Triazole antifungal agents have been in widespread clinical use for the treatment of various fungal infections, not only high-risk diseases such as acute invasive aspergillosis(1) but also low-risk pathological conditions such as nail infections.(2) Although fungal infections are generally regarded as nonfatal clinical conditions, immunocompromised patients with human immunodeficiency virus (HIV) infection or undergoing cancer chemotherapy are susceptible to life-threatening fungal diseases.(3) The representative triazole antifungal agent fluconazole (Figure 1), which was developed by Pfizer, is effective orally against a range of fungal infections.(4) However, it is poorly active against Aspergillus spp., which cause life-threatening infections in immunocompromised patients, and fluconazole resistance has been reported in patients receiving long-term treatment.(5) To address these issues, an advanced triazole antifungal agent, voriconazole (Vfend), was developed.(6) This agent is active against all Candida spp., including fluconazole-resistant Candida albicans, Candida glabrata, and Candida krusei, as well as several Aspergillus spp., including the amphotericin B-resistant Aspergillus terreus.(7) Therefore, voriconazole is a primary drug in the first-line treatment of invasive aspergillosis, as either an intravenous or oral formulation.(1a, 7f, 8) The high potency of voriconazole has inspired extensive efforts devoted to the development of a variety of synthetic derivatives.(9) The replacement of one of the triazole rings in fluconazole with heteroaromatics (e.g., 5-fluoropyrimidine for voriconazole) and the installation of a methyl group next to the tetrasubstituted stereogenic center have been proved to be beneficial structural modifications,(6a) leading to the identification of advanced triazole antifungal agents. As these modifications break the molecular symmetry of achiral fluconazole, the development of an efficient enantioselective synthetic route is in high demand.
Figure 1. Structures of fluconazole and advanced triazole antifungal agents.
Considering the common substructures that are shared in these antifungal agents, the enantiomerically pure epoxide 1 bearing 2,4-difluorobenzene, 1,2,4-triazole, and a methyl group is a rational intermediate for their divergent syntheses (Scheme 1). The enantioselective construction of the consecutive tetra- and trisubstituted stereogenic centers represents a formidable task in the synthesis of epoxide 1. Because of the pivotal role of 1 in the efficient synthesis of these antifungal agents, various synthetic approaches toward epoxide 1 have been investigated.(10-12) Almost all of the synthetic routes rely on the use of d- or l-lactic acid as a chiral pool.(11a-11c, 12) In particular, Bristol-Myers Squibb has developed an excellent approach toward epoxide 1 in six steps in 25% overall yield, leading to the scalable synthesis of ravuconazole.(10) Although other approaches utilizing Sharpless asymmetric epoxidation(11e) or enzymatic resolution(11f) have been accomplished, there remains room for improvement in terms of the number of synthetic steps. The exploitation of a catalytic asymmetric C–C bond-forming reaction is a viable option for the integration of the construction of a molecular skeleton with a stereogenic center. We recently demonstrated that the catalytic asymmetric cyanosilylation of a ketone is particularly useful for the construction of the elusive tetrasubstituted stereogenic center, culminating in the enantioselective synthesis of voriconazole.(13)
Scheme 1. Enantioselective Synthesis of Epoxide 1, a Key Intermediate for Various Antifungal Agents
Herein we report a new route, the shortest reported to date, to access epoxide 1 in four steps from commercially available ketone 2 in 29% yield. The key step features the catalytic asymmetric cyanosilylation to construct the tetrasubstituted stereogenic center of 3. The utility of this synthetic approach has been demonstrated by the efficient syntheses of the significant antifungal agents ravuconazole(9b, 10) and efinaconazole (Jublia).(9d, 14) Ravuconazole, which bears a functionalized thiazole, features a broad antifungal spectrum as well as the longest half-life and has completed P2 clinical trials. Efinaconazol (Jublia), which possesses a 4-methylenepiperidine moiety and has recently received approval in Canada,(15) is the first external-use antifungal agent for the treatment of onychomycosis.(16)
Our synthesis commenced with the catalytic asymmetric cyanosilylation of ketone 2, a key reaction promoted by a Gd-based asymmetric catalyst to construct the tetrasubstituted stereogenic center (Scheme 2).(17-19) A putative Gd-based polymetallic catalyst composed of Gd and the sugar-derived chiral ligand 4(20, 21) in a 2:3 ratio, as suggested by ESI-MS analysis in the presence of TMSCN,(17b) was generated by mixing Gd(HMDS)3 and 4 in a 2:3 ratio at −30 °C. The polymetallic catalyst (2 mol% based on Gd) promoted the catalytic asymmetric cyanosilylation of 2 with TMSCN at −30 °C in propionitrile to afford the desired cyanohydrin 3 with TMS protection in 92% yield with 80% ee. Because of its instability under acidic and basic conditions and silica gel column chromatography, 3 was immediately submitted to DIBAL reduction to give corresponding aldehyde 5. Our next focus was the diastereoselective installation of a methyl group and 1,2,4-triazole. Initially, we faced several undesired transformations. After the formation of secondary alcohol 6 using organometallic reagents, quenching with acidic or basic aqueous solutions led to partial migration of the TMS group to provide a complicated mixture of 6 and 7 and their diastereomers. The secondary TMS group of 7(22) was prone to deprotection under either acidic or basic conditions as well as silica gel column chromatography. Even when 7 was isolated via laborious purification and subjected to 1,2,4-triazole introduction under basic conditions at room temperature, deprotection of the TMS group occurred and the subsequent formation of epoxide 9 proceeded partially. The suppression of these unwanted transformations was intractable, and the complicated reaction mixtures made the purification in each step fruitless. Given that all of the byproducts could be converted into diol 10, we anticipated that the sequential manifestation of these undesired transformations in one-pot would allow direct access to 10. After extensive manipulations of the reaction conditions, we found that the installation of the methyl group, the deprotection of the TMS group, the formation of epoxide 9, and the installation of 1,2,4-triazole could be carried out in a one-pot operation. The initial Grignard addition to 5 gave secondary alcohol 6. Other organometallic reagents resulted in low yield or low diastereoselectivity.(23) Treatment of the reaction mixture with a 3 N aqueous NaOH solution in the same flask converted 6 into tertiary alcohol 7 via intramolecular migration of the TMS group. Successive addition of 1,2,4-triazole and TBAB as a phase-transfer catalyst initially induced the removal of TMS to give diol 8, which eventually cyclized to afford epoxide 9 under basic conditions. Ring opening of epoxide 9 proceeded slowly to furnish diol 10 in favor of the desired diastereomer in an 86:14 ratio. The diastereomers were easily separable using silica gel column chromatography to provide the requisite diol 10 as a single diastereomer in 65% yield from aldehyde 5.
