Structures of Methanomethylophilus alvus pyrrolysine tRNA-synthetases support the need for de novo selections when altering substrate specificity

Abstract

A recently developed genetic code expansion (GCE) platform based on the pyrrolysine amino-acyl tRNA synthetase (PylRS)/tRNA^Pyl pair from Methanomethylophilus alvus (Ma) has improved solubility and lower susceptibility to proteolysis compared with the homologous and commonly used Methanosarcina barkeri (Mb) and M. mazei (Mm) PylRS GCE platforms. We recently created two new Ma PylRS variants for the incorporation of the fluorescent amino acid, acridonyl-alanine (Acd), into proteins at amber codons: one based on “transplanting” active site mutations from an established, high efficiency Mb PylRS and one that was de novo selected from a library of mutants. Here, we present the crystal structures of these two Ma PylRS variants with Acd/ATP bound to understand why the “active site transplant” variant (Acd-AST) displayed 6-fold worse Acd incorporation efficiency than the de novo selected PylRS (called Acd-RS1). The structures reveal that the Acd-AST binding pocket is too small and binds the three-ringed aromatic Acd in a distorted conformation, whereas the more spacious Acd-RS1 active site binds Acd in a relaxed, planar conformation stabilized by a network of solvent-mediated hydrogen bonds. The poor performance of the AST enzyme is ascribed to a shift in the Ma PylRS β-sheet framework relative to that of the Mb enzyme. This illustrates a general reason why “active site transplantation” may not succeed in creating efficient Ma PylRSs for other non-canonical amino acids. The work also provides structural details that will help guide the development of future Ma PylRS/tRNA^Pyl GCE systems via de novo selection or directed evolution methods.

Graphical Abstract

INTRODUCTION

In an accompanying paper (20), we describe a genetic code expansion (GCE) system derived from the Methanomethylophilus alvus Candidatus (Ma) pyrrolysine (Pyl) aminoacyl tRNA synthetase (PylRS)/tRNA^Pyl pair for the efficient encoding of the fluorescent amino acid acridonyl-alanine (Acd, Fig. 1A) at amber stop codons in E. coli and human cells. The Ma PylRS/tRNA^Pyl platform has garnered notable attention among pyrrolysine encoding systems in that it lacks the N-terminal tRNA recognition domain found in other PylRSs used for GCE, such as those from the widely used Methanosarcina barkeri (Mb) and Methanosarcina mazei (Mm) Pyl PylRS/tRNA^Pyl pairs (Fig. 1B)^1–5. Mm/Mb GCE platforms have been used to encode hundreds of non-canonical amino acids (ncAAs) in both E. coli and eukaryotic cells, however the overall efficiency of these systems can be hampered by RS insolubility and instability caused by their N-terminal domains, and also by an internal nuclear localization signal. Conversely, the Ma PylRS is fully functional without this N-terminal domain, providing increased solubility, stability, and retention in the cytoplasm of eukaryotic cells^1,2. Thus, strategies to create Ma PylRS variants effectively with selectivity for targeted ncAAs are becoming increasingly important.

Figure 1. Comparison of the Mm/Mb and Ma PylRSs.

(A) Chemical structures of the amino acids referenced in this study, Acd (top) and Pyl (bottom). (B) Comparison of the domain architecture of the Mm/Mb and Ma PylRSs. (C) An overlay of the Mm PylRS catalytic domain (PDB: 2E3C) in orange and the apo Ma PylRS (PDB: 6JP2) in grey shows the similar overall structure of the enzymes. (D) Structural alignment of the Mm (orange) and Ma (gray) PylRS active sites with pyrrolysine from the Mm structure (PDB 2ZCE). Side chains are shown for the residues that were subjected to saturation mutagenesis in our companion study for the selection of new Ma PylRSs. Residue numbering is from the Mb PylRS for consistency. (E) Active site mutations in each PylRS referenced in this study. Differences in the Acd-RS1 are colored in red. (F) In-cell fluorescence of cells expressing sfGFP-150TAG in the presence and absence of Acd using the Mb Acd-RS 41, the Ma Acd AST, and the Ma Acd RS1 derived from life/death selections (data shown is from our accompanying manuscript, 20). Data shown is the average of triplicates and error bars represent the standard deviation.

