Filling the Void: Introducing Aromatic Interactions into Solvent Tunnels To Enhance Lipase Stability in Methanol

Enzymatic synthesis in organic solvents holds increasing industrial opportunities in many fields; however, one major obstacle is the limited stability of biocatalysts in such a denaturing environment. Aromatic interactions play a major role in protein folding and stability, and we were inspired by this to redesign enzyme voids. The rational protein engineering of solvent tunnels of lipase from Geobacillus stearothermophilus is presented here, offering a promising approach to introduce new aromatic interactions within the enzyme core. We discovered that longer tunnels leading from the surface to the enzyme active site were more beneficial targets for mutagenesis for improving lipase stability in methanol during biodiesel biosynthesis. A structural analysis of the variants confirmed the generation of new interactions involving aromatic residues. This work provides insights into stability-driven enzyme design by targeting the solvent channel void.

ABSTRACT

An enhanced stability of enzymes in organic solvents is desirable under industrial conditions. The potential of lipases as biocatalysts is mainly limited by their denaturation in polar alcohols. In this study, we focused on selected solvent tunnels in lipase from Geobacillus stearothermophilus T6 to improve its stability in methanol during biodiesel synthesis. Using rational mutagenesis, bulky aromatic residues were incorporated to occupy solvent channels and induce aromatic interactions leading to a better inner core packing. The chemical and structural characteristics of each solvent tunnel were systematically analyzed. Selected residues were replaced with Phe, Tyr, or Trp. Overall, 16 mutants were generated and screened in 60% methanol, from which 3 variants showed an enhanced stability up to 81-fold compared with that of the wild type. All stabilizing mutations were found in the longest tunnel detected in the “closed-lid” X-ray structure. The combination of Phe substitutions in an A187F/L360F double mutant resulted in an increase in unfolding temperature (T_m) of 7°C in methanol and a 3-fold increase in biodiesel synthesis yield from waste chicken oil. A kinetic analysis with p-nitrophenyl laurate revealed that all mutants displayed lower hydrolysis rates (k_cat), though their stability properties mostly determined the transesterification capability. Seven crystal structures of different variants were solved, disclosing new π-π or CH/π intramolecular interactions and emphasizing the significance of aromatic interactions for improved solvent stability. This rational approach could be implemented for the stabilization of other enzymes in organic solvents.

IMPORTANCE Enzymatic synthesis in organic solvents holds increasing industrial opportunities in many fields; however, one major obstacle is the limited stability of biocatalysts in such a denaturing environment. Aromatic interactions play a major role in protein folding and stability, and we were inspired by this to redesign enzyme voids. The rational protein engineering of solvent tunnels of lipase from Geobacillus stearothermophilus is presented here, offering a promising approach to introduce new aromatic interactions within the enzyme core. We discovered that longer tunnels leading from the surface to the enzyme active site were more beneficial targets for mutagenesis for improving lipase stability in methanol during biodiesel biosynthesis. A structural analysis of the variants confirmed the generation of new interactions involving aromatic residues. This work provides insights into stability-driven enzyme design by targeting the solvent channel void.

INTRODUCTION

The utilization of enzymes in nonaqueous media has been an ongoing aspiration in synthetic chemistry, as such biotransformations exhibit several advantages over those in conventional aqueous media. Some of the benefits include an increased solubility of hydrophobic substrates, the elimination of microbial contamination, and a suppression of water-dependent catabolic side reactions (1–4). These advantages become more imperative when combining biological catalysts together with chemocatalysts to improve the yields and efficiency of processes (5).

Despite the great interest in biocatalysis in organic solvents, the nontrivial combination of water-based enzymes within nonaqueous media presents many challenges, most importantly, the relatively low stability of enzymes compared to that under their natural habitat conditions (2, 6, 7). Conformational changes in the structure of the enzyme are the main reason for deactivation by organic solvents, due to an impaired balance of hydrophilic-hydrophobic interactions. Moreover, polar (hydrophilic) solvents can penetrate the hydrophilic core of the enzyme, affecting secondary and tertiary conformational changes while stripping structured water molecules from the protein hydration shell (2, 8–15).