Scheme 2. Short Synthesis of the Key Intermediate Epoxide 1
Diol 10 is a crystalline solid, and enantioenrichment was attempted at this stage. When a concentrated acetonitrile solution oversaturated at 60 °C was submitted to rapid nucleation with stirring at −20 °C, a nearly racemic solid (6.3% ee) appeared, and the filtrate was enriched to 97% ee.(24) The second cycle of an identical procedure (but with stirring at 0 °C) afforded the enantiopure diol 10 (>99% ee) in 74% recovery yield after two cycles.(25) With the optically pure diol 10 in hand, we examined its transformation to epoxide 1. Regioselective mesylation of diol 10 proceeded smoothly at 0 °C to provide the transient intermediate 11, which was subsequently treated with a 3 N aqueous NaOH solution and TBAB to afford the key intermediate epoxide 1 in 86% yield in one pot.
Next, we turned our attention to the synthesis of efinaconazole (Scheme 3). According to the literature procedure,(9d) epoxide 1 was subjected to a ring-opening reaction with 4-methylenepiperidine(26) at 80 °C. However, 51% of epoxide 1 remained unchanged after 24 h, and efinaconazole was isolated in 44% yield. Microwave irradiation at 120 °C solved this problem, affording efinaconazole in 90% yield.(27) The spectroscopic data of the synthesized sample were identical to those of the reported one.(9d) Moreover, we also demonstrated the synthesis of ravuconazole according to the literature procedure (Scheme 4).(9b, 10) Ring opening of epoxide 1 using Et2AlCN provided cyanide 12 in 76% yield. The nitrile functionality of 12 was transformed into a primary thioamide with diethyl dithiophosphate to give 13 in excellent yield. Treatment of 13 with 2-bromo-4′-cyanoacetophenone furnished ravuconazole in 78% yield.
Scheme 3. Synthesis of Efinaconazole (Jublia)
Scheme 4. Synthesis of Ravuconazole
In conclusion, we have developed a new route, the shortest reported to date, to access the key intermediate epoxide 1 in 29% overall yield in four steps from the commercially available ketone 2. The key step features a catalytic asymmetric cyanosilylation using Gd(HMDS)3 and a sugar-derived chiral ligand to construct the tetrasubstituted stereogenic center that is essential in advanced triazole antifungal agents. This streamlined synthetic approach led us to demonstrate enantioselective efficient syntheses of two significant antifungal agents.
The reactions were performed in a round-bottom flask with a Teflon-coated magnetic stirring bar and a three-way glass stopcock under an Ar atmosphere, unless otherwise noted. Air- and moisture-sensitive liquids were transferred via a gastight syringe and a stainless steel needle. All workup and purification procedures were carried out using reagent-grade solvents under ambient atmosphere. Flash chromatography was performed using silica gel 60 (230–400 mesh). Chemical shifts (δ) for protons are reported in units of parts per million downfield from tetramethylsilane and are referenced to residual protons in the NMR solvent (CDCl3, 7.24 ppm). For 13C NMR, chemical shifts are reported on the scale relative to the NMR solvent (CDCl3, 77.0 ppm) as an internal reference. For 19F NMR, chemical shifts are reported on the scale relative to trifluoroacetic acid (76.5 ppm) as an external reference. NMR data are reported as follows: chemical shifts (multiplicity, coupling constant in Hz, integration). Multiplicities are denoted as follows: s, singlet; d, doublet; dd, doublet of doublets; t, triplet; q, quartet; sep, septet; m, multiplet; br, broad signal. Optical rotation was measured using a 2 mL cell with a 1.0 dm path length. Compounds 1, 3, 5, 8, 9, 10, 11, 12, and 13 are known compounds (CAS registry numbers 127000-90-2, 861718-83-4, 861718-85-6, 832151-94-7, 126918-35-2, 133775-25-4, 133775-26-5, 170862-36-9, and 170863-34-0, respectively).
