Development of Biotransamination Reactions towards the 3,4-Dihydro-2 H -1,5-benzoxathiepin-3-amine Enantiomers

: The stereoselective synthesis of chiral amines is an appealing task nowadays. In this context, biocatalysis plays a crucial role due to the straightforward conversion of prochiral and racemic ketones into enantiopure amines by means of a series of enzyme classes such as amine dehydrogenases, imine reductases, reductive aminases and amine transaminases. In particular, the stereoselective synthesis of 1,5-benzoxathiepin-3-amines have attracted particular attention since they possess remarkable biological proﬁles; however, their access through biocatalytic methods is unexplored. Amine transaminases are applied herein in the biotransamination of 3,4-dihydro-2 H -1,5-benzoxathiepin-3-one, ﬁnding suitable enzymes for accessing both target amine enantiomers in high conversion and enantiomeric excess values. Biotransamination experiments have been analysed, trying to optimise the reaction conditions in terms of enzyme loading, temperature and reaction times. (M + H) + 299.0953, found 299.0955. Anal. Calcd for C 14 H 18 O 5 S: C, 56.36; H, 6.08; S, 10.75. Found: C, 56.45; H, 5.89; S, 10.55.


Results and Discussion
The synthesis of the benzo-fused seven-membered ketone 6 is depicted in Scheme 1. 2-Mercaptophenol was alkylated with two equivalents of ethyl bromoacetate in refluxing acetone in the presence of dry potassium carbonate to give diester 7 (83%). Examination of the Dieckmann reaction of 3 showed that the reaction occurred smoothly when sodium ethoxide/ethanol was used as a base in dry tetrahydrofuran (THF) to give ethyl 3-oxo-3,4-dihydro-2H-1,5-benzothiepin-4-carboxylate 8 as the sole cyclised product in 90% yield. Decarboxylation of the β-ketoester 4 in boiling acetic acid containing aqueous sulfuric acid gave the 3,4-dihydro-2H-1,5-benzothiepin-3-one (6, 60%). Regioselectivity of the Dieckmann cyclisation was deduced based on the 1 H-NMR (CDCl 3 ) spectral data of the resulting product 8, which exhibited two doublets (integrating each one for 1H) at δ 4.88 and 4.59 ppm (J = 17.5 Hz) assignable to the geminal methylene protons adjacent to the oxygen atom in the seven-membered ring. Compounds 7 and 8 have not been described previously, whilst ketone 6 was reported formerly by Sugihara et al. [44]. Due to the amine transaminases catalytic mechanism, which involves two pairs of ketones and amines in equilibrium, the reductive amination of the substrate must be thermodynamically favoured in order to obtain high yields of the desired product [45,46]. In order to displace the equilibrium towards amine formation, the removal of the generated co-products by coupling different multienzyme networks is often required [20], but also worth noting is the use of sacrificial substrates, which normally range from the use of a large excess of a commercially available amine donor, typically isopropylamine [47], to "smart cosubstrates", mainly diamines, in a stoichiometric amount that are able to drive equilibrium by spontaneous cyclisation or aromatisation reactions [31,[48][49][50]. Promisingly, we have found a favourable ΔG of -31.0 kJ/mol (calculated at M06-2X/6-311++G(3df,2p) level; see Section 3.8) for the transamination of 6 to 9 when using isopropylamine and acetone is formed as a by-product, probably due to ring strain instability. Figure 2 shows the charge density of the optimised geometry of the ketone 6, where steric and electronic differences between the two substituents of the carbonyl group can be observed. This prompted us to study the biocatalytic process in depth. The biotransamination of 3,4-dihydro-2H-1,5-benzoxathiepin-3-one (6, 20 mM) was then studied in standard conditions previously employed in our research group [46,51]. These settings include the use of a large excess of isopropylamine as amine donor (1 M), pyridoxal 5'-phosphate (PLP, 1 mM) as cofactor, a 100 mM phosphate buffer pH 7.5 with acetonitrile (5% v/v) as cosolvent to favour the ketone solubility, at 30 °C and 250 rpm for 20 h (Scheme 2). Three different types of enzymes were Scheme 1. Chemical synthesis of 3,4-dihydro-2H-1,5-benzoxathiepin-3-one 6.
