Engineering the thermostability of lysine hydroxylase for scalable production of (2S)-hydroxy-1,5-pentanediamine from L-lysine

Abstract

The development of robust enzymes with ideal stability plays a critical role in the construction of practical biotransformation processes. In this work, a native L-lysine hydroxylase from Kineococcus radiotolerans was rationally engineered to improve both thermodynamic and kinetic stability, representing the first case of thermostability engineering of an L-lysine hydroxylase. First, in silico screening based on the calculation of folding free energy (ΔG) was conducted to predict potentially beneficial mutations for stability enhancement in a labor-saving manner. Then, screening of the mutagenesis library revealed key mutations affording improved thermostability without compromising activity. Subsequent combination of the positive mutations enabled further enhancement of enzyme stability with t_1/2 (at 40 °C) up to 10.9 times higher and T_m improved by up to 5 °C as well as retained its activity in comparison with the wild-type. Structural analysis revealed that the stability enhancement could be comprehensively attributed to the change in intramolecular interactions, involving interior hydrophobic interactions, VDW forces, and hydrogen bonding, caused by the key mutations at the hotspots. Importantly, the best mutant was applied to a pilot scale dual-enzymatic cascade reaction (300 L) for production of (2S)-hydroxy-1,5-pentanediamine from L-lysine, which achieved an impressive product titer of 70.1 g L⁻¹ with nearly complete conversion, well demonstrating the great practicality of the developed robust mutant. This study provides useful guiding information for the rational design of native enzymes to efficiently develop thermostable mutants for practical industrial manufacturing.

---

Green foundation

1. Chemical hydroxylation (limited by toxic oxidants, rare metal dependence, and poor selectivity) is replaced with a robust engineered enzyme cascade, enabling industrial biocatalysis using air as an oxidant, water as a solvent, and mild reaction conditions.

2. Engineered L-lysine hydroxylase (10.9× longer half-life at 40 °C, +5 °C T_m) retained its activity and achieved 70.1 g L⁻¹ of target amino alcohol via a 300 L scale cascade reaction with near-complete conversion, eliminating hazardous reagents and protecting-group steps.

3. Future improvements include optimization of enzyme immobilization/reuse for continuous-flow manufacturing, further extension of the catalyst lifetime at higher temperatures, and integration with bio-based downstream processing to enhance overall sustainability and atom economy.

Introduction

Amino alcohols, containing active nucleophilic groups –NH₂ and –OH, are a significant class of structural motifs prevalent in natural and synthetic molecules.^1,2 (2S)-Hydroxy-1,5-pentanediamine (2-OH-PDA), as a new kind of aliphatic amino alcohol, has increasing applications in the pharmaceutical, chemical, and materials industries. In particular, as the structural analogue of 2-(2-aminoethylamino) ethanol, 2-OH-PDA could be used as a potential curing agent or chain extender for the synthesis of various materials, which may exhibit excellent properties such as low toxicity and high reactivity at ambient temperature.³

The introduction of –OH, also known as hydroxylation, is the key step in the preparation of 2-OH-PDA. Currently, hydroxylation via C–H bond activation without any harmful oxidizing reagents is technically difficult in synthetic chemistry and typically suffers from low yields, poor regio- and stereo-selectivity, use of rare transition metals, the need for group protection, and risks of environmental pollution.^4,5 As an alternative, production of 2-OH-PDA via biocatalytic oxidation offers significant green chemistry advantages—high yields with good regio- and stereo-selectivity, avoidance of protection and de-protection steps, and the use of benign oxidants such as air, aqueous conditions, and mild reaction conditions, which makes it a very attractive and green approach. Previous studies have shown that 2-OH-PDA can be produced from low cost and abundant amino acid L-lysine via a dual enzymatic cascade of hydroxylation and decarboxylation, i.e.L-lysine undergoes initial conversion to (2S,3S)-hydroxylysine (3-OH-lysine) catalyzed by L-lysine hydroxylase and then is transformed into 2-OH-PDA by L-lysine decarboxylase.⁶

L-Lysine hydroxylase, as a nonheme iron(II) α-ketoglutarate (αKG)-dependent oxygenase, can activate the C–H bond to yield –OH via oxygenation and is the rate-limiting enzyme in the metabolic synthesis of 3-OH lysine (Fig. 1A).^7–12 Recently, several potential αKG-dependent oxygenases have been identified from bacteria such as Marine actinobacterium, Chitinophaga pinensis, Kineococcus radiotolerans, and Flavobacterium johnsoniae,^13–15 which can catalyze the conversion of L-lysine into hydroxylysines (3-OH-lysine, 4-OH-lysine and 3,4-OH-lysine) with various –OH sites.¹⁰ However, these reported enzymes often showed poor thermostability when the catalytic temperature was above 30 °C.¹⁴ In our previous work, we found that L-lysine 3-hydroxylase from K. radiotolerans (K3H), catalyzing C3 hydroxylation, shows poor thermostability, which greatly hampers its application in practical manufacturing.¹⁰ Hence, improving the thermostability of K3H is highly desirable for the efficient production of 2-OH-PDA.

Fig. 1 (A) Reaction formula of the two-step biocatalytic cascade for the synthesis of 2-OH-PDA from L-lysine. (B) Workflow schematic for the rational design of K3H to enhance its thermostability.

