In situ epoxide generation by dimethyldioxirane oxidation and the use of epichlorohydrin in the flow synthesis of a library of β-amino alcohols

The flow coupling of epichlorohydrin with substituted phenols, while efficient, limits the nature of the epoxide available for the development of focused libraries of β-amino alcohols. This limitation was encountered in the production of analogues of 1-(4-nitrophenoxy)-3-((2-((4-(trifluoromethyl)pyrimidin-2-yl)amino)ethyl)amino)propan-2-ol 1, a potential antibiotic lead. The in situ (flow) generation of dimethyldoxirane (DMDO) and subsequent flow olefin epoxidation abrogates this limitation and afforded facile access to structurally diverse β-amino alcohols. Analogues of 1 were readily accessed either via (i) a flow/microwave hybrid approach, or (ii) a sequential flow approach. Key steps were the in situ generation of DMDO, with olefin epoxidation in typically good yields and a flow-mediated ring opening aminolysis to form an expanded library of β-amino alcohols 1 and 10a–18g, resulting in modest (11a, 21%) to excellent (12g, 80%) yields. Alternatively flow coupling of epichlorohydrin with phenols 4a–4m (22%–89%) and a Bi(OTf)3 catalysed microwave ring opening with amines afforded a select range of β-amino alcohols, but with lower levels of aminolysis regiocontrol than the sequential flow approach.

The flow coupling of epichlorohydrin with substituted phenols, while efficient, limits the nature of the epoxide available for the development of focused libraries of β-amino alcohols. This limitation was encountered in the production of analogues of 1-(4-nitrophenoxy)-3-((2-((4-(trifluoromethyl)pyrimidin-2-yl) amino)ethyl)amino)propan-2-ol 1, a potential antibiotic lead. The in situ (flow) generation of dimethyldoxirane (DMDO) and subsequent flow olefin epoxidation abrogates this limitation and afforded facile access to structurally diverse β-amino alcohols. Analogues of 1 were readily accessed either via (i) a flow/microwave hybrid approach, or (ii) a sequential flow approach. Key steps were the in situ generation of DMDO, with olefin epoxidation in typically good yields and a flowmediated ring opening aminolysis to form an expanded library of β-amino alcohols 1 and 10a-18g, resulting in modest (11a, 21%) to excellent (12g, 80%) yields. Alternatively flow coupling of epichlorohydrin with phenols 4a-4m (22%-89%) and a Bi(OTf) 3 catalysed microwave ring opening with amines afforded a select range of β-amino alcohols, but with lower levels of aminolysis regiocontrol than the sequential flow approach.

Introduction
As part of our medicinal chemistry programme we identified β-amino alcohol 1 as a lead compound of interest. Analogues of this nature are known to be biologically active and have  been used as key intermediates in the synthesis of natural products [1][2][3][4][5]. β-Amino alcohols are typically accessed through amine-mediated ring opening of epoxides, catalysed by Lewis acids [6][7][8][9]. In keeping with an epoxide ring opening strategy, 1 can be accessed via aminopyrimidine 2, racemic epichlorohydrin 3 and 4-nitrophenol 4, enabling robust focused library development (figure 1).
Herein we report a flow chemistry approach to epoxide installation using epichlorohydrin and the in situ epoxidation of olefins with dimethyldioxirane (DMDO) and the subsequent epoxide aminolysis, facilitating access to a focused library of β-amino alcohols based on 1 [44].

