Directed Evolution of Aerotolerance in Sulfide-Dependent Thiazole Synthases

Abstract

Sulfide-dependent THI4 thiazole synthases could potentially be used to replace plant cysteine-dependent suicide THI4s, whose high protein turnover rates make thiamin synthesis exceptionally energy-expensive. However, sulfide-dependent THI4s are anaerobic or microoxic enzymes and hence unadapted to the aerobic conditions in plants; they are also slow enzymes (k_cat < 1 h^–1). To improve aerotolerance and activity, we applied continuous directed evolution under aerobic conditions in the yeast OrthoRep system to two sulfide-dependent bacterial THI4s. Seven beneficial single mutations were identified, of which five lie in the active-site cleft predicted by structural modeling and two recapitulate features of naturally aerotolerant THI4s. That single mutations gave substantial improvements suggests that further advance under selection will be possible by stacking mutations. This proof-of-concept study established that the performance of sulfide-dependent THI4s in aerobic conditions is evolvable and, more generally, that yeast OrthoRep provides a plant-like bridge to adapt nonplant enzymes to work better in plants.

Introduction

Plants and fungi synthesize the thiazole moiety of thiamin using a suicide THI4 thiazole synthase (EC 2.4.2.60) in which an active-site Cys residue donates the required sulfur atom; this irreversibly inactivates the THI4 after just one reaction cycle (Figure 1).^1,2 The energy cost of replacing an inactivated plant THI4 is equivalent to a biomass yield penalty of up to 2–4%.³ Exchanging a Cys-dependent suicide THI4 for a catalytic THI4 (EC 2.4.2.59) that mediates multiple reaction cycles using sulfide as sulfur donor (Figure 1)^2,4 could therefore potentially increase biomass yield, making this a rational synthetic biology strategy for crop improvement.^3,5 However, there are obstacles to doing this. Sulfide-dependent THI4s, which have an Fe(II) cofactor,⁴ come from organisms whose environments are anoxic or deeply hypoxic, sulfide-rich, and in some cases, hot (≥60 °C)^4,6−8 and so are ill-adapted to plants.⁸ One challenge is thus to modify sulfide-dependent THI4s to work well in air (vs. < 1% oxygen), low-μM sulfide⁹ (vs. high-μM to mM⁸), and mild temperatures. Another challenge is that sulfide-dependent THI4s are slow enzymes: Methanococcus jannaschii THI4 mediated only two catalytic cycles in vitro in 3.5 h⁴ (i.e., k_cat < 1 h^–1, a catalytic turnover rate so low that it precludes biochemical characterization). Similarly, while sulfide-dependent THI4s complemented an Escherichia coli thiazole auxotroph in aerobic conditions,⁸ the THI4 activity per unit protein needed for this was very low since the THI4s were ∼30% of soluble protein,⁹ and even suicide THI4s complement E. coli if expressed at such levels.^2,10

Figure 1.

Biosynthesis of adenylated carboxythiazole (ADP-thiazole) by cysteine-dependent and sulfide-dependent THI4s. ADP-thiazole is formed from NAD⁺, glycine, and a sulfur atom. Cysteine-dependent THI4s take the sulfur atom from an active-site cysteine residue, leaving a dehydroalanine residue and causing inactivation. Sulfide-dependent THI4s use sulfide as sulfur donor.

This study aimed to adapt sulfide-dependent THI4s to aerobic conditions using continuous directed evolution (CDE), which has unmatched capacity for mutational depth and scale.^11,12 In CDE, the enzyme gene is hypermutated in vivo, the enzyme’s activity is coupled to host cell growth, and improved variants are obtained by selecting for faster growth.¹² We chose the yeast (Saccharomyces cerevisiae) OrthoRep CDE system because (i) it durably mutates a target gene at ∼10⁵-fold the natural rate;^13,14 (ii) sulfide-dependent prokaryote THI4s complement a yeast thiazole auxotroph;⁸ and (iii) yeast is a closer facsimile of plants than Escherichia coli,¹⁵ the main alternative CDE platform.¹² OrthoRep uses two linear plasmids (p1 and p2) from Kluyveromyces lactis, where the target gene to be evolved is cloned into p1 and mutations are introduced by an error-prone DNA polymerase (TP-DNAP1) that is specific to p1 and borne on a nuclear plasmid.¹³ To avoid overtaxing the initially low activity of the target THI4s, we supported their function by using a host strain (met15Δ) with a high internal sulfide level.¹⁶ Since cytosolic conditions in aerobic yeast, as in plants, differ from those in anaerobic bacteria with respect to, e.g., NADH/NAD⁺ ratio¹⁷⁻¹⁹ and protein folding and degradation systems,¹⁵ selection in yeast might improve adaptation to such factors as well as to oxygen. Structure modeling and comparative genomics were applied to distinguish these possibilities.

