Kinetic compartmentalization by unnatural reaction for itaconate production

Physical compartmentalization of metabolism using membranous organelles in eukaryotes is helpful for chemical biosynthesis to ensure the availability of substrates from competitive metabolic reactions. Bacterial hosts lack such a membranous system, which is one of the major limitations for efficient metabolic engineering. Here, we employ kinetic compartmentalization with the introduction of an unnatural enzymatic reaction by an engineered enzyme as an alternative strategy to enable substrate availability from competitive reactions through kinetic isolation of metabolic pathways. As a proof of concept, we kinetically isolate the itaconate synthetic pathway from the tricarboxylic acid cycle in Escherichia coli, which is natively separated by mitochondrial membranes in Aspergillus terreus. Specifically, 2-methylcitrate dehydratase is engineered to alternatively catalyze citrate and kinetically secure cis-aconitate for efficient production using a high-throughput screening system. Itaconate production can be significantly improved with kinetic compartmentalization and its strategy has the potential to be widely applicable.


Background
Diverse intracellular biochemical reactions have evolved to e ciently supply energy and synthesize essential metabolites required for organism survival [1][2][3][4] . For example, highly orchestrated enzymatic reactions involving substrate channeling by multienzyme complex or consecutive enzyme reactions, have evolved to rapidly consume intermediates, avoiding the formation of byproducts that undermine the e ciency of a desired pathway [5][6][7][8][9] . However, such an e cient metabolic reaction chain may limit metabolic engineering when the newly introduced metabolic pathway must use the metabolic intermediate as a substrate, as the accessibility of the intermediate is limited by kinetic competition with the native reaction chain 10 . For example, in itaconate biosynthesis, substrate availability can limit the biotechnological production of valuable compounds 11 . Itaconate is a dicarboxylic acid used in the resin and plastic industry 12 and can be synthesized from a decarboxylation reaction of cis-aconitate by cisaconitate decarboxylase 11,13 , with cis-aconitate as an intermediate molecule that is transiently generated in the tricarboxylic acid (TCA) cycle via the enzyme aconitase, which converts citrate into isocitrate 8 . In addition, the higher a nity for cis-aconitate leads aconitase to rapidly convert this compound into isocitrate 8 , reducing the cis-aconitate availability for itaconate biosynthesis catalyzed by cis-aconitate decarboxylase 11,13,14 .
One solution used by nature to avoid competing reactions is spatial compartmentalization 13,15,16 , which utilizes the physical separation of an intermediate to block kinetic competition between diverse chemical reactions 13,16 . For example, Aspergillus terreus, a native itaconate producer, facilitates spatial compartmentalization of cis-aconitate by pumping out mitochondrial cis-aconitate into the cytosol by exchanging cytosolic oxaloacetate via an antiporter, MttA (Supplementary Fig. 1) 13 . This physical separation allows the fungus to accumulate cis-aconitate in the cytosol and e ciently produce itaconate by avoiding competition with the TCA cycle. Further, itaconate production can be e ciently regulated via MttA activity 17 . However, because of the insu ciently implemented genetic engineering tools and complicated culture conditions of A. terreus, research focused on itaconate production has been performed in well-known workhorse cells such as Escherichia coli 11,14 . Mimicking the spatial compartmentalization in E. coli would be an e cient approach for itaconate production; however, prokaryotic hosts lack a membranous system, leading to challenges in building a compartmentalized system 11,14,15 .
To overcome the inapplicability of spatial compartmentalization in bacteria, we propose kinetic compartmentalization for kinetically separating competitive reactions by releasing the intermediate from the native reaction chain without creating a physical barrier. Particularly, we kinetically compartmentalized the itaconate production reaction from the TCA cycle by introducing a non-natural enzyme that can separate consecutive biochemical reactions catalyzed by aconitase. We hypothesized that introducing a non-natural biochemical reaction into E. coli cells, which can effectively synthesize cisaconitate from citrate, would increase the intracellular cis-aconitate level, thus triggering its conversion into itaconate through catalysis by cis-aconitate decarboxylase.
In this study, we successfully engineered an endogenous 2-methylcitrate dehydratase, PrpD 18 , to catalyze the conversion of citrate into cis-aconitate. PrpD was evolutionarily engineered to have high speci city toward citrate by computational simulation design and high-throughput screening [19][20][21] . Based on the non-natural enzyme, we constructed a novel itaconate production pathway that was kinetically isolated from the TCA cycle (Fig. 1). Using our approach, we obtained a signi cantly increased itaconate yield of up to 9.81-fold compared to the parental strain at 43.0% of the theoretical maximum. This kinetic compartmentalization concept allows bacteria to overcome the limitation of being a spatially inseparable host by introducing a new biochemical reaction to e ciently synthesize an intermediate molecule. This method shows potential for kinetic compartmentalization in prokaryotes, considering the substantial number of promiscuous enzymes that can be engineered using this approach.

