A Hydroxyquinoline‐Based Unnatural Amino Acid for the Design of Novel Artificial Metalloenzymes

Abstract We have examined the potential of the noncanonical amino acid (8‐hydroxyquinolin‐3‐yl)alanine (HQAla) for the design of artificial metalloenzymes. HQAla, a versatile chelator of late transition metals, was introduced into the lactococcal multidrug‐resistance regulator (LmrR) by stop codon suppression methodology. LmrR_HQAla was shown to complex efficiently with three different metal ions, CuII, ZnII and RhIII to form unique artificial metalloenzymes. The catalytic potential of the CuII‐bound LmrR_HQAla enzyme was shown through its ability to catalyse asymmetric Friedel‐Craft alkylation and water addition, whereas the ZnII‐coupled enzyme was shown to mimic natural Zn hydrolase activity.

Metalloproteins represent more than one-third of natural enzymes, catalysing many of the important chemical reactions that sustain life. [1] This has inspired the creation of artificial metallo-enzymes, which are hybrids of proteins and synthetic transition metal complexes, ultimately aiming to develop biocatalysts for reactions that are new to nature. [2] Synthetic transition metal complexes have been incorporated into protein scaffolds using different methods to bind the metal-ligand complexes. [3] A recent approach to introduce metal catalysts involves noncanonical amino acids (ncAAs), either serving as the ligand which directly chelates the metal, or by binding an external metal-binding ligand. The ncAAs are introduced using expanded genetic code methods, especially stop codon suppression. [4] Several metal-binding ncAAs have been incorporated into proteins in vivo by expanded genetic code methodology to create artificial metalloenzymes. [5] For example, the amino acid analogue of the bipyridine ligand, (2,2'-bipyridin-5-yl)alanine (BpyA), has been incorporated into different biomolecular scaffolds to bind different bivalent metals and catalyse a variety of reactions when complexed with Cu II . [6] Likewise, the incorpo-ration of noncanonical N-methylhistidine and 3,4-dihydroxyphenylalanine into metallo-enzymatic scaffolds has been shown to both boost enzymatic turnover, and facilitate reutilisation of an alcohol dehydrogenase active site to bind Zn II , respectively. [7] The metal ligand 8-hydroxyquinoline (HQ) is one of the earliest analytical reagents and has been shown to bind > 20 transition metals. [8] More recently, the strong NÀ O bidentate binding mode of HQAla has been used for a variety of Cu II , Zn II and Rh mediated reactions, highlighting the versatility of HQ. [9] The ncAA that employs 8-hydroxyquinoline as its functional group, 2-amino-3-(8-hydroxyquinolin-3-yl)propanoic acid (HQAla) (Figure 1), was first incorporated as unnatural amino acid into the Z-domain protein.
Here it was used as a fluorescent probe and to bind heavy metals for crystallography. [10] The constitutional isomer of this ncAA, 2-amino-3-(8-hydroxyquinolin-5-yl)propanoic acid, has been described and used in protein electron transfer and Zn II sensing in vitro and in vivo. [11] So far, expansion of the genetic code with HQAla has not resulted in designer biocatalysts. Here, we introduce HQAla into lactococcal multidrug-resistance regulator (LmrR) [12] and complex it with a variety of transition metal ions such as Cu II , Zn II and Rh III . The catalytic propensity of the Cu II bound enzymes was demonstrated in a vinylogous Friedel-Crafts alkylation and an enantioselective hydration reaction. The Zn II bound artificial metalloenzyme was shown to hydrolyse amide bonds.
