Scale-up and optimization of the synthesis of dual CBP/BRD4 inhibitor ISOX-DUAL†

ISOX-DUAL is a dual inhibitor of CBP/p300 (IC50 = 0.65 μM) and BRD4 (IC50 = 1.5 μM) bromodomains, and a useful chemical probe for epigenetic research. Aspects of the published synthetic route to this compound and its analogues are small-scale, poor-yielding or simply unamenable to scale-up without optimization. Herein we describe the development of a refined synthesis that circumvents the challenges of the original report, with notable improvements to several of the key synthetic transformations. Moreover, a general Suzuki Miyaura protocol for the late stage installation of alternative dimethyl-isoxazole acetyl-lysine (KAc) binding motifs is presented.


ISOX-DUAL ([3-[4-[2-[5-(dimethyl-1,2-oxazol-4-yl)-1-[2-(morpholin-4-yl)ethyl]-1H-1,3-benzodiazol-2-yl]ethyl]phenoxy]
propyl]dimethylamine) 1 is a dual inhibitor of BRD4 and CBP/p300 bromodomains. 1 It is structurally related to a series of CBP/p300 and BRD4 bromodomain inhibitors through a benzimidazole central scaffold with an adjoining 3,5-dimethylisoxazole group, which acts as an acetylated lysine (KAc) mimic. 2,3 In previous work, differential activity and overall potency were tuned through structure-based design, yielding compounds with selectivity for CBP/p300 over BRD4, together with a probe (ISOX DUAL) possessing balanced potency against both targets ( Fig. 1). ISOX-DUAL is a useful tool compound for exploring transcriptional regulation. 1,2,4 Our interest in functionalised bromodomain inhibitors necessitated a feasible, scalable route to 1 and its analogues. However, the reported synthetic strategy (Scheme 1) is, in places, both small-scale in nature and inviting of optimisation. 1 The key transformation is an elegant, albeit low-yielding, one-pot nitro-reduction and in situ cyclisation sequence, in which exposure of nitroaniline 4 to sodium dithionite in the presence of aldehyde 5 facilitated the installation of the benzimidazole framework in one synthetic operation. Of additional note, the late-stage alkylation of phenol 7 to furnish 1 was also of modest yield: these two crucial stages thereby being significant contributors to an overall isolated yield of just 1%.
Our modified retrosynthetic analysis identified that independent reduction of nitroaniline 4 to bis-aniline 8, followed by an amide coupling with subsequent acid-catalysed benzimidazole formation, might furnish the target 1 in higher yield (Scheme 2). Employing carboxylic acid 9 instead of aldehyde 5 offers the prospect of shortening the synthetic route by two chemical stages in comparison to the original, and further provides the distinct advantage of enabling an alternative earlystage phenolic alkylation, the comparative transformation having been a significant limitation to the output of the published synthesis.
The initial stage in both synthetic approaches involves an S N Ar reaction between commercially available 4-bromo-1fluoro-nitrobenzene 2 and 4-(2-aminoethyl)morpholine. In our hands, this transformation, to prepare bromide 3, was readily replicated on large scale in near-quantitative yield, and further, was demonstrated to proceed rapidly under microwave (MW) conditions with commensurate output.
Next, in sequence, we turned to elaboration of 3 to install the requisite pendant isoxazole functionality. This necessitated examination of the Suzuki-Miyaura cross-coupling reaction, where we hypothesised that competing hydrodehalogenation and protodeboronation could be potential contributory factors to a drop-off in expected product yield in the original report. [5][6][7][8] A screen for substrate scope was performed under microwave conditions using a range of palladium catalysts and bases, exploring their effectiveness as boronic acid, boronate ester and MIDA boronate coupling partners, with selected results displayed in Table 1. From this screen, we found that employ-ment of a boronate ester bearing the isoxazole moiety performed the best, with potassium phosphate and PdCl 2 (dppf )·DCM acting as base and catalyst, respectively.   This reaction could be conducted on small scale in a microwave and was successfully demonstrated thermally on a larger scale to give almost 40 g of product, in near quantitative yield.
To demonstrate the synthetic utility and reliability of this now optimised procedure, a small series of analogues (10)(11)(12)(13)(14) were synthesised employing a variety of heterocyclic boronate esters ( Table 2). This demonstrates the potential for late-stage KAc bio-isostere modification to these and similar scaffolds, which is of interest to ongoing work in our laboratory. Yields tend to be good to excellent and all analogues were further characterised in the solid state by X-ray studies.
With key precursor nitro-aniline 4 in-hand, we turned our attention to the reduction of the nitro group, whereupon standard exposure to palladium on carbon under a balloon of hydrogen, the desired product 8 was readily obtained in reasonably high isolated yield. However, analysis of the profile of the reaction indicated the formation of a minor by-product, presumed to be generated by cleavage of the labile isoxazole N-O bond. 10,11 This was established by repetition of the hydro-Scheme 2 Retrosynthetic approach to ISOX-DUAL 1. Organic & Biomolecular Chemistry Paper genation reaction under pressure, which, solely delivered the ring-opened product. Indicative confirmation was obtained by the changes observed in the 1 H NMR spectrum where the signals corresponding to the two methyl groups moved upfield from δ = 2.26 and 2.40 ppm to δ = 1.72 and 1.87 ppm respectively. The crude product was then carried forward through an amide coupling and cyclisation sequence to afford benzimidazole 15, whose structure was solved by single crystal X-ray analysis (Scheme 3). We opted to replicate and scale-up the reported reduction which can be induced by employment of sodium dithionite as the reducing agent 2 affording aniline 8 in high yield (81%) and purity (Scheme 4).
Independent construction of the requisite amide coupling partner 9 successfully allowed circumvention of the low yielding alkylation step from the original synthesis. As such, after conventional phenolic alkylation to give methyl ester 16 in high yield (87%), a subsequent ester hydrolysis gave rise to carboxylate 9 in quantitative yield, which proved most convenient to isolate and manipulate as the lithium salt (Scheme 4). To complete the synthesis, a HATU-mediated amide coupling between 8 and 9, followed by acetic acid treatment of the crude product induced cyclisation to 1 in a one-pot procedure with good yield (55%) and high purity.

