Combining Sanford Arylations on Benzodiazepines with the Nuisance Effect

Abstract 5‐Phenyl‐1,3‐dihydro‐2H‐1,4‐benzodiazepin‐2‐ones react under palladium‐ and visible light photoredox catalysis, in refluxing methanol, with aryldiazonium salts to afford the respective 5‐(2‐arylphenyl) analogues. With 2‐ or 4‐fluorobenzenediazonium derivatives, both fluoroaryl‐ and methoxyaryl‐ products were obtained, the latter resulting from a SNAr on the fluorobenzenediazonium salt (“nuisance effect”). A computational DFT analysis of the palladium‐catalysed and the palladium/ruthenium‐photocalysed mechanism for the functionalization of benzodiazepines indicated that, in the presence of the photocatalyst, the reaction proceeds via a low‐energy SET pathway avoiding the high‐energy oxidative addition step in the palladium‐only catalysed reaction pathway.


Introduction
There is a growing impetus for atom economical routes to high value end products employing late stage functionalization (LSF) processes. [1] These are particularly desirable in medicinal chemistry since they increase diversity and chemical space and enable rapid SAR (structure activity relationship) and ADME-Tox (Absorption, distribution, metabolism, eliminationtoxicity) feedback that is key to costly, high attrition, drug development. Late stage CÀH activation is a powerful tool in generating novel compounds for biological evaluation. [2] We recently described a palladium-catalyzed ortho-arylation of benzodiazepines employing iodonium salts in acetic acid under microwave irradiation. [3] The harsh conditions, relatively high commercial cost, and multistep synthesis of iodonium salts [4] (ArIAr' + ), coupled with a poor atom economy (ArÀI is a byproduct) prompted us to consider a visible-light photocatalyzed Pd-mediated protocol involving diazonium salts. [5] 2 Results and Discussion Our initial reaction trials were performed on the benzodiazepine 1 a, using the 2-fluoro-benzenediazonium salt 2 a under reflux (external oil bath temperature set at 70 8C). To our surprise, in addition to the expected product 3 a, we were able to isolate the ether product 4 a. However, reaction of the 3-isomer 2 b led exclusively to the fluorobiaryl derivative 3 b, whereas the 4-isomer 2 c afforded a mixture of fluorobiaryl 3 c and methoxy product 4 c (Scheme 1). Repeating the reaction in ethanol led to the ethyl ether 4 d whose xray structure is displayed (Scheme 1).
Characterization of 4 c was enabled by determination of its solid state x-ray structure [6] (Scheme 1) and by its unequivocal synthesis starting from 4-methoxybenzenediazonium tetrafluoroborate 2 d ( Table 1) where we found slightly better yields under reflux (Entry 1 vs. 2) compared to either ambient temperature or to the absence of photocatalyst (Entry 5). Moreover, a palladium catalyst was essential (Entry 4) for achieving a good yield. Microwave-mediated chemistry, in the absence of light and photocatalyst, gave little conversion of product.
To explain the formation of the ether products we propose a competing S N Ar, termed "nuisance effect," which has historically been observed for halogensubstituted benzenediazonium salts, given the strong electron withdrawing effects of the diazo group, notably operating on the 2-and 4-substituted isomers. [7] Indeed, simple alcoholysis of compound 2 c was achieved in the appropriate alcohol solvent at 70 8C (Scheme 2).
The CÀH activation reaction was also applied to aryldiazoniums incapable of undergoing such a Fsubstitution and, hence derivatives 4 e-4 i' were synthesized in good to excellent yields (Scheme 3) and the structure of 4h was determined by x-ray crystallography. Indeed, yields tend to be either similar or higher than those reported for the corresponding reactions involving iodonium salts, e. g. 4 e (60% vs. 56%), 4 f (54% vs. 35%), 4 g (71% vs. 55%) and 4 i (64% vs. 63%).
In the synthesis of 4 i, relatively large amounts of the diarylated adduct 4 i' were also observed. Such diarylations were previously reported by us. [3b] The current and previous library of benzodiazepines (Scheme 1) was tested for GABA binding. [8] Scheme 1. Benzodiazepine library synthesis. None of the current benzodiazepines displayed any appreciable biological activity although 7-chloro-benzodiazepines, as expected, had reasonable activity, although were ca. 7-10 fold less active than nordazepam and diazepam controls (Entries 1 and 2 respectively, Table 2) and were not pursued any further.