The artificial metalloenzymes presented in this study were prepared by using amber stop codon methodology, with a plasmid carrying the orthogonal tRNA-synthetase/tRNA pair specific for the incorporation of HQAla (pEVOL-HQAla). [10] HQAla was synthesised according to a previously reported method with a few minor improvements. [10] The positions M89 and V15 of LmrR were selected for the incorporation of HQAla, as these positions have been previously shown to allow efficient incorporation of ncAAs. [6e,f,13] Both residues are pointing towards the inside of the hydrophobic pore of the protein, with M89 located at the far edge and V15 more to the middle of the pore. The M89 position was shown to be the optimal position for studies with BpyA, [6e,f] whereas position V15 has been recently described as the optimal position for p-aminophenylalanine incorporation. [13] The pEVOL-HQAla plasmid and a pET17b plasmid containing either LmrR_V15TAG or LmrR_M89TAG constructs were cotransformed into Escherichia coli C43(DE3). The LmrR variant used in this study contains two mutations, K55D and K59Q, which are introduced to remove the DNA binding ability of LmrR, and a Cterminal Strep-Tag. [14] After addition of HQAla to the medium, the cells were induced to produce LmrR_V15HQAla or LmrR_ M89HQAla ( Figure S1 in the Supporting Information). The expression yields were 18-22 mg/L for LmrR_V15HQAla and 4-10 mg/L for LmrR_M89HQAla. The lower yield of LmrR_ M89HQAla is mainly attributed to faulty translation, that is, failure to suppress the stop codon UAG, resulting in truncated LmrR . The incorporation of HQAla was confirmed by electrospray ionisation mass spectrometry (ESI-MS; Figure S2). The quaternary structure of LmrR was studied by analytical sizeexclusion chromatography to determine the effect of unnatural amino acid incorporation. It was found that the structure was preserved and the proteins were eluted as single peaks at 11.4 (� 0.1) mL, which represents a molecular weight of approximately 30 kDa, consistent with dimeric LmrR ( Figure S3).
The ability of LmrR_V15HQAla and LmrR_M89HQAla to bind metal salts was studied by UV-Vis titrations, measuring the change in the UV-Vis absorption spectrum upon addition of the metal salts. The titrations were performed with Cu(NO 3 ) 2 , Zn (NO 3 ) 2 , Cp*RhCl 2 , RhCl(COD), Rh 2 (AcO) 4 and Rh 2 Cl 2 (CO) 4 . Titration of Cu(NO 3 ) 2 to LmrR_V15HQAla or LmrR_M89HQAla caused hypochromic shifts to the ligand-centred (LC) charge transfer band at 249 and to a lesser extent the LC band at 320, arising from π-π* and n-π* transitions respectively. Simultaneously, hyperchromic shifts of ligand-to-metal charge transfer (LMCT) transitions at 269 nm and to a lesser extent at~390 nm were observed (Figures 2ab and S4). [15] These shifts are in agreement with complex formation of metals to the quinoline moiety involving deprotonation of the phenolic group. [16] When approximately one equivalent of Cu(NO 3 ) 2 with respect to the protein monomer was added, no more changes were observed in the spectrum. Furthermore, LmrR without HQAla showed no changes in absorption upon addition of Cu(NO 3 ) 2 (Figure 2c). This suggests that the HQAla residue is the preferred Cu II binding site, albeit that some unspecific binding to other parts of the protein, which is not detectable by UV-Vis measurements, cannot be excluded. Using the same method, Zn(NO 3 ) 2 and the rhodium complexes were titrated against LmrR_V15/M89HQAla and LmrR (Figures S4 and S5). Complexing of HQala with Zn (NO 3 ) 2 and Cp*RhCl 2 caused similar spectral changes until one equivalent of metal was added, indicating binding accompanied by deprotonation. Again, no changes in absorption were observed when titrating either metal against LmrR, indicating binding of Zn II and rhodium-bound cyclopentadienyl to the HQ moiety in LmrR specifically. Rh 2 (AcO) 4 and Rh 2 Cl 2 (CO) 4 did not seem to bind LmrR or HQAla, as the absorption spectra of both LmrR_V15HQAla and LmrR did not change upon titration with these rhodium salts. RhCl(COD) behaved different, altering the absorbance of LmrR_V15HQAla and LmrR beyond the addition of 1 equivalent, and was therefore thought to bind unspecifically to the scaffold. Taken together, the versality of HQAla to bind metal ligands proved to be effective in capturing Cu II , Zn II or Cp*Rh III into the dimeric LmrR scaffold.
The catalytic potential of these novel artificial metalloenzymes was studied in several reactions. Previous functionalisation of LmrR with a covalently attached Rh catalyst was successful in the catalytic hydrogenation of CO 2, [17] whereas the water stable Cp*Rh complex was used to accelerate aromatic CÀ H activation when bound by streptavidin. [18] However, testing LmrR_V15/M89HQAla complexed with Cp*Rh for the same aromatic CÀ H activation did not give rise to detectable product formation, most likely due to the limited number of free coordination sites when Cp*Rh is bound to HQAla.