Selection of target enzyme for compartmentalization
To separate cis-aconitate production from the TCA cycle, we rst reviewed the detailed reaction mechanisms of cis-aconitate synthesis in E. coli. The precursor of itaconate, cis-aconitate, can be synthesized by aconitase encoded by acnA and acnB in E. coli. Dehydration of citrate is sequentially coupled to rehydration of cis-aconitate by the same enzymes, which appear to catalyze the isomerization of citrate to isocitrate. Thus, cis-aconitate was temporarily synthesized as a reaction intermediate ( Supplementary Fig. 2a). Furthermore, the catalytic e ciency (k cat /K m ) of aconitase is much higher for cis-aconitate than for citrate 8 . Consequently, cis-aconitate accumulates at very low levels in native E. coli 11,14 , resulting in low itaconate production. Aconitase may only conduct a dehydration reaction. However, the dehydration and rehydration reactions are coupled with two main catalytic residues, H444 and S244 22 . H444 functions as a proton donor, whereas S244 acts as a proton acceptor in the rst dehydration reaction 22 , after which these residues switch roles. In the rehydration reaction, H444 acts as a proton acceptor and S244 functions as a proton donor 22 . After these two reactions, the catalytic residues return to the initial state to conduct another enzyme reaction. Thus, it was di cult to isolate the dehydration reaction.
Thus, we screened enzymes with dehydration activity; their known substrates are similar to those of citrate structures. Among all enzyme candidates, the endogenous enzyme of E. coli, PrpD, converts 2methylcitrate into 2-methyl-cis-aconitate in a one-step dehydration reaction without rehydration to 2methylisocitrate ( Supplementary Fig. 2b) 18 . This enzyme has a much lower a nity and catalytic e ciency for citrate and cis-aconitate compared to aconitases 18 . Engineering PrpD to alternatively catalyze citrate may be a useful strategy for enhancing the pool of cis-aconitate for itaconate production.

