Preparation of cysteine adducts by regioselective ring-opening reactions of phenyloxirane

Adducts with blood proteins, mainly albumin and globin, are formed by reactions of numerous xenobiotics entering human body and/or their electrophilic metabolites. They found applications in biological monitoring of cumulative exposure to environmental and occupational mutagens and carcinogens [1, 2]. The main site of attack in both globin and albumin is the highly nucleophilic SH group of cysteine [3]. For bioanalytical studies on protein adducts, S-substituted cysteines are needed as analytical standards as well as other research tools. The nucleophilic cysteine SH group of glutathione also plays a fundamental role to prevent cell damage by various electrophiles as well as by free radicals [4, 5]. In connection with our ongoing research on protein adducts, we needed to prepare cysteine adducts of phenyloxirane (styrene oxide, 1), a mutagenic and carcinogenic metabolite of styrene. The direct reaction of l-cysteine (2) with 1 in strongly alkaline solution (EtOH/EtONa) afforded a 1:2 mixture of S-(2-hydroxy-1-phenylethyl)cysteine (3) and S-(2-hydroxy-2-phenylethyl)cysteine (4) as reported by Cundari et al. [6] (Scheme 1). These authors also attempted to achieve the regioselectivity of this reaction by using unprotected 2-bromo-2-phenylethanol and 2-bromo-1-phenylethanol as synthetic equivalents of 1. However, under the alkaline reaction conditions, these bromohydrines underwent dehydrobromination to give 1 so that mixtures 3 and 4 were always formed. This mixture was converted to (t-butoxycarbonyl)amino (N-BOC) derivatives 3a and 4a [6]. Small samples of 3 and 4 were previously obtained by the direct reaction of 1 with cysteine. This reaction gave a 2:1 mixture of 3 and 4, each of them consisting of two diastereomers. Individual isomers were then separated by high-performance liquid chromatography. However, it was not possible to scale up this procedure to obtain sufficient quantities of the individual products [7]. As a rule, a direct attack by nucleophiles proceeds on the less substituted (β) carbon of the oxirane ring. By contrast, in an acidic environment, where the oxirane oxygen is either protonated or coordinated to a Lewis acid, the nuclephilic attack is redirected to the more substituted (α) carbon [8]. However, 1 shows a rather anomalous regioselectivity of ring-opening reactions. In reactions of 1 with thiols in slightly alkaline aqueous solutions [9, 10] or in ionic liquids [11, 12] mainly but not exclusively, the products of α-attack were observed. However, with a wide range of substituted oxiranes, the same reactions gave exclusively products of β-attack. This may be due to a combination of the resonance stabilization of α-carbocation by adjacent aromatic ring and of the electron withdrawing effect of the phenyl group. The resonance stabilization of the carbocation at α-position is expected to play a crucial role if the reaction proceeds by S_N1 mechanism, which should be preferred in acidic solutions. However, when 1 was allowed to react with 2 in acidic aqueous solutions at pH ranging from 3 to 6, we observed the formation of both adducts 3 and 4 in the ratio ranging from 2:1 to 3:4. Less α-adduct was formed with decreasing pH but the ratio did not differ very much from that observed at pH 7. Apparently, at this pH range, the reaction does not proceed by S_N1 mechanism. The reaction is shown in Scheme 1.

Scheme 1

Observed minor changes in regioselectivity may be attributed to decreasing ratio of S^-/SH in cysteine (Table 1). Ring-opening reactions of 1 were further tested with two protected cysteine derivatives, namely, N-t-butoxycarbonyl- and N-benzyloxycarbonylcysteine methyl esters, 2a and 2b, respectively. Ring opening proceeding in the presence of N-bromosuccinimide (5 mol%), which for various benzenethiols and alkanethiols was reported to give mainly products of α-attack [13], gave low yields of expected products with the protected cysteines 2a and 2b. These products were accompanied by the corresponding cysteine derivatives (Table 2).