The growing understanding of enzyme inactivation mechanisms, using both structural and computational tools, has led to development of techniques, such as protein engineering and immobilization, for the stabilization of enzymes in organic solvents (2, 13, 16, 17). Protein engineering methods include (i) random mutagenesis, (ii) rational design, and (iii) semirational design (2, 8, 18–20). It was previously shown that these approaches can be applied separately or in combination to tailor enzymes to enhance their stability in organic solvents (1, 9, 12, 18, 21, 22).

Protein engineering by rational design requires structural information, and the precise regions for mutagenesis are identified usually by computational tools or prior knowledge (17, 22–24). One recently developed concept for enzyme stabilization in organic solvents is the modification of residues buried in tunnels within the protein structure. Globular enzymes are composed of clefts, pockets, channels, and cavities, which offer a unique microenvironment for biological functions, such as ligand binding or enzymatic catalysis. The tunnel properties (diameter, length, hydrophobicity, etc.) may alter substrate specificity or improve organic solvent resistance (2, 25–28). The identification of these networks requires computational engines, such as CAVER or MOLE generator, for detecting cavities and tunnels as well as a known structure or model of the target protein (29–31). The profound work of Damborsky and coworkers on stabilizing haloalkane dehalogenase (DhaA) has demonstrated the potential for the saturation mutagenesis of residues found in solvent channels. The random substitutions established in these works revealed the effects of small versus bulky residues on both the activity and stability of DhaA (32, 33).

There is a long history of the utilization of lipases in organic solvents (34, 35). They are ubiquitous hydrolytic enzymes possessing two unique features: (i) an “interfacial activation” phenomenon of enhanced activity at an oil-water interphase, and (ii) a helix “lid” that gates substrate accessibility and exposes the active site toward catalysis (“open/close” conformations) via a conformational change (36). In a microaqueous environment, lipases carry out synthesis reactions, such as the transesterification of a wide range of natural and unnatural substrates (36–38), and the synthesis of fatty acid methyl esters (FAMEs), also known as biodiesel (39–41). Biodiesel is a sustainable and renewable alternative to petroleum-based fossils that can be produced from a wide range of feedstocks (edible and nonedible animal fats and plant oils). In most cases, methanol serves as a second substrate in FAME production (42, 43). Compared with traditional chemical pathways to synthesize FAMEs, enzymatic routes are preferred with regard to energy consumption and downstream operations. The efficiency of converting oil feedstocks into biodiesel by lipases is mainly restrained by alcohol-induced inactivation; thus, methanol-stable enzymes are desired (10, 44–46).

The thermophilic bacterial lipase from Geobacillus stearothermophilus T6 (LipT6), was previously used for biodiesel synthesis, and protein engineering increased the stability of the wild-type recombinant enzyme (LipT6_WT) in methanol (47, 48). The crystal structure of LipT6_WT (PDB 4X6U) revealed tight side chain packing and a relatively rigid structure, whereas the structures of the methanol-stable variants confirmed the enhancement of the surface hydrogen bond network (47, 48). Furthermore, the helix lid was identified (α9, residues F177 to A192) (48) and is expected to have a significant interphase-triggered conformational change, as was reported by Carrasco-López et al. for a similar lipase from Geobacillus thermocatenulatus analyzed in an open-lid conformation (95% sequence identity) (49). Further stabilization of the methanol-stable H86Y/A269T/R374W variant was obtained through immobilization in sol-gel (50).

To date, the solvent tunnel engineering of lipases has not been applied for improving stability in methanol. Most works have aimed to alter enzyme selectivity and substrate specificity (26, 31, 51–54). While the previous report on stabilizing DhaA employed a saturation mutagenesis of tunnel residues (33), herein, a more systematic rational approach was practiced, incorporating bulky aromatic amino acids for introducing new interactions within the LipT6 inner hydrophobic core. Tighter packing of a protein lipophilic core was previously shown to result in higher stability with a correlation to thermophilic nature (55). In particular, aromatic interactions have a major contributing role in protein folding nucleation, membrane anchoring, and thermodynamic stability (56–60). Solvent tunnels were selected as the target regions for mutagenesis due to their void volume and accessibility to the enzyme centroid. Each solvent tunnel was carefully and logically examined, and Phe, Tyr, or Trp substitutions were incorporated. Single variants, as well as their double and triple combinations, were purified and evaluated for stability features, kinetic parameters, and biodiesel synthesis. The X-ray structures of 7 mutants revealed the changes induced by the inclusion of bulky residues. Interestingly, we discovered nontrivial correlations among three neighboring stabilizing Phe mutations.