Protein engineering, including directed evolution, (semi)rational design, and ancestral sequence reconstruction, has been proved to be an effective strategy for the improvement of enzyme thermostability (with increases in apparent melting temperature (T_m) in the range of 2–15 °C), which promotes the application of enzyme catalysis in industrial production.^16–19 For instance, loop ring reconstruction was used to improve the half-life (t_1/2) and T_m of a trans-proline 4-hydroxylase,²⁰ and the introduction of disulfide bridges enhanced the thermostability and catalytic activity of an L-isoleucine hydroxylase.²¹ However, the thermostability engineering of K3H has not yet been reported to date.

Herein, to obtain robust enzymes for the practical manufacturing of 2-OH-PDA, we developed a rational design for enzyme K3H to enhance its thermostability (Fig. 1B). In silico analysis based on protein folding free energy ΔG calculations was performed to identify key mutations at hotspots, which enabled the generation of a mutagenesis library and subsequent discovery of variants with thermostability enhancement. Structural modelling, molecular docking, and molecular dynamics simulations were then conducted to unravel the mechanism underlying the thermostability enhancement. Furthermore, the optimal conditions were investigated for biocatalytic transformation of L-lysine into 3-OH-lysine using the best biocatalyst, followed by cascading with L-lysine decarboxylase for 2-OH-PDA production via two-step whole cell catalysis. Finally, the biocatalytic reaction was successfully scaled up to 300 L with the 2-OH-PDA product prepared on a kilogram scale.

Results and discussion

The temperature performance of wild-type L-lysine 3-hydroxylase K3H

Previous studies have shown that the optimum temperature for various Fe(II)/αKG-dependent oxygenases ranged from 15 to 40 °C (Table S3). For instance, the optimum temperature of L-lysine 3-hydroxylase (ABS05421), L-lysine 4-hydroxylase (ABQ06186), and L-lysine 4-hydroxylase (AEV99100) was 40 °C.¹³ The optimal temperature of L-lysine 3-hydroxylase (EAR24255) and L-pipecolate trans-4-hydroxylase (BAU50539) was 15 °C.²² In this study, a screened L-lysine 3-hydroxylase from Kineococcus radiotolerans (K3H) with relatively high reactivity was expressed and purified (Fig. S1). Its optimal temperature was determined as 40 °C, and the activity decreased to 50% at 45 °C and was completely lost at 50 °C (Fig. 2A), which is in agreement with the previous study.²³ Noteworthily, the half-life (t_1/2) of K3H at the optimum temperature (40 °C) was only approximately 35.4 min (Fig. S3A), implying poor thermostability, which could hamper its practical application. Generally, enzymes derived from thermophilic microorganisms tend to exhibit good thermostability as a result of the coevolution of the in vivo enzymes to adapt to a high ambient temperature. Nevertheless, many Fe(II)/αKG-dependent dioxygenases from thermophiles did not present good thermostability.^20,24,25 Recently, computer-assisted protein engineering has established itself as a powerful driving force to enhance the thermostability of enzymes. Loop ring reconstruction, introduction of intramolecular disulfide bridges, and ancestral sequence reconstruction have been successfully applied to elevate the thermostability of Fe(II)/αKG-dependent dioxygenases. For example, Qiao et al. found that the introduction of disulfide bridges into the flexible regions of an L-leucine hydroxylase reduced the folding entropy,²¹ which led to a 10.3-fold increase of t_1/2 at 50 °C. Liu et al. reported that after loop reconstruction and mutagenesis at key sites, the T_m and t_1/2 of a trans-proline 4-hydroxylase were significantly increased to 46 °C and 359 min, respectively.²⁰ Inspired by these, semi-rational design based on structural analysis was performed for K3H to improve its thermostability.

Fig. 2 Alanine-scanning of K3H to identify key active site residues. (A) Activity vs. temperature profile of K3H. (B) Predicted protein structure of K3H. (C) The residues surrounding the two docked molecules within 5 Å. (D) Activity assessment of the alanine-screening variants.

Alanine-scanning for identifying key active site residues of K3H

The AlphaFold 2.0 model of K3H (Fig. 2B) provided a structural blueprint for rational engineering. It confirmed a conserved double-stranded β-helix core²⁶ but, more critically, identified flexible lid regions and loops. Since such loops are known to mediate substrate recognition^12,27 but can confer instability,²³ we hypothesized that engineering these regions would be a strategic path to improve the enzyme's robustness for practical applications.

The predicted protein structure of K3H was docked with two substrates L-lysine and α-ketoglutaric acid using AutoDock Vina, and the complex structure with the lowest binding free energy was selected. To probe the effect of the residues around the substrate-binding pocket on reactivity, the residues within 5 Å surrounding the two small molecules were mutated to alanine for activity examination (Fig. 2C). As a result (Fig. 2D), all 31 mutants displayed reduced activity compared to the wild-type. Two mutants E211A and I303A could afford comparable activity to the wild-type with relative activity levels of 96.4% and 83.1%, respectively. Among the other mutants, 15 mutants retained 10–80% relative activity, while the other 14 mutants were completely inactive, indicating that the above sites are key amino acids for catalysis. Besides, according to previous research studies, a large proportion of mutation sites for stability enhancement were identified out of the layer constituting the active site pocket of the enzyme. Hence, to avoid impairing the reactivity, the sites of alanine-scanning around the active pocket with relative enzyme activity below 80% were not considered in thermostability-enhancement engineering.

