BlastP searches were performed using the amino acid sequences of previously characterized TyrA homologs from plants [soybean PDH; GmPDH1 (Schenck et al., 2015) and Arabidopsis ADH; AtADH2 (Rippert and Matringe, 2002)] and microbes [Synechocystis sp. PCC6803 ADH (Legrand et al., 2006), and E. coli PDH (Hudson et al., 1984)] as the query in the NCBI database. This yielded only closely-related plant and microbial TyrA orthologs (e.g., algae and, γ-proteobacteria), which were then used as the query to perform additional BlastP searches. Every 5th BlastP hit was selected to provide sequences from various microbial lineages and limit bias in sample selection. Data S1 contains all the sequence information for the TyrA orthologs used in Figure 2 and Figure S1. A structure-guided amino acid alignment was performed in PROMALS3D (Pei and Grishin, 2007) using the default parameters with structures of TyrA enzymes from plants and microbes with varying substrate specificities (G. max TyrA_p; GmPDH1; PDB # 5T8X, H. influenzae TyrA_p; HiPDH; 2PV7, and Synechocystis sp. PCC6803 TyrA_a; SynADH; PDB # 2F1K). The amino acid alignment from PROMALS3D was used to construct phylogenetic trees using MEGA7 (Kumar et al., 2016). The full amino acid alignment can be found in Data S2. The analyses involved 130 amino acid sequences and all sites with <75% coverage were eliminated from the analysis. A neighbor-joining method (Figure S1A; Saitou and Nei, 1987) was used to estimate evolutionary history using 1,000 bootstrap replicates (values shown at branches). The tree in Figure 2 is a representative tree using a subset of the sequences found in Figure S1. Additional phylogenetic analyses were performed using the Maximum Likelihood method based on the Jones-Taylor-Thornton (JTT) matrix-based model (Jones et al., 1992), which gave overall similar results (Figure S1B). All phylogenetic trees are drawn to scale, with branch lengths measured in the number of substitutions per site.

Full length coding sequences from Ochrobactrum intermedium LMG 3301 (EEQ93947.1; OiTyrA), Sediminispirochaeta smaragdinae DSM 11293 (ADK80640.1; SsTyrA), and Methanosaeta harundinacea (KUK94425.1; MhTyrA) were codon optimized for expression in E. coli, gene synthesized (Biomatik), and inserted into pET28a vector using EcoR1 and Nde1 sites in frame with an N-terminal 6x-His tag using a previously described cloning method (Schenck et al., 2015).

For site directed mutagenesis, plasmid template was diluted 100-fold, mixed with 0.04 U/μL Phusion DNA polymerase (Thermo), 0.2 mM dNTP's, 0.5 μM forward and reverse mutagenesis primers and 1x Phusion reaction buffer (Thermo), and then placed in a thermocycler for 98°C for 30 s followed by 20 cycles of 10 s at 98°C, 20 s at 70°C, 4.5 min at 72°C with a final extension at 72°C for 10 min. The sequence of primers used for mutagenesis were (5′-CATTCTGGCCGAAAGCCCGGAACTGTATAGTAGC-3′) and (5′-GTTCCGGGCTTTCGGCCAGAATGCGGCCCACAAAATC-3′) for MhTyrA and (5′-GTAACAATCCACTTCAGCTGTTTATAGACTTGCAAC-3′) and (5′-CTGAAGTGGATTGTTACACGTTTGTTCGCGCACCTG-3′) for SsTyrA (mutated codons are italicized). The PCR products were purified with QIAquick Gel Extraction Kit (Qiagen), treated with DpnI (Thermo) to digest methylated template DNA for 30 min at 37°C, and then transformed into E. coli XL1-Blue cells. Plasmids were sequenced to confirm that no errors were introduced during PCR and cloning.