Development of itaconate-speci c screening system
Recently, a itaconate-responsive LysR-type transcription factor, ItcR, from Yersinia pseudotuberculosis was shown to regulate the expression of the itaconate degradation pathway based on the presence and amount of itaconate 23 . We exploited ItcR and its cognate promoter, P ccl , to construct an itaconateresponsive screening system. The screening system was designed to regulate the expression of the antibiotic resistance gene 24,25 according to itaconate concentration. Speci cally, ItcR was constitutively expressed using a synthetic promoter (P BBa_J23106 ), and the tetracycline resistance gene (tetA) was controlled under P ccl . Collectively, the system was intended to provide a growth advantage under tetracycline pressure in environments with high itaconate concentrations or in high-producing strains (Fig.  2a).
To validate the screening system, itaconate-responsive screening system was transformed into an acidtolerant E. coli W strain 14 to produce the WS strain (Supplementary Table 1). We demonstrated the effectiveness of the screening system and further improvement in itaconate production using a previously developed itaconate-producing E. coli strain with acetate as the sole carbon source 14 . The tetracycline concentration as the selection pressure was varied up to 15 mg/L and growth retardation were con rmed in accordance with the increased tetracycline concentration, indicating tight regulation of ItcR/P ccl (Fig. 2b). In addition, the gradual increase in the speci c growth rate was validated by extracellular addition of itaconate under selection pressure. Speci cally, the speci c growth rate increased as the concentration of itaconate increased to 2 g/L. Additionally, the growth rate according to itaconate showed different tendencies depending on the tetracycline selection pressure (Fig. 2b). Collectively, these results indicate that the itaconate-responsive screening system was successfully constructed and is widely applicable for itaconate production.
Screening PrpD mutant with altered substrate speci city To select enzyme variants that can produce cis-aconitate, PrpD protein engineering was conducted using the itaconate-responsive screening system. Target residues for mutagenesis were selected based on structural analysis of tartrate-bound MmgE (PDB code: 5MUX), the homologue of PrpD derived from Bacillus subtilis 26 (Supplementary table 2). To enhance the catalytic activity toward citrate rather than 2methylcitrate, which has no methyl group on the second carbon, residues near the methyl group position of 2-methylcitrate were selected as targets: W110, G111, and I331. A mutant library of PrpD (W110, G111, and I311 residues) was constructed for expression in moderate strength (BBa_P J23108 ) based on structural analysis (see Methods) with theoretical 20 3 variants numbers in size and transformed into the WAICS strain.
To increase the intracellular level of citrate and ensure the activity of PrpD mutants, the catalytic e ciency of the competing enzyme, AcnB, was decreased through site-directed mutagenesis (AcnB W482R , Table 1) 27 . Mutagenesis of AcnB e ciently lowered the catalytic e ciency to citrate by 3.75-fold (Table  1). Notably, the glyoxylate shunt pathway was activated to increase anaplerosis and enhance itaconate production by inactivating the IclR-encoding transcriptional repressor 14,28 . The resulting WAIC strain with AcnB W482R showed a lower cell biomass (2.16 g DCW/L) and itaconate production (0.26 g/L) by 1.17and 1.22-fold, respectively, compared to the WCI strain ( Supplementary Fig. 3). Nevertheless, the accumulation of citrate was signi cantly enhanced by up to 0.51 g/L, as expected. Overall, an environment was created in which the PrpD mutant with altered substrate speci city produced increased itaconate and showed a su cient growth advantage with the screening system. Screening was conducted by increasing the selection pressure from 7-15 mg/L of tetracycline over four rounds.

Characterization of enriched PrpD mutants
After enrichment, 10 isolated mutants were analyzed, and ve types of mutants were characterized (Supplementary Table 3 Supplementary Fig. 4). The WAICP VTL strain (PrpD VTL with W110V, G111T, and I331L) showed a 1.50-fold increase in itaconate production, whereas most mutants showed a decreased level of itaconate production compared to the WAICP strain with wild-type PrpD.
An additional culture was conducted at the ask-scale to validate the WAICP VTL strain ( Fig. 3a and 3b).
Interestingly, the cell biomass was remarkably reduced in the WAICP VTL strain (Fig. 3b) compared to that of the WAICP strain (Fig. 3a) by 1.80-fold (1.01 g DCW/L), and citrate accumulation was reduced by 2.00fold (0.15 g/L), similar to that of the WCI strain ( Fig. 3b and Supplementary Fig. 3a). These results indicate that PrpD VTL exhibits increased catalytic e ciency toward the citrate to produce cis-aconitate, which induced kinetic separation and redirected more citrate for use in itaconate production. Indeed, the WAICP VTL strain showed a 2.56-fold increase in itaconate production to 0.86 g/L compared to the WAICP strain. Moreover, the itaconate yield was signi cantly enhanced in the WAICP VTL strain by up to 4.76-fold (0.27 g/g), indicating that kinetic compartmentalization strategy is a powerful approach for enhancing the cis-aconitate availability and itaconate production.
To validate the effect of each residue on PrpD VTL , mutants at single residues and a combination of double mutants were generated as the WAICP V , WAICP T , WAICP L , and WAICP TL strains (Supplementary Table 1); these mutants showed no signi cant difference in itaconate production compared to the WAICP strain ( Supplementary Fig. 5). In contrast, the WAICP VT and WAICP VL strains showed 3.55-(1.19 g/L) and 3.84-fold (1.28 g/L) increases in the itaconate titer at 48 h ( Supplementary Fig. 5 and Fig. 3c), respectively, which are even higher than that of WAICP VTL (Fig. 3b). These results agree with those of structural analysis, as W110 is the closest residue among the target residues to the methyl group of the second carbon on 2-methylcitrate (Supplementary Table 2), combined with the rationale for the initial multi-residue library design.