In silico screening based on folding free energy calculations

Compared with random mutations and site-directed saturation mutations, computer-aided redesign of protein structure via prediction of beneficial mutations has been used as an efficient and labor-saving methodology to improve the thermostability or activity of various enzymes, such as PETase, transaminase, and carbonyl reductase.²⁸ Importantly, after in silico screening, the identified beneficial amino acid substitutions usually resulted in a synergistic or additive effect on protein thermostability enhancement. To explore beneficial mutations for enhancing the thermostability of K3H, FireProt, a robust computational tool for designing stability-enhanced mutants, was used to predict potential mutation sites based on the constructed 3D model, and the ΔΔG values were calculated using the STABILITY module of FoldX. Consequently, a total of 54 potential mutations (Fig. 3) were finally selected for mutagenesis due to their relatively low values of change in Gibbs free energy (ΔΔG) calculated using FoldX or Rosetta.

Fig. 3 Analysis of potential single mutants based on the change in Gibbs free energy (ΔΔG).

Activity and thermostability screening of single site mutants

The 54 predicted mutants were constructed and expressed, and most of them exhibited good soluble expression levels similar to that of the wild-type. Only mutants A65F, E67R, Q106M, and Q107L showed low soluble expression, which may result from the incorrect folding of the protein. Subsequently, the mutants were purified to characterize their activity and thermostability (Fig. S1).

The results of the activity assay (Fig. 4A) show that 43 of the mutants could maintain ≥80% relative activity. Then, they were incubated at 40 °C for 2 h to determine the residual activity (Fig. 4B). As a result, A20P, A62M, A140F, H191L, T206F, S227P, D243A, S304I, and S304 retained more than 50% of the initial activity, while the wild-type and the other mutants were completely inactive (Fig. S2). To explore the possible synergistic effect on thermostability enhancement, the nine single-site mutants were then subjected to combinatorial mutagenesis.

Fig. 4 Assay of the activity and thermostability of the wild-type and the mutants. (A) Relative activity (%). (B) Change in residual activity at 40 °C. (C) Melting temperature (T_m). (D) Thermal inactivation half-lives (t_1/2) at 40 °C.

Activity and thermostability screening of combinatorial mutants

For the assembly of 9 single-site mutations (derived from 8 mutation sites), a mutagenesis library of 35 possible two-site combinatorial mutants was constructed. Except for mutant A140F/S304I which was expressed in the insoluble form, all the other mutants could be well expressed and were then purified (Fig. S1). Furthermore, the activity assay showed that the activity levels of the combinatorial mutants were similar to that of the wild-type.

The melting temperature (T_m) values of the 9 single-site and 34 combinatorial mutants were measured for thermostability comparison (Fig. 4C). Consequently, the melting temperatures (T_m) of almost all the mutants were higher than that of the wild-type (40.27 °C), with 31 mutants displaying an increase of ≥1.5 °C in T_m. In particular, the T_m values of T206F/S304I (46.06 °C), A62M/S304V (46.13 °C), T206F/S304V (46.43 °C), D243A/S304V (47.55 °C), and A62M/T206F (46.03 °C) were remarkably increased by 5.79 °C, 5.86 °C, 6.16 °C, 7.28 °C, and 5.76 °C, respectively, indicating significantly improved thermostability.

The half-lives (t_1/2) of 10 mutants with significantly improved T_m values were then determined at 40 °C (Fig. 4D). The t_1/2 of the wild-type was determined as 0.59 h (35.4 min), whereas all 10 mutants displayed greatly improved t_1/2 values ranging from 2.40 h to 6.53 h. Noteworthily, the average t_1/2 values of the top four mutants A62M/S304V, T206F/S304V, S227P/S304V, and D243A/S304V were 10.5 fold higher than that of the wild-type.

The kinetic parameters of the wild-type and mutant enzymes were experimentally determined. The K_m, k_cat and k_cat/K_m values of the wild-type were determined as 1.57 mM, 49.74 s⁻¹, and 31.68 s⁻¹ mM⁻¹, respectively (Table 1). The S304V mutant displayed a slightly decreased K_m (1.23 mM) with an improved k_cat/K_m (41.46 s⁻¹ mM⁻¹), which was 31% higher than that of the wild-type. However, D243A/S304I displayed an increased K_m (2.09 mM) with k_cat/K_m (41.46 s⁻¹ mM⁻¹) decreased by 30% in comparison with the wild-type. For the other mutants, k_cat, K_m, and k_cat/K_m were maintained at the same level, suggesting that the mutations did not pose a significant effect on the catalytic efficiency.