For recombinant protein expression, E. coli Rosetta2 (DE3) cells (Novagen) transformed with the above plasmids were cultured as previously reported (Schenck et al., 2017). For protein purification, 20 mL of the E. coli supernatant expressing the appropriate plasmid was applied to a 1 mL HisTrap FF column for purification of the His-tagged recombinant protein using an ÄKTA FPLC system (GE Healthcare). After loading the supernatant, the column was washed with 20 column volumes of 90% buffer A (0.5 M NaCl, 0.2 M sodium phosphate, and 20 mM imidazole) and 10% buffer B (0.5 M NaCl, 0.2 M sodium phosphate, and 0.5 M imidazole) followed by elution with 100% buffer B. Fractions containing purified recombinant enzymes were pooled and desalted by Sephadex G50 column (GE Healthcare) size-exclusion chromatography into lysis buffer (Schenck et al., 2017). The purity of purified proteins were analyzed by SDS-PAGE using ImageJ software (Schneider et al., 2012). All protein purification steps were performed at 4°C unless stated otherwise.

ADH and PDH assays were performed using purified recombinant enzymes for SsTyrA Wild-type (Wt) and D208N mutant, and MhTyrA Wt and Q227E mutant, while the E. coli cell lysate was used for OiTyrA as expression and purification of this enzyme was unsuccessful. Reactions contained 0.8 mM substrate (arogenate or prephenate) and 0.8 mM cofactor (NADP⁺ or NAD⁺) together with reaction buffer [25 mM HEPES pH 7.6, 50 mM KCl, 10% (v/v) ethylene glycol]. For OiTyrA assays containing cell lysates, reactions were incubated for 45 min and analyzed using HPLC as previously reported (Schenck et al., 2015). For pure enzymes, reactions were monitored every 10–15 s for reduced cofactor at A_340nm using a microplate reader (Tecan Genios) in a reaction volume of 30 μL. Kinetic parameters of purified recombinant enzymes were determined from assays containing varying concentrations of arogenate (39.1 μM−5 mM) or prephenate (46.9 μM−6 mM) substrate, 0.8 mM of the preferred cofactor and monitored every 10–15 s for production of reduced cofactor at A_340nm using a microplate reader (Tecan Genios). Kinetic parameters were determined by fitting initial velocity data to the Michaelis–Menten equation using Origin software (OriginLab) from technical replicate assays (n = 3). Arogenate substrate was prepared by enzymatic conversion of prephenate (Sigma-Aldrich) as previously reported (Maeda et al., 2010). Enzyme assays were quantified using the A_340nm of a standard curve of reduced cofactor (NADPH or NADH) and activity is expressed as Kat/mg (moles of product produced per second per mg protein). All enzyme assays were conducted at a reaction time and protein concentration that were in the linear range and proportional to reaction velocity.

Homology models were made using SWISS-MODEL (Biasini et al., 2014) with default parameters to predict the structures of divergent TyrA enzymes. Enzymes that are more closely-related to plants (e.g., SsTyrA and MhTyrA) were modeled using the GmPDH1 structure as the template, though this resulted in a poor model for BdTyrA, which falls within the outgroup. BdTyrA was additionally modeled using Synechocystis sp. PCC6803 ADH. Homology models were visualized using PyMOL.