Kinetic and structural analysis of PrpD mutants
The enzyme kinetics of wild-type PrpD, PrpD VTL , PrpD VT , and PrpD VL were characterized to evaluate the enhancement of itaconate production ( Table 2). As demonstrated previously, wild-type PrpD shows promiscuity for catalyzing citrate 18 . However, this enzyme showed 12.21-fold lower catalytic e ciency compared to AcnB (Tables 1 and 2), supporting the slight decrease in citrate and increase in itaconate of the WAICP strain (Fig. 3a) compared to the WAIC strain ( Supplementary Fig. 3b). In contrast, the higher a nity for citrate (15.81 mM of K m ) was con rmed in PrpD VL compared to wild-type PrpD (66.39 mM), as expected. However, the mutant displayed a 1.35-fold lower turnover rate (k cat ) for citrate (7,804.72 s -1 ), improving the catalytic e ciency by 3.11-fold (Table 2).
A noticeable decrease in the a nity for cis-aconitate was also observed in PrpD VL (1.52 mM of K m ) compared to wild-type PrpD (0.70 mM, Table 2). The change in the binding pocket following mutagenesis may have altered the a nity for cis-aconitate and citrate 25,29 . In addition, the turnover rate of PrpD VL was reduced (96.34 s -1 ), leading to a 2.68-fold decrease in catalytic e ciency. Collectively, these results indicate that PrpD VL e ciently extracted citrate from the TCA ux and induced kinetic compartmentalization by converting it into cis-aconitate to facilitate itaconate production (Fig. 1).
Computational simulations of 2-methylcitrate into PrpD enzyme showed that in the wild-type, single mutated (PrpD V , PrpD T , and PrpD L ), and double mutant (PrpD TL ) enzymes, the methyl group faced the residues W110 and G111 (wild type or mutated), and the conformation shifted in the opposite direction ( Supplementary Fig. 6). Moreover, docking simulations revealed a slight decrease in hydrogen bonds between 2-methylcitrate and PrpD residues in the double mutants PrpD VT and PrpD VL and triple mutant PrpD VTL compared to simulation into the active pocket of wild-type (Table 3). In contrast, docking predictions of citrate interactions showed an increased number of hydrogen bonds between this ligand and PrpD residues in all mutants compared to in the wild-type enzyme ( Supplementary Fig. 7, Table 3).
Interestingly, mutants showing a higher substrate speci city for citrate (PrpD VT , PrpD VL , and PrpD VTL ) exhibited a greater increase in hydrogen bonds between citrate and PrpD residues, with W110V potentially acting as the most important residue determining the substrate shift, which is supported by the fact that mutating this single amino acid increased the number of hydrogen bonds, as for the previously mentioned double and triple mutants (Table 3). Finally, docking on the double mutant PrpD TL gave the same amount of hydrogen bonds with 2-methylcitrate and citrate, in accordance with the data showing the lowest enzymatic activity of this mutant towards citrate.