Table 1 The kinetic parameters of the wild-type and the mutant enzymes

Mutant	k cat (s^−1)	K m (mM)	k cat/Km (s^−1 mM^−1)	Fold increase
Wild type	49.74 ± 1.57	1.57 ± 0.23	31.68	1.00
S304I	49.11 ± 1.20	1.83 ± 0.19	26.84	0.85
S304V	51.00 ± 1.32	1.23 ± 0.17	41.46	1.31
A62M/S304I	52.89 ± 1.23	1.73 ± 0.19	30.57	0.96
T206F/S304I	50.37 ± 1.20	1.86 ± 0.21	27.08	0.85
D243A/S304I	46.59 ± 1.01	2.09 ± 0.20	22.29	0.70
A20P/S304V	53.52 ± 1.19	1.74 ± 0.18	30.76	0.97
A62M/S304V	51.63 ± 1.21	1.73 ± 0.19	29.85	0.94
T206F/S304V	47.85 ± 1.39	1.46 ± 0.18	32.78	1.03
S227P/S304V	56.04 ± 1.70	1.67 ± 0.26	33.56	1.06
D243A/S304V	52.26 ± 1.32	1.38 ± 0.25	37.87	1.20

---

Structural insights into the enhancement of thermostability

To explore the structural stability and flexibility, the wild-type and the representative mutants (S304V, S304I, A20P/S304V, A62M/S304V, T206F/S304V, S227P/S304V, and D243A/S304V) with highly increased t_1/2 values were chosen for further MD simulations. Consequently, compared with the wild-type, the mutants showed reduced average values of root mean square deviation (RMSD) and root mean square fluctuation (RMSF), indicating that the overall structural rigidity of the mutants was elevated (Fig. S7 and S8).

Further structural analysis to unravel the mechanism underlying the increased thermostability was conducted by investigation of intramolecular interactions including VDW forces, hydrophobic interactions, and hydrogen bond forces as well as solvent accessible surface area (SASA). According to previous research, thermophilic proteins were observed to have higher intramolecular interaction forces than mesophilic ones based on the structural analysis of a large dataset of protein families.²⁹ In this study, single site mutants S304V and S304I displayed significantly increased hydrophobic interactions, VDW forces, and hydrogen bond forces compared to the wild-type (Fig. S9). Val304 (S304V) and Ile304 (S304I) with the substitution of residue S304 with a hydrophobic residue formed a more hydrophobic cavity along with the surrounding hydrophobic amino acids to reduce contact with water (Fig. 5). Hydrophobic interactions were present in mutant enzymes S304V (Val304_Ala104, Val304_Leu129, Val304_Leu172) and S304I (Ile304_Leu103, Ile304_Ala104, Ile304_Leu129, Ile304_Leu172), whereas no hydrophobic interactions were observed between the residues in the wild-type. The hydrophobic effect was widely believed to be important for maintaining the protein's thermodynamic and dynamic stability as well as functional quaternary structure,³⁰ and thus a more hydrophobic cavity required additional free energy to alter protein folding.³¹ Besides, compared with the wild-type, increased VDW force could also be beneficial for the stability enhancement of mutants S304V (Val304_Leu169) and S304I (ILE304_Gly99, ILE304_Leu169).³²

Fig. 5 Location of the hydrophobic cavities at the 304th site of the wild-type (A), S304V (B) and S304I (C) and the overall structural scaffold of the wild-type (D). The hydrophobic cavity is shown in TV-blue, and the mutated residues S304V and S304I are presented as red and green spheres, respectively. The overall structural scaffold of the wild-type contains 11 β strands (magenta strands) and several α helices lying on the protein surface (light blue surface).

Thermostability was further improved by combinatorial mutations (A62M/S304V, T206F/S304V, and D243A/S304V), in comparison with the single mutant S304V. Structural analysis revealed that the backbones of Ala62 in the wild-type and Met62 in mutant A62M/S304V both formed four hydrogen bonds with the surrounding residues Glu58, Val59, Ala65, and Phe66 (Fig. 6A and B). However, the substitution of Ala62 with a sterically more demanding residue Met could enhance the hydrophobic interaction of the residue with residue Phe16 nearby. To probe the effect of mutation T206F on the thermostability of the enzyme, the interactions of the 206th residue in the wild-type and mutant T206F/S304V were inspected (Fig. 6C and D). In the wild-type, residue T206 formed three hydrogen bonds with His246 on a proximate β-sheet, whereas the substitution with a hydrophobic residue Phe206 in mutant T206F/S304V formed only two hydrogen bonds with the surrounding residue, which could be favorable to improve the flexibility of the local structure. As for mutation D243A, a comparison of the structures of the wild-type and mutant D243A/S304V revealed that no hydrogen bonds were observed for residue Asp243 in the wild-type with the surrounding residues, whereas residue Ala243 in the mutant formed one hydrogen bond with Asp241 (Fig. 6E and F). The increased intramolecular hydrophobic interactions and hydrogen bonds by the newly introduced mutations (A62M and D243A) could pose beneficial effects on the thermostability enhancement of the enzyme. In a protein fold, 70% of peptide bond groups and 65% of polar side chains are embedded within the protein, blocking their interaction with exterior water,³³ and buried hydrogen bonds are believed to make large contributions to protein stability as hydrophobic interactions do.³⁴ Huang found that hydrogen bonds may contribute to the thermostability increase of an amine transaminase.¹³ In addition, the substitution of Ala20 located in the N-terminus of the enzyme with proline in A20P/S304V resulted in slightly improved local rigidity (RMSF), compared with the wild-type. Proline has a pyrrolidine ring on the side chain, and the rotation of N-Cα is more restricted, thus contributing to a smaller degree of conformational freedom and an increase in conformational rigidity.³⁵

Fig. 6 Analysis of intramolecular interactions formed at the mutated positions (Ala62, Thr206, and Asp243). (A) Ala62 of the wild-type; (B) Met62 of mutant A62M/S304V; (C) Thr206 of the wild-type; (D) F206 of mutant T206F/S304V; (E) Asp243 of the wild-type; and (F) A243 of mutant D243A/S304V. The hydrogen bonds between atoms are presented as yellow dotted lines with distance numbers (Å).