RESULTS

Tunnel analysis and selection of mutations.

An analysis of the LipT6_WT structure (PDB 4X6U) with MOLE 2.0 generator yielded 9 tunnels, and several residues near (4 to 5 Å) these channels were identified and selected for mutagenesis (Fig. 1). Overall, 10 positions were selected for rational design, and 16 single mutants were generated as presented in Table 1. The general considerations were (i) the obstruction of solvent tunnels with bulky side chains, (ii) the maintenance of the native hydrogen bond networks, and (iii) the avoidance of changes to catalytic and metal binding residues. The geometrical and biochemical properties of each residue dictated the choice of the specific amino acid used for site-directed mutagenesis (Phe, Tyr, or Trp). The different primers used for mutagenesis are listed in Table 2.

FIG 1.

Visualization of the LipT6_WT solvent tunnels as generated by the computational algorithm MOLE 2.0. The α-helix lid (α9) is marked in red. Calcium and zinc metal ions are shown as green and gray spheres, respectively. Solvent tunnels are shown in blue, and target residues intended for mutagenesis are shown as magenta sticks. Catalytic Ser114 is presented in cyan. Numbers in black indicate the tunnel numbering according to the MOLE job report.

TABLE 1.

Residues selected for mutagenesis on the basis of LipT6_WT tunnel analysis

Tunnel no.a	Tunnel length (Å)b	Residue	Generated mutation
1	21	L184	F, Y
		A187	F, Y
		L360	F, Y
2 and 3 (Y shape)	13.5 and 13.9	R215	F, Y
4	11.5	H154	Y, W
5	7.8	I11	W
7	8.5	F226	Y
		K330	Y, W
8	8.8	L380	F
		F268	Y

^aAccording to MOLE 2.0 automatic numbering; tunnels 6 and 9 were unchanged to maintain their rich hydrogen bond network.

^bAccording to MOLE 2.0 job report.

TABLE 2.

Primers for site directed mutagenesis of pET9a-LipT6_WT plasmid

Position	Original aaa	Substitution	Nucleotide sequence 5′→3′b
11	I	W	GCTAACGATGCGCCATGGGTACTTCTCCACGGG
154	H	Y	TTTGAAGGCGGACATTATTTTGTGTTGAGCGTG
		W	TTTGAAGGCGGACATTGGTTTGTGTTGAGCGTG
184	L	F	GATCGCTTTTTTGACTTCCAGAAGGCGGTGTTG
		Y	CGATCGGTTTTTTGACTATCAGAAGGCGGTGTTG
187	A	F	GACTTGCAAAAATTCGTGTTGAAAGCAGCGGC
		Y	GACTTGCAAAAATACGTGTTGAAAGCAGCGGC
215	R	F	GACCAATGGGGACTGTTTCGCCAGCCAGGTGAA
		Y	GACCAATGGGGACTGTATCGCCAGCCAGGTGAA
226	F	Y	GAATCATTCGACCAATATTATGAACGGCTCAAACGG
268	F	Y	CGAATACGTATTATTTGAGCTATGCCACAGAACGGACG
330	K	W	ATGAACGGACCATGGCGAGGATCGACAGAT
		Y	ATGAACGGACCATATCGAGGATCGACAGATCGG
360	L	F	ACAATGTAGATCATTTCGAAGTCATCGGCGTTG
		Y	CGTACAACGTAGATCATTATGAAGTCATCGGCGT
380	L	F	GCCTTTTATTTGCGATTTGCAGAGCAGTTGGCG
L184F/A187Fc	L	F	CGCTTCTTCGACTTCCAAAAATTCGTGTTG

^aaa, amino acid.

^bPositions of alterations in mutagenesis primers are indicated in bold.

^cUsed for the combination of L184F/A187F, while A187F (underlined) was used as the template.

Screening for enhanced stability in 60% methanol.

