Photoreactive Bicyclic Amino Acids as Substrates for Mutant Escherichia coli Phenylalanyl-tRNA Synthetases

Unnatural amino acids carrying reactive groups that can be selectively activated under non-invasive biologically benign conditions are of interest in protein engineering as biological tools for the analysis of protein-protein and protein-nucleic acids interactions. The double ring system phenylalanine analogues benzofuranylalanine and benzotriazolylalanine were synthesized, and their photolability was tested by UV irradiation at 254, 320, and 365 nm. Although both showed photo reactivity, benzofuranylalanine appeared as the most promising compound because this amino acid was activated by UVA (long wavelength) irradiation. These amino acids were also tested for in vitro charging of tRNA(Phe) and for protein mutagenesis via the phenylalanyl-tRNA synthetase variant alphaA294G that is able to facilitate in vivo protein synthesis using a range of para-substituted phenylalanine analogues. The results demonstrate that benzofuranylalanine, but not benzotriazolylalanine, is a substrate for phenylalanine tRNA synthetase alphaA294G, and matrix-assisted laser desorption ionization time-of-flight analysis showed it to be incorporated into a model protein with high efficiency. The in vivo incorporation into a target protein of a bicyclic phenylalanine analogue, as described here, demonstrates the applicability of phenylalanine tRNA synthetase variants in expanding the scope of protein engineering.

The use of unnatural amino acids for protein engineering is a rapidly developing technology that adds new dimensions to conventional mutagenesis by allowing introduction of novel chemical and biological functionality into proteins (1). Escherichia coli has been the organism most widely used for in vivo unnatural amino acid incorporation (e.g. [2][3][4][5], but eukaryotic host systems (6 -8) have also recently been described. The development of all such systems is dependent on engineering the substrate specificity of the aminoacyl tRNA synthetases or of the tRNA-aminoacyl tRNA synthetase pair. The aminoacyl tRNA synthetases charge tRNA with cognate amino acid before delivery of the aminoacyl-tRNA by elongation factor Tu to the ribosome for incorporation into nascent polypeptide (9). One approach to the development of unnatural amino acid recogni-tion involves selective mutagenesis of the relevant aminoacyl tRNA synthetase to recognize amino acid analogues that are close structural mimics of the cognate amino acid (2,3). For instance, an E. coli phenylalanine tRNA synthetase (PheRS) 1 variant carrying a single Ala 3 Gly amino acid substitution at ␣-subunit residue 294 (PheRS-␣A294G) (10) displays relaxed substrate specificity in vivo toward a number of para-substituted phenylalanine mimics. Examples include several halogenated phenylalanines including p-chlorophenylalanine (2), pbromophenylalanine (11) and p-iodophenylalanine as well as p-cyanophenylalanine, p-ethynylphenylalanine, p-azidophenylalanine, and 2-,3-, and 4-pyridylalanine (12). Introducing the additional mutation Thr 3 Gly in position 251 (PheRS-␣T251G/A294G) further enlarges the amino acid binding pocket, which provides space for phenylalanine analogues carrying modifications on the benzene ring and allows activation of still larger unnatural amino acids such as p-acetylphenylalanine (3).
Based upon previous studies of proteinogenic photoreactive amino acid analogues with bicyclic structures containing a benzene ring "scaffold" (13), we set out to test such compounds as substrates for incorporation into a protein via PheRS-␣A294G. Other unnatural amino acids shown to be substrates for this aminoacyl tRNA synthetase variant display distinct yet limited structural divergence from phenylalanine. In contrast, the presently tested analogues, benzofuranylalanine and benzotriazolylalanine ( Fig. 1), contain an additional five-membered ring system fused to the benzene ring of phenylalanine. Although PheRS-␣A294G (and PheRS-␣T251G/A294G) appear rather flexible with respect to modifications at the para position of the benzene side chain, it is far from obvious that this mutant would accommodate amino acids carrying double ring systems involving joint substitutions at the meta and para position of the phenylalanine benzene side chain. Nevertheless, based on growth inhibition, we have previously proposed that benzofuranylalanine is indeed a substrate for PheRS-␣A294G, opening up the possibility that such bicyclic amino acid analogs could be developed as substrates for protein synthesis (13).
We now present in vitro results showing that benzofuranylalanine is a substrate for aminoacylation of tRNA Phe by PheRS-␣A294G and directly demonstrate the in vivo incorporation of this amino acid analogue into a model protein using E. coli strains expressing mutant PheRS ␣-subunits and wild type tRNA Phe . In addition, another photoreactive and structurally * This work was supported by grants from the Lundbeck Foundation (to T. B.), the Alfred Benzon Foundation (to M. I.), the Novo Nordisk Foundation (to M. I.), and the Danish Research Agency (to T. B., M. I., and P. E. N.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed. E-mail: pen@ imbg.ku.dk. related amino acid, benzotriazolylalanine, was synthesized and tested as a substrate for activation and aminoacylation.
