Enzyme-Catalysed Conversion of Atranol and Derivatives into Dimeric Hydrosoluble Materials: Application to the Preparation of a Low-Atranol Oakmoss Absolute

Oakmoss absolute, a solvent extract from Evernia prunastri, is a valuable fragrance ingredient widely used in fine fragrance for almost two centuries. Some minor components of oakmoss absolute, such as atranol and chloroatranol, are attested contact allergens and their presence in fragrance and cosmetic products should be as low as possible. In this context, we have developed an enzyme-based protocol upon which these undesirable molecules are converted in a hydrosoluble dimeric material, and thus easily separated from the absolute by liquid–liquid extraction. Analytical and sensory analyses were performed to confirm the specificity of the process, the absence of alteration of the olfactory quality of the absolute, and the final titles of atranol and chloroatranol, which eventually were observed in the ppm range. This highly sustainable process is a viable alternative to conventional time-, energy-, and manpower-consuming techniques to produce very low-atranol oakmoss absolute.


Introduction
Oakmoss is a lichen of the species Evernia prunastri, and is different to tree moss, which is mostly obtained from lichens belonging to Evernia furfuracea [1,2]. Dry oakmoss can be extracted by organic solvents such as hexane to yield an odorant resinoid. Upon treatment with hot ethanol and further cooling, insoluble waxes precipitate, and an absolute is obtained after filtration and ethanol removal by distillation. The use of oakmoss absolute in perfumery was popularized by the fragrance Chypre (Guerlain, Paris, France, 1840), and has become one the main ingredients constitutive of the 'chypre' and 'fougere' accords. Symbiotic organisms, such as lichens, develop secondary metabolites for which the biosynthetic origin is not always known with accuracy [3]. Common metabolites include monoand polyphenolics [4], dibenzofurans [5], quinones/anthraquinones, and xanthones. The chemical composition of oakmoss absolute has been the subject of several analytical studies [6,7]. It is mostly constituted of monoaromatic compounds and their dimers, called depsides. These are typically formed by an ester link between two monoaromatic units. Depsides are not supposed to confer an odor to the extract, but monoaromatic compounds such as methyl atratate 1, formed upon hydrolysis or transesterification by EtOH of some depsides, are considered odor impact constituents of oakmoss absolute ( Figure 1). Unfortunately, toxic atranol 2 and chloroatranol 3, responsible for allergic contact dermatitis, are also formed upon solvolytic cleavage of depsides [8]. Complex natural products obtained by extraction techniques are mixtures of dozens of individual compounds encompassing a relatively large number of chemical families. Evernia prunastri (L.) extracts may thus contain more than 170 individual chemicals [2] including 14 depsides, such as atranorin and evernic acid; 18 monoaromatic compounds, such as methyl atratate 1, atranol 2, and ethyl orsellinate; 9 chlorinated monoaromatic compounds, such as chloroatranol 3 and methyl chloroatratate; 25 triterpenes, or steroids, such as hopanoid derivatives [9]; and usnic acid, which is characteristic of Evernia prunastri (L.) [5]. However, being insoluble in EtOH, usnic acid is generally present in the resinoid but not in the absolute. Volatile compounds, including those formed upon depsides' degradation, are also present [10].
Although oakmoss absolute has been used for decades, it is known to contain undesirable molecules such as atranol 2 and chloroatranol 3, which are severe allergens by contact [11][12][13][14][15]. Sensitization and elicitation in allergic contact dermatitis are complicated cellular and molecular processes [16], and the issue of skin irritation has been taken very seriously by perfume manufacturers and regulatory bodies. In 1999, the Scientific Committee on Consumer Safety (SCCS) of the European Community issued a list of 26 allergens. Manufacturers were thus required to identify these ingredients individually in cosmetic and fragrance products containing them in more 0.01% in rinsed products and 0.001% in non-rinsed products [17]. The list was expanded to 82 allergens in 2012 [18]. In the particular case of atranol 2 and chloroatranol 3, which are not simple allergens but severe irritation elicitors, IFRA recommendations require their presence to be limited to 100 ppm in oakmoss extracts used in perfumery [19], and SCCS recommends banning these compounds for cosmetic products. Various studies on the frequency of oakmoss absolute contact allergy in large panels in European countries show, on average, 1% frequency in the general population and 2% in unselected dermatitis patients, upon exposure to 2% oakmoss absolute solutions in petrolatum [20]. The European Union (EU) published Commission Regulation 2017/1410 on 2 August 2017 prohibiting the use of atranol and chloroatranol in cosmetic products. From 23 August 2019, cosmetic products containing these substances shall not be placed on the Union market, and from 23 August 2021, cosmetic products containing these substances shall not be made available on the Union market.