Conclusions
In summary, we detail an optimised synthetic route to ISOX-DUAL (1) and related analogues, with improved yields at every stage. By repositioning the phenolic alkylation to earlier in the plan, and by implementing a redesigned stepwise construction approach to the benzimidazole core, we now report an improved synthesis, shortened from eight stages to six, with a significant overall isolated yield increase from 1% to 42%. Additionally, development of the Suzuki coupling invoking a boronate ester establishes prospective in-roads to structural modification and opportunity to employ late stage Pd coupling process in the installation of alternative KAc bioisosteres in bioactive scaffolds.

General methods
All reagents and solvents were purchased from commercial sources and used without further purification. Nuclear magnetic resonance spectra were recorded on a Bruker Avance III HD spectrometer operating at 400 MHz or a Varian VNMRS 500 or VNMRS 600 spectrometer operating at 500 MHz or 600 MHz for 1 H NMR and 126 MHz or 151 MHz for 13 C NMR, respectively. 19 F NMR spectra were recorded on a Varian VRMS 400 spectrometer operating at 376 MHz. 1 H NMR and 13 C NMR chemical shifts (δ) are reported in parts per million ( ppm) and are referenced to residual protium in solvent and to the carbon resonances of the residual solvent peak, respectively. DEPT and correlation spectra were run in conjunction to aid assignment. 19 F NMR chemical shifts are reported in ppm and are uncorrected. Coupling constants ( J) are quoted in Hertz (Hz), and the following abbreviations were used to report multiplicity: s = singlet, d = doublet, dd = doublet of doublets, t = triplet, q = quartet, m = multiplet. Purification by flash column chromatography was carried out using Fisher Scientific silica gel 60 Å (35-70 μm), or with silica gel or C18 columns using a Combi flash RF 75 PSI or 150 PSI, Teledyne ISCO unit. Analytical thin layer chromatography was performed on glass plates pre-coated with silica gel (250 μm/UV254), with visualization being achieved using UV light (254 nm) and/or by staining with alkaline potassium permanganate dip. Reaction monitoring LC-MS analyses were conducted using a Shimadzu LC-MS 2020, on a Gemini 5 m C18 110 Å column. High resolution mass spectral (HRMS) data was collected in the laboratories of the University of Sussex Chemistry Department using a Bruker Daltonics Apex III (Apollo ESI ion source). Single crystal X-ray measurements were recorded in the laboratories of the UK National Crystallography Service at the University of Southampton.