Dimerization and comments on the reactivity of homophthalic anhydride

Abstract

Homophthalic anhydride (HPA) dimerizes under the influence of base to provide, sequentially, the (3-4’)-C-acyl dimer, a pair of chiral diastereomeric bis(lactones), 3-(2-carboxybenzyl)isocoumarin-4-carboxylic acid, and finally, 3-(2-carboxybenzyl)isocoumarin. The structures of the bis(lactones) were misassigned in 1970 based on the (presumed) cis thermal decarboxylative elimination reaction of the lower melting one. The preferred pathway should be trans-anti, however, and crystallographic analysis of one of the bis(lactones) reverses the earlier assignment. The formal cycloaddition reaction of HPA with imines occurs in preference to HPA dimerization; the mechanistic implications of this reactivity difference are discussed.

1. Introduction

New antimalarials continue to be a first line of defense against infection by the Plasmodium agent and its ability to develop resistance to even highly effective older drugs such as those based on artemisinin.¹ We recently developed a new class of antimalarial agents based on the 1,2,3,4-tetrahydro-1-isoquinolone 4-carboxanilide framework.² The synthesis of this ring system depends on the formal cycloaddition reaction of imines such as 2 with homophthalic anhydride (HPA, 1).³ This reaction is facilitated by N-methylimidazole (NMI), and evidence has been presented that the role of NMI is to intercept the putative Mannich intermediate and promote ring closure, while suppressing Knoevenagel-type elimination.⁴ Because HPA is known to dimerize under the influence of base,^5,6 and because this latter process would detract from the efficiency of the desired conversion, it was prudent to examine the control reaction in which the imine is not present. We report the results of that experiment, as well as a re-assignment of the stereochemistry of two of the reported HPA dimeric products. Furthermore, the mechanism of our NMI-promoted HPA-imine formal cycloaddition can now be clarified based on observations of the behavior of HPA in the control (imine-free) reaction.

2. Results and Discussion

2.1. The NMI promoted reaction of HPA with imines

Addition of 2 equiv of NMI to the reaction of certain imines (e. g., 2, Scheme 1) with HPA (1) improves both the yield and the selectivity of the reaction.⁴ One explanation for the role of NMI is that it intercepts an intermediate Mannich-type amino-anhydride 4, and thereby promotes the N-cyclization process leading to lactam (here, 3) at the expense of a yield-reducing Knoevenagel-type elimination from 4. However, for optimization of the imine reaction it was important to establish that parallel reaction of 1 with NMI doesn’t lead to side products that remove 1 from the desired pathway.

Scheme 1.

The NMI promoted formal cycloaddition of HPA (1) with (E)-2,2,2-trifluoro-N-(pyridine-3-ylmethylene)ethanamine, 2.

2.2 Dimerization reactions of HPA

HPA (1) has been reported to undergo dimerization when treated with base. Bogdanov et al.⁵ described the isocoumarin derivatives 5 and 6 (Scheme 2) that resulted from heating HPA (1) in pyridine solution for 1 h (5 is the major product) or 3 h (6 is the major product). Analogous reactions of benzo-substituted HPA derivatives have also been reported.^7,8

Scheme 2.

Dimeric HPA products at higher temperature.

Schnekenburger and Kaiser⁶ reported that, under much milder conditions (triethylamine, −15 °C), HPA (1) dimerizes to give, after acid quench, a pair of diastereomeric bis(lactones) 7 and 8 (Scheme 3). The lower melting isomer was assigned the (3S*,4R*) structure 8 (arbitrary HPA numbering, “isomer b”) based on its thermal decarboxylative elimination to give 6. The assumption was made that this is a cis-elimination. However, ample literature precedent^9,10,11 suggests that, in a decarboxylative elimination, the departing heteratom (here, the lactone ring carboxylate oxygen atom, as shown in 9) is aligned most favorably in a trans-anti arrangement with respect to the breaking carbon-carbon bond to CO₂. Only the (3R*,4R*) bis(lactone) can achieve this geometry (namely, 9), suggesting that the actual structure of “isomer b” is that shown in the box (Scheme 3).