Considering both the importance of oakmoss absolute in fine perfumery and the deleterious effect of atranols, various processes, of varying efficacies, have been evaluated to remove them, resulting in low-atranol or atranol-free products that could eventually be used in formulas compliant with the latest regulation. However, the body of knowledge on the subject available in the scientific Complex natural products obtained by extraction techniques are mixtures of dozens of individual compounds encompassing a relatively large number of chemical families. Evernia prunastri (L.) extracts may thus contain more than 170 individual chemicals [2] including 14 depsides, such as atranorin and evernic acid; 18 monoaromatic compounds, such as methyl atratate 1, atranol 2, and ethyl orsellinate; 9 chlorinated monoaromatic compounds, such as chloroatranol 3 and methyl chloroatratate; 25 triterpenes, or steroids, such as hopanoid derivatives [9]; and usnic acid, which is characteristic of Evernia prunastri (L.) [5]. However, being insoluble in EtOH, usnic acid is generally present in the resinoid but not in the absolute. Volatile compounds, including those formed upon depsides' degradation, are also present [10].
Although oakmoss absolute has been used for decades, it is known to contain undesirable molecules such as atranol 2 and chloroatranol 3, which are severe allergens by contact [11][12][13][14][15]. Sensitization and elicitation in allergic contact dermatitis are complicated cellular and molecular processes [16], and the issue of skin irritation has been taken very seriously by perfume manufacturers and regulatory bodies. In 1999, the Scientific Committee on Consumer Safety (SCCS) of the European Community issued a list of 26 allergens. Manufacturers were thus required to identify these ingredients individually in cosmetic and fragrance products containing them in more 0.01% in rinsed products and 0.001% in non-rinsed products [17]. The list was expanded to 82 allergens in 2012 [18]. In the particular case of atranol 2 and chloroatranol 3, which are not simple allergens but severe irritation elicitors, IFRA recommendations require their presence to be limited to 100 ppm in oakmoss extracts used in perfumery [19], and SCCS recommends banning these compounds for cosmetic products. Various studies on the frequency of oakmoss absolute contact allergy in large panels in European countries show, on average, 1% frequency in the general population and 2% in unselected dermatitis patients, upon exposure to 2% oakmoss absolute solutions in petrolatum [20]. The European Union (EU) published Commission Regulation 2017/1410 on 2 August 2017 prohibiting the use of atranol and chloroatranol in cosmetic products. From 23 August 2019, cosmetic products containing these substances shall not be placed on the Union market, and from 23 August 2021, cosmetic products containing these substances shall not be made available on the Union market.
Considering both the importance of oakmoss absolute in fine perfumery and the deleterious effect of atranols, various processes, of varying efficacies, have been evaluated to remove them, resulting in low-atranol or atranol-free products that could eventually be used in formulas compliant with the Cosmetics 2018, 5, 69 3 of 12 latest regulation. However, the body of knowledge on the subject available in the scientific literature and in patent databases is very limited, possibly because it can be more profitable for knowledge to remain confidential than for patented technology to be made public. There are two broad approaches to the removal of atranol and related compounds from a natural complex mixture such as oakmoss absolute. The first is based on specific physicochemical interactions, while the second involves the reactivity of aromatic aldehydes. Thus, combined preparative chromatographic methods (including column chromatography, GPC, HPTLC, and HPLC) [21], are known methods used to remove atranol from oakmoss absolute. Distillation techniques, in particular molecular distillation, which is already used for discoloration purposes in the case of oakmoss absolute [2], might also be considered for atranol removal, but to the best of our knowledge, this is not documented in the literature. Techniques based on molecular approaches are mostly aimed at converting aromatic aldehyde moieties into their corresponding benzyl alcohols. This limits their ability to form imines with skin proteins, which triggers the immune response [22]. Protocols involving catalytic hydrogenation [21]; reaction with hydrides, such as sodium and lithium borohydrides [23]; or alkaline treatment (mostly for hematommates and atranorin) [21], have thus been used. Imination reactions with water-soluble amino acids, such as leucine or lysine, have also been used to facilitate the removal of atranol and other aldehydes by liquid-liquid extraction [24,25]. The trapping of allergens contained in natural complex substances using aminoalkyl resins has been reported, but was not applied to atranol nor other oakmoss absolute allergens [26,27]. Selective filtration through molecular imprinting polymers has been developed to remove safrole from nutmeg oil, but this approach has not been applied to oakmoss extracts neither [28].
