Reaction design and optimization
We initiated our study using 2-naphthol (2) as a nucleophile, and p-aminobenzoic acid (1) as the amine source (Figure 2A). We first questioned whether there is a possibility of halting the activity of AzoC at the nitroso-producing step, which might be cascaded with electron-rich phenol compounds for the offspring reactions. Therefore, we expressed and purified AzoC, and performed the nitroso generating reaction and control reactions. With glucose dehydrogenase (GDH, supplementary methods) as the NADH regenerating system (to replace the 1 mM NADH usage of the source literature) and phenazine methosulfate (PMS) as an electron mediator to regenerate the iron center of diiron oxygenase (Figure 4B) 44, nitroso intermediate 3 was formed as the primary product (Figure 2B, i and ii). Interestingly, when 2 was added to the reaction system, a new peak with a retention time of 15 min was observed (Figure 2B, iii). The incubation of pro-boiled AzoC with 1 and 2 showed no production of 3, 4 and 5 (Figure 2B, iv). We then performed a scale-up reaction and isolated 5. NMR and single-crystal X-ray diffraction data revealed that 5 is a phenol-diarylamine product (Figure 2C). According to the literature, 5 is an efficient inhibitor of HIV integrase 45 and amyloid aggregation46.
We then optimized the reaction condition. Temperature, additive, pH, and buffer solution were analyzed on the yield of 5 (Table 1, entries 1–15, the yields were first determined using HPLC, and the best yield was confirmed by column isolation). The yield of 5 increased from 36% (Table 1, entry 2) to 82% (Table 1, entry 7) when the reaction was performed at 25 °C in PBS buffer (pH 8) for 1 h. No 5 was observed at pH values above 11 or below 6; thus, pH is regarded as the crucial factor in this reaction. Under this reaction system, 4 was also isolated. According to a previous study, 4 was non-enzymatically dimerized from 3 in a radical manner.36 The non-enzymatic diarylamine and the azoxy synthesis are competing in the consumption of 3 (Figure S3). Hence, 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO), a radical scavenger and polymerization inhibitor, was used to block the azoxy synthetic shunt and to improve the yield of 5. Indeed, the yield of 5 increased to 97% when 10 mol% of TEMPO was added to the reaction system (Table 1, entry 13).
Substrate scope
With the optimal reaction condition, we then tested the aniline substrate scope. We first tested the p-carboxy substituted aniline substrates, and found all of them were converted to diarylamine products with good to excellent yield (Figure 3, 5 and 5-1a to 5a-1e). We then tested p- substituted trifluoromethyl (strong electron-withdrawing group) and methoxy (strong electron-donating group) substrates, and obtained the diarylamine products 5-1f and 5-1g in good yields (Figure 3). However, the m- and o- sites substituted aniline substrates, are not accepted by AzoC. We then performed protein engineering to expand the substrate scope. According to the crystal structure of non-heme diiron N-oxygenase AurF (PDB ID: 3-CHT) 47, which shows a broader substrate scope and 39% identity to AzoC, the Y93, V97, T100, N200 and R302 residues are involved in substrate binding (Figure S2). Hence, we built a mutant library on the corresponding Y94, T98, L101, N203, Q304 residues of AzoC, and tested the activity against p-, m- and o- sites substituted aniline substrates. A mutant: AzoC-T98V-L101T (AzoC-II), which recognizes both the p- and m- substituted substrates, was obtained (Figure 3, 5, 5-1a to 5-1i). The bearing of broad electron inductive substitution on the p- and m- sites by AzoC-II suggest wide compatibility of the aniline substrates.