Scheme 3.

Bis(lactone) dimeric HPA products at lower temperature and (mis)assignment of their structures.

By following the procedures shown in Schemes 2 and 3, we have synthesized authentic samples of HPA dimers 5, 6, 7, and 8. Crystals of the higher melting “isomer a” (initially assigned as 7), obtained by slow evaporation of a solution in propionitrile, were analyzed by X-ray crystallography. The actual structure of “isomer a” (see ORTEP representation, Figure 1) was thereby shown to possess the (3R*,4S*) stereochemistry, with the CO₂ and lactone ring O substituents cis. The actual structure of “isomer b” must be (3R*,4R*) (Scheme 3, in box), with the CO₂ and lactone ring O substituents trans), in harmony with its decarboxylative elimination reaction to give 6.

Figure 1.

ORTEP representation of higher melting “isomer a”.

2.2. The control reaction of HPA with NMI

The reaction of HPA (1, 1 equiv) with NMI (2 equiv) in dichloromethane-d₂ solution, equivalent to conditions under which 1 reacts rapidly with 2,⁴ was monitored by ¹H NMR spectroscopy (Scheme 4). After 2 min, small amounts of 7 and 8 were apparent, as evidence by their respective diagnostic signals at 4.31 (s) and 4.22 (s) ppm. Over the course of 40 min, the signals for 8 intensified, those of 7 less so, and the pair reached a respective ratio (8:7) of about 4:1. The broad singlet for unreacted HPA at 4.05 was still present but partially diminished. At 40 min, the diagnostic singlet for diacid 5 was also apparent at 4.62 ppm. Signals for the probable initial dimer 10 were not readily identifiable, nor were signals attributable to its enolate 11 apparent. No major change, other than an increase in 5, occurred over the next 2 days, and the diagnostic signal for HPA dimer 6 at 6.24 (s) ppm was not seen.

Scheme 4.

The reaction of HPA with NMI in dichloromethane solution.

A preparative scale (1 g of 1) version of the reaction shown in Scheme 4 was quenched after 40 min with cold 5% aqueous sulfuric acid, and then filtered (the solid was collected) and the filtrate extracted with additional dichloromethane; this extract in turn was dried and concentrated. The collected solid, which is mostly 7, and the organic soluble products were recombined – together these comprise an 80% mass recovery. Analysis of this mixture by ¹H and ¹³C NMR spectroscopy as their solution in acetone-d₆ indicates the presence of five products, shown in the box in Scheme 4. Integration (per H) of the diagnostic singlets attributable to each of these products (NMR spectra of the authentic compounds in acetone-d₆ solution were compared) gives the apparent molar ratios shown. All of the signals attributable to carbonyl carbons in the product mixture (161–172 ppm) can be assigned to the 13 carbonyl groups of the five products.

About half of the product mixture consists of recovered 1 and its hydrolysis product homophthalic acid (12), and some of the latter was lost to the aqueous layer. The 8:7 ratio and combined molar conversion is similar to that seen by ¹H NMR monitoring, suggesting that these products might form, at least in part, by double cyclization of 10 or 11 prior to the quench. One path to the diacid 5 before quench would be NMI mediated cyclization of E-11 (NMI is an acyl transfer promoter⁴).

The downstream (e. g., 24 h) product of HPA dimerization pathways under the conditions of Scheme 4 (2 equiv of NMI, 23 °C) is the diacid 5. However, no 5 is detected (i. e., the diagnostic singlet at 4.73 ppm is absent) among the products in the formal cycloaddition of imine 2 with HPA monitored by ¹H NMR spectroscopy over 48 h. In fact, no dimeric HPA products (5, 6, 7, 8) at all are observed in the reactions of imine 2 under the conditions of Scheme 1. Therefore, the reaction of HPA with itself under these conditions is distinctly slower than its reaction with 2, which is complete within 2.5 h at −30 °C, and 2.5 min at 23 °C.