Due to the amine transaminases catalytic mechanism, which involves two pairs of ketones and amines in equilibrium, the reductive amination of the substrate must be thermodynamically favoured in order to obtain high yields of the desired product [45,46]. In order to displace the equilibrium towards amine formation, the removal of the generated co-products by coupling different multienzyme networks is often required [20], but also worth noting is the use of sacrificial substrates, which normally range from the use of a large excess of a commercially available amine donor, typically isopropylamine [47], to "smart cosubstrates", mainly diamines, in a stoichiometric amount that are able to drive equilibrium by spontaneous cyclisation or aromatisation reactions [31,[48][49][50]. Promisingly, we have found a favourable ∆G of -31.0 kJ/mol (calculated at M06-2X/6-311++G(3df,2p) level; see Section 3.8) for the transamination of 6 to 9 when using isopropylamine and acetone is formed as a by-product, probably due to ring strain instability. Figure 2 shows the charge density of the optimised geometry of the ketone 6, where steric and electronic differences between the two substituents of the carbonyl group can be observed. This prompted us to study the biocatalytic process in depth. Due to the amine transaminases catalytic mechanism, which involves two pairs of ketones and amines in equilibrium, the reductive amination of the substrate must be thermodynamically favoured in order to obtain high yields of the desired product [45,46]. In order to displace the equilibrium towards amine formation, the removal of the generated co-products by coupling different multienzyme networks is often required [20], but also worth noting is the use of sacrificial substrates, which normally range from the use of a large excess of a commercially available amine donor, typically isopropylamine [47], to "smart cosubstrates", mainly diamines, in a stoichiometric amount that are able to drive equilibrium by spontaneous cyclisation or aromatisation reactions [31,[48][49][50]. Promisingly, we have found a favourable ΔG of -31.0 kJ/mol (calculated at M06-2X/6-311++G(3df,2p) level; see Section 3.8) for the transamination of 6 to 9 when using isopropylamine and acetone is formed as a by-product, probably due to ring strain instability. Figure 2 shows the charge density of the optimised geometry of the ketone 6, where steric and electronic differences between the two substituents of the carbonyl group can be observed. This prompted us to study the biocatalytic process in depth. The biotransamination of 3,4-dihydro-2H-1,5-benzoxathiepin-3-one (6, 20 mM) was then studied in standard conditions previously employed in our research group [46,51]. These settings include the use of a large excess of isopropylamine as amine donor (1 M), pyridoxal 5'-phosphate (PLP, 1 mM) as cofactor, a 100 mM phosphate buffer pH 7.5 with acetonitrile (5% v/v) as cosolvent to favour the ketone solubility, at 30 °C and 250 rpm for 20 h (Scheme 2). Three different types of enzymes were The biotransamination of 3,4-dihydro-2H-1,5-benzoxathiepin-3-one (6, 20 mM) was then studied in standard conditions previously employed in our research group [46,51]. These settings include the use of a large excess of isopropylamine as amine donor (1 M), pyridoxal 5'-phosphate (PLP, 1 mM) as cofactor, a 100 mM phosphate buffer pH 7.5 with acetonitrile (5% v/v) as cosolvent to favour the ketone solubility, at 30 • C and 250 rpm for 20 h (Scheme 2). Three different types of enzymes were employed: Initially, for the biotransamination experiments made in house ATAs were used, all of them overexpressed in Escherichia coli. Some of them, such as the ones from Chromobacterium violaceum [52] or Arthrobacter species [53], displayed very low activity (<5%), while others such as Arthrobacter citreus [54] or the Arthrobacter species evolved variant named ArRmut11 [55] provided almost quantitative conversion but moderate (74% ee) or negligible stereoselectivity, respectively. Trying to improve both activity and selectivity values, commercially available ATAs were employed from two different commercial sources (Codexis Inc. and Enzymicals AG).