Although the sequence identity of the Ma PylRS and Mm/Mb PylRS catalytic domain is low (~35%), the tertiary structures of the Mm catalytic domain and Ma PylRS are very similar (Cα RMSD=1.2 Å) (Fig. 1C) and the substrate recognition elements in the amino acid binding pocket appear well conserved (Fig. 1D)². Consequently, it has been proposed that specificity for a particular ncAA can be “transplanted” from Mm/Mb PylRS variants to the Ma PylRS by recreating the same constellation of active site mutations^1,2. This “active site transplant” (AST) method has been used successfully to create Ma PylRS variants for incorporating lysine-like derivatives with bulky, hydrophobic head groups by transplanting up to two mutations rationally designed to enlarge the back end of the ncAA binding pocket (e.g. Y126A/Y206F)^1,2,5. Similarly, mutations designed to open space in the front end of Mm/Mb PylRSs amino acid binding pocket (e.g. N166S or N166A/C168A) were sufficient to alter Ma PylRS substrate selectivity for singly-substituted phenylalanine derivatives^2,6,7. However, attempts to transplant the five Mm PylRS active site mutations that confer specificity for acetyl-lysine into Ma PylRS were not successful⁸. These observations suggest the AST method can be effective for creating Ma PylRS variants where only one or two mutations are required, but more extensively remodeled active sites may not be readily transferred across the platforms.

To explore the limits of the AST strategy, we evaluated the efficiency of Ma PylRS ASTs harboring 3 or more mutations relative to wild-type for incorporating three aromatic ncAAs: acridonyl-alanine (Acd), 3-nitro-tyrosine and m-methyl-tetrazinyl-phenylalanine (Tet3.0-Me) (20). In these cases, the AST strategy yielded Ma PylRS variants that were able to encode the targeted ncAAs but were inefficient compared to the Mb PylRSs from which they were derived, whereas much more efficient Ma PylRS variants were generated from de novo selections. Specifically, for Acd, sequence analysis of the top selected Ma Acd-PylRS (referred to as Acd-RS1) revealed three residue differences compared to the Ma Acd-AST (Fig. 1E). These differences provided Ma Acd-RS1 with improved in vitro kinetic properties and yielded a 6-fold improvement in Acd incorporation efficiency in E. coli compared to the Ma Acd-AST (Fig.1F).

Here, we sought to better understand at the structural level why the AST strategy for generating new Ma PylRSs with ncAA selectivity was less effective than de novo selections. We report the crystal structures of both the evolved Ma Acd-RS1 and the Ma Acd-AST bound to Acd and adenosine nucleotides. These structures reveal the basis for the improved performance of the evolved Ma Acd-RS1 compared to the Acd-AST and offer strategies for future engineering of the PylRS for acridone derivatives with altered fluorescence properties as well as non-fluorescent ncAAs. More broadly, this work supports the conclusion that evolution/selection-based methods rather than AST or rational design will likely prove the most effective methods to produce Ma PylRS variants able to encode ncAAs with efficiencies more closely matching that of the wild-type enzyme.