Zn II serves as the catalytic metal in numerous different hydrolases. [19] Therefore, we decided to probe the hydrolytic potential of LmrR_HQAla_Zn II on five different substrates. First, the hydrolysis of ester bonds was studied by using two model substrates: p-nitrophenyl acetate and p-nitrophenyl butyrate. Although LmrR_HQAla_Zn II hydrolyses these substrates with rates 3-3.5 times higher than the uncatalysed reaction, the fact that unbound Zn(NO 3 ) 2 with LmrR yielded similar rate accelerations indicated that the specific active site was not required for these reactions to proceed. The hydrolysis of p-nitrophenylphosphate was tested, however no hydrolysis was observed, with or without the metal bound.
Next, the ability of LmrR_V15/M89HQAla_Zn II to hydrolyse amide bonds was examined using substrates 1 and 2 ( Figure 3). Both of these substrates are small peptides with two or three amino acids bound to p-nitrophenylalanine. Hydrolysis of these substrates produces p-nitroaniline, which can be spectrophotometrically followed at 410 nm. Surprisingly, he hydrolysed product of 1 was not observed, but we did observe (slow) hydrolysis of substrate 2 ( Figure S7). Although the LmrR_V15/ M89HQAla_Zn II enzymes were active on substrate 2, no pnitroaniline was produced when LmrR with Zn(NO 3 ) 2 was added. Therefore, the hydrolysis of 2 is thought to rely on the Zn II containing active site in LmrR_HQAla.
Finally, the catalytic activity of Cu II /HQAla-containing artificial metalloenzymes was evaluated in the vinylogous Friedel-Crafts alkylation reaction of 5-methoxy-1H-indole (4) with 1-(1methyl-1H-imidazol-2-yl)but-2-en-1-one (3; Scheme 1a), [6e,20] and the 1,4-addition of water to α,β-unsaturated 2-acyl pyridine 6 resulting in the corresponding β-hydroxy ketone product 7 (Scheme 1b). [6f,14,21] Both reactions were carried out using 9 mol % of Cu(NO 3 ) 2 (90 μM) with a small excess of LmrR_V15HQAla or LmrR_ M89HQAla (112.5 μM of the monomer). The uncatalysed Friedel-Crafts alkylation proceeds with 4 % conversion, whereas the unbound Cu(NO 3 ) 2 catalysed this reaction to give 99 % conversion (Table 1, entries 1, 2). The reaction catalysed by LmrR_V15HQAla_Cu II yielded similar results to those with Cu (NO 3 ) 2 in solution, suggesting that although Cu II binds to HQAla  in the hydrophobic pore (Figure 2a), the metal catalyst is situated too far from the chiral scaffold to influence chirality of the reaction. ( Overall, our novel artificial copper enzymes were successful in the catalysis of both studied reactions. Notably, similar trends were observed between the V15 and M89 mutants. With the LmrR_V15HQAla mutant, high conversions were obtained. These results can be explained either by ideal localisation of the central tryptophanes contributing to substrate binding, or by the HQAla moiety being more solvent exposed at this position. The latter seems more likely because of the low enantioselectivity of the reactions with this mutant. However, this hypothesis should ideally be confirmed with X-ray structures. The low conversions with LmrR_M89HQAla suggest that HQAla incorporated at this position is less accessible, or does not form interactions favourable for catalysis, however this more constrained position does facilitate enantioselective product formation. In summary, we have presented novel artificial metalloenzymes containing the noncanonical amino acid HQAla as metal-binding moiety. The catalytic potential of these newly created artificial metalloenzymes was evaluated in various reactions. HQAla was successfully incorporated into the structure of LmrR at two different positions: V15 and M89. Both mutants of LmrR showed a good affinity for different metal salts, such as Cu(NO 3 ) 2 , Zn(NO 3 ) 2 and Cp*RhCl 2 . The novel Zn IIcontaining artificial metalloenzymes showed activity in the hydrolysis of peptide bonds, while Cu II -containing variants showed activity in Friedel-Crafts alkylation reaction of 5-methoxyindole with α,β-unsaturated-2-acyl imidazole and wateraddition reaction to α,β-unsaturated 2-acyl pyridine, although with low-to-moderate enantioselectivities. Overall, due to the high affinity of LmrR_HQAla towards the different metal salts and complexes, this artificial metalloenzyme can provide a platform for a range of different, currently unexplored, metalcatalysed reactions. We believe that HQAla, together with other known metal-binding ncAAs, open up the way for a variety of (new-to-nature) biotransformations .  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