To start with, 30 Codexis enzymes were employed (Table 1), and we found that 19 of them led to the complete disappearance of the starting ketone. Remarkably, four enzymes from this kit provided the desired amine 9 in optical purities over 80% ee, the ATA-200 conducting to the (S)-9 (entry 8), while the TA-P1-B04, TA-P1-F03 and TA-P1-G05 gave access to its amine antipode (entries 23, 24 and 26). Initially, for the biotransamination experiments made in house ATAs were used, all of them overexpressed in Escherichia coli. Some of them, such as the ones from Chromobacterium violaceum [52] or Arthrobacter species [53], displayed very low activity (<5%), while others such as Arthrobacter citreus [54] or the Arthrobacter species evolved variant named ArRmut11 [55] provided almost quantitative conversion but moderate (74% ee) or negligible stereoselectivity, respectively. Trying to improve both activity and selectivity values, commercially available ATAs were employed from two different commercial sources (Codexis Inc. and Enzymicals AG).
To start with, 30 Codexis enzymes were employed (Table 1), and we found that 19 of them led to the complete disappearance of the starting ketone. Remarkably, four enzymes from this kit provided the desired amine 9 in optical purities over 80% ee, the ATA-200 conducting to the (S)-9 (entry 8), while the TA-P1-B04, TA-P1-F03 and TA-P1-G05 gave access to its amine antipode (entries 23, 24 and 26). Using the best found enzyme, TA-P1-G05 (entry 26, >99% conversion and 93% ee), the transamination of 6 was followed over time using two enzyme loadings (90% and 45% w/w enzyme vs. ketone); we observed a very fast conversion in the first 2 h and then a slower rate until complete depletion of the substrate occurred, after 6 h or 24 h, respectively ( Figure 3). Using the best found enzyme, TA-P1-G05 (entry 26, >99% conversion and 93% ee), the transamination of 6 was followed over time using two enzyme loadings (90% and 45% w/w enzyme vs. ketone); we observed a very fast conversion in the first 2 h and then a slower rate until complete depletion of the substrate occurred, after 6 h or 24 h, respectively ( Figure 3). Eight enzymes from Enzymicals AG were employed (Table 2), finding in three cases an amine with over 90% ee (entries 3, 7 and 8). Interestingly, the ATA08 from Silicibacter pomeroyi allowed the quantitative conversion of the ketone into the amine (R)-9 (entry 7).   Using the best found enzyme, TA-P1-G05 (entry 26, >99% conversion and 93% ee), the transamination of 6 was followed over time using two enzyme loadings (90% and 45% w/w enzyme vs. ketone); we observed a very fast conversion in the first 2 h and then a slower rate until complete depletion of the substrate occurred, after 6 h or 24 h, respectively (Figure 3). Eight enzymes from Enzymicals AG were employed (Table 2), finding in three cases an amine with over 90% ee (entries 3, 7 and 8). Interestingly, the ATA08 from Silicibacter pomeroyi allowed the quantitative conversion of the ketone into the amine (R)-9 (entry 7).   Using the best found enzyme, TA-P1-G05 (entry 26, >99% conversion and 93% ee), the transamination of 6 was followed over time using two enzyme loadings (90% and 45% w/w enzyme vs. ketone); we observed a very fast conversion in the first 2 h and then a slower rate until complete depletion of the substrate occurred, after 6 h or 24 h, respectively (Figure 3). Eight enzymes from Enzymicals AG were employed (Table 2), finding in three cases an amine with over 90% ee (entries 3, 7 and 8). Interestingly, the ATA08 from Silicibacter pomeroyi allowed the quantitative conversion of the ketone into the amine (R)-9 (entry 7).  Eight enzymes from Enzymicals AG were employed (Table 2), finding in three cases an amine with over 90% ee (entries 3, 7 and 8). Interestingly, the ATA08 from Silicibacter pomeroyi allowed the quantitative conversion of the ketone into the amine (R)-9 (entry 7). In order to improve the conversion values towards the amine (S)-9 the ATA03 Neosartorya fischeri (entry 3) and ATA07 Mycobacterium vanbaalenii (entry 7) were selected for optimization studies. So, new experiments were developed that includes the decrease of the substrate concentration, the use of longer reaction times, higher temperatures and enzyme loadings, and the performance of the biotransaminations without an organic cosolvent (Table 3). Interestingly, the best results were found when no cosolvent was employed, suggesting a deactivation