To a solution of 5 (1.24 g, 4.23 mmol) in THF (6.70 mL) was added 0.92 M MeMgBr solution in THF (6.44 mL, 5.92 mmol) at −78 °C, and the reaction mixture was stirred at the same temperature for 35 min. The reaction mixture was quenched with saturated aqueous NH4Cl, and the resulting mixture was warmed to room temperature and stirred for 20 min. The aqueous layer was extracted twice with EtOAc. The combined organic layers were washed with H2O, dried over MgSO4, filtered, and concentrated under reduced pressure. The resulting residue was purified using silica gel column chromatography (n-hexane/EtOAc = 85:15) to give 740 mg of (2R,3S)-6 (57% yield) as a colorless oil and 117 mg of (2S,3S)-6 (9% yield) as a colorless oil. 1H NMR for (2R,3S)-6 (400 MHz, CDCl3) δ 7.53–7.47 (m, 1H), 6.90–6.85 (m, 1H), 6.77–6.71 (m, 1H), 4.21 (d, J = 12.1 Hz, 1H), 4.20–4.15 (m, 1H), 4.12 (dd, J = 12.1, 1.1 Hz, 1H), 0.90 (d, J = 6.0 Hz, 3H), 0.29 (s, 9H); 13C NMR for (2R,3S)-6 (100 MHz, CDCl3) δ 162.4 (dd, J = 249, 13 Hz), 158.2 (dd, J = 246, 12 Hz), 131.0 (dd, J = 9.1, 6.2 Hz), 124.8 (dd, J = 13, 3.8 Hz), 111.0 (dd, J = 21, 3.4 Hz), 103.9 (dd, J = 29, 26 Hz), 83.2 (d, J = 6.7 Hz), 71.4 (d, J = 3.8 Hz), 50.1 (d, J = 6.7 Hz), 18.6, 2.41; 19F NMR for (2R,3S)-6 (376 MHz, CDCl3) δ −108.7, −111.6; 1H NMR for (2S,3S)-6 (400 MHz, CDCl3) δ 7.47–7.41 (m, 1H), 6.90–6.85 (m, 1H), 6.81–6.75 (m, 1H), 4.27 (d, J = 12.1 Hz, 1H), 3.97 (q, J = 6.3 Hz, 1H), 3.88 (dd, J = 12.1, 1.4 Hz, 1H), 1.10 (d, J = 6.3 Hz, 3H), 0.24 (s, 9H); 13C NMR for (2S,3S)-6 (100 MHz, CDCl3) δ 162.5 (dd, J = 249, 13 Hz), 159.1 (dd, J = 248, 12 Hz), 131.3 (dd, J = 9.6, 5.8 Hz), 124.0 (dd, J = 13, 4.8 Hz), 110.8 (dd, J = 22, 4.3 Hz), 104.3 (dd, J = 29, 25 Hz), 83.2 (d, J = 4.8 Hz), 72.8 (d, J = 1.9 Hz), 48.6 (d, J = 7.7 Hz), 17.6, 2.33; 19F NMR for (2S,3S)-6 (376 MHz, CDCl3) δ −107.1, −111.2; IR for (2R,3S)-6 (CHCl3, cm–1) ν 3588, 3467, 2956, 1615, 1498, 1419, 1253; HRMS for (2R,3S)-6 (ESI-TOF) calcd for C13H19O2ClF2SiNa [M + Na]+m/z 331.0703, found 331.0702.
To a solution of (2R,3S)-6 (21.4 mg, 0.0693 mmol) in THF (115 μL) was added 3 N NaOH (46.0 μL, 0.139 mmol) at room temperature, and the reaction mixture was stirred at the same temperature for 10 min. The reaction mixture was quenched with saturated aqueous NH4Cl, and the aqueous layer was extracted twice with EtOAc. The combined organic layers were dried over MgSO4, filtered, and concentrated under reduced pressure. The resulting residue was purified using preparative TLC (n-hexane/EtOAc = 7:1) to give 12.4 mg of 7 (58% yield) as a colorless oil. 1H NMR (400 MHz, CDCl3) δ 7.72–7.65 (m, 1H), 6.92–6.87 (m, 1H), 6.78–6.72 (m, 1H), 4.31 (q, J = 6.2 Hz, 1H), 4.04 (d, J = 11.4 Hz, 1H), 3.84 (d, J = 11.4 Hz, 1H), 3.12 (s, 1H), 0.90 (d, J = 6.2 Hz, 3H), 0.15 (s, 9H); 13C NMR (100 MHz, CDCl3) δ 162.5 (dd, J = 249, 13 Hz), 158.6 (dd, J = 247, 13 Hz), 130.7 (dd, J = 9.6, 6.7 Hz), 123.4 (dd, J = 13, 3.8 Hz), 111.3 (dd, J = 21, 3.4 Hz), 103.8 (dd, J = 27, 25 Hz), 77.9 (d, J = 5.8 Hz), 70.7 (d, J = 4.8 Hz), 51.5 (d, J = 5.8 Hz), 18.5, 0.21; 19F NMR (376 MHz, CDCl3) δ −109.7, −111.2; IR (CHCl3, cm–1) ν 3545, 2959, 1619, 1503, 1422, 1254; HRMS (ESI-TOF) calcd for C13H19O2ClF2SiNa [M + Na]+m/z 331.0703, found 331.0703.
To a solution of (2R,3S)-6 (31.7 mg, 0.103 mmol) in THF (343 μL) was added 1.0 M TBAF solution in THF (113 μL, 0.113 mmol) at 0 °C, and the reaction mixture was stirred at the same temperature for 15 min. The reaction mixture was quenched with saturated aqueous NH4Cl, and the aqueous layer was extracted twice with EtOAc. The combined organic layers were dried over MgSO4, filtered, and concentrated under reduced pressure. The resulting residue was purified using preparative TLC (CHCl3/MeOH = 10:1) to give 18.3 mg of 8 (75% yield) as a colorless crystal. Mp 86–87 °C; 1H NMR (400 MHz, CDCl3) δ 7.63–7.57 (m, 1H), 6.93–6.88 (m, 1H), 6.81–6.75 (m, 1H), 4.27–4.14 (m, 3H), 3.09 (brs, 1H), 2.17 (brs, 1H), 0.96 (d, J = 6.4 Hz, 3H); 13C NMR (100 MHz, CDCl3) δ 162.7 (dd, J = 250, 12 Hz), 158.6 (dd, J = 247, 12 Hz), 130.1 (dd, J = 9.6, 6.7 Hz), 123.8 (dd, J = 13, 3.8 Hz), 111.4 (dd, J = 21, 3.8 Hz), 104.2 (dd, J = 28, 25 Hz), 77.8 (d, J = 4.8 Hz), 70.0 (d, J = 4.8 Hz), 51.7 (d, J = 5.8 Hz), 18.6; 19F NMR (376 MHz, CDCl3) δ −109.2, −110.7; IR (CHCl3, cm–1) ν 3433, 3266, 2979, 1617, 1500, 1272; HRMS (ESI-TOF) calcd for C10H11O2ClF2Na [M + Na]+m/z 259.0308, found 259.0310.