Charging and Competition Analyses-Wild type and ␣A294G E. coli PheRS were purified as previously described (10). E. coli tRNA Phe was from Sigma. Pyrophosphate (PP i ) exchange and aminoacylation reac-tions were performed as described (10), with unlabeled benzofuranylalanine included at 2 and 4 mM for the determination of inhibition constants during aminoacylation. The direct attachment of benzofuranylalanine and benzotriazolylalanine to in vitro transcribed tRNA was monitored by direct 32 P labeling of tRNA Phe using E. coli tRNA-terminal nucleotidyltransferase (15) followed by aminoacylation and product visualization as previously described (16).
Expression Analyses-In vivo incorporation of benzofuranylalanine into recombinant murine dihydrofolate reductase (DHFR) was performed as previously described using the phenylalanine auxotrophic   (11). An overnight culture was diluted into fresh M9 minimal media supplemented with the 20 amino acids at 20 g/ml and the antibiotics ampicillin (100 g/ml) and chloramphenicol (40 g/ml) and grown to an A 600 of 0.6 -1.0. The cells were harvested by centrifugation, washed twice in 20 ml of ice-cold 0.9% NaCl, and resuspended in fresh M9 media containing antibiotics and all natural amino acids except phenylalanine. The cells were separated into aliquots of 1-2-ml fractions and supplemented with benzofuranylalanine (2 mM) and phenylalanine (0.1 mM) as indicated, grown for 10 min at 37°C, induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside, and then grown for 4 -5 h at 30°C. 100 l of cells grown in the presence of phenylalanine (and the A 600 /ml normalized amount of cells grown with unnatural amino acid) were harvested by centrifugation. The pellet was resuspended in 20 l of B-PER (Pierce) and 20 l of 2x SDS loading buffer and heated to 90°C for 5-10 min, and 10 l was analyzed using 12% acrylamide-SDS page minigels followed by Coomassie Brilliant Blue staining.
Mass Spectrometry-Mass spectrometry analyses were performed using a SpectraChrom Kompact MALDI II instrument from Kratos. MALDI-TOF samples were prepared by mixing 1 l of protein sample (0.2-0.25 g) per 4 l of matrix (␣-cyano-4 hydroxycinnamic acid at 10 mg/ml in 50% acetonitrile 50% H 2 O). The samples were spotted onto MALDI-TOF slides using 4.5 l per application. Chymotrypsinogen A (25,657 Da) and ␤-lactoglobulin A (18,364 Da) were used as calibration standards. The slide was examined for "sweet spots" before running the experiment. 250 profiles were accumulated per sample.

RESULTS AND DISCUSSION
Synthesis of Bicyclic Amino Acids-Benzofuranylalanine was synthesized as previously reported (13), and the novel benzotriazolylalanine was prepared as shown in Scheme 1 starting from commercially available 5-(hydroxymethyl)triazole using standard chemistry. Enantiomeric resolution was achieved using Carlsberg subtilisin (13) for specific hydrolysis of the L-form of an ester intermediate.

Photochemistry of the Unnatural Amino Acids-The
[2ϩ2] photocycloadditions of alkenes are extremely specific and often of high (quantum) yields. These could therefore be very useful for sequence-specific post-modification of proteins, provided an activated alkene with appropriate absorbance characteristics can be site specifically incorporated into a protein, since no other functionality of similar reactivity is naturally present in proteins.