The mutants were expressed in Escherichia coli, and the soluble cell extract was used for stability evaluation in 60% methanol, as previously described (47, 48). Relative hydrolysis activity values of the designed mutants compared with those of LipT6_WT are presented in Table 3. A threshold of a 4-fold increase in stability was used to determine which mutations to further combine to investigate a potential additive effect. Of the 16 variants evaluated, only the L184F, A187F, and L360F variants had significant improvements of 81.2-, 5.3-, and 4.5-fold, respectively, compared to the wild type (WT). All three mutations were found in the vicinity of the longest solvent tunnel (tunnel 1), which is 2.9 Å away from the catalytic Ser114 as presented in Fig. 2. L184 and A187 are located on the LipT6 helix lid, while L360 is part of a flexible internal loop. The three mutations were combined in all possible rearrangements to explore their additive effect on LipT6 stability (Table 3). Only two double mutants, the L184F/A187F and A187F/L360F variants, were found to be more stable than LipT6_WT but inferior to the L360F mutant, while the triple mutant was a very poor catalyst, suggesting a complex association between these three residues.

TABLE 3.

Relative activity of LipT6 variants in 60% methanol

Varianta	Relative activity ratiob
Single mutants
L360F	81.2 ± 8.51
A187F	5.3 ± 0.72
L184F	4.5 ± 0.99
F268Y	3.46 ± 0.22
R215F	2.40 ± 0.35
H154Y	1.89 ± 0.25
L184Y	1.51 ± 0.19
R215Y	1.31 ± 0.29
A187Y	1.22 ± 0.43
F226Y	1.08 ± 0.45
H154W	0.94 ± 0.04
K330Y	0.49 ± 0.15
K330W	0.30 ± 0.08
I11W	0.21 ± 0.03
L360Y	0.19 ± 0.09
L380F	0.10 ± 0.05
Double mutants
A187F/L360F	26.5 ± 7.48
L184F/A187F	19.9 ± 0.89
L184F/L360F	0.46 ± 0.05
Triple mutant
L184F/A187F/L360F	0.29 ± 0.06

^aEach variant was expressed in E. coli, and the soluble cell extract (CE) was used for the screen. SDS-PAGE analysis ensured an appropriate expression level, and hydrolysis activity in buffer ensured no drastic activity loss.

^bThe relative hydrolysis activity of pNPL was calculated as (E/E₀)/(E/E₀)_WT by comparing the activity in the CE from each variant before (E₀) and after (E) a 30-min incubation in 60% methanol divided by the same value for LipT6_WT (E/E₀)_WT. The results represent the averages from duplicates.

FIG 2.

Residues found to influence stability in methanol by tunnel engineering. (A) Closeup view of tunnel 1. (B) Surface visualization. The α-helix lid (α9) is marked in red. The solvent tunnel is shown in blue and target residues are shown as magenta sticks. Catalytic serine is presented in cyan.

Stability measurements of purified enzymes in 70% methanol.

To validate the screening results, single variants and their double mutant combinations were purified, and their stability in 70% methanol was measured while incubating for up to 6 h (47). The relative activity values (compared with those under stress-free conditions in buffer) are presented in Fig. 3. An expected decrease in activity occurred in all variants after 1 h, while LipT6_WT lost more than 70% of its activity after 6 h of incubation. The L184F, A187F, and L360F single variants maintained 37%, 47%, and 73%, respectively, of their hydrolytic activity after 6 h, as was also inferred from the screening results (Table 3). In addition, the L184F/A187F and A187F/L360F double mutants showed increases in stability compared to that of LipT6_WT, preserving more than 48% and 58%, respectively, of their initial activity. Nevertheless, the L184F/L360F double mutant presented a lower stability after losing more than 80% of its initial activity after 6 h. As found in the screening stage, only 5 of 6 mutants were more stable than LipT6_WT.

FIG 3.

Relative residual activity of LipT6 variants after incubation in 70% methanol. Purified LipT6 mutants were incubated in 70% methanol for various durations, and their remaining activity was measured and compared with that under native conditions (marked as 100% in cyan bars). Samples for pNPL hydrolysis assay were collected after 1, 4, and 6 h of incubation.

Unfolding temperature of LipT6 mutants.

The unfolding temperatures (T_ms) of the stable mutants (in their purified form) were characterized in buffer (native environment) and in organic solvents (denaturing environment). The assays were conducted using differential scanning fluorimetry (50) with 60% methanol solutions, resembling biodiesel synthesis reaction conditions (Table 4). Furthermore, T_m values were also measured in 50% to 70% (vol/vol) solutions of methanol, ethanol, acetonitrile, and dimethyl sulfoxide (DMSO) (see Table S1 in the supplemental material).