To a solution of (2R,3S)-6 (33.6 mg, 0.109 mmol) in THF (363 μL) was added 1.0 M TBAF solution in THF (272 μL, 0.272 mmol) at room temperature, and the reaction mixture was stirred at the same temperature for 23 h. The reaction mixture was quenched with saturated aqueous NH4Cl, and the aqueous layer was extracted twice with EtOAc. The combined organic layers were dried over MgSO4, filtered, and concentrated under reduced pressure. The resulting residue was purified using preparative TLC (CHCl3/MeOH = 15:1) to give 7.4 mg of 9 (34% yield) as a colorless oil. 1H NMR (400 MHz, CDCl3) δ 7.42–7.36 (m, 1H), 6.89–6.84 (m, 1H), 6.81–6.76 (m, 1H), 4.07 (qd, J = 6.6, 1.6 Hz, 1H), 3.28 (d, J = 5.3 Hz, 1H), 2.78 (dd, J = 5.3, 0.5 Hz, 1H), 1.14 (dd, J = 6.6, 1.1 Hz, 3H); 13C NMR (100 MHz, CDCl3) δ 162.8 (dd, J = 249, 13 Hz), 160.4 (dd, J = 249, 12 Hz), 130.6 (dd, J = 10, 6.2 Hz), 120.6 (dd, J = 15, 3.8 Hz), 111.4 (dd, J = 21, 3.8 Hz), 103.7 (dd, J = 25, 25 Hz), 68.4 (d, J = 1.9 Hz), 60.6, 51.9, 19.1; 19F NMR (376 MHz, CDCl
Early flow processing approaches
The first published examples of flow chemistry applied to the synthesis of pharmaceutically active molecules emerged in the early 2000s when several research groups reported on specific flow transformations that enabled a new synthesis of these known pharmaceuticals. Examples of these early endeavours include the syntheses of efaproxiral (1) and rimonabant (2) using a AlMe3-mediated direct amidation in flow , an improved metalation step in the scaled synthesis of NBI-75043 (3) , a continuous dehydration process to deliver over 5 kg of dehydropristane 4, a precursor of the immunoactivating agent pristane  or the flow synthesis of hydroxamic acids by a procedure that was also applied to the preparation of suberoylanilide hydroxamic acid (5, SAHA, Figure 1) . Another early application of microreactor technology was reported in 2005 detailing the assembly and subsequent decoration of the fluoroquinolinone scaffold 6 resulting in the synthesis of a library of analogues including the well-known antibiotic ciprofloxacin (6, R1 = cyclopropyl, R2 = piperazinyl) .
An important early industry-based example was disclosed by scientists at Bristol-Myers Squibb in 2008 detailing a flow approach towards converting the psychotropic agent buspirone (7) into its major active metabolite, 6-hydroxybuspirone (9) . This work comprised three consecutive flow steps including a low-temperature enolisation of buspirone (7). The subsequent reaction of the enolate with gaseous oxygen in a trickle-bed reactor was coupled to a direct in-line quench of the reaction mixture to yield 6-hydroxybuspirone (Scheme 1).
This approach furthermore made use of in-line analysis techniques like FTIR (for the monitoring of the enolisation step) and was successfully run at steady state for 40 h generating the target compound at multi-kilogram scale. As this paper states, the main advantages of a continuous approach over batch processing in this scale-up campaign were found to be related to safety, isolated purity and economics.
The successful outcome of the above study can in part be ascribed to the use of a static mixing device which allowed for the selective and clean mono-deprotonation under scale-up conditions. This was in stark contrast to the related batch scenarios which were difficult to control. Owing to the importance of efficient micro-mixing attainable in continuous processing another interesting reactor design coined as a ‘continuous oscillatory baffled reactor’ (COBR) was introduced. In this set-up the reactor stream being processed is directed into a tubular reactor which contains periodically spaced annular baffles thereby creating a series of eddies through oscillatory motion simultaneously applied to the reactor (Figure 2) . The resulting vigorous axial and radial mixing results in very sharp residence time distributions and excellent heat and mass transfer. Consequently, long batch processes (including crystallisations, fermentations, polymerisations or waste water treatments) can be translated into a continuous process. In an early example such COBRs were applied to the flow synthesis of aspirin showcasing the effectiveness of this reactor type during a week long campaign delivering the target compound at scale with very high product purity (99.94%) and minimal loss of product during cleaning (<0.005%) .
In 2009 a flow synthesis of a high volume pharmaceutical was reported by the McQuade group describing a three step approach towards ibuprofen (16) using microreactor technology . A fully continuous process was aspired to, in which only final purification was to be performed off-line at the end of the sequence. Each of the individual steps were first optimised in flow being mindful of the reagents used in order to avoid downstream incompatibilities. The initial step was a Friedel–Crafts acylation of isobutylbenzene (10) with propionic acid (11) in the presence of excess triflic acid (12). The transformation was found to work very effectively and the acid catalyst was also tolerated in the subsequent 1,2-aryl migration step. This was mediated by a hypervalent iodine reagent, PhI(OAc)2 (13), conducted in trimethyl orthoformate (14, TMOF) and methanol (Scheme 2). The direct saponification of the resulting rearranged methyl ester with an excess of base thus completed the telescoped flow synthesis of ibuprofen (16). Work-up, via acidification and repeated washes with ether, water and brine, followed by filtration, evaporation, treatment with active carbon and finally recrystallisation was performed manually to eventually yield pure ibuprofen product (99%). Overall this pioneering work allowed for the synthesis of ibuprofen in only ten minutes residence time albeit in a yield of only 51% equating to a productivity of 9 mg/min.