RESULTS

To solve the structures of the Ma Acd-RS1 and Acd-AST, proteins were overexpressed in E. coli and purified to homogeneity (Supporting Figure 1). Size-exclusion chromatography coupled to multi-angle light scattering (SEC-MALS) confirmed the purified proteins were monodisperse dimers (Supporting Fig. 2, Supporting Table 1), consistent with the obligate dimeric nature of the Mm/Mb Pyl PylRSs⁹. Sparse matrix crystallization screens were carried out after pre-incubating the proteins with ATP (or its non-hydrolyzable analog, AMPPNP) and Acd in the presence of Mg²⁺, producing a variety of crystal growth conditions which were further optimized. From these efforts, we report four structures at ~1.5 Å resolution (Supporting Table 2): (1) Acd-RS1 bound to ATP + Acd after 1 day and (2) two weeks of crystal growth, (3) Acd-RS1 bound to AMPPNP + Acd, and (4) Acd-AST bound to ATP + Acd after 4 weeks of crystal growth.

Overview of liganded structures.

Each crystal belonged to the I4 space group, and each structure was solved by molecular replacement using the apo Ma wild-type PylRS as a search model (PDB 6JP2). The asymmetric units contained two PylRS molecules constituting the functional dimer (Fig. 2A). In all the structures, one molecule of the dimer (chain A) contained clear electron density for the ATP (or AMPPNP) and Acd (Supporting Fig. 3). The second molecule (chain B) had clear density only for ATP/AMPPNP in the Acd-RS1 structures, and AMP (that we surmise arose from hydrolysis of ATP during crystal growth) for the Acd-AST (Supporting Fig. 4). No evidence was found for Acd binding in the second chain (chain B) for any structure even though Acd was present at saturating conditions (5 mM vs K_D ~ 0.3 mM) (20). Trapping of the dimer in this asymmetric state likely arose as a consequence of crystallization, as two symmetrical protein molecules were not compatible with crystal packing (Supporting Fig. 5).

Figure 2. Overall structure of the Ma Acd-RS1.

(A) The asymmetric unit contains one RS functional dimer. Chain A with ATP and Acd bound is colored in light pink and chain B with ATP bound is colored in dark pink. Model statistics for all the structures are shown in Supporting Table 2. (B) Chain A of the Ma Acd-RS1 with the conserved class II aaRS motifs highlighted in yellow, cyan, and green. A feature enhanced maps¹⁹ (2F_o-F_c) for the bound ligands (contoured to 2.00σ) is shown.

The global structures are very similar to that of the Ma apo enzyme (PDB 6JP2, overall Cα RMSD=0.59 Å) sharing the typical β-sheet core flanked by α-helices and other highly conserved class II PylRS elements such as motifs I, II, and III (Fig. 2B)^10,11. As expected, motif I is located at the dimer interface and motifs II and III compose part of the active site near the bound ATP nucleotide¹⁰. The Acd in chain A is in a hydrophobic pocket, similar to the location of bound amino acid in structures of other PylRSs^9,12. In each structure, ATP/AMPPNP is in a bent conformation as is expected for class II PylRSs, with three well-coordinated magnesium ions stabilizing it (Supporting Fig. 6 and 7). The assignment of these atoms as Mg²⁺ was based on the short distances to the oxygens of the phosphate groups, and their 6-coordinate octahedral interaction geometry.

Conformational changes upon substrate binding.

With these structures we can compare snapshots of the Ma PylRS at three different stages of substrate binding: apo, ATP bound and ATP/Acd bound (Fig. 3A). The conformational changes described in this section are consistently seen across the Ma Acd-RS1 and Acd-AST structures, and so for simplicity we use the Ma Acd-RS1 structure with AMPPNP bound for comparative analyses with Mm PylRS structures. The binding of AMPPNP causes the loop between β3 and β4, part of motif II (residues 149–161, blue residues in Fig. 3A), to swing about 10 Å outward from the active site to accommodate the bound ATP molecule (Supporting Fig. 8A). This induced fit upon the binding of ATP is required, because in the apo structure this loop clashes with the location of the adenine of ATP. Similar conformational changes were observed for the Mm PylRS upon binding of AMPPNP (Fig. 3B, Supporting Fig. 8B)¹³.

Figure 3. Conformational changes upon ligand binding.