Results and Discussion

Target Choice and Prescreening

We chose three bacterial THI4s that have detectable activity under aerobic, high-sulfide conditions, as shown by complementation of an E. coli ΔthiG thiazole auxotroph.⁸ These were: MhTHI4 from Mucinivorans hirudinis (leech gut mesophile); SfTHI4 from Saccharicrinis fermentans (marine mud mesophile); and TaTHI4 from Thermovibrio ammonificans (deep-sea vent thermophile). S. fermentans and T. ammonificans can be classed as microaerobes.⁸ As OrthoRep expresses target genes at relatively low levels due to low copy number and a low-to-moderate strength promoter,¹² we prescreened MhTHI4 and SfTHI4 for the ability to complement a yeast met15Δ thi4Δ strain when more highly expressed from a CEN6/ARS4 plasmid, as previously done for TaTHI4.⁸ (If complementation fails this prescreen it will fail in OrthoRep.) The THI4s were tested plus or minus yeast THI4’s 28 N-terminal residues, a weakly predicted mitochondrial targeting peptide for which proteomic data provide no support.²⁰ All three THI4s complemented and so went forward to OrthoRep. The yeast N-terminus had little effect on complementation (Figure S1) and was not added in further experiments.

CDE in OrthoRep

We transformed a met15Δ thi4Δ strain with the nuclear ArEc-TDH3 plasmid harboring the error-prone polymerase TP-DNAP1_611, then introduced plasmids p1 and p2 by protoplast fusion.¹⁵ p1 harbors the THI4 target gene that the error-prone polymerase hypermutates (Figure 2A). We ran three independent clones of each THI4 through three selection schemes:¹⁵ 1, serial passages on thiamin-free medium; 2, initial passages on medium with limiting thiamin, then transfer to thiamin-free medium; and 3, initial passages on medium with luxury thiamin (to build a mutant library), then transfer to scheme 2 (Figure 2B–D). Six MhTHI4 and three SfTHI4 populations showed growth in one or more schemes; growth improved with passaging until it neared that of a control whose p1 harbored yeast THI4 (Figure 2B–D and Figure S2). As this control marks the upper limit of the selection window, we ended campaigns at this point and sequenced individual clones from each population. No TaTHI4 population showed any growth, possibly because this thermophilic THI4 had too little activity at 30 °C.

Figure 2.

Continuous directed evolution of MhTHI4 and SfTHI4 in OrthoRep. (A) Overview of Ortho-Rep. An error-prone version of the terminal protein-primed DNA polymerase (TP-DNAP1) of the p1 plasmid is transferred to a nuclear plasmid with a HIS3 selection marker in strain BY4741 thi4Δ met15Δ his3Δ leu2Δ ura3Δ. TP-DNAP1 on p1 is replaced by the MhTHI4 or SfTHI4 gene and a LEU2 marker. The error-prone polymerase replicates p1, mutating the THI4. (B) Selection scheme 1. MhTHI4 and SfTHI4 were evolved without thiamin for up to 21 passages. Growth (OD₆₀₀) of 3 mL cultures was monitored and is shown for passages 1, 6, and 14. (C) Selection scheme 2. MhTHI4 and SfTHI4 were evolved for eight passages in 10 nM thiamin, then for 12 passages without thiamin. Growth is shown for passages 1 and 8 (10 nM thiamin) and passage 14 (no thiamin). (D) Selection scheme 3. MhTHI4 and SfTHI4 were grown for 17 passages in 300 nM thiamin, in 10 nM thiamin for seven passages, and without thiamin for eight passages. Growth is shown for passages 25 and 29 (no thiamin). All data are means ± s.e.m. of 2–3 replicates.