Further ux optimization for increasing production
The WAICP VL strain was further optimized. First, PrpD VL expression was optimized by employing constitutive promoters (Supplementary Fig. 8) 30 . Itaconate production was signi cantly affected by PrpDVL expression, as expected, indicating that e cient kinetic compartmentalization can allow metabolic ux to be regulated by changing the activity of PrpD VL in a manner similar to spatial compartmentalization in A. terreus, where the precise ux distribution is regulated by MttA activity. The WAICP100 VL strain with the highest PrpD VL expression showed the highest itaconate production (1.35 g/L) after 48 h of cultivation. Next, the TCA cycle and glyoxylate shunt were additionally activated by overexpression of citrate synthase and isocitrate lyase encoded by gltA and aceA for further ux ampli cation and to facilitate the anaplerotic reaction to maximize itaconate production, respectively 14,31 ( Fig. 4a and b). The expression of aceA was varied to determine the optimized ux distribution between the glyoxylate shunt and TCA cycle 14,31 . The itaconate titer was increased along with increased expression of aceA (Fig. 4a), and the WAICPG5 strain with the highest expression level of aceA showed a 1.52-fold increase in itaconate production (2.03 g/L) compared to the WAICP100 VL strain after 48 h of cultivation. Finally, itaconate production reached up to 3.13 g/L at 96 h; the yield was maintained at a signi cant level (0.31 g/g, 43.0% of theoretical maximum yield) throughout cultivation and was driven by the substantially increased acetate uptake rate (Fig. 4b) 14 . These results indicate that in addition to the strongly kinetically compartmentalized ux toward the itaconate, the ux distribution could be optimized for itaconate production.

Discussion
In nature, e ciently designed spatial compartmentalization systems have been introduced to preserve substrate availability 13 or prevent damage caused by toxic intermediates 15,32 . Naturally assembled proteinaceous organelles including carboxysomes 33 and metabolosomes 34 have been detected even in some prokaryotes. However, in most prokaryotes, simultaneous reactions involving numerous substrates within a single space are determined only by the kinetic properties of the enzymes, which have evolved to be well-coordinated to maximize cell growth with high precision. Based on these characteristics, we achieved kinetic compartmentalization in prokaryotes by developing a non-natural enzymatic reaction. Given that around 40-50% of enzymes with known functions have multiple substrates and 10-20% of these multi-substrate speci c enzymes can mediate consecutive reactions ( Supplementary Fig. 9), our approach can be applied to other consecutive reactions to enable metabolic engineering of unreachable intermediates.
Numerous studies of the heterologous production of itaconate have consistently focused on substrate availability 11,14,35 but could not imitate or recapitulate the spatial compartmentalization strategy of its native producer, A. terreus, pumping out cis-aconitate into an independent space 13 . Therefore, we introduced kinetic compartmentalization to provide an e cient supply of cis-aconitate. Rather than using existing aconitase which is highly reactive to cis-aconitate, the promiscuous enzyme PrpD was semi-rationally engineered and applied in itaconate production. Especially, we applied the itaconate production from acetate, which is a non-preferred carbon source, and generally, intensive engineering should be essential for its e cient conversion 14 . The e cient itaconate production using PrpD mutant is also expected to be widely applicable to production systems from glucose 11 or glycerol 36 .
In addition to the modi ed catalytic characteristic of PrpD, the non-natural cis-aconitate synthesis reaction can be regulated by the expression level of the PrpD mutant. Generally, the itaconate titer was increased when a stronger promoter was used to express PrpD VL ( Supplementary Fig. 8), indicating that more carbon ux was kinetically compartmentalized according to PrpD VL expression. Surprisingly, the resulting strain WAICPG5 produced up to 3.13 g/L itaconate at 43.0% of the theoretical maximum yield from acetate (Fig. 4b), which showed a 9.81-fold increase in yield compared to the parental strain (WCI strain, Table 1), indicating the e cient kinetic compartmentalization.
An itaconate-responsive screening system was successfully constructed in this study. The initial library was constructed to su ciently cover all combinations; however, while determining the effect of each residue, double mutants with more desired characteristics were identi ed. In addition, the enriched population clearly contained mutants that improved itaconate production; however, not all mutants showed the desired phenotype. After the PrpD VTL mutant dominates during the initial enrichment, hitchhikers may have been enriched together through crosstalk effects 37 . Nonetheless, we screened for effective mutant in four rounds of enrichment and observed an unintended decrease in catalytic e ciency toward cis-aconitate, supporting the effectiveness of the high-throughput screening system 21,24,25,38,39 . This method can be further applied in diverse strategies for itaconate production, including adaptive laboratory evolution, the evolution of cis-aconitate decarboxylase, and even the de novo itaconate production pathway.