These methodologies generally suffer from a lack of selectivity, resulting in an alteration of the olfactory quality of the modified extract; excessive consumption of energy, resources and manpower; or weak practical applicability.
We have recently been using enzymatic approaches to specifically modify the chemical composition of natural complex mixtures, such as essential oils, in order to improve their properties [29][30][31][32][33]. In particular, we have described a procedure which involves use of horseradish peroxidase (HRP) to selectively convert 2-methoxy-4-(2-propenyl)-phenol (eugenol) into insoluble dimeric material, followed by simple separation of the mixture by filtration [34]. Regardless of the allergenic potentials of atranorin, evernic acid, perlatolic acid, divaricatic acid, and fumarprotocetraric acid, we have been interested in developing an enzymatic procedure based on peroxidases to selectively remove atranol 2 and chloroatranol 3 from oakmoss extracts, considering that they remain strong contact allergens and that only these two compounds are formally targeted by health agencies' decisions and IFRA recommendations. Our strategy consisted of the use of HRP (EC 1.11.1.7) in the presence of H 2 O 2 to catalyze the dimerization of atranols and facilitate their separation from the absolute.

Results and Discussion
Atranol 2 was synthesised by formylation of commercially available orcinol 4. Our initial attempts to perform the direct formylation by the action of POCl 3 and DMF following the standard procedure led to a regioisomer of 2 by formylation in ortho position relative to the methyl group. However, upon protection of the hydroxyl functions by methoxymethyl groups, ortho-lithiation could be performed efficiently, and the target atranol 2 was obtained in 85% isolated yield (39% in four steps from 4) [35]. Alternate synthesis should be performed upon methylation/formylation/demethylation by BBr 3 [36]. With this starting material, we started our study by testing HRP on pure 2 at pH 7 and 8, with and without cosolvent, to ensure substrate solubilisation (Table 1) [34,37,38]. The use of a cosolvent miscible with water was likely to modify the conformation of the enzyme, and therefore aspects of its activity, such as how it accommodated substrates. However, at a low 2% v/v, we considered this effect limited. The purpose of the solvent was to ensure substrate solubilisation. None of these conditions allowed the conversion of 2, which was recovered quantitatively in almost all cases. In order to activate the phenolic functions of 2, the enzymatic reaction was performed in carbonate buffer at pH 9 (20 mM). Surprisingly, 5-methylpyrogallol 5 was the only observed product, albeit in low yields (Table 2). The product 5 was characterised by 1 H and 13 C NMR as well as MS (see supplementary materials). It is likely that at such basic pH, the aromatic aldehyde substrate underwent a Dakin oxidation by H 2 O 2 to the corresponding phenol. A series of control experiments was performed to confirm this result (Table 3).
In the absence of H 2 O 2 , 2 was recovered unchanged upon extraction with AcOEt, either with or without HRP, used in 1% w/w (entries 1, 2). In the presence of H 2 O 2 but without HRP, 5 was obtained in 84% isolated yield upon extraction with AcOEt (entry 3). Surprisingly, in the presence of HRP (1% w/w) and H 2 O 2 (2 equiv.), only 2% of 5 was isolated by extraction with AcOEt. Evaporation of the aqueous phase, however, led to the isolation of a hydrosoluble dimer 6 isolated in 75% yield. in 84% isolated yield upon extraction with AcOEt (entry 3). Surprisingly, in the presence of HRP (1% w/w) and H2O2 (2 equiv.), only 2% of 5 was isolated by extraction with AcOEt. Evaporation of the aqueous phase, however, led to the isolation of a hydrosoluble dimer 6 isolated in 75% yield.