Sanford et al. proposed a possible mechanism to explain their Pd/Ru photocatalysed CÀH arylation. [5a] Here we present a computational study of a Pdcatalysed and a Sanford-derived Pd/Ru photocalysed mechanism for the functionalization of 1 a to 4 g (Scheme 4) to rationalise the increased yield in the presence of light and a Ru photocatalyst.
The detailed mechanism is shown in Scheme 5 and the reaction profile (relative to the reactants) in Figure 1. The reaction mechanism, with and without the Ru(II)-photocatalyst, essentially follows the same path except that the oxidative addition step in the presence of just the Pd(II)-catalyst (path shown in green, Scheme 5 and Figure 1), is replaced by a singleelectron-transfer (SET) process when the Ru(II)photocatalyst is added (shown in red, Scheme 5 and Figure 1).
The initial step of the catalysed mechanism involves the coordination to Pd(OAc) 2 by a N atom on the un-functionalised benzodiazepine to provide Int1, followed by the formation of an agostic complex Int2 prior to CÀH activation. The atomic distance between Pd and the agostic H in Int2 is 1.903 Å , which is in good agreement with similar agostic interactions in the literature: PdÀH = 1.91 Å [9] and RhÀH (1.95 Å ). [10] The barrier to CÀH bond activation is 41.4 kJ mol À1 , and involves H migration from C to O via a six-membered ring (TS2-3). Prior to coordination with the p-nitrobenzenediazonium (Ar-N 2 + ) the complex undergoes an isomerisation step (TS3-4), which involves a change in the C1ÀPdÀO3 angle from 132.0 to 172.0 degrees with an energy barrier of 27.8 kJ mol À1 to form Int4.
In the absence of the photocatalyst, Ar-N 2 + interacts with the Pd(II) complex and follows an oxidation addition (OA) pathway, (highlighted in green, Scheme 5 and Figure 1). The oxidative addition via TS5-6(OA) has an energy barrier of 127.1 kJ mol À1 and involves the formation of an ArÀPd(IV) complex. The N 2 is then eliminated leading to Int7. When the Ru(II)-photocatalyst is present, the nitrobenzene radical (Ar*) is generated from Ar-N 2 + (via oxidative quenching of Ar-N 2 + by the photoexcited [Ru(bpy) 3 ] 2 + complex to form [Ru(bpy) 3 ] 3 + ) [11] and follows a single-electron-transfer (SET) pathway, (in red, Scheme 5 and Figure 1). The square planar geometry of the Pd(II) complex Int4 becomes a Pd(III) distorted-octahedral structure when the Ar binds to the Pd centre in Int5(SET); this is consistent with the crystal structure of other Pd(III)-complexes although we did not consider bimetallic species. [12] Int7 is formed directly from Int5(SET) by the transfer of an electron to the [Ru(bpy) 3 ] 3 + complex to recover  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57 the photocatalyst. The Gibbs free energy barrier for single electron transfer (SET) resulting in the formation of the Pd(IV) complex Int7 was calculated to be 2.5 kJ mol À1 using Marcus and Savé ant theory. [13] The details of this calculation are provided in the Computational Method section. This barrier is very small but similar to literature values that range from 0.4-15.1 kJ mol À1 . [14] Both mechanisms (OA and SET) result in the same Pd(IV) structure for Int7. At this stage reductive elimination occurs via TS7-8 with a barrier of 43.2 kJ mol À1 . This step involves the formation of a CÀC bond to facilitate the functionalization of the benzodiazepine and the oxidation state of the Pdcenter changes from Pd(IV) to Pd(II) (Int7!Int8). The geometry Int8, involves an h 2 (C=C) interaction with Pd. A similar interaction was observed by Ariafard et al. [15] and Canty et al. [16] in their DFT calculations and in a palladium complex crystal structure. [17] It is clear from Figure 1 that, in the presence of the Pd-catalyst, the oxidative addition step is rate determining with a considerable energy barrier. However, in the presence of both the Pd(II)-catalyst and the Ru(II)-photocatalyst this OA step, and hence large energy barrier, is avoided as the reaction proceeds via a very low-energy single-electron-transfer process. This provides a rationale for the increased yield in the presence of a photocatalyst.

