Tandem dinucleophilic cyclization of cyclohexane-1,3-diones with pyridinium salts

The cyclization of cyclohexane-1,3-diones with various substituted pyridinium salts afforded functionalized 8-oxa-10-aza-tricyclo[7.3.1.02,7]trideca-2(7),11-dienes. The reaction proceeds by regioselective attack of the central carbon atom of the 1,3-dicarbonyl unit to 4-position of the pyridinium salt and subsequent cyclization by base-assisted proton migration and nucleophilic addition of the oxygen atom to the 2-position, as was elucidated by DFT computations. Fairly extensive screening of bases and additives revealed that the presence of potassium cations is essential for formation of the product.

During the past decade the reaction of dinucleophiles with dielectrophiles has been a major research subject in both our laboratories. For example, we have studied reactions of quinolinium [36,37], isoquinolinium [38][39][40], quinazolinium [41,42] and quinoxalinium [43] salts with 1,3-bis(silyl enol ethers), i.e., masked 1,3-dicarbonyl compounds, which provided facile access to a number of bicyclic systems. Some of these reactions, such as the cyclization with quinazolinium salts, proceeded as one-pot cyclizations. Other cyclizations, such as the reaction with isoquinolinium salts, had to be carried out in two steps (formation of open-chained condensation products, which are subsequently cyclized by the addition of acid). In some cases the cyclization step failed, e.g., the reaction of 1,3-bis(silyl enol ethers) with pyridine in the presence of methyl chloroformate resulted in the formation of 1,4-dihydropyridines, which however, could not be cyclized by the addition of acid due to decomposition [36]. Additionally, we have shown a broad application of the quinolinium [44][45][46][47][48] and isoquinolinium [49] salts for the synthesis of a wide variety of alkaloid-like frameworks.
This efficient cycloaddition encouraged us to investigate this process, to improve the reaction conditions, and to extend the reaction with other pyridinium salts and diketones. Herein we report the results of that study. First, we studied the reaction of pyridinium salt 2a with dimedone (1a) using various bases in different solvents (Table 1). At the same time, different ratios of 1a and 2a were examined. To reach the optimal yields, dimedone and pyridinium salt 2a should be used in a ratio of two to one. In addition, the optimization study showed the essential role of a non-nucleophilic base as well as that of the potassium cation.
This optimization study afforded a general procedure and optimal reaction conditions: 2.0 mmol of diketone, 1.0 mmol of the appropriate pyridinium salt and 1 mmol of K 2 CO 3 in 7 mL of CH 3 CN are stirred at room temperature for 24 hours.
1,3-diketones were investigated, but these failed for all mentioned cases. For 3-cyanopyridinium salts a departure from the general trend was observed in terms of reaction conditions regarding the base used; here the highest efficiency was observed for NaHCO 3 , and the reaction took 4 days to reach completion. In general, the reaction proceeded with high regioand diastereoselectivity.
Additionally, the reaction was monitored by HPLC, and for most pyridinium salts investigated no byproducts were detected. However, in the case of 3-cyanopyridinium salts the formation of additional products, namely the constitutive isomers of type 5 and 8, was observed. These compounds were isolated and identified by using standard 2D NMR methods and X-ray analysis [51]. At the same time, to our great disappointment 4-substituted pyridinium salts appeared to be unreactive, and testing different reaction conditions as well as bases with various combinations of time and temperature did not deliver any positive results.