of both enzymes in the presence of even low amounts of the organic solvent (MeCN, 5% v/v). In particular, the reduction of the substrate concentration from 20 to 10 mM of ketone 6 allowed higher conversions, although this limited its practical application. In addition, prolonged reaction times led to better conversions, while the use of higher temperatures led to a significant deactivation of the enzyme. Table 3. Optimisation of the biotransamination of ketone 6 using selected enzymes a . In order to improve the conversion values towards the amine (S)-9 the ATA03 Neosartorya fischeri (entry 3) and ATA07 Mycobacterium vanbaalenii (entry 7) were selected for optimization studies. So, new experiments were developed that includes the decrease of the substrate concentration, the use of longer reaction times, higher temperatures and enzyme loadings, and the performance of the biotransaminations without an organic cosolvent (Table 3). Interestingly, the best results were found when no cosolvent was employed, suggesting a deactivation of both enzymes in the presence of even low amounts of the organic solvent (MeCN, 5% v/v). In particular, the reduction of the substrate concentration from 20 to 10 mM of ketone 6 allowed higher conversions, although this limited its practical application. In addition, prolonged reaction times led to better conversions, while the use of higher temperatures led to a significant deactivation of the enzyme. Focusing on the scaling up of the biotransformations, we decided to move to higher substrate concentrations (50 mM of ketone) in order to produce a significant amount of the optically active amine (R)-9, which is a precursor of organic molecules with interesting biological profiles [2][3][4][5]. In this case, 225 mg of 6 were used, selecting TA-P1-G05 (entry 26, Table 1) as the ideal candidate since in standard conditions the amine (R)-9 was formed in complete conversion and good selectivity (93% ee). The enzyme loading was reduced from an initial 90% w/w enzyme vs. substrate ratio to 33% to improve the economy of the process, and after 22 h quantitative conversion was also achieved, maintaining the selectivity and isolating the desired amine in 98% yield after a simple liquid-liquid extraction protocol (Scheme 3). Measurement of the optical rotation for the pure amine and its corresponding hydrochloride salt allowed us to unequivocally assign the absolute configuration by comparison with previously reported data [4,5]. Focusing on the scaling up of the biotransformations, we decided to move to higher substrate concentrations (50 mM of ketone) in order to produce a significant amount of the optically active amine (R)-9, which is a precursor of organic molecules with interesting biological profiles [2][3][4][5]. In this case, 225 mg of 6 were used, selecting TA-P1-G05 (entry 26, Table 1) as the ideal candidate since in standard conditions the amine (R)-9 was formed in complete conversion and good selectivity (93% ee). The enzyme loading was reduced from an initial 90% w/w enzyme vs. substrate ratio to 33% to improve the economy of the process, and after 22 h quantitative conversion was also achieved, maintaining the selectivity and isolating the desired amine in 98% yield after a simple liquid-liquid extraction protocol (Scheme 3). Measurement of the optical rotation for the pure amine and its corresponding hydrochloride salt allowed us to unequivocally assign the absolute configuration by comparison with previously reported data [4,5].
Melting point of compound 6 was measured in an open capillary in an Electrothermal digital melting point IA9200 apparatus (Cole-Parmer, Stone, UK) and is uncorrected. Elemental analyses were performed on a Thermo Scientific Flash 2000 analyzer (Thermo Flash & Carlo Erba Analyzers, Pennsauken, NJ, USA) and the measured values were indicated with the symbols of the elements or functions within ± 0.4% of the theoretical values. NMR spectra were recorded on a Bruker AV300 MHz spectrometer (Bruker Co., Faellanden, Switzerland). All chemical shifts (δ) are given in parts per million (ppm) and referenced to the residual solvent signal as internal standard. High-resolution mass spectroscopy (HRMS) was performed on a VG AutoSpec Q high-resolution mass spectrometer (Fision Instrument, Milford, MA, USA). Measurement of the optical rotation values was carried out at 590 nm on an Autopol IV Automatic polarimeter (Rudolph Research Analytical, Hackettstown, NJ, USA).