Although this work nicely demonstrates the feasibility of constructing a continuous process it is mainly an academic proof of principle based upon an important well known molecule. We state this not to detract from the work but to comment here about the different approaches and considerations that generally focus the minds of academics and industrialists and use this example as illustration.
This route would certainly not constitute an economically viable approach compared to the existing manufacturing routes which have been highly refined and optimised [49-51]. Although modern reagents such as hypervalent iodine and triflic acid represent very valuable additions to the chemists’ repertoire they are also inherently expensive and difficult to source at scale. In addition the waste streams generated through their use would also be difficult and costly to dispose. This aptly leads to an interesting relationship that is often encountered in innovative work employing new technologies where a general mind set exists to also test the limits of modern reagent equivalents in addition to the equipment. From an academic perspective this is a positive and beneficial contribution to the progression of the subject, however, this can significantly restrict the translational value of the methodology with respect to adoption or convenient uptake by industry. Commonly industry cites cost, unacceptable solvent combinations and limited availability of new reagents (metal ligand combinations) at scale as the main hindrances to uptake. This message is certainly being acknowledged with many of the more recent publications originating from academia using industry evaluation metrics and reagent selection guides to influence their route selection.
However, it is not only academia which is in the firing line, industry scientists are often heavily criticised as being too reliant on existing reactions/reagents and therefore being too conservative and resistant to change. Although this is often a corporate promoted strategy resulting from being risk adverse it can bias mind sets to fall back on the proven rather than innovate and explore. The additional pressures of meeting regulatory compliance, which is often easier based upon precedent, and the constant ‘time = money’ equation also compound the effect. Again such perceptions are changing with many companies creating specialist innovation groups dedicated to exploration and exploitation of new technologies. Fledgling innovations are in-house tested, monitored and if viable rolled out more expansively throughout the company. An excellent illustration would be the adoption of microwave reactors which have become primary heating methods in many medicinal chemistry labs. This is also being seen in the adoption of flow processing technologies where all the major pharmaceutical companies have internal teams working on business critical projects as well as longer term objectives. Furthermore the generation of various consortia between academia and industry is also influencing the transfer of knowledge, reasoning and importantly expectations. All these considerations are helping to drive the area of flow chemistry.
Flow processing scenarios
Recently, the Jamison laboratory reported on an improved flow synthesis of the ibuprofen sodium salt (17) that delivers the target compound in only three minutes residence time with an improved productivity of about 135 mg/min . As the key steps are the same as in McQuade’s approach (Friedel–Crafts acylation, 1,2-aryl migration and saponification) this report focuses on improved output by intensifying the overall sequence (Scheme 3). As such an in-line extraction is performed after the Friedel–Crafts acylation step, followed by dissolving intermediate 18 in trimethyl orthoformate and DMF. This stream is then combined with a stream of ICl (21) to affect the 1,2-aryl migration in a heated flow reactor (1 min, 90 °C) followed by treatment of the stream with NaOH, 2-mercaptoethanol, MeOH and water in order to hydrolyse the intermediate methyl ester and quench residual ICl. After collection of the crude reaction mixture an extractive work-up was performed off-line, in which ibuprofen was generated upon acidification from its sodium salt 17.
Another high profile pharmaceutical for which a flow synthesis has been developed is imatinib (23), the API of Novartis’ tyrosine kinase inhibitor Gleevec [53-55]. Reported by the Innovative Technology Centre (ITC) in 2010, this landmark synthesis was realised as a continuous process featuring an amide formation, a nucleophilic substitution and a Buchwald–Hartwig coupling as key synthesis steps performed in flow (Scheme 4).
Further highlights of this approach were the use of scavenger resins for intermediate purification and solvent switching operations as well as the use of in-line UV-monitoring needed to orchestrate the various reagent streams. Although the low solubility of various intermediates proved challenging, the designed route was able to successfully deliver sufficient quantities of imatinib (23) and several of its analogues (~30–50 mg each) in high purity within one working day allowing subsequent testing of new derivatives. Although this approach was conducted as a fully integrated telescoped continuous flow sequence its capacity to run as an uninterrupted process is certainly limited by the solid-phase scavengers employed as purification aids. The stoichiometric scavenging capacity of many of these species coupled with their limited loadings does restrict the quantities of material which can be generated in a run. As a consequence this approach is better suited to the rapid formation of small quantities of directly purified material for screening purposes but does not constitute a viable mode of performing direct large scale manufacture.
In the same year the ITC also reported on their efforts towards the flow syntheses of two lead compounds reported earlier by AstraZeneca. The first one details the flow synthesis of a potent 5HT1B antagonist (28) that was assembled through a five step continuous synthesis including a SNAr reaction, heterogeneous hydrogenation, Michael addition–cyclisation and final amide formation (Scheme 5) . This sequence again makes use of in-line scavenging resins for purification purposes and demonstrates the utility of in-line solvent switching protocols and high temperature reactor coils operating at 130–245 °C, well above the boiling points of the solvents employed.
In the second study, the flow synthesis of the selective δ-opioid receptor agonist 33 was discussed (Scheme 6) . Again, a strategy of integrating each of the three synthetic steps with a sequenced cascade of scavenger agents to perform the aspects of work-up and purification was used.