Conclusion
The CÀH activation of benzodiazepines with 2-or 4fluorobenzene diazonium salts under Pd catalysis with a Ru photocatalyst, in alcohol solvent, under reflux, leads to a mixture of both fluoroaryl and alkoxyaryl products. Reaction temperature is a key factor in determining the ratio of expected vs. "nuisance effect" (S N Ar) products. At ambient temperature trace amounts of the S N Ar product are detected whereas significant amounts can be obtained after prolonged heating under reflux. This process can also be extended to other aryl diazonium salts affording ortho-arylated benzodiazepines. These were tested for biological activity but were found to be significantly less active than e. g. nordazepam and diazepam controls. Density functional theory (DFT) has been used to provide a detailed mechanistic understanding of the functionalization of the benzodiazepines and to offer an explanation for the increased yield in the presence of a Ru(II)-photocatalyst. The Pd/Ru catalytic cycle follows the mechanism proposed by Sandford et al. [5a] The increased yield in the visiblelight photocatalysed Pd-mediated protocol is attributed to the transformation step leading to the formation of the Pd(IV) complex. In the presence of the photocatalyst the reaction proceeds via a lowenergy SET pathway and avoids the high-energy oxidative addition step in the Pd-only catalysed reaction pathway.
NMR spectra were recorded on a Varian V NMRS 500 ( 1 H: 500 MHz, 13 C: 126 MHz) spectrometer and prepared in deuterated solvents such as CDCl 3 and DMSO-d 6 . 1 H and 13 C chemical shifts were recorded in parts per million (ppm). Multiplicity of 1 H-NMR peaks are indicated by s -singlet, d -doublet, dd -doublets of doublets, t -triplet, pt -pseudo triplet, q -quartet, m -multiplet and coupling constants are given in Hertz (Hz).
Electrospray ionisation -high resolution mass spectra (ESI-HRMS) were obtained using a Bruker Daltonics Apex III where Apollo ESI was used as the ESI source. All analyses were conducted by Dr A. K. Abdul-Sada at Sussex. The molecular ion peaks [M] + were recorded as mass to charge (m/z) ratio.
LC-MS spectra were acquired using a Shimadzu LC-MS 2020, on a Gemini 5 mm C18 110 Å column and percentage purities were run over 30 minutes in water/acetonitrile with 0.1% formic acid (5 min at 5%, 5%-95% over 20 min, 5 min at 95%) with the UV detector at 254 nm. Purifications were performed by flash chromatography on silica gel columns or C18 columns using a Combi flash RF 75 PSI, ISCO unit. The following CCDC deposition numbers have been obtained, in parentheses; for 4 c (1518056), 4 d (1551609) and 4 h (1551610).

4-Methoxybenzenediazonium Tetrafluoroborate (2 d)
A stirred suspension of 4-fluorobenzenediazonium tetrafluoroborate (0.10 g, 0.48 mmol) in methanol (2 mL) was heated at 70 8C by using an external oil bath for 1 hour. The reaction was allowed to cool to ambient temperature and concentrated under reduced pressure. The residue was precipitated by the addition of diethyl ether and collected by filtration, affording 2 d as a white solid (0.090 g, 85%). The spectral data were concurrent with those reported. [18]

Computational Details
Density functional theory (DFT) calculations were performed at the wB97XD/6-311 + + G(2df,2p)[SDD]//PBE/6-31 + G(d,p)[SDD] level of theory, using the Gaussian09 program. [20] The Pople basis sets were used on all atoms except Pd and Ru for which the SDD relativistic effective core potentials were used. [21] The PBE functional [22] was used for the geometry optimisation and frequency analysis as it combines good accuracy for Pd complexes with computational speed. [23] The long-range corrected hybrid functional wB97XD, [24] which includes empirical dispersion corrections, was used for energies to ensure accurate energetics. [25] Methanol solvent energy corrections were applied using the conductor-like polarisable continuum model (CPCM). [26] Accordingly, the Gibbs free energies presented in Figure 1 were obtained by adding the thermal free energy corrections obtained at the PBE/6-31 + G(d,p)[SDD] level of theory to the solvent-corrected electronic energies obtained at the wB97XD/6-311 + + G(2df,2p)[SDD] level of theory. All stationary states were verified as minima or transition states by the absence or presence, respectively, of a single imaginary vibrational frequency. Eigenvector following was used to ensure transition states connected the desired minima.