CDE Outcomes

Sixteen variants carrying one or two nonsynonymous ORF mutations were recovered, mostly from MhTHI4 (Figure 3A), as well as variants with synonymous mutations and promoter mutations (Figure S3). ORF mutations were 75% T > C, 15% A > G, and 10% G > A, which are similar to those reported.¹³ Four single MhTHI4 mutations recurred in independent populations: V124A in three, and V28A, Y122C, and V138A in two each. Likewise, the SfTHI4 mutation D168G recurred in three populations. Such convergent evolution of mutations implies fitness benefits. To check that benefits were due to mutations in the THI4 ORF and not elsewhere in p1 or the yeast genome, each mutant ORF was recloned into p1, inserted into fresh cells, and growth in thiamin-free medium was measured for independent clones. Wildtype MhTHI4 or SfTHI4 and yeast THI4 served as benchmarks to gauge improvement (Figure 3B and Figure S4). In interpreting these checks, account should be taken of two built-in sources of variability in the OrthoRep system: the error-prone DNA polymerase itself (which enables populations to improve while experiments are running) and the initial, temporary presence of a low, variable level of wild-type p1 plasmids carrying the wildtype DNA polymerase.¹³ These intrinsic sources of variability within and between experiments are particularly at play among wildtype THI4s, which sporadically enable growth to start within the ten-day test period. Taking MhTH4 and SfTHI4 together, seven mutations were beneficial, four were neutral, and four were deleterious, i.e., acted alone or canceled a beneficial mutation (Figure 3B). Consistent with expectation, all convergent mutations were beneficial. The deleterious mutations probably arose after beneficial ones had partly swept the population (Figure 3A) and survived because the cells harboring them were cross-fed thiamin by thiamin-producers in the rest of the population.¹⁵ Because sulfide-dependent THI4s are labile and have extremely low k_cat values,^4,8 the biochemical characteristics of wildtype and evolved enzymes cannot be compared using in vitro assays. We therefore turned to comparative genomics and structure modeling to interpret the observed mutations.

Figure 3.

Mutations in six evolved populations of MhTHI4 and three of SfTHI4. (A) Promoter (10B2) and ORF mutations. V124A and D168G each occurred in three populations; V28A, Y96H, Y122C, and V138A each occurred in two populations. (B) Classification of mutations based on growth relative to that supported by wildtype MhTHI4 or SfTHI4 and yeast THI4 as benchmarks. Beneficial (positive) mutations grew approximately as well as the yeast benchmark. Deleterious (negative) mutations prevented growth when alone or when combined with a beneficial mutation. Neutral mutations had no measurable effect. Double mutants are classified according to how the second mutation (black) affected the first positive mutation (gray). (C) Natural variation in residues at mutation sites. Numbers are frequencies (%) of naturally occurring amino acids at each site among 199 diverse sulfide-dependent THI4s;⁸ the most frequent amino acid is in red. White spaces indicate 0% frequency. The observed mutations are highlighted in beige.

Mutations in Relation to Natural Variation

We assessed natural variation among residues at the mutated positions using a set of 199 diverse sulfide-dependent THI4s⁸ (Figure 3C). Of the seven beneficial mutations, all except Y122C and D168G are conservative or semiconservative, and all except Y122C occur naturally at frequencies of 1–22%. The four neutral mutations are also conservative and occur naturally. In contrast, two of the deleterious mutations are nonconservative and none are natural. The outcomes of the CDE campaigns are thus broadly similar to the outcomes of natural selection.

Structure Modeling

The mutations’ positions were mapped onto the modeled structure of MhTHI4 (which is very similar to that of SfTHI4) (Figure 4A). The five convergent, beneficial mutations (V28A, Y122C, V124A, V138A, and D168G) are in the active-site cleft, near the adenine ring of the bound intermediate. The other two beneficial mutations, A106T and M157V, are in or near the dimer interface. Two of the four deleterious mutations (I171T and V151A) are in or near the active-site and two (I198T and S229P) are on the protein surface. The neutral mutations are all in or near the dimer interface or on the surface. Interestingly, beneficial mutations V28A, V124A, and V138A line a hydrophobic pocket via which oxygen might access the Fe(II) center. Replacing any of them with a smaller Ala residue could allow closer packing that impedes oxygen access and thus lessens damage to the catalytic metal. Another possibility is that the beneficial mutations near the bound intermediate reduce discrimination between the NAD⁺ substrate and NADH, which might compete with NAD⁺ for binding at the active-site. Such discrimination may be less necessary when MhTHI4 and SfTHI4 are expressed in yeast instead of their anaerobic native hosts, whose cytosolic NADH/NAD⁺ ratios are likely at least 10-fold that in yeast.^17,18

Figure 4.