(A) Surface models of the wildtype apo RS (PDB: 6JP2), Acd-RS1 chain B with AMPPNP bound, and Acd-RS1 chain A with AMPPNP and Acd bound. Regions with major conformational changes between the structures are highlighted in light blue (residues 149–161), yellow (201–211), and light green (223–230). (B) Surface models of the M. mazei WT PylRS apo (PDB: 2E3C), AMPPNP bound (PDB: 2Q7E), and AMPPNP and mAzZLys bound (PDB: 6AAC). Regions corresponding to those that undergo major conformational changes in the Ma structures are highlighted in light blue, yellow, and light green.

When both ATP and acridone bind, the hairpin connecting β5 – β6 (residues 201–211, yellow residues in Fig. 3A) swings about 9 Å towards the bound Acd into a closed conformation, which allows a stacking interaction between Y206 and the Acd side chain as well as an H-bond to the α-carboxylate of the acridone (Supporting Fig. 8C). Similar conformational changes have been shown in structures of the Mm PylRS catalytic domain (Fig. 3B, Supporting Fig. 8D); these changes have been seen in response to AMPPNP binding (PDB 2Q7E), to Pyl-adenylate reaction product binding (PDB 2Q7H), and to ncAA/ATP binding (PDB 6AAC), confirming this conformational change is a conserved feature of the PylRS family¹³.

In addition, we observed conformational changes in Ma Acd-RS1 residues 224–230 that occur upon Acd binding that help shape the ncAA binding pocket and, to our knowledge, have not been observed in the Mm PylRS. In the published structure of the apo Ma PylRS², His227 points inward toward the ncAA binding pocket, while Leu229 and Pro226 point away (Supporting Fig. 9A). Upon binding of Acd/ATP-bound, these residues flip orientation such that Leu229 and Pro226 point inward and line the ncAA binding cavity, while His227 points toward solvent. As these regions are not involved in crystal contact interfaces in either structure (apo Ma RS and Acd/ATP-bound), we infer that each structure represents the dominant conformation of that loop that is adopted in solution. For the apo Mm PylRS (residues 402–408), the equivalent loop conformation is identical to that of the apo Ma PylRS, and it does not change upon Pyl/ATP binding (Supporting Fig. 9B). These observations imply that varying residues in the 224–230 loop of Ma PylRS should be considered in the design of future libraries and directed evolution approaches to create new Ma synthetase functionality.

Acd binding to the Ma Acd-AST.

In the Acd-AST structure, clear electron density is observed for Acd and ATP (Supporting Figs. 3A and 10A) revealing an off-angle positioning and puckering of the Acd ring, as well as a bending of the carbonyl group out of plane (Fig. 4A and 4C). To support the modeling for this distorted ring conformation, Acd was also refined into the AST density with strict planarity restraints, but this resulted in a slightly higher R-free value compared to using relaxed planarity restraints (data not shown). This distortion of the Acd is caused by a steric clash with the side chain of C168, a library site positioned at the back of the amino acid binding pocket and directly underneath the Acd ring system (Fig. 4A and 4C). The amine group within the Acd ring system does not appear to have any ordered hydrogen bonding acceptors, and Ser166 is poorly positioned for hydrogen bonding to the carbonyl of the Acd ring (Fig. 4A and 4C). Overall, the limited hydrogen bonding interactions suggests Acd binding is driven by shape complementarity and hydrophobic interactions with the Acd-AST active site binding pocket and indeed there is little additional room in the pocket to accommodate the Acd ring system in other orientations, especially if it were to maintain an orientation that leads to catalysis. Inspection of the crystal lattice contacts for the regions “above and below” the catalytic domain—where lattice interactions might be sufficient to force the described distortions—indicates these regions are not involved in protein crystal contacts. It is possible that thermal fluctuations could alleviate the strain in solution, though in the absence of a full RS/Acd/tRNA^Pyl complex structure this is difficult to evaluate. Nevertheless, we infer from these data that the core of the catalytic domain is adopting a structural conformation that closely resembles its native state and the contortion of the Acd three-ring system is necessary for it to bind to the Acd-AST amino acid binding pocket.