Benzofurans undergo [2ϩ2] photocycloaddition with a variety of alkenes (e.g. Ref. 17), although the furan double bond is only weakly activated by the adjacent oxygen. However, benzofurans also undergo other photochemical reactions. To study the photosensitivity of benzofuranylalanine, the compound was irradiated at different wavelengths, and the corresponding UV absorption spectra were recorded over time (Fig. 2). Upon irradiation at 254 nm the absorption peak at 245 nm declined, and a "shoulder" peak with a maximum of 342 nm appeared with similar kinetics (estimated t1 ⁄2 , ϳ50 min) compatible with a photochemical reaction ( Fig. 2A). However, because most proteins (due to the presence of aromatic amino acids) and all nucleic acids as well as other cell components have strong absorption at 254 nm, many unwanted side reactions are bound to occur both in vitro and especially in vivo upon such irradiation. Conversely, considerably fewer photoreactions take place in biological systems upon irradiation at longer wavelengths. Consequently, we also irradiated the amino acids at 320 and 365 nm, where the benzofuran chromophore exhibits a weak absorption ( max ϭ 342 nm), now focusing on the UV spectra in the range 300 -400 nm (Fig. 2, B and C). At 320 nm of irradiation, both the spectral changes as well as the time course of the change in absorbance at the initial absorbance maximum of 342 nm (Fig. 2B) clearly show that at least two photochemical transformations take place, a fast reaction with an estimated t1 ⁄2 of ϳ 6 min and a subsequent slow reaction over many hours. Strikingly, at 365 nm of UVA irradiation only the fast component was observed. We have not attempted to analyze the actual products at this stage, but these results clearly demonstrate the photosensitivity of the benzofuran ligand, and thus, its potential utility as a photochemical handle upon incorporation into proteins. A potential caveat of using UVA irradiation (, ϳ335 nm) involves the formation of photo adducts between thiouridine and cytidine at positions 8 and 13 in some tRNAs of E. coli (18), causing an amino acid starvation response. Once the cells are removed from UV exposure, however, the effects are reversed, and they recover by synthesis of new functional tRNAs (19).
Benzotriazoles undergo photochemical reactions reminiscent of the highly photosensitive azides. The predominant photoreaction involves release of molecular nitrogen (N 2 ) and formation of a radical species that may either undergo radical insertion reactions or nucleophilic addition reactions (e.g. Ref. 20), both giving rise to chemical cross-linking. Upon incorporation FIG. 4. PheRS-␣A294G aminoacylates tRNA Phe with benzofuranylalanine. tRNA Phe was labeled as previously described (16). The reaction contained 100 mM Hepes, pH 7.2, 10 mM MgCl 2 , 30 mM KCl, 2 mM ATP, 1 M E. coli tRNA Phe (Roche Applied Science), and a trace (3.5 nCi) of 3Ј 32 P-labeled tRNA Phe , and aminoacylation was performed at 1 M tRNA and 1 nM wild type PheRS (diamonds and squares) or ␣A294G mutant (triangles and circles) with 2 mM Phe (diamonds and triangles) or 0.2 mM benzofuranylalanine (squares and circles).  1-4) or AF-1Q/pQE-15 ( lanes 5-8). The cells were incubated as described under "Experimental Procedures" with the indicated amino acids. Positions for a protein molecular weight marker are indicated on the left side of the gel. Bzf, benzofuranylalanine; wt, wild type. into a protein, the benzotriazole ligand would constitute a site for specific photochemical cross-linking to an interacting biological ligand, e.g. another protein or a nucleic acid.
Irradiation at 254 and 320 nm of the benzotriazolylalanine resulted in pronounced spectral changes, fully compatible with a single photochemical conversion taking place (isosbestic points are observed at least at shorter irradiation times), with an estimated t1 ⁄2 of 60 min (Fig. 3, A and B). These results support the potential use of this amino acid for site-specific photosensitization of proteins if incorporated. However, at 365 nm of irradiation (at which wavelength the benzotriazole does not show any significant absorption) no spectral changes were observed, indicating that this wavelength cannot be utilized for induction of photochemical reactions using the benzotriazolylalanine.
Steady-state Kinetics for Bicyclic Amino Acids-To investigate charging with unnatural amino acids, we first tested the ability of benzofuranylalanine and benzotriazolylalanine to inhibit aminoacylation of tRNA Phe . Benzofuranylalanine did not show any inhibition of the wild type enzyme (data not shown) but acted as a competitive inhibitor for PheRS-␣A294G (Table  I). In contrast, the addition of benzotriazolylalanine had no effect on aminoacylation by either form of PheRS. To further investigate whether benzofuranylalanine might be a substrate for activation by the mutant form of the enzyme, PP i exchange analysis was performed. The PP i exchange reaction showed that benzofuranylalanine is a substrate for PheRS-␣A294G, with a K m of 2.3 Ϯ 0.4 mM and a k cat of 923 Ϯ 65 min Ϫ1 . PP i exchange with the wild type enzyme showed some low but inconsistent activity, suggesting that benzofuranylalanine might be poorly activated. As expected, phenylalanine showed a higher reaction rate, and no activity at all was seen when the reaction was performed without substrate.
We also tested benzofuranylalanine and benzotriazolylalanine directly as substrates for the aminoacylation of 3Ј-[ 32 P]-tRNA Phe (Fig. 4). Benzofuranylalanine was found to be a substrate for aminoacylation by ␣A294G but not wild type PheRS, and benzotriazolylalanine was a substrate for neither enzyme.
These results are consistent with the kinetic analyses and confirm that benzofuranylalanine, but not benzotriazolylalanine, can be attached to tRNA Phe by PheRS-␣A294G. In addition they indicate that benzofuranylalanine, in contrast to natural non-cognate amino acids, is not a substrate for the hydrolytic proofreading activity of PheRS (10).