The report also showcased the generation and use of organometallic species (i.e., Grignard reagents) in flow synthesis as well as in-line React-IR monitoring in order to precisely control the onset of late stage flow streams that are affected by dispersion effects thus marking the first use of this now commonly incorporated analysis technique.
As the safe use of organometallic reagents has emerged as a key facet of flow chemical synthesis , the ITC reported on the design and implementation of a dual injection loop system that could deliver solutions of organometallic reagents (i.e., LiHMDS or n-BuLi) as a pseudo-continuous process . This protocol enables loading of a second loop with the unstable organometallic reagent whilst the first loop (previously filled with the same solution) is being directed to the intended flow transformation. Once this first reagent loop is empty, an automated protocol switches the valves so that the second loop transfers the reagent, whilst the first one is being recharged.
This concept was successfully applied to the flow synthesis of a 20-member library of casein kinase I inhibitors (38) that also demonstrate the selective mono-bromination, heterocycle formations and high temperature SNAr reactions as key flow steps in the sequence (Scheme 7).
One of the early published examples of industry-based research on multi-step flow synthesis of a pharmaceutical was reported in 2011 by scientists from Eli Lilly/UK and detailed the synthesis of fluoxetine 46, the API of Prozac . In this account each step was performed and optimised individually in flow, with analysis and purification being accomplished off-line. The synthesis commences with the reduction of the advanced intermediate ketone 47 using a solution of pre-chilled borane–THF complex (48) to yield alcohol 49 (Scheme 8). Conversion of the pendant chloride into iodide 51 was attempted via Finckelstein conditions, however, even when utilising phase-transfer conditions in order to maintain a homogeneous flow regime the outcome was not satisfactory giving only low conversions. Alternatively direct amination of chloride 49 utilising high temperature flow conditions (140 °C) allowed the direct preparation of amine 50 in excellent yield. Flow processing using a short residence time (10 min) at the elevated temperature allowed for a good throughput; in addition, the handling of the volatile methylamine within the confines of the flow reactor simplifies the practical aspects of the transformation, however, extra precautions were required in order to address and remove any leftover methylamine that would pose a significant hazard during scaling up.
The final arylation of 50 was intended to be performed as a SNAr reaction, however, insufficient deprotonation of the alcohol 50 under flow conditions (NaHMDS or BEMP instead of using a suspension of NaH as used in batch) required a modification to the planned approach. To this end a Mitsunobu protocol based on the orchestrated mixing of four reagent streams (50, 54 and reagents 52 and 53) was developed and successfully applied to deliver fluoxetine (46) in high yield. Overall, this study is a good example detailing the intricacies faced when translating an initial batch synthesis into a sequence of flow steps for which several adaptations regarding choice of reagents and reaction conditions are mandatory in order to succeed.
The flow synthesis of the high profile antimalaria agent artemisinin (55) was reported by the Seeberger group in 2012 [61,62]. This intriguing approach represents one of the few examples where photochemistry has been employed in the synthesis of a pharmaceutical. For this endeavour dihydroartimisinic acid (56), an advanced building block that is available via chemoselective batch reduction of bioengineered artemisinic acid (57), was chosen as the starting point. The key transformations to yield artemisinin thus demanded a reaction cascade including a singlet oxygen mediated ene-reaction, a Hock cleavage of the resulting hydroperoxide 58 followed by oxidation with triplet oxygen and a final peracetalisation (Scheme 9).
Based on previous work by the Seeberger group and others [63-65] a simple flow photoreactor set-up comprising of a layer of FEP-polymer tubing wrapped around a cooled medium pressure mercury lamp was used to efficiently generate and react the singlet oxygen in the presence of tetraphenylporphyrin (TPP) as a photosensitizer. Upon exiting the photoreactor, the reaction stream was acidified by combining with a stream of TFA in order to enable the remaining reaction cascade to take place in a subsequent thermal reactor unit. After off-line purification by silica gel chromatography and crystallisation artemisinin was isolated in 39% yield equating to an extrapolated productivity of approximately 200 g per day.
More recently, Seeberger and McQuade reported on further improvements of this strategy enabled by the development of a NaBH4-based flow reduction procedure of artemisinin (55) to yield dihydroartemisinin (61) as well as in-line purifications and derivatisations to also generate several related malaria medications (i.e., β-artemether (62), β-artemotil (63) and α-artesunate (64)) in an efficiently telescoped manner (Scheme 10) [66,67].
As the authors mention, their work is related to an earlier study by researchers from the Universities of Warwick and Bath describing a continuous reduction protocol of artemisinin using LiBHEt3 in 2-Me-THF as a greener solvent . Although this reductant is more expensive than NaBH4 this approach convinces through its simplicity and superior productivity (~1.6 kgh−1L−1).
Beside the use of photochemical processing towards the synthesis of artemisinin and its derivatives, this strategy has also been employed in the flow synthesis of a carprofen analogue  as well as in the regioselective bromination towards a rosuvastatin precursor  showcasing how continuous flow photochemistry is receiving a significant level of interest. This is not least because of the perceived green reagent concept of photons and the ability to overcome the inherent dilution problems encountered in batch. The ability to control residence times and hence decrease secondary transformations whilst using the small dimensions of the microreactor flow streams to enhance the photon flux has been claimed to increase productivity. However, it should be noted that many of the articles promoting the use of flow photochemistry do not currently adequately quantify or describe the systems in sufficient detail in order to fully justify such statements . This is a general consideration but especially pertinent to the use of low power LED’s which are becoming increasingly popular. The calibration and quantification of the incident light from such devices is not normally evaluated or even commented upon in many of these studies hence reproducibility is therefore a major issue. Considering one of the main drivers of flow chemistry is an increase in reproducibility this seems a rather negative trend.