Figure 4. Comparison of Acd binding to the Acd-RS1 and Acd-AST active sites.

(A) The active site of the Ma Acd-AST with Acd and ATP bound. Hydrogen bonds are shown with dotted lines. (B) The active site of the Ma Acd-RS1 with Acd and AMPPNP bound. (C) Overlay of the AST (cyan carbon atoms) and RS1 (pink carbon atoms) active sites. The side chain for Acd-AST S166 and C168, and Acd-RS1 A166 are shown as sticks. In the AST, S166 is poorly positioned for an H-bond with bound Acd. Also in the AST, C168 clashes with the bound Acd leading to puckering of the rings. Capped lines show the poor H-bond geometry between S166 and nearby H-bond acceptors as well as the clash of C168 with Acd.

While the aromatic system Acd is well-resolved, the amino acid backbone portion is not, and this weak density implies that multiple conformations could be present. To best match the electron density, we have modelled the backbone in a pre-catalytic conformation with the carboxylate flipped away from the α-phosphate of ATP, similar to how other Mb PylRS structures with amino acid substrate have been modeled¹²(Fig. 4A).

Acd binding to Ma Acd-RS1.

Structures of Acd-RS1 from crystals grown in the presence of ATP + Acd after 2 weeks of growth, and Acd + AMPPNP (a non-hydrolyzable analog of ATP) were solved. No notable differences were found between these two structures in which the Acd side chain is well-resolved but the backbone is modelled in a pre-catalytic conformation with the carboxylate flipped away from the α-phosphate of ATP, just like the Acd-AST (Fig. 4B, Supporting Figs. 3B, 3C and 10B). In contrast to the Acd-AST structure, the Acd aromatic ring system binds in a perfectly planar conformation in the Acd-RS1 structure, which is made possible by the additional space at the back end of Acd binding pocket created by the Cys to Gly change at position 168 (Fig. 4C). Since the crystals were isomorphous, we could compute a difference density map between the two data sets (i.e. |F_obs^AST| - |F_obs^RS1|), and this confirmed the validity of the differences in Acd aromatic ring conformations between the RS1 and AST structures, showing well-defined positive and a negative difference peaks positioned above and below the Acd ring as expected for planar ring binding in RS1 and contorted binding in the AST structure (Supporting Fig 11A). This difference map also revealed that the β-sheets of the Acd-AST that line the bottom of the active site pocket are flexed downward by 0.5–0.6 Å relative to the same β-sheets of Acd-RS1 (Supporting Figs. 11B and 11C). In other words, for Acd to bind to the more constricted Acd-AST active site pocket, the Acd ring not only puckers upward but also the floor of the active site is pushed downward. Thus, the Cys to Gly mutation at position 168 identified by de novo selections enlarges the Acd-RS1 active site sufficiently to permit a more favorable, or less constrained, binding of Acd in a fully planar conformation. We also found that the change from Thr to Cys at position 239 frees up the carboxyl group of D230 in the Acd-RS1 to support a well-defined hydrogen bonding network with water molecules that connect to the amino group of Acd (Fig. 4B). In the Acd-AST, D230 is not available to build a similar stabilizing water-mediated hydrogen bonding bridge between the active site and Acd because of its interaction with the hydroxyl of T239 (Fig. 4A). Taken together, the Acd-RS1 structure reveals differences that provide a plausible basis for the improved activity of the de novo selected Acd-RS1.

Acd binding to Ma Acd-RS1 in crystals grown for 1 day.