Positions of beneficial, neutral, and deleterious mutations within the modeled MhTHI4 structure, and the association of the natural V124A variant with oxygen-tolerance. (A) The observed mutations mapped onto the structure of MhTHI4, modeled from the crystal structure of Thermovibrio ammonificans THI4 (PDB: 7RK0). Two subunits of the octamer are shown (subunit A is gray, subunit B is cyan) with mutations indicated in blue (beneficial), gold (neutral, double), khaki (neutral, single), and red (deleterious). The bound substrate is indicated in purple and the metal ion in orange. (B) The occurrence of V124 and A124 variants in natural THI4 sequences in relation to the oxygen adaptation of the host organism, based on habitat data and on the types of cytochrome encoded by the genome, i.e., aa₃-type and cbb₃-type cytochrome c oxidases and cytochrome bd, which are adapted to operate at progressively lower oxygen levels.²¹ Habitat oxygen is shown schematically on a gray scale, from zero (white) to air-level (black). See Figure S5 for a larger panel of oxygen metabolism genes and more detailed information.

Comparative Genomics of Beneficial Mutations

Comparative genomics suggests rationales for three of the beneficial mutations. The first is V124A, a fairly common natural variant (Figure 3C) that comparative genomics implicates in oxygen-tolerance, based on the presence or absence of genes for oxygen-metabolizing enzymes (Figure S5). Most simply, if A at position 124 confers more oxygen tolerance than V, then V124 should be prevalent in low-oxygen organisms and A124 in high-oxygen ones. This is observed (Figure 4B and Figure S5). Most A124-THI4s are from organisms of hypoxic or microoxic habitats whose genomes encode aa₃-type and cbb₃-type cytochrome c oxidases, which are adapted to high and low oxygen levels, respectively.²¹ Conversely, most V124-THI4s are from anaerobes that lack cytochrome c oxidases and whose sole oxygen-dependent terminal oxidase is cytochrome bd, which is adapted to ultralow oxygen levels and may serve to scavenge trace oxygen.²¹

The non-natural MhTHI4 Y122C mutation might also improve oxygen tolerance. Changing Y122 to C creates, with C121, a CC motif near the Fe(II) center that mimics one in the aerobic suicide THI4 of Ustilago trichophora and other fungi. As Cys residues can provide sacrificial thiol groups to protect critical sites from oxidants,²² the Y122C mutation might shield the Fe(II) center from oxidation. An alternative rationale is that a small hydrophobic residue is preferred at position 122 (Figure 3C), and Cys is the only such residue accessible from the Y122 codon (TAC) by an OrthoRep point mutation.¹³

The SfTHI4 D168G mutation could confer a benefit unrelated to oxygen. D168 is ∼10 Å from the Fe(II) center and is followed by C. DC motifs can be hotspots for spontaneous isomerization of aspartate to isoaspartate, which alters the peptide backbone structure and can reduce enzyme activity.²³ Like other Gram-negative bacteria as well as plants and animals, S. fermentans (the source of SfTHI4) has the protein isoaspartyl methyltransferase (PIMT) that repairs isoaspartyl residues; yeast does not.²⁴ Thus, despite being a radical mutation, D168G could benefit yeast, but not necessarily other organisms, by reducing isomerization damage that yeast cannot repair.

Performance of Beneficial Mutations in E. coli

Given that THI4s cannot be biochemically characterized due to extremely low specific activity and in vitro instability,⁴ we turned instead to physiological characterization in E. coli as an orthogonal platform. We first tested the above three mutations (i.e., those with strongly beneficial effects that could be rationalized) for complementation activity in a thiazole auxotroph (ΔthiG) cultured in 3 mL tubes. Consistent with the above inference about aspartate isomerization, the D168G variant was less active than wildtype SfTHI4 in E. coli, which has PIMT (Figure S6), i.e., the D168G mutation was deleterious rather than beneficial in a host that repairs isoaspartyl residues. The non-natural Y122C mutation was also deleterious in E. coli whereas the V124A mutation was apparently neutral (Figure S6), i.e., conferred no growth advantage in E. coli in 3 mL tube culture conditions. That the V124A mutant did not outperform the wildtype could be because shake cultures of E. coli differ in oxygen tension from those of yeast.²⁵