A high number of non-polar amino acids in contact with a polar solvent is generally not conducive to the structural stability and conformational unfolding of the protein.³⁶ By contrast, the burial of non-polar side-chains could contribute to the formation of more hydrophobic effects for the tight packing in the folded protein interior. Hence, the average solvent accessible surface area (SASA), a factor used to estimate the contribution of the hydrophobic effect to protein folding, was calculated for the wild-type and the mutants over 100 ns simulations (Fig. 7). Consequently, the SASA values of single mutants (S304V, S304I) and combinatorial mutants (A20P/S304V, A62M/S304V, T206F/S304V, S227P/S304V, and D243A/S304V) were maintained at 150–155 nm² and 140–155 nm², respectively, while that of the wild-type was maintained at 155–160 nm². The decrease in SASA indicated an increased packing efficiency and an improved rigidity of the protein structure, which is favorable to maintain the thermostability and catalytic activity of the protein, especially in a high-temperature catalytic environment.³⁷

Fig. 7 Analysis of the SASA (nm²) for the wild-type and mutant enzymes over a 100 ns molecular dynamics simulation. (A) Wild-type; (B) S304I; (C) S304V; (D) A20P/S304V; (E) A62M/S304V; (F) T206F/S304V; (G) S227P/S304V; (H) D243A/S304V.

Comparison of the efficiency of wild-type K3H and mutant D243A/S304V cells for L-lysine hydroxylation

To evaluate the hydroxylation ability of optimal K3H mutant D243A/S304V, the whole cell-catalyzed transformation of L-lysine was, respectively, conducted at 30 °C and 40 °C in a 5 L bioreactor. As shown in Fig. 8A, the catalytic efficiency of wild-type K3H began to deteriorate after 36 h at 30 °C and 6 h at 40 °C, respectively, and the maximum titer reached 45.8 g L⁻¹ (molar yield of 47.2%) at 72 h and 37.2 g L⁻¹ (molar yield of 38.4%) at 42 h respectively. SDS-PAGE confirmed no significant difference in the soluble expression between wild-type K3H and the D243A/S304V mutant (Fig. S10). The main reason for the low concentration and yield could be attributed to the poor thermostability of the wild type K3H enzyme, which makes it difficult for the enzyme to adapt to long-term catalysis.

Fig. 8 Comparison of the catalytic efficiency of wild-type K3H cell and mutant D243A/S304V cell for L-lysine hydroxylation at 30 °C and 40 °C. (A) Time-course of 3-OH-lysine production and L-lysine consumption. (B) The volumetric productivity of 3-OH-lysine production.

By contrast, mutant D243A/S304V maintained high catalytic efficiency, and the L-lysine was almost entirely converted into 3-OH-lysine both at 30 °C and at 40 °C, achieving maximum concentrations of 96.5 g L⁻¹ (molar yield of 99.2%) and 96.7 g L⁻¹ (molar yield of 99.8%), respectively, which were 2.11-fold and 2.59-fold that of wild-type K3H respectively. Although previous studies have reported that 86.1 g L⁻¹ 3-OH-lysine with a molar yield of 88% was obtained at 30 °C by whole-cell catalysis in a 40 mL system,^13,38 the titer and molar yield both are lower in comparison with this study. Noteworthily, the reaction using D243A/S304V at 40 °C reached the maximum titer at a shorter time (48 h) compared to 30 °C (72 h), indicating that improved thermostability is beneficial for the catalytic activity of K3H at its optimal temperature (40 °C). In addition, the productivity of wild-type K3H and D243A/S304V was also compared. As a result, the productivity of D243A/S304V reached 2.08–4.12 g L⁻¹ h⁻¹ from 6 h to 48 h, which were both significantly higher than those of wild-type K3H both at 30 °C and at 40 °C (Fig. 8B).

Large-scale production of 2-OH-PDA using two-step whole-cell catalysis

To investigate the feasibility of scaled-up biocatalytic 2-OH-PDA production, two-step whole-cell catalysis was carried in a 500 L bioreactor. The hydroxylation process of L-lysine using D243A/S304V in the form of whole cells was conducted at 40 °C. The trend of the reaction time-course was similar to that of the aforementioned reaction in the 5 L bioreactor (Fig. 9A and Table 2). Finally, 96.7 g L⁻¹ 3-OH-lysine was obtained from 87.6 g L⁻¹L-lysine, with a nearly 100% molar yield. The intermediate 3-OH-lysine was purified and validated by ¹H NMR (Fig. S11). Notably, the catalytic time (36 h) was slightly shorter compared with that in the 5 L bioreactor (42 h), which might be attributed to the higher oxygen supply in the 500 L bioreactor.