In an effort to resolve the ambiguous density for the Acd backbone atoms, we solved the structure of the Acd-RS1 from crystals grown only for 1 day with Acd + ATP. Interestingly, density for the backbone was observed with the carboxylic acid flipped toward the α-phosphate of ATP so that its oxygen atom is only 1.6 Å away from the α-phosphorus atom as would be expected during the adenylation of Acd (Supporting Figs. 3D and 12A). Further, the bond angles of the α-phosphate, between the phosphorous and oxygens, more closely resemble a trigonal bi-pyramidal structure as might be expected for a reaction intermediate rather than the classical tetrahedral bond angles of the α-phosphate group seen in the ATP bound and 2-week crystal growth structures (Supporting Fig. 12C). Whether this structure represents a trapped catalytic intermediate, and why it is different from the other Acd-RS1 structures, requires further characterization.

Comparison of the M. mazei and M. alvus RS amino acid binding pockets.

The above observations provide compelling reasons why the Ma Acd-AST is notably less efficient than the Ma Acd-RS1. We next asked why the Mb Acd-RS 41 was among the top Mb variants selected for Acd incorporation, and why none of the de novo selected Ma Acd-RSs had the same set of active site mutations as any de novo selected Mb Acd-RS (20). To do this, we overlayed the structures of M. mazei PylRS bound to Pyl + AMPPNP onto the Ma Acd-RS1 bound to Acd +AMPPNP. We use the Mm PylRS structure here because no structure is available for the Mb PylRS and these two orthologs are of sufficiently high similarity (83% identity across the catalytic domains). This overlay revealed that β-sheets β4 and β8, forming the floor of the amino acid binding pocket, sit roughly ~1.4 Å lower in the Mm structure than the Ma structure, while the ATP molecules and the helices that line the ceilings of those pockets are superimposable (Fig. 5A and 5B). This observation shows that the β-sheet framework that forms the amino acid binding pocket of Ma and Mm RSs are inherently distinct: the Ma active site is more constricted having physical space that can be manipulated via mutagenesis compared to the Mm active site (and presumably, by extension, the Mb active site as well). As a result, the Ma active site will likely favor smaller amino acid side chains in its active site compared to the Mm/Mb active sites to bind the same ncAA. This seems to explain why the top performing Mb Acd-RS can tolerate, and possibly prefers, a Cys at position 313 (the equivalent to 168 in Ma) for efficient Acd incorporation, while the top performing Ma Acd-RS1 has Gly at this position. However, if smaller amino acids in the Ma RS active site cannot sufficiently accommodate the targeted ncAA, a different ncAA binding mode may be required. In our accompanying paper (20), we found that for nitroY and Tet3.0-Me, the constellation of Ma active site mutations were sufficiently different from Mb that a different mode of binding was likely adopted. These observations yet further emphasize the utility of de novo selections over AST.

Figure 5, Constriction of the Ma Acd-AST amino acid binding pocket.

(A) The back end of the Ma Acd-RS1 active site pocket (pink), which is lined at the bottom by β-sheets β4, β8 and β7 and into which the amino acid substrate side chain extends, is constricted compared to the Mb PylRS (orange) as shown by the Mb b-sheets residing ~1.4 Å lower. As a result, more space is created between e.g. Cys348 of the Mm RS and Pyl, compared to Cys166 and Acd. Because the Mm sheets sit lower, so does the a-helix a6 underneath. (B) Zoom in of the β4 and β8-sheets demonstrating the shift in positioning in the Mm backbone compared to the Ma backbone.