Performance of the V124A Mutation in Yeast at Different Shaking Speeds

To compare the oxygen tolerance of wildtype and V124A mutant MhTHI4, we took advantage of the effect of shaking speed on aeration.^25,26 At higher aeration (240 rpm, the speed used in the directed evolution campaigns), the V124A mutant performed as expected, i.e., far better than the wildtype and about as well as the yeast THI4 benchmark (Figure S7). At lower aeration (180 rpm), all cultures grew more poorly as expected.²⁵ The growth of the yeast THI4 benchmark was modestly slowed; growth of wildtype MhTHI4 was severely slowed, whereas growth of the V124A mutant stopped completely, i.e., the V124A mutation lost its advantage and possibly became deleterious (Figure S7). This result is consistent with the inference that the V124A mutation increases oxygen tolerance drawn from the comparative genomic and structure modeling evidence above. A deleterious effect of V124A at lower oxygen as well as a beneficial one at higher oxygen would also fit with the natural occurrence patterns of V or A at position 124, i.e., A is underrepresented in organisms from extremely oxygen-poor environments (Figure S5). More generally, a change from plus to minus in the effect of V124A would conform to the principle that a beneficial mutation in one environment can be deleterious in a different environment, not just neutral.²⁷

Deleterious Mutations

Because they are common and a priori unlikely to specifically affect oxygen tolerance, detrimental mutations were not examined in depth. Two nevertheless stood out: I198T and V151A (Figures 3B and 4A). The I198T mutation abolished activity when present alone in SfTHI4 and when combined with the beneficial Y122C mutation in MhTHI4. As I198 is a surface residue that interacts with the adjacent monomer’s N-terminus, replacement by Thr could impact dimerization. The V151A mutation in MhTHI4 abrogated the benefit of V124A and abolished activity. Because V151 is next to an Fe(II)-liganding Asp,⁸ its replacement by Ala could disrupt metalation.

Conclusions

Our evolution campaigns delivered single mutant THI4s that drove growth rates similar to yeast native THI4, so that only the sequence space within one mutational step of the wildtype sequence could be explored. There is thus much space left to explore to improve activity in aerobic, plant-like conditions. The THI4s evolved so far are a priori likely to be improvable. First, our improved THI4s at best equaled the performance of yeast THI4. This is a low bar: as yeast THI4 is a suicide enzyme¹ with a half-life of ∼9 h,²⁸ improved THI4s need make only a few turnovers per day to match its activity. The median enzyme turnover number (k_cat) is 10⁵-fold more than this.²⁹ Second, we have not yet selected for activity in low-sulfide conditions, or for Co(II) vs Fe(II) as cofactor, which could reduce oxygen-sensitivity.^4,8

Beyond the quest for oxygen-tolerance per se, this study sought proof-of-concept for OrthoRep in two new applications: (i) as a plant-like bridge to evolve microbial enzymes to function better in plants, for which there is an unmet need;^15,30 and (ii) as a platform to increase the number of catalytic cycles that enzymes mediate in vivo before being replaced (catalytic-cycles-until-replacement, CCR³¹). The effectiveness of OrthoRep demonstrated here supports its use in both these applications.

Methods

Plasmid Construction

THI4 cloning in E. coli strain DH10B/TOP10 was as previously described.^14,15 Bacterial THI4s were codon-optimized for yeast and synthesized by GenScript (Piscataway, NJ); the recoded sequences are given in Table S1. Native yeast THI4 was not recoded. The recoded bacterial THI4s were used with or without addition of the N-terminal 28 codons of yeast THI4 plus a GG linker (encoding MSATSTATSTSASQLHLNSTPVTHCLSDGG). Plasmids used in this study are listed in Table S2. Primers were purchased from Eurofins Genomics (Louisville, KY) and are listed in Table S3. Enzymes for PCR and cloning were obtained from Thermo Fisher Scientific (Miami, FL).

Yeast Strains and Media

Strains are listed in Table S4. Yeast was grown aerobically in YPD or selection media as described.¹⁵ Selection media were synthetic complete (SC) minus the amino acids used as selection markers, e.g., yeast GA-Y319 containing p1_THI4 was grown in SC-Leu and yeast BY4741 his3Δ leu2Δ met15Δ ura3Δ thi4Δ containing p1_THI4 and the error-prone TP-DNAP was grown in SC -His -Leu -Trp. Before cloning THI4s in OrthoRep, complementation in BY4741 thi4Δ was confirmed by expression in ArEC-TDH3.¹⁵ Due to its met15Δ deletion, the BY4741 his3Δ leu2Δ met15Δ ura3Δ thi4Δ strain has an elevated sulfide level.¹⁶

