucleophilic attack from the substrate does not originate from the N(1)-lone pair of the substrate enolate in the s-plane, but rather the p-system of the extended enolate. This is in contrast to the analogous (disfavoured) NHC-TS-VI-N, where nucleophilic attack is predicted to occur from the N(1)-lone pair of the substrate enolate in the s-plane. As yet, the exact origins of this p vs. s reactivity are unknown. Our working hypothesis is that the relatively sterically unencumbered conjugated DMAP promotes p-p electrostatic interactions, 42 allowing the p-system of the extended enolate to be an energetically more competent nucleophile compared with the s-N(1)-lone pair.

Conclusion
In conclusion, the regiodivergent O-to C-or N-carboxyl transfer of pyrazolyl carbonates has been investigated, with DMAP giving preferential N-carboxylation and triazolinylidenes promoting selective C-carboxylation (both with up to 99 : 1 regioselectivity). An enantioselective O-to C-carboxyl variant using NHC catalysis is demonstrated (up to 92% ee), while mechanistic and DFT studies outline the pathways operative in this system and delineate insight into the structural reasons for the observed selectivity. Current investigations from within our groups are focused upon the demonstration of further Lewis base-mediated organocatalytic transformations and developing further computational insight into these transformations.