DISCUSSION

The diversity of natural PylRS orthologs provides a large inventory of potential GCE platforms that are, or can easily be made, orthogonal to each other^1,5. Yet at the same time, this diversity complicates direct transfer of PylRS functionality and evolvability from one PylRS to another. In our accompanying manuscript, as well in work by others^2,8, this complication became evident by our inability to make robust Ma PylRSs using prior knowledge of Mb PylRS specificity and the active site transplant strategy. To understand the origins of this failure, we took advantage of the increased stability and solubility of the Ma PylRS and report here crystal structures of Ma PylRS variants with Acd and ATP bound that shed light on why a Ma Acd-AST derived from a highly functional Mb PylRS variant was not as efficient as a de novo selected Ma Acd-RS1 variant. These structures revealed that in order to maintain the position of the ncAA backbone atoms that is required for catalysis, the too-small Acd-AST active site caused the aromatic Acd ring to bind in a distorted conformation and simultaneously displaced the β-sheets that line the bottom of the active site. Specifically, the Acd ring clashed with residue C168 of the Acd-AST. On the other hand, the selected Acd-RS1 had a sufficiently more spacious active site with Gly at position 168, thereby allowing the Acd aromatic rings to remain perfectly planar upon binding, resulting in an ~6-fold improve incorporation efficiency of Acd into sfGFP in E. coli.

These results should help provide strategies for expanding the ncAA set encodable by the Ma GCE platform. Notably, we observed a subtle but important difference in the backbone positioning of the β-sheets that line the Mm and Ma PylRS amino acid binding pockets, resulting in a slightly smaller volume architecture in the Ma PylRS active site. For Acd, de novo selections were effective at overcoming this issue by identifying a constellation of active site residues that compensated for this decreased size. Yokoyama and colleagues also noted the Ma PylRS active site was smaller than the Mb PylRS, but this observation was based on the differences in amino acid side chains lining the active site pocket rather than the positioning of β-sheet backbones lining the active site². They were able to rationally design mutations in the 2^nd shell of the active site to expand its size (H227I/Y228P), though we found these mutations did not improve the incorporation efficiency of aromatic ncAAs (Acd, tetrazine3.0, and 3-nitro-tyrosine) in their respective ASTs (20). Our structural comparisons of the apo Ma PylRS and the Acd/ATP bound Acd-RS1 structures show that these two amino acids point toward solvent in the bound state, and thus provide a rationale for why little effect was observed. Nevertheless, what ncAA size limits the Ma PylRS platform may have, and whether they will be more restrictive than the Mm/Mb PylRS system, has yet to be determined. The crystal structures presented here provide a blueprint for exploring these limits with the design of new libraries.

Moving forward, we found the Ma PylRS to be particularly amenable to structural characterization due to ease of expression and purification, unlike the Mb/Mm RS which is challenging to purify in its active state in large quantities and for which no structures of the full-length enzyme have been reported^14–16. Also, while the functional, full-length Mm PylRS can only be concentrated to ~4 mg/mL, the functional Ma RS protein is much more soluble (easily concentrated to over 30 mg/mL), and this feature was leveraged to improve yields for cell free protein synthesis², and facilitated our structural characterizations here as well as our in vitro kinetic characterizations described in the accompanying manuscript. Thus, the tractability of the Ma PylRS/tRNA^Pyl pair for in vitro studies should prove useful for expanding the utility of this GCE platform, particularly for amino acids that differ from a lysine-like scaffold. Further, to our knowledge, the structures reported here are the first RS structures with a fluorescent ncAA bound, highlighting the necessity of large pockets to accommodate the large, rigid, multi-ring aromatic systems characteristic of fluorescent functional groups.

MATERIALS AND METHODS

Cell lines.

DH10b and BL21ai strains of E. coli were obtained from Invitrogen. The PPY strain used to generate SLiCE extract for cloning was a kind gift from Yongwei Zhang (Albert Einstein College of Medicine).

Molecular Biology reagents.

Oligonucleotide primers and double stranded DNA fragments were synthesized by Integrated DNA Technologies (Coralville, IA). Molecular biology reagents including restriction enzymes and polymerases were purchased either from Thermo Fisher Scientific or New England Biolabs. DNA Miniprep, PCR cleanup and gel extraction kits were purchased from Machery Nagel.

Cloning.

Genes encoding the Acd-AST and Acd-RS1 Ma PylRSs were cloned into a pET28 expression plasmid containing an N-terminal 6x His-tag followed by a cleavable bdSUMO solubility tag using SLiCE¹⁷. All expression plasmids were confirmed by Sanger sequencing prior to protein expression.