In 2012 researchers from AstraZeneca (Sweden) reported upon a scale-up campaign for their gastroesophageal reflux inhibitor programme. Specifically, flow chemical synthesis was used to efficiently and reliably provide sufficient quantities of the target compound AZD6906 (65), which had been prepared previously in batch. From these earlier batch studies concerns had been raised regarding exothermic reaction profiles as well as product instability which needed to be addressed when moving to larger scale synthesis. Flow was identified as a potential way of circumventing these specific problems and so was extensively investigated. The developed flow route  started with the reaction of methyl dichlorophosphine (66) and triethyl orthoacetate (67), which in batch could only be performed under careful addition of the reagent and external cooling using dry ice/acetone. Pleasingly, a simple flow setup in which the two streams of neat reagents were mixed in a PTFE T-piece maintained at 25 °C was found effective in order to prepare the desired adduct 68 in high yield and quality showcasing the benefits of superior heat dissipation whilst also safely handling the toxic and pyrophoric methyl dichlorophosphine reagent (Scheme 11).
As the subsequent Claisen condensation step was also known to generate a considerable exotherm, a similar flow setup was used in order to allow the reaction heat to dissipate. The superiority of the heat transfer process even allowed this step to be performed on kilogram quantities of both starting materials (68, 69) at a reactor temperature of 35 °C giving the desired product 72 within a residence time of only 90 seconds. Vital to the successful outcome was the efficient in situ generation of LDA from n-BuLi and diisopropylamine as well as the rapid quenching of the reaction mixture prior to collection of the crude product. Furthermore, flow processing allowed for the reaction of both substrates in a 1:1 ratio (rather than 2:1 as was required in batch) as the immediate quenching step prevented side reactions taking place under the strongly basic conditions. Having succeeded in safely preparing compound 72 on kilogram scale, the target compound 65 was then generated by global deprotection and subsequent recrystallisation where batch was reverted to as the conditions had been previously devised and worked well.
As seen above, avoiding detrimental exotherms in scale up campaigns is a common reason for developing a continuous flow process. This approach is also demonstrated in the synthesis of the pyrrolotriazinone 73 via a exothermic oxidative rearrangement from 75, a key intermediate towards brivanib alaninate (74) that was reported by researchers at BMS in 2014 (Scheme 12) .
Another application that undoubtedly benefits from performing scale up processes continuously concerns the generation and use of the Vilsmeier reagent (76). An early study by scientists at Roche (UK) demonstrated an approach in which Auto-MATE equipment combined with reaction simulation software was used to predict heat flow data for making and using Vilsmeier reagent at scale . Using this information the formylation of 3,5-dimethoxyphenol was then performed at multi-kilo scale showing good agreement of the results with the devised simulations. More recently, scientists at Novartis (Switzerland) extended this study by developing a semi-continuous flow approach for the synthesis of the oral antidiabetic vildagliptine (77) using in situ generated Vilsmeier reagent (Scheme 13) .
Neat streams of DMF and POCl3 were mixed in a simple Teflon T-piece before entering a tubular reactor maintained at 22 °C (4.5 mL, tres = 30 s). Upon exiting this reactor the crude stream of the Vilsmeier reagent 76 was combined with a stream of amide 80 in DMF that was prepared in situ in a batch reactor from proline amide and chloroacetyl chloride. The crude nitrile product 81 was then collected in a batch vessel and isolated in pure form after crystallisation and washing with n-heptane. Alkylation of 81 with the corresponding amino-adamantane derivate in the presence of excess K2CO3 following an existing batch protocol completed the synthesis of vildagliptine (77). Again, it was highlighted that the control of the exothermic Vilsmeier reagent formation and subsequent handling of this toxic and unstable intermediate was ideally suited to a continuous production and consumption in flow protocol.
Gaseous reagents in flow
Another example in which flow chemical synthesis was used as the key step in an industrial setting was reported by scientists from Eli Lilly (USA) in 2012. An asymmetric high-pressure hydrogenation towards LY500307 (82)  was demonstrated (Scheme 14). As this campaign aimed to produce the key intermediate 83 at pilot-scale, a flow-based asymmetric hydrogenation was chosen as an economically more viable option compared to establishing a high-pressure batch process.
As depicted in Scheme 14, solutions of the substrate 84 and a zinc triflate additive were combined with the rhodium precatalyst (85, 0.025 mol %, and Josiphos ligand 86) before being mixed with hydrogen gas and entering a plug flow tubular reactor (volume 1.46 or 73 L, hydrogen pressure 70 bar, 70 °C, residence time 12 h). Several campaigns were run over periods of several days (e.g., campaign 1: 282 hours total cumulative reagent feed time) in order to evaluate this hydrogenation process. The process proved robust allowing reproducible and safe generation of the desired product in both high yield and enantiomeric excess. Additionally, semi-continuous liquid–liquid extraction, in-line distillation and product crystallisation were coupled to this hydrogenation step allowing for a total of 144 kg of the product 83 to be produced, purified and isolated using equipment that fits into existing laboratory fume hoods and hydrogenation bunkers. As the authors point out, this flow process not only delivered the hydrogenation product 83 with an improved safety profile at pilot-scale in a cost-effective manner, but moreover gave the same weekly throughput as a 400 L plant module operating in batch mode.
As the preceding examples clearly illustrate flow chemistry has quickly proven a viable means to assemble complex target molecules in a continuous and more modern fashion thus starting to satisfy claims regarding its advantageous nature compared to batch synthesis. Whilst some of these early examples can be seen as proof of concept studies, others have already demonstrated the application of further strategic elements including in-line purification and in-line analysis, both being crucial in order the achieve multistep flow synthesis. As the reader will see in the following part of this review, further advancements are geared towards more readily scaled processes and will also include the development of new devices allowing safe and efficient use of gaseous reagents as well as more effective ways of quickly transitioning between very low and very high temperatures that are key for streamlining modern flow synthesis routes.