Gas chromatography (GC) analyses were performed for the determination of conversion values using a Hewlett-Packard HP-6890 chromatograph (Hewlett Packard, Palo Alto, CA, USA). A nonchiral HP-1 column (Agilent Technologies, Inc., Wilmington, DE, USA) was used with the following temperature programme: 90 °C (2 min) then 10 °C/minutes and finally 180 °C (0 min). The reaction crudes were analysed, obtaining the following retention times: 9.3 min for ketone 6 and 10.5 min for amine 9.
High-performance liquid chromatography (HPLC) analyses were performed for enantiomeric excess value measurements using an Agilent 1260 Infinity chromatograph with UV detector at 210 nm (Agilent Technologies, Inc., Wilmington, DE, USA). A Chiralpak IA (25 cm × 4.6 mm) was used as chiral column at 30 °C (Chiral Technologies, Mainz, Germany), employing a mixture of n-hexane/2propanol (90:10) as eluent with a 0.8 mL/min flow. The reaction crudes were derivatised as acetamides, obtaining the following retention times: 11.2 min for the (R)-10 and 12.6 min for the (S)-10 enantiomer (Figure 4).
Melting point of compound 6 was measured in an open capillary in an Electrothermal digital melting point IA9200 apparatus (Cole-Parmer, Stone, UK) and is uncorrected. Elemental analyses were performed on a Thermo Scientific Flash 2000 analyzer (Thermo Flash & Carlo Erba Analyzers, Pennsauken, NJ, USA) and the measured values were indicated with the symbols of the elements or functions within ±0.4% of the theoretical values. NMR spectra were recorded on a Bruker AV300 MHz spectrometer (Bruker Co., Faellanden, Switzerland). All chemical shifts (δ) are given in parts per million (ppm) and referenced to the residual solvent signal as internal standard. High-resolution mass spectroscopy (HRMS) was performed on a VG AutoSpec Q high-resolution mass spectrometer (Fision Instrument, Milford, MA, USA). Measurement of the optical rotation values was carried out at 590 nm on an Autopol IV Automatic polarimeter (Rudolph Research Analytical, Hackettstown, NJ, USA).
Gas chromatography (GC) analyses were performed for the determination of conversion values using a Hewlett-Packard HP-6890 chromatograph (Hewlett Packard, Palo Alto, CA, USA). A non-chiral HP-1 column (Agilent Technologies, Inc., Wilmington, DE, USA) was used with the following temperature programme: 90 • C (2 min) then 10 • C/minutes and finally 180 • C (0 min). The reaction crudes were analysed, obtaining the following retention times: 9.3 min for ketone 6 and 10.5 min for amine 9.
High-performance liquid chromatography (HPLC) analyses were performed for enantiomeric excess value measurements using an Agilent 1260 Infinity chromatograph with UV detector at 210 nm (Agilent Technologies, Inc., Wilmington, DE, USA). A Chiralpak IA (25 cm × 4.6 mm) was used as chiral column at 30 • C (Chiral Technologies, Mainz, Germany), employing a mixture of n-hexane/2-propanol (90:10) as eluent with a 0.8 mL/min flow. The reaction crudes were derivatised as acetamides, obtaining the following retention times: 11.2 min for the (R)-10 and 12.6 min for the (S)-10 enantiomer (Figure 4).