Reaction mechanism
In order to shed light on the reaction mechanism we used DFT computations at the B3LYP/6-31G(d,p) [52][53][54] level, as implemented in the Gaussian 09 program [55], and acetonitrile was included implicitly within the IEFPCM method (more details on inclusion of the solvent can be found in Supporting Information File 2). For simplicity 1,3-cyclohexanedione rather than dimedone was included in the computations.
Our ability to calculate the mechanism as well as to make any further predictions about reaction products is limited by the obscure structure of the real reaction product. In fact the reaction of pyridinium salt 2a with cyclic 1,3-diketones in the presence of either K 2 CO 3 or DIPEA/KI yields an orange precipitate in acetonitrile. Redissolved in DMSO-d 6 , this precipitate gave rise to broad peaks in the 1 H NMR spectra; thus its structure was not defined and is supposed to be of polymeric nature. Potassium with a counter-ion (HCO 3 − or I − ) is also included in the precipitate since that cannot be burned completely and yields an unburned material. On the other hand the orange precipitate is quickly decomposed by silica gel liberating the desired product 3a or 6a, which was finally eluted from silica. The formation of a precipitate must be a driving force of the reaction promoted by DIPEA/KI (see Supporting Information File 2, thermodynamic analysis).
The reaction starts with 1,3-diketone deprotonation and nucleophilic attack of potassium 3-oxocyclohex-1-enolate on pyridinium cation, which yields an intermediate with a new C-C bond, and this is a key stage in the reaction. The C-O bond formation can be excluded in the initial stage (see Supporting Information File 2 for the explanation). Bearing this in mind, we computed the local softness of 3-acetyl-1-methylpyridinium (2a + ), 2-cyano-1-methylpyridinium (2b + ) and 3-cyano-1methylpyridinium (2e + ) cations towards nucleophilic attack. According to the local softness indexes (see Supporting Information File 2) with 2b + cation we should observe a regioselective formation of product 4a or 7a, while for cations 2a + and 2e + a regioisomeric mixture should be expected (but is really only observed for 2e + ). In reality, regioisomeric mixtures of compounds 3/5 or 6/8 are produced by cation 2e + , while for cations 2a + and 2b + regioselective reaction is inherent. Intrigued by this discrepancy, we computed the full reaction mechanism for cation 2a + with potassium 3-oxocyclohex-1enolate in the presence of potassium cation and hydrogencarbonate anion as base. We included HCO 3 − in our computation, both because it has only five atoms and because this base is presumed to operate; it is formed from neutral carbonate when it deprotonates the cyclic diketone. However, the mechanism involving any other base should not differ significantly.
A plausible mechanism is presented in Figure 2. After initial deprotonation of the diketone and formation of the hydrogencarbonate anion, the latter is exchanged with iodine in the salt of    between 22a and 19).
Other TSs (K-S1a-2, K-S1a-3, see Supporting Information File 2), inherent for the initial stage of new C-C bond formation turned out to be unimportant, because in both cases reactions run again through TS 22a (rate limiting), as was shown by additional computations. On the other hand, we have computed four conformers of TS 22: the conformation of the cyclohexane ring and that of the acetyl group were changed. While the cyclohexane-ring conformation has only a negligible effect on the TS energy (TS 22b, see Supporting Information File 2), the change of the acetyl group conformation prevents potassium from coordinating to it and makes the TS energy somewhat higher (TSs 22c and 22d from Supporting Information File 2).
In the proposed mechanism the regioselective formation of product 6a is not explained (see Supporting Information File 2 for discussion). We attribute this to a higher thermodynamic stabilization of compound 6a by a potassium cation (Scheme 3), as compared with that of regioisomeric 6aa (its formation was not observed); the energetic distance between 6a-K + and 6aa-K + is 2.5 kcal/mol at the B3LYP/6-31G(d,p) level and 2.0 kcal/mol at the B3LYP/6-311++G(d,p)//B3LYP/6-31G(d,p) level. As mentioned before, the assumption is hard to prove due to lack of information about the real product composition.

Scheme 3:
The influence of K + on the product free energy.

Conclusion
In summary, we have studied the dicomponent binucleophilic cycloaddition of 1,3-diketones with diverse substituted pyridinium salts. The reaction took place under mild conditions and provided an efficient route to 8-oxa-10-aza-tricyclo-[7.3.1.0 2,7 ]trideca-2(7),11-dienes with a wide range of substituents. Further studies to extend the scope of the synthetic utility of these cyclizations are in progress in both of our laboratories.

Experimental General Information
Solvents were purchased from ACROS and directly used without further purification. Analytical thin-layer chromatography was performed on 0.20 mm 60 Å silica gel plates. Column chromatography was performed by using 60 Å silica gel (60-200 mesh). Pyridinium salts were synthesized according to the published procedure [56].
The density functional B3LYP [52][53][54] with 6-31G(d,p) basis set for all atoms except iodine and solvation by acetonitrile within the IEFPCM method as implemented in the Gaussian 09 package [55], was employed for the calculations. For cases when iodine was included, the LANL2DZ basis set and core potential were assigned. All compounds were computed in the closed-shell singlet state (confirmed by "stable" calculation). Frequency calculations were performed on all optimized geometries to ensure only positive eigenvalues for minima and one imaginary frequency for transition states. Free energies at 298.15 K and 1 bar pressure were calculated according to the recommendation of Gaussian, Inc. [57,58], and the frequency scaling factor (0.9608) was applied. For every located transition state an IRC calculation at the same level of theory confirmed that this TS connects the corresponding educts/intermediates/products. The methodology of local softness computation [59,60] uses Fukui functions; for their approximation we used Mulliken, ESP, NPA and Hirshfeld charges. However, only the Fukui functions approximated by NPA and Hirshfeld charges were unambiguous and taken into consideration.
General procedures for the synthesis of compounds 3-8

Procedure (A)
In a 25 mL Schlenk flask, under argon flow, 2.0 mmol of diketone, 1.0 mmol of the appropriate pyridinium salt, and 1.0 mmol (138 mg) of K 2 CO 3 were loaded. The flask was covered with a septum stopper and 7 mL of absolute CH 3 CN was added by syringe. The reaction mixture was left under intensive stirring at room temperature for 24 hours. Then the solvent was removed under reduced pressure and the crude material was subjected to column chromatography.

Procedure (B)
In the case of the 3-cyanopyridinium salt, NaHCO 3 (2.0 mmol, 168 mg) was used as a base, and the reaction mixture was left over 4 days.

Procedure (C)
In the case of 2-cyanopyridinium salts the reaction was completed within 1 hour.

Supporting Information
Supporting Information File 1 Details on synthetic procedures, list of pyridinium salts, characterization of new compounds, copies of NMR spectra, X-ray structures of compounds 6d, 7c and 8.

Supporting Information File 2
Computational results, optimized structures (atomic coordinates as reported by Gaussian 09), charges and local softness indexes (nucleophilic attack) for cations 2a + , 2b + and 2e + , additional discussions related to computations.