2.3. Mechanistic implications

Qualitatively, the uncharged imine carbon of 2 ought to be a weaker electrophilic site than the C-3 anhydride carbonyl of HPA (Figure 2), since the former is considerably less electropositive. Furthermore, the C-1 anhydride carbonyl, which is conjugated with the HPA benzo portion, should also be less reactive than C-3. Reactions of HPA preferentially at C-3 with various nucleophiles such as tert-butylamine,¹² methoxide,¹³ and a wide variety of anilines,^14,15 besides HPA itself (Scheme 4), confirm the greater C-3 reactivity.

Figure 2.

Relative reactivity of electrophilic sites in 1 and 2

On the other hand, the (protonated) iminium salt (i. e., 2 · H⁺) ought to be more reactive than 1, inasmuch as the iminium carbon bears a considerable amount of positive charge.¹⁶ The fact that imine 2 under these conditions is apparently much more reactive toward HPA than the C-3 carbonyl group of HPA itself suggests that the reaction does not involve the anion of HPA adding to the imine carbon, but rather, the reaction could feature a protonated iminium electrophile derived from 2.

Figure 3 shows the approximate pK_a’s of species relevant to the reaction in Scheme 1. The pK_a’s of HPA,¹⁷ 2,2,2-trifluoroethylammonium,¹⁸ and NMI¹⁹ have been measured experimentally. Pyridinium species, such as 3-methylpyridinium (pK_a = 5.8) and pyridine-3-carboxylic acid (pK_a = 4.8), typically have pK_a’s ~5,²⁰ so the pyridine nitrogen of 2 and the product 3 probably lie in that vicinity. The pK_a of the iminium species derived from 2 may be estimated at roughly 2 by using the guideline²¹ that acetimine pK_a’s are typically about 3 units lower than the corresponding amine, and aldimines lower still.

Figure 3.

Approximate pK_a’s

Under the conditions of Scheme 1, partial formation of the enolate of HPA is not precluded; in fact, it is a possible intermediate in forming the HPA dimers in Scheme 4. However, the imine nitrogen of 2 is unlikely to be protonated to any significant extent, because there are multiple sites that are more basic, including NMI, which is present in excess. Therefore, the iminium species per se can be excluded as a participant in the formal cycloaddition. Nucleophilic addition of the imine nitrogen acting as a nucleophile toward the (less reactive) C-1 carbonyl of HPA or its enol tautomer has been proposed,^2,22 but more recently, calculations have been reported that suggest that this pathway is too high in energy.²³ Concerted cycloaddition of imine with a diene-like enol of HPA (13) or its derived enolate have been considered,^24,25 but imines are normally regarded as weak dienophiles in a thermal Diels-Alder reaction,²⁶ and 1 reacts rapidly with 2 even at mild temperatures.

A concerted, cyclic, Mannich-like reaction²³ of the enol of HPA (13) accounts for the reactivity of imine 2 despite its low basicity and nucleophilicity (Scheme 5). A hydrogen bond between enol 13 as the donor and the imine nitrogen of 2 as the acceptor can organize the two coupling partners into a transition state geometry in which either the N-2,2,2-trifluoroethyl substituent of 2 lies over the HPA ring oxygen (a “boat”, 14), or the pyridine ring of 2 lies over the HPA C-1 carbonyl group (a “chair”, 15). After C-C bond formation, cyclization of the Mannich adducts 16 and 18 leads respectively to 3,4-cis-substituted 1,2,3,4-tetrahydro-1-isoquinolone carboxylic acid product 17, and 3,4-trans-substituted isomer 19. For the reaction of HPA and 2 at 23 °C, the ratio of cis-3 to trans-3 is 1.2:1.⁴ If proposed transition states 14 and 15 control the stereochemistry of the products, then their energies are comparable. The intermediacy of the Mannich adducts 16 and 18 is also implicated by the earlier demonstration of the effect of NMI in promoting cyclization rather than elimination, both of which can occur from the Mannich intermediate 4.⁴

Scheme 5.

Proposed mechanism for the formal cycloaddition of 1 and 2.