Results and discussion
Method development commenced with the use of a flow instrument equipped with peristaltic pumps; passing two reagent streams, epichlorohydrin 3 (neat) and 4-nitrophenol 4 (0.1 M) in dimethylformamide (DMF), through an Omnifit ® column packed with Cs 2 CO 3 and acid washed sand (1 : 1) at 75°C, increasing in 10°C increments to 105°C, with the reaction monitored by UPLC-MS (scheme 1; table 1). The system pressure was maintained at 5 bar using the third pump in an active back pressure regulation mode, to circumvent the blockage issues caused by the precipitation of the product we had previously observed. This screen identified an enhanced coupling efficiency with 3 and 4 at higher temperatures, with DMF affording 97% at 105°C. Solvent switching to CH 3 CN/Cs 2 CO 3 or DMF/diisopropylethyl amine (DIPEA) returned only unreacted 3 and 4. DMF/1,8-diazabicyclo(5. 4.0)undec-7-ene (DBU) and DMF/N(Bu) 4 OAc gave 90% and 100% conversion to 5a, respectively. With DMF/N(Bu) 4 OAc, poor product recovery was a function of difficulties observed in isolating the desired product in the presence of N(Bu) 4 OAc; thus Cs 2 CO 3 was used for on-going optimizations. Despite the ready recycling of epichlorohydrin, the use of reagents as solvents seriously limited future library generation and poses rsos.royalsocietypublishing.org R. Soc. open (table 2). Use of 0.1 M of 4-nitrophenol 4 and 5, 2.5 and 0.5 M 3 in DMF gave 66%, 39% and 13% conversion to 5a, respectively. At 2.5 M of 3, a 20 min residence time saw increased conversion to 5a from 39% to 70%. At 115°C, unwanted diol 6a was evident, but this was reduced through the use of a 4 Å sieves/acid washed sand/Cs 2 CO 3 (2 : 1 : 1 v/v/v) loaded Omnifit ® column. Quantitative conversion to epoxy ether 5a was noted at 105°C, with 5 M 3 and a 20 min residence time (scheme 2).
consumption of phenol 4 but the presence of the undesired diol (5c, 5e and 5m), incomplete consumption of 4 (5j, 5k and 5l) or incomplete consumption of 4 and undesired diol (5f-5i) (table 3). In our hands the use of DMF and solid Cs 2 CO 3 bypassed any potential issue with bi-phasic solutions, and the boiling point difference between DMF and epichlorohydrin facilitated easy removal and recycling of the latter reagent [29]. Racemic epichlorohydrin was used; while moderately inexpensive, the recycling procedure was considered necessary to reduce the environmental effects of excess epichlorohydrin, as well as demonstrate the use of the protocol if enantiomerically pure products were required.
Concerned about the possible depletion of the Cs 2 CO 3 from the Omnifit ® column, we scaled up the synthesis of 5a, with a column packed with 16 g 1 : 1 w/w clean sand : Cs 2 CO 3 , examining the conversion rates as a function of time with analysis conducted at 30 min time points (figure 2). From this data, we were able to generate 1.48 g 5a in 3 h (84% isolated yield; 0.49 g h −1 ). During the course of this continuous production the observed conversion rate remained essentially constant for the first 2.5 h, only showing a slight diminution after 3 h, to 94%. There was no evidence of channels of favoured flow paths developing within the Omnifit ® column.
To nullify the limited availability of novel epoxides, we explored the in situ epoxidation capability of DMDO with selected aromatic and aliphatic olefins (table 4) [45,46]. The production of water-soluble by-products, as well as leveraging the known advantage of flow approaches in handling potentially sensitive reagents, made this method more attractive than traditional epoxidizing approaches [31,[41][42][43]. No epoxidation of 4-allylanisole 7 with NaHCO 3 /Oxone ® (1 : 1; in situ production of DMDO) was observed at 19°C, but at 30°C 16% conversion to epoxide 8a was noted by gas chromatography mass spectrometry (GCMS) analysis (scheme 3). This increased to 61% at 60°C. Quantitative conversion to epoxide 8a was accomplished using 2 equivalents of Oxone ® at 60°C (electronic supplementary material, table S1). To further improve the safety of this reaction, the optimized reaction was repeated with the addition of a 0.4 M sodium sulfite (aq) stream introduced after the back pressure regulator to quench any unreacted DMDO. Analysis and subsequent yield calculation showed no reduction in conversion or isolated yields.
The DMDO flow epoxidation protocol was explored with an auto-sampler and fraction collector equipped flow system and provided an effective and robust approach to the semi-automated synthesis and collection of epoxides via aromatic and aliphatic olefins (scheme 3 and table 4). Simple aromatic vinyl scaffolds showed good conversion (greater than 70%), with excellent isolated yields of benzyl epoxide 8a, styrene oxides 8b and 8c, as did naphthoquinone and cinnamyl alcohol, yielding 8d and 8e (80% and 70%, respectively). By contrast, cinnamic acid 7f, chromeone 7g, indole 7h and cyclohex-2-en-one 7i scaffolds were resistant to DMDO epoxidation, with epoxides 8f-8i at best observed at low levels (5% or less). Allyl bromide 7j showed quantitative conversion to 8j, while the highly substituted mesityl oxide 7k showed a 79% conversion to mesityl epoxide 8 k (table 4).
Having established two robust flow protocols to install epoxides via epichlorohydrin and DMDO, we next examined the amine-mediated ring opening of the freshly installed epoxide moiety. As we had previously reported similar amine mediated ring openings using bismuth (III) chloride as catalyst [8], we explored this approach; but to suppress the formation of chlorinated side products that can occur with this catalyst, bismuth (III) triflate (Bi(OTf) 3 ) was examined. However, under flow conditions, employing 30 mol% Bi(OTf) 3 , 120°C, 5 bar and a residence time of 20 min, only a small amount (17%) of the amino alcohol was detected. Increasing the amount of bismuth catalyst to 50% increased the formation of product to 75%; however, as excess catalyst is undesirable for purification, we shifted focus to the application of microwave approaches (scheme 4) [44,47,48].
Coupling of epoxide 5a with aniline and Bi(OTf) 3                 While the Bi-catalysed approach did allow access to the desired amino alcohols, the poor regioselectivity observed severely hampered compound access. We turned our attention to a recent report from the Seeberger laboratory that reported the model epoxide aminolysis of styrene with t-butyl and isopropyl amine by flow, [29] as well as a previous publication by Jamison and coworkers. [30]. While Seeberger's work reported the successful synthesis of a number of APIs under catalyst free conditions, the aminolysis step was limited to the aforementioned amines and the scope of the reaction was not defined. We endeavoured to extend the reach of these approaches. Application of the former protocols in our systems revealed that the optimum epoxide aminolysis conditions for aryl substituted epoxides ranged from 5-7 equivalents of t-butylamine, 120-150°C, 5-10 bar back pressure (scheme 5) and 1 : 1 v/v mixture of toluene : ethanol. These conditions afforded predominately the desired secondary amino alcohols (table 5).
Having established standard reaction conditions for aryl, aryl methyl and aryloxy epoxides, we applied this methodology across a range of primary, secondary and tertiary amines, as well as aniline, benzyl and N-benzylmethyl amines. Regioselective conversion to secondary alcohols 10b-10g was observed with epoxide 5a and amines 9b-9g (table 6), as with the microwave protocol, and flow aminolysis yields ranged from modest to excellent (figure 2, 36%-80%). The exceptions to this were reactions of 5a with 9a and 9b. Reaction of N-propylamine 9b afforded a mixture of primary (10b), secondary alcohols (13b) and bis-alkylation (16b), while reaction of epoxide 5a and t-butylamine rsos.royalsocietypublishing.org R. Soc. open Table 6. Flow aminolysis 5a, 8a and 8b leading to the formation of amino alcohols 1 and (10a-g)-(18a-g          9a furnished a 20% impurity of the bis-alkylated by-product (16a). Treatment of 5a with N 1 -(4-(trifluoromethyl)pyrimidin-2-yl)ethane-1,2-diamine 2 regioselectively afforded exclusively 1 in a 65% isolated yield. This protocol delivered higher selectivity than microwave and batch methods, with no evidence of bis-alkylated side-product; [9] it has been reported that flow methodologies offer better selectivity over other methods owing to improved mixing [49]. Aminolysis of epoxide 8a with amines 9a-9g furnished 11a-11g with ≥90% regioselectivity for the secondary alcohol, in principle simplifying purification of the desired regioisomer compared to the microwave methodology. However, the isolated yields of 11a-11g ranged from 21% to 79%. LCMS analysis revealed product contamination with residual amine, and this was most apparent with aliphatic amines 9a and 9c.
Reaction of styrene epoxide 8b with amines 9a-9g showed reduced regioselectivity. The coupling of styrene epoxide 8b with 9a and 9f afforded mixtures of primary (12a and 12f) and secondary alcohols (15a and 15f), as well as bis-alkylation (18f). Aniline 9e afforded a 1 : 1 mixture of primary (12e) and secondary alcohols (15e), while the secondary amines 9b and 9c, as well as morpholine 9d, benzyl amine 9f and N-benzylmethylamine 9 g, furnished the secondary alcohol in ≥82% regioselectivity. Yields of 12a-12f ranged from modest to excellent (35-78%). As seen with epoxides 5a and 8a, reaction of Npropylamine 9b resulted in a mixture of primary (12b), secondary (15b) and bis-alkyated products (18b). Compared with the microwave protocol, purification was improved because of improved regioselectivity (see Experimental section for regioisomer ratios). The notable exception to this proved to be 12g, which was inseparable from 15g by column chromatography, but offered comparable selectivity to the desired secondary alcohol as the microwave protocol.