Oligonucleotides and reagents
Oligonucleotides were obtained from Cosmo Genetech (Seoul, South Korea). For routine DNA manipulation, Takara PrimeSTAR™ HS DNA Polymerase (Shiga, Japan) was used for DNA ampli cation. Plasmid DNA and genomic DNA were extracted using the AccuPrep R Nano-Plus Plasmid Mini Extraction Kit (Bioneer, Daejeon, Korea) and GeneAll R Exgene TM Cell SV Kit (GeneAll, Seoul, Korea), respectively. Enzymes, including ligase and restriction enzymes, and Gibson assembly were purchased from New England Biolabs (Ipswich, MA, USA). Other chemical reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA).

Bacterial strains and plasmids
All bacterial strains and plasmids used in this study are listed in Supplementary Table   1. Escherichia coli Mach1-T1 R (Thermo Fisher Scienti c, Waltham, MA, USA) was utilized as a cloning host, and the acid-tolerant E. coli W strain was utilized as an expression host 14 and source of wildtype prpD. The WA and WAI strains were constructed using the Lambda-Red recombination system with plasmids pM_FKF, pKD46, and pCP20 40 . The pCDF_CAD plasmid was used to construct pCAD, and variants of the pCOGA plasmid (pCOGA, pCOGA1,3,4,5) were used to amplify the glyoxylate shunt from our previous study 14 . The taconate-responsive screening system, a codon-optimized itcR fragment from Y. pseudotuberculosis, was synthesized and assembled with ampli ed pETduet-1 and tetA fragments 23,24 . pPRPD was constructed by assembling ampli ed pACYCduet-1 and prpD from E. coli W.
Genetic variants of PrpD were obtained by site-directed mutagenesis and the Gibson assembly method 41 . The terminators and promoters for all vectors were obtained from the Registry of Standard Biological Parts (http://parts.igem.org). Synthetic 5′ untranslated regions were computationally designed using UTR Designer (http://sbi.postech.ac.kr/utr_designer) 42 .
For itaconate production, single colonies of each strain were inoculated into 3 mL of medium in a 15 mL test tube. After 12 h, the cell cultures were inoculated into 3 or 20 mL of fresh medium in the test tube or 300-mL Erlenmeyer asks, respectively, to an optical density at 600 nm (OD 600 ) of 0.05. Isopropyl-β-Dthiogalactopyranoside was initially added to a nal concentration of 0.1 mM for induction. Cultures were performed in biological triplicate with continuous shaking (200 rpm) at 30℃ 14 . The pH was adjusted to 7.0 by adding an appropriate amount of 5 M HCl solution. Culture samples were periodically collected and stored at −80℃ for analysis.
To validate the screening system and enrichment for screening PrpD mutants, we exerted selection pressure using tetracycline. The concentration of tetracycline was varied from 0-50 mg/L to determine the initial selection pressure. The toxicity of itaconate was determined using various concentrations of itaconate from 0-2 g/L. The mutant library was initially enriched with 7 mg/L tetracycline and increased to 15 mg/L over four rounds of enrichment. All cultures except for the enrichment culture were conducted in triplicate.

Analytical methods to detect cellular metabolites
Cell biomass (OD 600 ) was measured using a UV-1700 spectrophotometer (Shimadzu, Kyoto, Japan), and the dry cell weight (DCW) was calculated by converting 1 unit of OD 600 to 0.31 g/L 43 .
Metabolites were measured using an Ultimate 3000 high-performance liquid chromatography system (Dionex, Sunnyvale, CA, USA). Filtered samples were analyzed using an Aminex HPX-87H column (Bio-Rad Laboratories, Hercules, CA, USA). As the mobile phase, 5 mM H 2 SO 4 was used at a ow rate of 0.6 mL/min; the temperature of the column oven was maintained at 14℃ 28 . The refractive index and absorbance at a UV wavelength of 210 nm were monitored using a Shodex RI-101 detector (Shodex, Klokkerfaldet, Denmark) and variable wavelength detector (Dionex).