In an attempt to improve regioselectivity of the reaction and/or yields of the products, we further tested several combinations of the activating agent and solvent, such as Et₃N, LiClO₄, or ZnCl₂ in dichloromethane or acetonitrile. Products of the α-attack always prevailed, and the yields were always low due to both low conversion and by-product formation. Efficient and regioselective cleavage of 1 via α-attack by benzenethiol, aniline, allyl alcohol, and propargyl alcohol was reported in the presence of tris(pentafluorophenyl)borane as a Lewis acid catalyst [14]. This process proved to be applicable for the reaction with 2a and 2b, which afforded the corresponding protected isomer 3 exclusively (Scheme 2). However, the reaction proceeded slowly, and conversion was still incomplete after 3 days at room temperature (Table 2). Increasing the reaction temperature did not lead to any improvement of the yield but resulted in lower product purity. Reaction with 2b afforded a better yield than that with 2a (Table 2). Acid-catalyzed ring-opening reactions of substituted oxiranes generally proceed by S_N1 mechanism, resulting in the racemization of the products [8]. Nevertheless, we observed stereospecific ring opening with the inversion of configuration at the α-carbon, so that the reaction of pure (R)-1 gave exclusively S,R- 3a or S,R-3b from 2a and 2b, respectively. In a similar way, pure substrate (S)-1 was transformed into the diastereomer R,R-3a or R,R-3b (Scheme 2). Absolute configuration of the products was confirmed by the comparison of ¹³C NMR shifts with those of the analogous N-acetyl derivatives [15].

Scheme 2

The inversion of configuration is characteristic for S_N2 reactions. Therefore, it can be concluded that, at least under the conditions used, oxirane ring opening catalyzed by B(C₆F₅)₃ proceeds by S_N2 mechanism, without formation of an intermediary carbocation but probably with a significant weakening of the C_α-O bond of 1 in the transition state due to the binding interaction of the Lewis acid with the oxirane oxygen.

Eventually, experiments with tetrabutylammonium fluoride (TBAF) as a base afforded predominantly products of β-attack. Thus, the reaction of 1 with 2a in the presence of 10 mol% of TBAF led to a 10:1 mixture of protected products 4a and 3a. After chromatographic separation, pure protected isomer 4a was obtained in a satisfactory yield (Table 2). In fact, the SH group of cysteine can be easily deprotonated (pK_a=8.2). Hence, the observed exceptional regioselectivity indicates that TBAF does not serve solely as a base because other bases direct the SH-attack mainly to the α-carbon. A commercially available solution of TBAF in THF used in these experiments contained approximately 5% of water so that the nucleophilic attack could be redirected to the β-carbon of 1 by the formation of complexes held together by hydrogen bonding. ‘Anhydrous’ TBAF was reported to contain about 0.3 equivalents of water [16]. Attempts to remove this residual water results in the degradation of TBAF due to Hofmann elimination [17]. Therefore, we also used tetramethylammonium fluoride (TMAF) as a catalyst that, unlike TBAF, can be obtained in an anhydrous form. The result, however, was very similar to that obtained with TBAF. Unlike the reaction catalyzed by Lewis acids, base-catalyzed reaction should be stereospecific so that a single diastereomer of 4a and 4b should be obtained from each enantiomer of 1. However, we observed the formation of 1:1 diastereomer mixtures of 4a and 4b in the reaction with pure enantiomers (R)-1 and (S)-1 as well as with racemic 1. Control experiments showed that in the presence of enantiomeric 1, the products rapidly racemized in the presence of TBAF. The deprotection of 3a and 4a was conducted by treatment of trifluoroacetic acid (TFA) in dichloromethane for 4 h at room temperature. As expected, acid-catalyzed racemization was observed after the deprotection of pure diastereomers of 3a. Attempts to remove the Z-protecting group of 3b and 4b by Pd-catalyzed hydrogenolysis were unsuccessful probably because of catalyst poisoning by sulfur-containing substrates. This protecting group can be removed by reaction with trimethylsilyl iodide (TMSI) [18]. With 3b, this reaction followed by alkaline ester hydrolysis afforded a mixture of 3 and 4. Pure diastereomers R,R- and S,R-3b gave predominantly but not exclusively R,R- and S,R-diastereomers, respectively, accompanied with a mixture of R,R- and S,R-diastereomer of 3, as determined by LC/MS analysis comparing retention times and mass spectra with those of the products obtained by the reaction of pure enantiomers of 1 with 2. Although this reaction was not a useful preparative route to pure 3, it helped us to confirm the absolute configuration of R,R-3 and S,R-3. The deprotection of 4b by TMSI also yielded a product mixture.

In conclusion, simple regioselective ring-opening methods were developed, enabling the preparation of pure regioisomeric adducts 3 and 4 of phenyloxirane with cysteine. These products are useful tools in the research on protein adducts and in the biological monitoring of long-term cumulative exposure to styrene. Surprisingly, the oxirane ring opening of 1 catalyzed by tris(pentafluorophenyl)borane proceeds by S_N2 mechanism with full inversion of configuration. By contrast, the ring opening catalyzed by TBAF, which was expected to proceed stereospecifically, always led to the mixture of diastereomers because of full racemization of 1 enantiomers in the presence of TBAF.