Yeast Transformation and Protoplast Fusion

Transformations were performed as done previously¹⁵ with minor modifications. ScaI-digested GR-306MP harboring ScTHI4, MhTHI4, or SfTHI4 was added to GA-Y319 competent cells with 10 μL of 11 mg/mL ssDNA and 600 μL of an 8:1:1 50% PEG:10× TE:1 M lithium acetate mixture and incubated for 45 min at 30 °C. Following an additional 20 min incubation at 42 °C, the DNA-cell mixture was centrifuged (5 min, 700g), resuspended in 1 mL of sterile Milli-Q water, and plated on SC -Leu. Plates were incubated at 30 °C for up to 5 days, and single colonies were selected for gDNA isolation and sequencing to confirm integration of the THI4 gene and the leucine selection marker in the p1 plasmid.¹⁵ Sequence-verified p1_THI4 in GA-Y319 was used as the donor strain for protoplast fusion and BY4741 thi4Δ as recipient. Protoplast fusions were made as before.¹⁵

Evolution Campaigns

Campaigns used the three schemes described previously¹⁵ and shown in Figure 2B–D. Two MhTHI4 populations were evolved by scheme 1, three MhTHI4 populations and two SfTHI4 populations were evolved by scheme 2, and one population each of MhThi4 and SfTHI4 was evolved by scheme 3. Cells were washed five times with thiamin-free medium when transitioning from thiamin-containing to thiamin-free medium. Cultures were started at OD₆₀₀ = 0.05. Cultures were analyzed for mutations by sequencing amplicons containing the THI4 ORF and promoter. Mutant THI4s were recloned in p1, introduced into fresh cells, and validated as described¹⁵ and outlined in the text.

Performance Tests in E. coli

Wildtype and variant MhTHI4 and SfTHI4 sequences (recoded for E. coli(8)) were cloned in pBAD24 and used to complement an MG1655 ΔthiG strain; the medium was MOPS medium³² plus 0.2% glycerol and specified supplements. Pilot experiments defined the l-arabinose concentrations that induced expression levels of MhTHI4 and SfTHI4 (30 and 10 μM, respectively) that gave growth rates approximately half those when expression was maximally induced with 1 mM l-arabinose. Single colonies were inoculated in 3 mL of medium containing 100 nM thiamin; after overnight incubation the cultures were used to inoculate (0.02 OD₆₀₀) 3 mL of medium containing 1 mM l-arabinose and 1 mM cysteine. After 24 h, these cultures were used to inoculate 3 mL of cultures containing 30 or 10 μM l-arabinose and 1 mM cysteine, which were in turn used to inoculate cultures containing the same medium; OD₆₀₀ was then monitored.

Comparative Genomics

Distributions of V124-THI4s and A124-THI4s and oxygen metabolism genes were analyzed using GenBank complete genomes and metagenome assembled genomes and Blastp. The query sequences used for oxygen metabolism genes are given in the legend of Figure S5.

Structure Modeling

The MhTHI4 model was generated using Swiss-Model (Expasy) with the TaTHI4 structure as template (PDB: 7RK0). The homology model had a global model quality estimate of 0.86.³³ Structure graphics were made using Chimera 1.15.³⁴

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssynbio.2c00512.

• Figure S1. THI4 prescreening in the CEN6/ARS4 plasmid. Figure S2. Growth of populations harboring MhTHI4 or SfTHI4 during evolution campaigns. Figure S3. Nonsynonymous and synonymous mutations in MhTHI4 and SfTHI4. Figure S4. Complementation by MhTHI4 and SfTHI4 mutants in yeast. Figure S5. Association of the A124 variant in natural THI4 sequences with oxygen adaptation of the host organism. Figure S6. Performance tests in an E. coli ΔthiG strain of wildtype MhTHI4 and SfTHI4 and mutants selected for improved performance in yeast OrthoRep. Figure S7. Responses to aeration (shaking speed) of yeast cultures harboring the MhTHI4 wildtype (WT) or V124A mutant sequence. Table S1. Sequences of Recoded Bacterial THI4 Genes Used in This Study. Table S2. Plasmids Used to Clone into OrthoRep. Table S3. Primers Used to Clone and Sequence ScTHI4, MhTHI4, and SfTHI4. Table S4. Yeast Strains Used in OrthoRep (PDF)

Author Contributions

A.D.H. devised the study. K.V.G., J.D.G.-G., and E.R.O.-F. ran experiments; K.V.G., Y.H., and S.D.B. built structure models. All authors analyzed data. A.D.H. and K.V.G. wrote the manuscript.

The authors declare no competing financial interest.