We then studied the phenol substrate scope using AzoC-II as the catalyst. We first investigated the inductive influence of 2 on this reaction. The strong electron-withdrawing bromine group- and the strong electron-donating methoxy group-modified analogs of 2 were compatible with the reaction, and good to excellent yields were obtained (Figure 3, 5a-5d). The yield of the bromine-modified substrates was lower than that of the methoxy-modified substrates, implying that the electronic state of phenols has a potential influence on this reaction. We then questioned whether other phenols were tolerable to this enzymatic diamine coupling reaction. We tested 1-naphthol, and 6 (CCDC 2124158) was successfully isolated in excellent yield. We further tested other bicyclic phenols. Quinolinol substrates 8-hydroxyquinoline and 6-hydroxyisoquinoline were compatible under our reaction conditions, and respective products 7 (CCDC 2124159) and 8 were generated in good yields. For the case of indole substrates as 4-hydroxyindole and 5-hydroxyindole, the reaction solutions turned red. However, unidentified insoluble sediments were generated instead of the diamine products 10 and 11, which might be caused by the active secondary amine on indole’s pyrrole ring. We thus used the N-Boc-protected 4-hydroxyindole as the phenol substrate, most of which was unconverted, and 10 was isolated in 23% yield (Boc is a deactivating group of phenol and was found deprotected during the silicon isolation), which indicates a compatibility of the indole as the phenol aromatic source. Single-ring phenols such as hydroquinone, 4-(dimethylamino)phenol, and phenol were also tested. Hydroquinone diamine product 9 was successfully isolated in good yield. For the case of 4-(dimethylamino)phenol and phenol, instead of isolating the expected diamines 12 and 13, only nitroso product 3 and azoxy product 4 were generated when the reactions time were prolonged up to 12 h. Since the electronic state of phenols influences the reaction, we presume that the relatively low electron density on the benzene ring hindered the formation of 12 and 13.
Mechanism studies
We then investigated the reaction mechanism. Previous study suggests that NADH can mediate the transformation of aromatic nitroso compound into its hydroxylamine analog, both of which are active nitrogen sources. 36 Therefore, we first determined the actual nitrogen source for the diamine product. Commercially available nitrosobenzene and phenylhydroxylamine (1f and 1g in Figure 4A) were used to conduct non-enzymatic reactions (1f and 1g were incubated respectively with 50 μM 2, 1 mM NADH, and 5 μM TEMPO in 20 mM pH 8 PBS buffer for 1 h at 25 °C), and 5f was isolated in 98% yield (1f as nitrogen source) and 11% yield (1g as nitrogen source), indicating that the nitroso compound is the authentic nitrogen source. A radical reaction mechanism for the non-enzymatic nitroso-phenol coupling reaction could be ruled out because adding a radical scavenger improved the yield of 5 (Table 1). Considering that pH is a crucial factor in our reaction system, we then focused on the ionic reaction mechanism. The nitrogen atom of nitroso group is partly positively charged. Thus, phenols have two possible routes: either the electron-rich aromatic ring or the phenoxide negative ion serves as the negative charge center. We tested other electron-rich ring systems, such as naphthalene, 2-methoxynaphthalene, and anthracene; and no products other than nitroso and azoxy compounds were formed. Thus, the first possibility could be ruled out. While alkaline buffers facilitated the diarylamine formation (Table 1 and Figure S4), the phenoxide ion mechanism seems more reliable. The NADH produced by GDH system might serve as the reducing agent in N-O to N-H reduction. To confirm this, we synthesized 3 48 and conducted the non-enzymatic reaction in PBS buffer (pH 8) without using GDH or NADH. The yield of 5 was significantly reduced to 41%; fortunately, a new product 5’ was isolated in 11% yield (Figure 4A). Based on these studies and the previous enzymatic mechanism study, we proposed a mechanism in Figure 4B.14b In the enzymatic process, the GDH system provides the persistent NADH to propel the cycling between PMS and reduced-PMS; then, the reduced PMS serves as a chemical electron mediator44 to regenerate the active center of AzoC; and the cycled iron center of AzoC recruit oxygen and catalyzes the oxidation of aniline to nitroso intermediate in a cycling manner. In the non-enzymatic process, the partially positively charged nitroso nitrogen atom couples with the negatively charged carbon center of phenoxide ions to form C-N bonds, followed by the reduction of the N-O bond to N-H bond by NADH to afford the diarylamine product (Figure 4B). This mechanism study also reviews an efficient non-enzymatic synthetic route to diarylamine in aqueous solution with aromatic nitroso compound and phenol as substrates.