Previous studies suggested that plant TyrAs are not derived from an eukaryotic ancestor or through cyanobacterial endosymbiosis because they are most similar to other microbes including some proteobacteria (Bonner et al., 2008; Reyes-Prieto and Moustafa, 2012; Dornfeld et al., 2014; Schenck et al., 2017); however, their precise origin was unclear. To resolve the phylogenetic relationship of TyrA orthologs from divergent organisms including plants and microbes, here we performed structure-guided phylogenetic analyses using PROMALS3D to achieve alignment of TyrA orthologs with low sequence similarities (see section Materials and Methods; Pei and Grishin, 2007). Three distinct clades were identified that contain: plant TyrAs together with those from algae, spirochaetes, α- and δ-proteobacteria (clade I, shaded blue in Figure 2, Figure S1), TyrA orthologs from some archaea, fungi, γ-proteobacteria, and chloroflexi (clade II, shaded green), and TyrA orthologs from various microbes, which formed the outgroup and contains previously characterized microbial TyrA orthologs from Synechocystis sp. PCC 6803 and Aquifex aeolicus having very low sequence similarity (~30%) to plant TyrAs (clade III, Figure 2, Figure S1). Interestingly, TyrAs from some spirochaetes lineages (some of which are known to cause harmful human diseases like Lyme disease; Pritt et al., 2016) formed a subclade with plant and algae TyrAs within clade I using various phylogenetic methods (Figure 2, Figure S1). These data suggest that Plantae TyrA may have been acquired through horizontal gene transfer (HGT) from an ancestor of one of these closely-related microbes.

The amino acid sequence alignment of TyrAs showed that the Asp222 residue, which is conserved across plant TyrA_a (Schenck et al., 2017) was also highly conserved in clade I (Figure 2). On the other hand, most sequences in clade II, including some archaea TyrA, have a non-acidic Gln residue at the corresponding 222 position (Figure 2), similar to legume TyrA_p enzymes (Schenck et al., 2017). Homology models of representative TyrA from clade I—Arabidopsis thaliana ADH (AtADH2, Plantea; Rippert and Matringe, 2002) and S. smaragdinae DSM 11293 (SsTyrA, spirochaetes)—and clade II—M. harundinacea (MhTyrA, archaea)—generated using GmPDH1 structure as the template indeed showed that their acidic and non-acidic residues, respectively, correspond to Asp222 in the active site of plant TyrA (Figure S2). These data together suggest that TyrAs from clade I are likely arogenate-specific TyrA_a enzymes, whereas more distantly-related microbial TyrAs from clade II are likely prephenate-specific TyrA_p enzymes.

To experimentally test if TyrAs from clade I have ADH activity, representative TyrA orthologs from two distinct subclades of clade I, spirochaetes (SsTyrA) and α-proteobacteria (O. intermedium; OiTyrA, Figure 2, Figure S1), were expressed in E. coli as recombinant enzymes and biochemically characterized. SsTyrA and OiTyrA were chosen as they are located at key phylogenetic boundaries within clade I and contain residues required for cofactor binding and catalysis (Figure S2, Data S2). Purified SsTyrA recombinant enzyme showed ADH activity with a slight preference for NAD⁺ over NADP⁺ cofactor; however, PDH activity was not detectable (Figure 3A). Similarly, the E. coli cell lysate expressing OiTyrA had ADH but not PDH activity and strongly preferred NAD⁺ over NADP⁺ cofactor (Figure 3B), although the purification of OiTyrA was not successful due to low expression. These results demonstrate that microbial TyrA orthologs from clade I, which contain an acidic residue at the corresponding 222 position (Figure 2, Data S2), are arogenate-specific TyrA_a enzymes.

To test if TyrA orthologs from clade II, which contain a non-acidic residue at the corresponding 222 position, are prephenate specific TyrA_p enzymes, a representative archaeon TyrA from M. harundinacea (MhTyrA) was biochemically characterized. MhTyrA was chosen as no TyrAs from its subclade of clade II have previously been characterized (Figure 2). Also, MhTyrA is a monofunctional enzyme, while some archaea, fungi, and γ-proteobacteria orthologs in clade II are bifunctional and have a chorismate mutase enzyme domain (Hudson et al., 1984; Shlaifer et al., 2017). MhTyrA was expressed in E. coli and the recombinant enzyme was purified to homogeneity using affinity-chromatography (Figure S3) and used for biochemical analyses. Unlike plant and microbial TyrA_a orthologs from clade I, MhTyrA showed strong PDH and very weak ADH activity (Figure 3C). Interestingly, MhTyrA strongly preferred NADP⁺ over NAD⁺ cofactor (Figure 3C), like plant TyrAs (Gaines et al., 1982; Connelly and Conn, 1986). These results suggest that TyrA orthologs from clade II that have a non-acidic residue at the corresponding 222 position are TyrA_p enzymes that strongly prefer prephenate over arogenate substrate.