Upon optimisation of the amounts of H2O2 and HRP, the isolated yield of 6 could be increased to 82%. When 5, prepared independently, was submitted to the same reaction conditions, the dimer 6 was formed similarly (Scheme 1). A mechanistic rationale to account for this unusual product formation has been described in a separate publication [35]. It was thus established that 5 could be converted into a water-soluble dimeric product 6 in an unprecedented and surprising way. The structure of this original product was determined by 1 H and 13 C NMR spectroscopy, and in particular COSY, HMBC, and HMQC, and MS. It was further confirmed by formation of its triacetylated product, which occurred together with a lactonization of the 3-hydroxyacid moiety, acetylation of the two hydroxyl groups, and acetylation of the dienol form of the α,β-unsaturated ketone through enolisation on the methyl group. Additional NMR analysis and HRMS on the triacetylated product 7 confirmed the proposed structure. Following a Dakin oxidation of 2 to 5 occurring in the presence of H2O2 at pH 9, an additional oxidation step led to an ortho-quinone intermediate, possibly dependent on HRP catalysis since in the absence of the enzyme, 5 was isolated in good yield. This activated species then engaged in a fast reaction with a second molecule of 5 in a 1,6-conjugate addition/enol addition to an oxocarbenium ion intermediate, to yield a bicyclo[2.2.2]octane core fused with a cyclohexadiene. This dimeric adduct then underwent an epoxidation of the double bond by peroxylate ions, easily formed in the reaction medium. Upon internal nucleophilic attack of the enol towards the epoxide, and subsequent transannulation reaction, the final product 6 was formed.  Upon optimisation of the amounts of H 2 O 2 and HRP, the isolated yield of 6 could be increased to 82%. When 5, prepared independently, was submitted to the same reaction conditions, the dimer 6 was formed similarly (Scheme 1). A mechanistic rationale to account for this unusual product formation has been described in a separate publication [35]. It was thus established that 5 could be converted into a water-soluble dimeric product 6 in an unprecedented and surprising way. The structure of this original product was determined by 1 H and 13 C NMR spectroscopy, and in particular COSY, HMBC, and HMQC, and MS. It was further confirmed by formation of its triacetylated product, which occurred together with a lactonization of the 3-hydroxyacid moiety, acetylation of the two hydroxyl groups, and acetylation of the dienol form of the α,β-unsaturated ketone through enolisation on the methyl group. Additional NMR analysis and HRMS on the triacetylated product 7 confirmed the proposed structure. in 84% isolated yield upon extraction with AcOEt (entry 3). Surprisingly, in the presence of HRP (1% w/w) and H2O2 (2 equiv.), only 2% of 5 was isolated by extraction with AcOEt. Evaporation of the aqueous phase, however, led to the isolation of a hydrosoluble dimer 6 isolated in 75% yield.
Upon optimisation of the amounts of H2O2 and HRP, the isolated yield of 6 could be increased to 82%. When 5, prepared independently, was submitted to the same reaction conditions, the dimer 6 was formed similarly (Scheme 1). A mechanistic rationale to account for this unusual product formation has been described in a separate publication [35]. It was thus established that 5 could be converted into a water-soluble dimeric product 6 in an unprecedented and surprising way. The structure of this original product was determined by 1 H and 13 C NMR spectroscopy, and in particular COSY, HMBC, and HMQC, and MS. It was further confirmed by formation of its triacetylated product, which occurred together with a lactonization of the 3-hydroxyacid moiety, acetylation of the two hydroxyl groups, and acetylation of the dienol form of the α,β-unsaturated ketone through enolisation on the methyl group. Additional NMR analysis and HRMS on the triacetylated product 7 confirmed the proposed structure. Following a Dakin oxidation of 2 to 5 occurring in the presence of H2O2 at pH 9, an additional oxidation step led to an ortho-quinone intermediate, possibly dependent on HRP catalysis since in the absence of the enzyme, 5 was isolated in good yield. This activated species then engaged in a fast reaction with a second molecule of 5 in a 1,6-conjugate addition/enol addition to an oxocarbenium ion intermediate, to yield a bicyclo[2.2.2]octane core fused with a cyclohexadiene. This dimeric adduct then underwent an epoxidation of the double bond by peroxylate ions, easily formed in the reaction medium. Upon internal nucleophilic attack of the enol towards the epoxide, and subsequent transannulation reaction, the final product 6 was formed. Scheme 1. Optimized conditions for the formation of the dimer 6 and its acetylated derivative 7 [35].