Fig. 9 Two-step whole-cell biocatalytic production of 2-OH-PDA in a 500 L bioreactor. (A) Time-course of 3-OH-lysine and 2-OH-PDA production. (B) HPLC chromatograms of the reaction solution after hydroxylation and decarboxylation.

Table 2 The scaling-up effect on (2S,3S)-OH lysine production

Reactor and catalytic volumes	Initial L-lysine concentration (g L^−1)	Catalytic time (h)	3-OH-lysine concentration (g L^−1)
5 L fermenter (3 L)	87.6 g L^−1	42	96.3 ± 1.2
500 L fermenter (300 L)		36	96.7 ± 0.76

---

For 3-OH-lysine decarboxylation, decarboxylase CadA cells were added after 36 h of hydroxylation. As shown in Fig. 9A and B, the 2-OH-PDA production and 3-OH-lysine consumption were efficient within the first 4 h, and this was followed by a gradual decline in production intensity. The 3-OH-lysine was completely converted into 70.1 g L⁻¹ 2-OH-PDA with 100% yield within 12 h.

The separation and purification of 2-OH-PDA

2-OH-PDA was purified via filtration, extraction, concentration, and molecular distillation, as described in the Experimental section (Fig. S12). Thus, 1.48 kg of 2-OH-PDA was obtained from 30 L of reaction mixture with a total yield of 70.3 wt% after the whole separation and purification process (Table 3). The viscosity and boiling point of the purified product were determined as 438 cP (20 °C) and 216–220 °C, respectively (Table 4). The mass spectra of the product showed two typical peaks at 119.1179 m/z and 141.1038 m/z, which, respectively, correspond to 2-OH-PDA (118.1800 Da) with a proton adduct and a sodium ion adduct (Fig. S13). The purity of the product was analyzed by liquid chromatography (LC) and gas chromatography (GC), both of which showed only one peak, confirming that the product has good purity (98.6%) (Fig. S14 and S15). The chemical structure and purity of 2-OH-PDA were also confirmed by ¹H NMR and ¹³C NMR analyses (Fig. S16).

Table 3 Yield of 2-OH-PDA during the separation process

Separation step	Yield (%)
Filtration	93.1
Extraction	85.3
Concentration	94.7
Molecular distillation	93.5
Total	70.3

---

Table 4 Physical and chemical properties of 2-OH-PDA

Parameter	Value
Viscosity	438 cP (20 °C)
Boiling point	216–220 °C

---

Notably, this biocatalytic process aligns with green chemistry principles, featuring an E-factor of 8.97 (Table S4), aqueous reaction conditions, >95% conversion, and no use of toxic reagents—demonstrating its sustainability for industrial-scale 2-OH-PDA production.³⁹

Experimental section

Gene expression and protein purification

The recombinant E. coli BL21(DE3) strain harboring K3H and CadA were cultured in LB medium containing 50 µg mL⁻¹ kanamycin at 37 °C and 200 rpm. When the cell density (OD₆₀₀) reached 0.6–0.8, IPTG was added at a final concentration of 0.5 mM and subsequently incubated at 25 °C and 200 rpm for 16 h.

For protein purification, the cells were collected, re-suspended in 50 mM PBS (pH 7.0) and disrupted by JY92-IIN ultrasonication (Ningbo Xinzhi Biotechnology, Ltd, Ningbo, China); the cell debris was removed by centrifugation at 8228g for 10 min at 4 °C. Proteins were initially purified using an FPLC system (ÄKTA Pure 150, GE Healthcare Co., Fairfield, USA) with a Ni-NTA affinity column. The proteins were loaded and eluted with 25 mM and 500 mM imidazole buffer, respectively. The eluted fractions were collected and passed through an ultrafiltration tube of 5 kDa (Millipore, USA) to remove the imidazole and concentrated.

Enzymatic activity assay of L-lysine 3-hydroxylase

The assay for K3H activity was carried out in a reaction mixture containing 25 mg L⁻¹ purified protein, 10 mM L-lysine hydrochloride, 15 mM α-ketoglutaric acid, 1 mM vitamin C sodium salt and 1 mM FeSO₄ in 200 µL of PBS (50 mM, pH 7.0) at 30 °C for 10 min. The reaction was terminated by boiling for 10 min to avoid the use of toxic chemical quenchers, leveraging a heat-based denaturation method. Subsequently, the samples were centrifuged at 13 523g for 5 min, filtered through a 0.22 μm membrane, and analyzed by HPLC.²³ One unit of enzyme activity (U) was defined as the amount of enzyme required to produce 1 mM product per minute.

Scale-up production of 2-OH-PDA using two-step whole-cell catalysis

The production of 2-OH-PDA was accomplished through an inherently greener whole-cell catalytic system, which bypasses the resource- and energy-intensive steps of enzyme purification. This two-step process integrates hydroxylation and decarboxylation in a streamlined manner. The high-density fermentation of optimal mutant D243A/S304V and CadA cells was, respectively, conducted in 500 L and 50 L bioreactors, and the conditions were the same as those for the 5 L bioreactor. The recombinant E. coli cells were harvested by tube centrifugation. The hydroxylation reaction of L-lysine using a whole-cell catalyst with a final cell density (OD₆₀₀) of 15 was routinely performed in the 500 L bioreactor containing 300 L reaction volume with an air speed of 0.5 m³ h⁻¹ and 150 rpm at 40 °C. When the hydroxylation finished, CadA cells with a final OD₆₀₀ of 5 and 0.1 mM PLP were added and further reacted for 12 h under sealed conditions to completely convert 3-OH-lysine into 2-OH-PDA.