Conclusion
Herein we have demonstrated that the flow coupling of phenols with epichlorohydrin provides a facile and highly scalable route to a wide variety of epoxides in moderate to excellent yields (22-89%). This access is scalable with multi-gram quantities of aryloxy epoxides accessible, e.g. 5a (0.49 g h −1 ; with our three hour flow synthesis realizing 1.48 g of material). A microwave/Bi(OTf) 3 -mediated epoxide opening aminolysis gave moderate control over β-amino alcohol regiochemistry. However, our refinement of previously published aminolysis protocols [29,30] enabled development of a robust flow protocol with aryl, arylmethyl and aryloxy epoxides and a range of primary, secondary and tertiary amines, as well as aniline, benzyl and N-benzylmethyl amines. With this approach we were able to access a diverse range of β-amino alcohols with high regioselectivity. Further, we abrogated the limitations of this protocol, i.e. the use of epichlorohydrin, through the first application of flow DMDO olefin epoxidation, which was applicable to a wide range of aromatic and aliphatic olefins with isolated epoxide yields of 60-98%. The in situ generation of the reactive DMDO species improves the safety of the reaction compared with traditional batch methodologies, and semi-automation of this protocol using an auto-sampler and fraction collector facilitates rapid library generation. The in-flow epoxidation of olefins and flow epoxide aminolysis protocols developed herein offer the medicinal chemist a simple pathway to novel epoxides and, consequently, highly decorated β-amino alcohols of potential biological importance.