Following a Dakin oxidation of 2 to 5 occurring in the presence of H 2 O 2 at pH 9, an additional oxidation step led to an ortho-quinone intermediate, possibly dependent on HRP catalysis since in the absence of the enzyme, 5 was isolated in good yield. This activated species then engaged in a fast reaction with a second molecule of 5 in a 1,6-conjugate addition/enol addition to an oxocarbenium ion intermediate, to yield a bicyclo[2.2.2]octane core fused with a cyclohexadiene. This dimeric adduct then underwent an epoxidation of the double bond by peroxylate ions, easily formed in the reaction medium. Upon internal nucleophilic attack of the enol towards the epoxide, and subsequent transannulation reaction, the final product 6 was formed.
Following our results on pure 2, we next turned our attention to crude oakmoss absolute for the HRP-based enzymatic treatment. Oakmoss absolute (100 mg) was dissolved in a carbonate buffer at pH 9 (20 mM, 110 mL) under sonication for 2 h. The reaction was initiated with the addition of HRP (1 mg) and H 2 O 2 (as a concentrated aqueous solution; several equivalents were tested). After two different reaction times, the reaction mixture was extracted with AcOEt, the modified absolute was analysed by HPLC-UV, and residual 2 was specifically titrated by HPLC-MS (external calibration method) and compared with the title in the untreated oakmoss absolute used in this study (43,000 ppm) [14,39]. The results are summarized in Table 4. Residual 2 below the regulatory threshold value of 100 ppm could be obtained when more than 2 equivalents of H 2 O 2 were used. With 4 equivalents of H 2 O 2 , and upon 4 h of reaction, a residual title as low as 7 ppm was measured. The reaction could be performed at the gram scale and yielded, after 2 h of reaction, a modified oakmoss absolute containing 60 ppm of atranol and in which chloroatranol was not detected (HPLC-MS, SIM mode). To assess the overall effect of the enzymatic reaction on the entire oakmoss absolute, HPLC-UV remained useful. The chemical composition globally remained the same, even if some compounds showed slight variation; by comparing chromatograms before and after enzymatic treatment, no significant mass losses were observed ( Figure 2). For example, orsellinic acid (10.8 min) and everninic acid (18.7 min) were simply washed out by liquid-liquid extraction during the work-up. The most odorant compound, methyl atratate, eluting at 21.2 min (confirmed by injection of an authentic sample), was unchanged under our conditions.
The olfactory quality of the modified oakmoss absolute was assessed by sensory analysis following the triangular testing methodology. Three identical vials containing two samples of oakmoss absolute (samples 1 and 2) and one sample of modified oakmoss absolute (sample 3), as solutions in EtOH (0.5% w/w), were submitted to a panel of 56 persons who were separately asked to identify the modified sample within the three samples. Statistical distribution will therefore be 33% for each sample, and a test should be used to determine whether our results fall in this statistical distribution or not. The following formulas were used, with n 1-3 being the value one should exceed to be sure that the result is not the statistical distribution for a given level of confidence, and N the number of panelists: For a panel of 56 persons, n 1 = 25.9, n 2 = 28.0, and n 3 = 30.9. Scores obtained during the triangular testing for samples 1, 2, and 3 were 18, 16, and 22, respectively. A score of 22 means that even at the lowest level of confidence, the panel was not able to identify the modified sample. The olfactory quality of the modified oakmoss absolute was assessed by sensory analysis following the triangular testing methodology. Three identical vials containing two samples of oakmoss absolute (samples 1 and 2) and one sample of modified oakmoss absolute (sample 3), as solutions in EtOH (0.5% w/w), were submitted to a panel of 56 persons who were separately asked to identify the modified sample within the three samples. Statistical distribution will therefore be 33% for each sample, and a test should be used to determine whether our results fall in this statistical distribution or not. The following formulas were used, with n1-3 being the value one should exceed to be sure that the result is not the statistical distribution for a given level of confidence, and N the number of panelists:

Materials and Methods
1 H NMR and 13 C NMR spectra were recorded on Bruker (Billerica, MA, USA) AC Avance spectrometers (200 and 400 MHz). 1 H NMR spectra are reported as follows: chemical shifts in ppm (δ) relative to the chemical shift of TMS at 0 ppm, multiplicities (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet and br = broad), coupling constants (Hz), and integration. 13 C NMR spectra chemical shifts are reported in ppm (δ) relative to CDCl 3 at 77.16 ppm. Identity was assessed by comparison with data from authentic samples or literature data.