Melting point of compound 6 was measured in an open capillary in an Electrothermal digital melting point IA9200 apparatus (Cole-Parmer, Stone, UK) and is uncorrected. Elemental analyses were performed on a Thermo Scientific Flash 2000 analyzer (Thermo Flash & Carlo Erba Analyzers, Pennsauken, NJ, USA) and the measured values were indicated with the symbols of the elements or functions within ± 0.4% of the theoretical values. NMR spectra were recorded on a Bruker AV300 MHz spectrometer (Bruker Co., Faellanden, Switzerland). All chemical shifts (δ) are given in parts per million (ppm) and referenced to the residual solvent signal as internal standard. High-resolution mass spectroscopy (HRMS) was performed on a VG AutoSpec Q high-resolution mass spectrometer (Fision Instrument, Milford, MA, USA). Measurement of the optical rotation values was carried out at 590 nm on an Autopol IV Automatic polarimeter (Rudolph Research Analytical, Hackettstown, NJ, USA).
Gas chromatography (GC) analyses were performed for the determination of conversion values using a Hewlett-Packard HP-6890 chromatograph (Hewlett Packard, Palo Alto, CA, USA). A nonchiral HP-1 column (Agilent Technologies, Inc., Wilmington, DE, USA) was used with the following temperature programme: 90 °C (2 min) then 10 °C/minutes and finally 180 °C (0 min). The reaction crudes were analysed, obtaining the following retention times: 9.3 min for ketone 6 and 10.5 min for amine 9.

General Procedure for the Biotransamination of 6 Using ATAs Overexpressed in Escherichia coli
The lyophilised cells of E. coli containing overexpressed transaminases (5 mg) were suspended in a 100 mM phosphate buffer pH 7.5 (475 µL) containing PLP (1 mM) and 2-propylamine (1 M). Then, a stock solution of ketone 6 in MeCN was added (25 µL of stock 0.4 M; final concentration 20 mM) and the mixture was shaken at 30 • C and 250 rpm for 20 h. After this time, the reaction was quenched by adding an aqueous 10 M NaOH solution (200 µL) and extracted with EtOAc (2 × 500 µL). The organic phases were combined and dried over anhydrous Na 2 SO 4 . The reaction crudes were analysed by GC to determine conversion values. Derivatisation were carried out in situ using acetic anhydride and K 2 CO 3 for the measurement of the enantiomeric excesses through HPLC.

General Procedure for the Biotransamination of 6 Using Commercial ATAs
Transaminases from Codexis or Enzymicals AG (2 mg, 90% w/w) were suspended in a 100 mM phosphate buffer pH 7.5 (475 µL) containing PLP (1 mM) and 2-propylamine (1 M). Then, a stock solution of ketone 6 in MeCN was added (25 µL of stock 0.4 M; final concentration 20 mM) and the mixture was shaken at 30 • C and 250 rpm for 20 h. After this time, the reaction was quenched by adding an aqueous 10 M NaOH solution (200 µL) and extracted with EtOAc (2 × 500 µL). The organic phases were combined and dried over anhydrous Na 2 SO 4 . Reaction crude was analysed by GC to determine conversion values and in situ derivatisation was carried out using acetic anhydride and K 2 CO 3 for the measurement of the enantiomeric excesses by HPLC.

Computational Methods
Calculations were performed using the Gaussian 09 package [57] at the M06-2X/6-311++G(3df,2p) level [58]. Molecular geometries of the studied compounds were optimised with tight convergence criteria and the frequencies were computed in order to obtain the thermal correction to the energy (298.15 K).

Conclusions
The synthesis of the 3,4-dihydro-2H-1,5-benzoxathiepin-3-amine enantiomers has been possible by means of the stereoselective biotransamination of the 3,4-dihydro-2H-1,5-benzoxathiepin-3-one. A broad panel of commercially available amine transaminases were employed, finding after optimisation of parameters that affect the enzyme catalysis suitable reaction conditions for the access to both amine antipodes in high conversions and good selectivities. A scale-up experiment considering 50 mM substrate concentration was successfully achieved for the formation of the (R)-3,4-dihydro-2H-1,5-benzoxathiepin-3-amine (R-9), a valuable precursor of anti-proliferative agents.