Although the widely used H-Cube system had provided a popular solution for safe and convenient hydrogenation reactions at lab scale [76-79], the safe utilisation of other gaseous reagents at above ambient pressure was a relatively neglected area in flow chemistry for a long time. Only a few examples of flow hydrogenations and carbonylations had been reported [80-83]. The redevelopment and commercialisation of a laboratory based tube-in-tube reactor by the Ley group in 2009 changed the playing field and popularised the wider use of gases and volatile components. The design of the tube-in-tube system is based on a semipermeable Teflon AF2400 tubing (1 mm o.d., 0.8 mm i.d.) being housed within a wider PTFE tube (3.2 mm o.d., 1.6 mm i.d.; Figure 3). Depending on the intended application the gas can be fed either into the inner or the outer tube and upon pressurisation penetrates into the reagent stream where the desired reaction occurs. It has also been shown that an applied vacuum can enable the extraction of gaseous substances from a flow stream.
This concept has since been studied in a variety of applications using for instance O3, CO, H2, CO2, O2, NH3 or syngas and has been reviewed very recently . One noteworthy application of the tube-in-tube system by the Ley group in 2013 details the flow synthesis of the anti-inflammatory agent fanetizole (87) , in which ammonia gas was fed into the inner tube, whilst the outer tube contains a solution of phenethylisothiocyanate (89) in DME. The tube-in-tube system was placed onto the cooling unit of a Polar Bear Plus system maintained at 0 °C in order to generate the urea adduct 90 in quantitative yield (Scheme 15). In order to prepare the target compound this flow stream was then combined with an additional stream of bromoacetophenone (91) and passed through a heated tubular reactor unit (100 °C, 15 min) furnishing the 2-aminothiazole core of fanetizole (87). Due to preceding studies on the use of ammonia gas in this tube-in-tube system including in-line titrations only a minimal excess of gas (1.06 equivalents) was necessary to obtain complete conversion in the initial reaction subsequently allowing safe scale-up with a productivity of 70 g fanetizole (87) in 7 h.
In 2013 the Jamison group reported the flow synthesis of the important H1-antagonist diphenhydramine·HCl (92) showcasing the potential of modern flow chemistry to adhere to green chemistry principles (minimal use of organic solvents, atom economy etc.) . The synthetic strategy relied on reacting chlorodiphenylmethane (93) with an excess of dimethylaminoethanol (94) via a nucleophilic substitution reaction (Scheme 16).
As both starting materials are liquid at ambient temperature the use of a solvent could be avoided allowing direct generation of the hydrochloride salt of 92 in a high temperature reactor (175 °C) with a residence time of 16 min. Conveniently at the same reaction temperature the product was produced as a molten paste (m.p. 168 °C) which enabled the continued processing of the crude product circumventing any clogging of the reactor by premature crystallisation. Analysis of the crude extrude product revealed the presence of minor impurities (<10%) even when stoichiometric amounts of 94 were used, consequently an in-line extraction process was developed. Additional streams of aqueous sodium hydroxide (3 M, preheated) and hexane were combined with the crude reaction product followed by passage through a membrane separator. The hexane layer was subsequently collected and treated with hydrochloric acid (5 M in IPA) leading to the precipitation of diphenhydramine hydrochloride (92) in high yield (~90%) and purity (~95%). Furthermore, options to further reduce waste generated during the purification sequence are presented by combining hot IPA with the crude flow stream leading to the isolation of the target compound (92·HCl) by direct crystallisation in the collection vessel (yield 71–84%, purity ~93%, productivity 2.42 g/h).
More recently, the Jamison group also reported upon a short flow synthesis of the antiepileptic agent rufinamide (95) . The 1,2,3-triazole ring was prepared via a dipolar cycloaddition between an in situ generated benzylic azide and propiolamide (also prepared in situ), which by maintaining a low inventory of the reactive intermediate reduced the safety concerns associated with the use of the azide. The choice of flow when handling hazardous materials like azides is a very frequently encountered driver and several publications detailing the associated benefits have emerged over the years [88-90]. Importantly in this study, a flow reactor consisting of copper tubing maintained at 110 °C (6.2 minutes residence time) was employed as this would release small amounts of copper salts catalysing the regioselective triazole formation. A cautionary note regarding the potential of generating copper azide within the reactor should be made here from a scale-up perspective as this was not explored in the paper. Overall, this small scale syringe pump based set-up enabled the preparation of rufinamide (95) within ~11 minutes processing time and with a productivity of ~0.22 g/h, although this does not take into account the time required for work-up and purification necessary to isolate the pure rufinamide (Scheme 17).
Although the above approach generates rufinamide (95) in a continuous fashion, a more convincing strategy towards rufinamide has been reported by the Hessel laboratory in 2013 . Their route focused upon a dipolar cycloaddition between azide 100 and (E)-methyl 3-methoxyacrylate (101) to yield triazole 102 that was converted into rufinamide (95) (Scheme 18).
The benefits of using this alternative dipolarophile 100 are that it is not only considerably cheaper and less toxic than 98, but it also delivers the desired 1,4-triazole regioisomer without the need for a metal catalyst that requires stringent purification afterwards. Due to the reduced reactivity of 100 the crucial cycloaddition step was conducted neat at elevated temperature (210 °C) yielding pure 102 within short residence times (5–30 minutes studied) on 20–200 mmol scale after crystallisation (70–83% yield).
As in the case of rufinamide (95), the choice of the flow reactor also plays a key role in the synthesis of meclinertant (SR48692, 103), which is a potent probe for investigating neurotensin receptor-1