Characterization of enzyme kinetics
Cells were cultivated for 9 h after induction, and cell pellets were resuspended in 40 mM Tris-HCl buffer (pH 8.0). The cells were lysed using a Qsonica sonicator (Sonics & Materials, Newtown, CT, USA) for 3 min. To avoid oxidation of an iron-sulfur cluster of enzymes, dithiothreitol and (NH 4 ) 2 Fe(SO 4 ) 2 were added to nal concentrations of 2.5 and 0.25 mM, respectively 8 . The cell lysates were centrifuged for 10 min at 13,000 x g at 4℃. The supernatants were utilized to purify the 6X His-tagged enzymes with a MagListo TM His-tagged protein puri cation kit (Bioneer) under anaerobic conditions to prevent inactivation of aconitase activity 8 . The elutes were treated with 1 mM dithiothreitol, 0.14 mM (NH 4 ) 2 Fe(SO 4 ) 2 , and 0.12 mM Na 2 S to prevent oxidization of iron-sulfur clusters. The amount of puri ed enzyme was quanti ed and adjusted to the same amount using the Bradford assay.
The enzyme assay was conducted in 2 mM Tris-HCl buffer with varying amounts of substrate: 1, 2, 5, 10, and 20 mM for citrate; 0.05, 0.1, 0.2, 0.5, and 1.0 mM for cis-aconitate. This reaction was conducted at 37℃ for 10 min followed by inactivation of the enzyme at 90°C for 5 min. The samples were analyzed using a high-performance liquid chromatography system with an XTerra R RP18 column. H 2 SO 4 (5 mM) was utilized as the mobile phase at a ow rate of 0.6 mL/min. Absorbance was monitored at a UV wavelength of 215 nm. All assays were conducted in triplicate.

Structural analysis for PrpD mutant library design
To identify the catalytic sites of PrpD, a structural model of PrpD was rst generated using the structure of the homologous protein MmgE (PDB code: 5MUX) 26 from B. subtilis. From the ligand of 5MUX, the catalytic site of MmgE was identi ed, and well-conserved residues were found in both PrpD and MmgE ( Supplementary Fig. 10) 44 . A 3D model of 2-methylcitrate was generated from the downloaded SDF format of (2S,3S)-2-methylcitrate from PubChem. The SDF le was converted to PDB format using the online SMILES translator and structure le generator (https://cactus.nci.nih.gov/translate/). Finally, the 3D model of 2-methylcitrate was aligned with the ligand of 5MUX using LS-Align. From this nal model structure, the distance between the catalytic sites of PrpD and nearby residue from the methyl group of 2methylcitrate was calculated (Supplementary Table 3).
Among the residues in the catalytic site, those key for catalytic reactions and important for the interaction with the carboxyl group of citrates were preserved to maintain catalytic activity. For example, histidine often acts as a proton donor and acceptor, which is critical in the catalytic reaction, and arginine is an open-form salt bridge with the anion form of a carboxyl group. Hydrophobic and small residues near the methyl group of 2-methylcitrate were selected as target residues for the mutant library.

Docking simulations of the mutants
Both the PrpD natural substrate 2-methylcitrate (PubChem ID: 5460420) and citrate (PubChem ID: 31348) were docked into the active pocket of the PrpD enzyme using the crystallographic structure of apo-protein 2-methylcitrate dehydratase from E. coli (PDB: 1SZQ) 26 using Chimera software 45 . This protein structure included two chains, one of which was deleted before docking simulation. The protein was prepared by eliminating water molecules and adding hydrogen and charge. The ligands were also charged before analysis. As the presence of the methyl group of 2-methylcitrate is an important factor in the substrate speci city of PrpD, the orientation of this group was evaluated in all docking simulations along with the number of hydrogen bonds between the ligand and protein residues resulting from each analysis.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. 211026SupplementaryInformation.docx