Separation and purification of 2-OH-PDA

A reaction solution of 2-OH-PDA (300 L) was centrifuged to remove cells and protein using microfiltration and ultrafiltration equipment. The obtained clear solution was concentrated to 100 L by evaporation at reduced pressure. The pH of the concentrated solution was adjusted to 13 using sodium hydroxide for freeing 2-OH-PDA. An equal volume of n-butanol was added to extract 2-OH-PDA twice, followed by collecting the n-butanol phase. Water and n-butanol were removed from the n-butyl-alcohol phase containing 2-OH-PDA using evaporation at reduced pressure, and then the concentrated solution was treated using a molecular distillation device at 170 °C to obtain a light phase (2-OH-PDA product).

Conclusion

In conclusion, the thermostability of a native L-lysine hydroxylase from Kineococcus radiotolerans was improved while maintaining its catalytic activity through virtual screening based on free energy calculations, which to the best of our knowledge is the first case for thermostability engineering of an L-lysine hydroxylase catalyzing the synthesis of 3-OH-lysine. Mutants (A62M/S304V, T206F/S304V, S227P/S304V, and D243A/S304V) exhibited remarkably improved T_m by up to 5 °C. The protein structure models were built and subjected to MD simulations to reveal the molecular mechanism of thermostability improvement. These results demonstrated that the ΔΔG-based rational design is a very effective strategy for the enhancement of enzyme thermostability. The folding free energy decrease suggested that the mutant structure tends to be more thermally stable. Furthermore, structural analysis demonstrated that intramolecular interactions including hydrophobic interactions, hydrogen bonding, and VDW forces were key factors affecting the thermostability at high temperatures. Besides, the whole-cell transformation by recombinant E. coli with the developed mutant is a feasible and highly effective approach for the large-scale production of 2-OH-PDA, affording an impressive product titer of 70.1 g L⁻¹. The research provides important insights into the molecular mechanism of the stability enhancement of Fe(II)/αKG-dependent dioxygenases and is favorable to promote the application of the enzyme class in the future. However, a final T_m of ∼45 °C and the residual activity of L-lysine hydroxylase may still limit its broader industrial adoption; therefore, further optimization for more demanding industrial environments is required in the future.

Author contributions

Alei Zhang: Conceptualization, Methodology, Software, Investigation, Formal analysis, Writing – original draft, Review & editing, Visualization, Funding acquisition. Zhijie Zheng: Conceptualization, Methodology, Software, Investigation, Writing – original draft. Yangyang Li: Data curation, Writing – original draft. Feifei Chen: Writing–review & editing, Visualization, Validation, Supervision. Chaoqiang Wu: Software, Data curation. Kequan Chen: Conceptualization, Supervision, Project administration, Review & editing, Funding acquisition. All authors have given approval to the final version of the manuscript.

Conflicts of interest

The authors declare no competing financial interest.

Data availability

All data supporting this study are included in the article and its supplementary information (SI)Supplementary information: additional experimental details, computational details, analytical methods, general tables and figures, HPLC chromatograms and NMR spectra. See DOI: https://doi.org/10.1039/d5gc04196a.

Acknowledgements

This work was supported by the National Key Research and Development Program of China (2021YFA0911400), the National Natural Science Foundation of China (U21B2097), the Young Elite Scientist Sponsorship Program by China Association for Science and Technology (2024QNRC368). We extend our gratitude to A. Prof. Zedu Huang of Fudan University for his guidance on this paper.