General methods
All reagents were purchased from Sigma-Aldrich, AK Scientific, Matrix Scientific or Lancaster Synthesis and were used without purification. All solvents were used as supplied. 1

Recycling procedure for epichlorohydrin
Epichlorohydrin was recovered and recycled by collecting post-reaction under vacuum (30 mbar, 50°C). The resulting crude mixture was then diluted with CH 2 Cl 2 (100 ml). The solution was washed with water (3 × 150 ml) and saturated brine (150 ml). The organic layers were separated, dried over Mg 2 SO 4 and concentrated in vacuo (40°C, 240 mbar) to afford a colourless oil.

General procedure 1
The Vapourtec easy-MedChem was charged with a 0.1 M solution of phenol (2 mmol) in anhydrous DMF (20 ml) at a flow rate of 0.5 ml min −1 . The solution was then passed through a 7.85 ml Omnifit column packed with a 1 : 1 mixture of Cs 2 CO 3 /acid washed sand (2.30 g, total weight) and mixed with 5 M solution of epichlorohydrin in anhydrous DMF (0.5 ml min −1 ) through a T-piece. The resulting mixture was subsequently flowed through two 10 ml perfluroalkoxy (PFA) reaction coils set up in series at 105°C, 5 bar and affording a total of 20 min residence time. The resulting reaction mixture was collected, concentrated in vacuo and subject to column chromatography. The up-scaled reaction was carried out with the above concentration and an Omnifit column packed with a 1 : 1 mixture of Cs 2 CO 3 /acid (1 M HCl) washed sand (16 g, total mass).

2-((4-Nitrophenoxy)methyl)oxirane (5a)
Compound 5a was prepared using general procedure 1 and 4-nitrophenol (0.278 g, 2 mmol). The product was obtained as a light yellow solid (347 mg, 89%), m.p.: 66-69°C. 1   The resulting reaction stream was then quenched in-line (immediately after the back pressure regulator) using a stream 0.4 M sodium sulfite (aq) . The solution was then collected, concentrated in vacuo to remove acetone and diluted up to 50 ml with water (pH 7). The pH of the solution was adjusted to pH ∼7 using saturated ammonium chloride (aq) . The aqueous solution was extracted with ethyl acetate (3 × 50 ml), the organic layers were combined and washed with brine (50 ml). The organic layer was separated, dried over magnesium sulfate and concentrated in vacuo to afford the desired product; no further purification was required unless stated.

General procedure 3
A suspension of epoxide (1.00 mmol), amine (1.10 mmol) and bismuth (III) trifluoromethane sulfonate (15 mol%) in CH 3 CN (3 ml) was subjected to microwave irradiation at the below stated temperature and time (typical graph of the temperature/power/pressure in the electronic supplementary material, figure S1). The reaction mixture was concentrated in vacuo and crude material subjected to column chromatography.

General procedure 4
Using a Vapourtec RS-400 equipped with fraction collection kit and auto-sampler, either a 0.5 ml (loop A), 1 . The solutions were flowed together and the resulting stream was then passed through two PFA coil reactors in series at 150°C, 10 bar back pressure and 1 ml min −1 (residence time 20 min). The resulting reaction mixture was collected and purified as described.