To test if the acidic residue at the corresponding 222 position in spirochaetes SsTyrA_a is involved in its substrate specificity, the D208N mutation that converts the active site acidic residue into a non-acidic residue was introduced in SsTyrA. The purified recombinant SsTyrA_a D208N enzyme (Figure S3) drastically decreased its original ADH activity (Figure 4A), but now exhibited PDH activity (Figure 4B), with a slight alteration in cofactor specificity in the mutant compared to Wt (Figure S4).

Kinetic analyses showed that SsTyrA_a Wt enzyme did not accept prephenate and had a K_m for arogenate of 901 μM (Table 1), which is substantially higher than previously reported plant and microbial TyrA enzymes (Rippert and Matringe, 2002; Bonvin et al., 2006; Ku et al., 2010; Schenck et al., 2015). The SsTyrA_a D208N mutant exhibited a seven-fold lower catalytic efficiency (k_cat/K_m) with arogenate than Wt, though the K_m was not altered (Table 1). SsTyrA_a D208N showed significant PDH activity that was absent in Wt, though its K_m was still much higher than previously reported TyrA enzymes (Rippert and Matringe, 2002; Bonvin et al., 2006; Ku et al., 2010; Schenck et al., 2015) and had poor catalytic efficiency (Table 1). These results suggest that a single mutation of the active site acidic residue to a non-acidic residue can alter the substrate specificity of spirochaetes TyrA, similar to plant TyrAs (Schenck et al., 2017).

To test if the non-acidic residue of MhTyrA_p at the corresponding 222 position (Gln227) is involved in prephenate substrate specificity, site-directed mutagenesis was performed on MhTyrA_p to replace Gln227 with acidic Glu and generate the MhTyrA_p Q227E mutant. The purified recombinant MhTyrA_p Q227E enzyme (Figure S3) showed decreased PDH activity (Figure 4D) with a substantial gain of ADH activity (Figure 4C, Table 1) without altering cofactor preference (Figure S4).

Further kinetic analyses showed that Wt MhTyrA_p had a K_m of 378 μM and turnover rate (k_cat) of 2.4 s⁻¹ using prephenate substrate and NADP⁺ cofactor (Figure 4, Table 1), which are comparable to previously characterized microbial TyrA_p enzymes (Bonvin et al., 2006; Ku et al., 2010). The very weak ADH activity of MhTyrA_p Wt (Figure 4, Table 1) precluded it from kinetic analysis using arogenate.

The Q227E mutant, on the other hand, exhibited almost 10-fold reduction in K_m for prephenate (2.7 mM), while the catalytic efficiency (k_cat/K_m) was reduced by 60-fold (0.1 vs. 6.4 mM⁻¹ s⁻¹, Figure 4D, Table 1). The Q227E mutant displayed substantial ADH activity compared to the Wt enzyme with a K_m for arogenate of 3.3 mM, similar to that of Q227E for prephenate (2.7 mM, Figure 4C, Table 1) though still 10-fold higher than that of the Wt enzyme for prephenate (Figure 4D, Table 1) and other previously characterized TyrA_a enzymes (Bonvin et al., 2006; Ku et al., 2010; Schenck et al., 2015, 2017). The Q227E mutant had roughly two-fold higher catalytic efficiency with arogenate than with prephenate (0.2 vs. 0.1 mM⁻¹ s⁻¹, Figure 3). These results demonstrate that the Q227E mutation can shift the substrate preference of MhTyrA_p from prephenate to arogenate, suggesting that the single residue is responsible for substrate specificity of archaea TyrA_p enzymes.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.