In Vivo Incorporation of Bicyclic Amino Acids-To test whether benzofuranylalanine and/or benzotriazolylalanine are in vivo protein synthesis substrates, we used a phenylalanine auxotrophic E. coli strain encoding PheRS-␣A294G on an episome (strain AF-1Q/pQE-FS) and the isogenic strain lacking the mutant tRNA synthetase-encoding gene (strain AF-1Q/ pQE-15) as a control. Both strains harbor an isopropyl-1-thio-␤-D-galactopyranoside-inducible expression cassette encoding murine DHFR carrying a short histidine tag (shtDHFR). As previously reported (11), both strains show abundant overproduction of shtDHFR upon the addition of isopropyl-1-thio-␤-Dgalactopyranoside in the presence of phenylalanine (Fig. 5,  compare lanes 1 with 2 and lane 5 with 6). However, only strain AF-1Q/pQE-FS produced large amounts of shtDHFR when substituting benzofuranylalanine for phenylalanine (compare lanes 3 and 7). In fact, in the presence of wild type PheRS alone shtDHFR production did not increase beyond background when conducting the experiment with benzofuranylalanine. These results indicate that benzofuranylalanine is indeed an in vivo substrate for PheRS-␣A294G and, consistent with the in vitro data, is not utilized by wild type PheRS. A slight but notable expression of shtDHFR was observed even in the absence of added phenylalanine or benzofuranylalanine (lanes 4  and 8). This was previously ascribed to residual cellular pools of phenylalanine remaining even after extensive washing of the cells (3,11). Similar analyses were conducted using the benzotriazolylalanine. Consistent with the in vitro data, no shtDHFR production above background was observed upon the addition of benzotriazolylalanine (data not shown). We also tested the strain AF-1Q/pQE-T251G/A294G carrying a double mutant PheRS ␣-subunit (3) for incorporation of unnatural amino acids FIG. 6. Mass spectrometry analysis of benzofuranylalanine incorporation into shtDHFR. Shown are examples of mass spectra showing shtDHFR expressed in the presence of benzofuranylalanine (A) and phenylalanine (B). The molecular masses obtained for this particular experiment are indicated. Protein mass determination was calculated as the mean of 4 -5 experiment repetitions. The molecular weights obtained Ϯ S.D. is 24,284 Ϯ 119 (shtDHFR containing benzofuran) and 23,957 Ϯ 76 (shtDHFR-Phe), and these masses are statistically significantly different (double-sided Student's t test for data with equal variance, p Ͻ 0.005; n ϭ 9). The calculated mass of shtDHFR-Phe is 24040. Sequencing of the shtDHFR-coding region revealed no amino acid alterations as compared with the published sequence (Qiagen, plasmid pQE15). The discrepancy of the found and calculated masses of shtDHFR-Phe are therefore ascribed to experimental inaccuracy. into shtDHFR. In accordance with the data obtained using strain AF-1Q/pQE-FS, the double mutant utilized benzofuranylalanine but showed no shtDHFR expression above background in the presence of benzotriazolylalanine (data now shown). Thus, benzofuranylalanine, but not benzotriazolylalanine, is a substrate for PheRS ␣T251/A294G.
Analysis of Benzofuran-containing shtDHFR-Because benzofuranylalanine (M r 205) has a larger mass as compared with phenylalanine (M r 165), incorporation of benzofuran into the reporter protein should be readily detectable as a significant mass increase of shtDHFR containing benzofuran. Mass spectrometry analysis of purified protein revealed molecular masses of 24,284 and 23,957 Da for shtDHFR expressed in the presence of benzofuranylalanine and phenylalanine, respectively (Fig. 6). This clearly demonstrates the incorporation of unnatural amino acid and corresponds to an average of ϳ8 benzofuranylalanine replacements in nine possible positions.
Conclusions-The data presented show that benzofuranylalanine, but not benzotriazolylalanine, is an efficient substrate for cellular protein synthesis as a result of its ability to be attached to tRNA Phe by PheRS-␣A294G. These findings are of potential interest in the development of novel protein crosslinking methodologies given that benzofuranylalanine was found to be photoreactive at UVA wavelengths. The effective application of such methodologies is dependent on site-specific replacement of particular residues with benzofuranylalanine rather than global insertion in response to all phenylalanine codons as described here. Although a completely site-specific system may require extensive mutagenesis of both PheRS and a tRNA Phe -derived suppressor species or perhaps another pair (4), existing heterologous codon-biased systems offer the potential to replace designated Phe residues with benzofuranylalanine in certain targets (21,22).