References

1. M. S. Zeng and S. B. Herzon, J. Org. Chem., 2015, 80, 8604–8618.
2. S. C. Bergmeier, Tetrahedron, 2000, 56, 2561–2576.
3. M. Shoaib and A. Bahadur, e-Polym., 2016, 16, 411–418.
4. M. Borrell, S. Gil-Caballero, M. Bietti and M. Costas, ACS Catal., 2020, 10, 4702–4709.
5. A. Palone, G. Casadevall, S. Ruiz-Barragan, A. Call, S. Osuna, M. Bietti and M. Costas, J. Am. Chem. Soc., 2023, 145, 15742–15753.
6. A. Fossey-Jouenne, C. Vergne-Vaxelaire and A. Zaparucha, J. Visualized Exp., 2018, 16, e56926.
7. S. Martinez and R. P. Hausinger, J. Biol. Chem., 2015, 290, 20702–20711.
8. C. Q. Herr and R. P. Hausinger, Trends Biochem. Sci., 2018, 43, 517–532.
9. C. R. Zwick and H. Renata, Nat. Prod. Rep., 2020, 37, 1065–1079.
10. R. Q. Xu, W. H. Jiang, Y. G. Liu, J. M. Hu, D. Liu, S. Zhou, Y. S. Zhong, F. X. Zhang and M. Zhao, Biochem. Biophys. Res. Commun., 2022, 602, 156–162.
11. N. Widderich, A. Höppner, M. Pittelkow, J. Heider, S. H. J. Smits and E. Bremer, PLoS One, 2014, 9, e93809.
12. K. Bastard, T. Isabet, E. A. Stura, P. Legrand and A. Zaparucha, Sci. Rep., 2018, 8, 16587.
13. R. Hara, K. Yamagata, R. Miyake, H. Kawabata, H. Uehara and K. Kino, Appl. Environ. Microbiol., 2017, 83, e00693.
14. D. Baud, P. L. Saaidi, A. Monfleur, M. Harari, J. Cuccaro, A. Fossey, M. Besnard, A. Debard, A. Mariage, V. Pellouin, J. L. Petit, M. Salanoubat, J. Weissenbach, V. de Berardinis and A. Zaparucha, ChemCatChem, 2014, 6, 3012–3017.
15. A. Amatuni and H. Renata, Org. Biomol. Chem., 2019, 17, 1736–1739.
16. H. J. Wijma, R. J. Floor and D. B. Janssen, Curr. Opin. Struct. Biol., 2013, 23, 588–594.
17. X. L. Wang, J. H. Du, B. C. Zhao, H. Y. Wang, S. Q. Rao, G. C. Du, J. W. Zhou, J. Chen and S. Liu, J. Agric. Food Chem., 2021, 69, 15268–15278.
18. H. Wu, Q. M. Chen, W. L. Zhang and W. M. Mu, Crit. Rev. Food Sci. Nutr., 2023, 63, 2057–2073.
19. H. P. Modarres, M. R. Mofrad and A. Sanati-Nezhad, RSC Adv., 2016, 6, 115252–115270.
20. C. Liu, J. Zhao, J. Liu, X. Guo, D. M. Rao, H. P. Liu, P. Zheng, J. B. Sun and Y. H. Ma, Appl. Microbiol. Biotechnol., 2019, 103, 265–277.
21. Z. N. Qiao, M. J. Xu, M. L. Shao, Y. X. Zhao, M. F. Long, T. W. Yang, X. Zhang, S. T. Yang, H. Nakanishi and Z. M. Rao, Eng. Life Sci., 2020, 20, 7–16.
22. M. Hibi, R. Mori, R. Miyake, H. Kawabata, S. Kozono, S. Takahashi and J. Ogawa, Appl. Environ. Microbiol., 2016, 82, 2070–2077.
23. Y. Y. Li, A. L. Zhang, S. W. Hu, K. Q. Chen and P. K. Ouyang, Microb. Cell Fact., 2022, 21, 142.
24. Y. Sato, K. Okano, H. Kimura and K. Honda, Microbes Environ., 2020, 35, ME20074.
25. R. S. C. Cordeiro, J. Enoki, F. Busch, C. Mügge and R. Kourist, J. Biotechnol., 2018, 284, 68–74.
26. E. L. Hegg and L. Que Jr., Eur. J. Biochem., 1997, 250, 625–629.
27. H. Tao, T. Mori, H. P. Chen, S. Lyu, A. Nonoyama, S. Lee and I. Abe, Nat. Commun., 2022, 13, 95.
28. Y. L. Cui, Y. C. Chen, X. Y. Liu, S. J. Dong, Y. E. Tian, Y. X. Qiao, R. Mitra, J. Han, C. L. Li, X. Han, W. D. Liu, Q. Chen, W. Q. Wei, X. Wang, W. B. Du, S. Y. Tang, H. Xiang, H. Y. Liu, Y. Liang, K. N. Houk and B. Wu, ACS Catal., 2021, 11, 1340–1350.
29. M. M. Gromiha, M. C. Pathak, K. Saraboji, E. A. Ortlund and E. A. Gaucher, Proteins:Struct., Funct., Bioinf., 2013, 81, 715–721.
30. G. S. Ratnaparkhi and R. Varadarajan, Biochemistry, 2000, 39, 12365–12374.
31. G. G. Ferenczy and M. Kellermayer, Comput. Struct. Biotechnol. J., 2022, 20, 1946–1956.
32. J. W. Li, Y. F. Wang, L. Y. An, J. F. Chen and L. S. Yao, J. Am. Chem. Soc., 2018, 140, 3194–3197.
33. G. J. Lesser and G. D. Rose, Proteins:Struct., Funct., Bioinf., 1990, 8, 6–13.
34. C. N. Pace, J. M. Scholtz and G. R. Grimsley, FEBS Lett., 2014, 588, 2177–2184.
35. A. S. Dotsenko, S. Pramanik, A. V. Gusakov, A. M. Rozhkova, I. N. Zorov, A. P. Sinitsyn, M. D. Davari and U. Schwaneberg, Biochem. Eng. J., 2019, 152, 107395.
36. D. R. Tompa, M. M. Gromiha and K. Saraboji, J. Mol. Graphics Modell., 2016, 64, 85–93.
37. C. Vieille and G. J. Zeikus, Microbiol. Mol. Biol. Rev., 2001, 65, 1–43.
38. F. H. Wang, M. L. Zhu, Z. Song, C. Li, Y. Y. Wang, Z. L. Zhu, D. Y. Sun, F. P. Lu and H. M. Qin, ACS Catal., 2020, 10, 13946–13956.
39. R. A. Sheldon, Green Chem., 2017, 19, 18–43.

---

Footnote

† These authors contributed equally to this study.

---