M. alvus pyrrolysine synthetase expression and purification.

pET28-His₆-bdSUMO-MaPyl-RS plasmids were chemically transformed into the E. coli strain BL21ai and grown in 2xYT media at 37 °C. The expression was induced at an OD₆₀₀ of ~0.7 with 1 mM IPTG and 0.2% (w/v) arabinose. Expression continued overnight at 20°C. The cells were harvested and lysed in buffer containing 500 mM NaCl, 50 mM Tris-HCl (pH 7.4), and 20 mM imidazole (buffer A). The lysate was spun at 29,000 RCF for 25 minutes to remove cell debris. The supernatant was then applied to a Ni-NTA resin column pre-equilibrated with buffer A. The column was washed with excess (>20 bed volumes) buffer A. Ma PylRS variants were eluted by on-column proteolytic cleavage using 50 nM His-tag free bdSENP1 in a buffer containing 150 mM NaCl, 50 mM Tris, and 20 mM imidazole (pH 7.4)¹⁸. The protein was concentrated and further purified by size-exclusion chromatography using a HiLoad 16/600 Superdex 200 column (Cytiva) equilibrated with 150 mM NaCl, 50 mM Tris (pH 7.4), and 10 mM MgCl₂. The eluted fractions were combined and concentrated with a 10,000 MW concentrator (GE healthcare) to 30–40 mg/mL. The protein was aliquoted, flash frozen in liquid N₂ and stored at −80°C.

Crystallization.

Prior to crystallization, the Ma synthetase proteins were incubated with 2 mM ATP or AMPPNP and 2–5 mM acridone for 1 hr on ice. Crystallization screens were performed with the sitting drop vapor diffusion method, by mixing 0.25 μL of protein solution with 0.25 μL of reservoir solution using an Art Robins Phenix liquid handling robot. Preliminary crystal growth conditions were identified and subsequently optimized using hanging drop vapor diffusion methods. Optimized crystals grew at room temperature by mixing 1 μL of protein solution (15–25 mg/mL) with 1 μL of reservoir solution containing 0.1 M sodium citrate tribasic dihydrate (pH 5.0–6.0), and 18–22% (w/v) PEG 3350. Crystals were soaked in cryo-protecting solution of 0.1 M Na citrate tribasic dihydrate (pH 5.5), 20% PEG 3350, 20% glycerol, 10 mM Acd, and 2 mM ATP or AMPPNP for 5 minutes before being mounted on nylon loops and flash frozen in liquid nitrogen.

Data Collection.

The diffraction datasets were collected at the Lawrence Berkeley National Laboratory Advanced Light Source (Berkeley, California) on beamline 5.0.3 at liquid nitrogen temperatures, and processed by XDS.

Structure Determination and Refinement.

Phases were identified by molecular replacement using Phenix Phaser and the apo structure of Ma PylRS (pdb 6JP2) as the search molecule. Two Ma PylRS molecules were found in each asymmetric unit. Cycles of refinements in Phenix and manual model building in Coot were performed. Final refinement of the AMPPNP bound structure was performed with anisotropic B-factors. Data collection and refinement statistics for each model are collected in Supporting Table 2.

Size exclusion chromatography coupled to multi-angled light scattering.

Experimental molecular weights were obtained by size exclusion chromatography (SEC) using an AKTA FPLC (GE Healthcare) coupled to a DAWN multi-angle light scattering (MALS) and Optilab refractive index system (Wyatt Technology). Size exclusions were conducted on a Superdex200 10/300 GL column (Cytiva Life Sciences) pre-equilibrated in 50 mM Tris pH 7.4, 150 mM NaCl, and 10 mM MgCl₂ at room temperature. Protein samples were prepared at 100 μM and injected at a flow rate of 0.8 mL/min. Duplicate datasets were analyzed using ASTRA software package, version 8 (Wyatt Technology).