TABLE 4.

Unfolding temperature of LipT6 variants in methanol

LipT6 variant	Tm (°C) in:
	Buffer	60% methanol
WTa	66.6 ± 0.1	38.9 ± 0.3
L184F	67.5 ± 0.1	41.0 ± 0.2
A187F	70.0 ± 0.2	43.0 ± 0.6
L360F	69.9 ± 0.2	45.6 ± 0.2
L184F/A187F	67.6 ± 0.1	46.1 ± 0.2
L184F/L360F	65.5 ± 0.1	38.1 ± 0.1
A187F/L360F	72.2 ± 0.1	46.2 ± 0.1

^aObtained from Gihaz et al. (50).

The results in Table 4 clearly indicate that all single mutants were more stable than LipT6_WT in both buffer and 60% methanol, while similar results were obtained for additional organic solvents (Table S1). Among the single mutants, the L360F variant presented the highest thermal stability improvements (increase in T_m) in buffer and in 60% methanol of more than +3°C and +6°C, respectively. Moreover, it displayed a higher thermal stability in all other solvents tested. The addition of A187F to form the A187F/L360F variant resulted in further T_m improvements of +5°C and +7°C in buffer and 60% methanol, respectively. Surprisingly, the L184F/L360F mutant exhibited a lower T_m than LipT6_WT, despite the stabilizing effect observed by L184F and L360F separately (the same trend was observed in the other organic solvents studied). However, the L184F/A187F double mutant had a minor T_m enhancement in buffer (+1°C) but a substantial stability with an increase in T_m of +7°C in 60% methanol.

Kinetic analysis.

To investigate effects of the mutations on enzyme kinetics, 4-nitrophenyl laurate (pNPL) hydrolysis was selected on the basis of its wide usage in lipase studies (61–64), including those with LipT6 (47, 48). The kinetic constants (Table 5) were calculated on the basis of activity under native conditions without methanol. In general, all variants displayed a decrease in k_cat values compared with that of LipT6_WT, with the L184F/L360F double mutant displaying a 70% decline. The A187L and L360F single mutants and L184F/A187L and A187F/L360F double mutants all had similar lower K_m constants, while the value for the L184F mutant was similar to that for LipT6_WT. The L184F/L360F double mutant had the lowest K_m value compared with that of LipT6_WT and also the lowest activity rate. Moreover, the values for the enzyme efficiency parameter k_cat/K_m for most of the variants were lower than that for LipT6_WT, except the L184F/L360F double mutant, which displayed a 2-fold increase. Generally, the L360F mutation had the largest negative impact on k_cat in the hydrolysis reaction.

TABLE 5.

Kinetic parameters of LipT6 variants in pNPL hydrolysis

LipT6 variant	Km (10−2 mM)a	kcat (103 s−1)	kcat/Km (103 s−1 · mM−1)
WTb	7.9 ± 0.6	4.7	59
L184F	8.4 ± 0.9	3.0	36
A187F	5.4 ± 0.6	3.0	55
L360F	5.1 ± 0.5	2.1	41
L184F/A187F	5.1 ± 0.4	2.1	41
L184F/L360F	1.1 ± 0.1	1.3	116
A187F/L360F	5.6 ± 0.7	1.6	30

^aValues are means ± standard errors of the means.

^bObtained from Dror et al. (47).

Biodiesel production from waste chicken oil.

Purified single and double variants were used as soluble biocatalysts for biodiesel synthesis from waste chicken oil, with 5:1 molar ratio of methanol to oil. The results presented in Fig. 4 emphasize the superior stability and yield of the A187F/L360F double mutant, which converted 88% of waste chicken oil into FAME (3-fold improvement over the LipT6_WT yield after 24 h) under the conditions tested. The L360F single mutant achieved the second-highest FAME yield of 59% under the conditions tested (2-fold improvement). The synthesis yields highly correlated with the increase in T_m (Table 4), the initial screening results of the soluble cell extracts (Table 2), and the results of the pure enzyme-based stability assay in 70% methanol (Fig. 3). Furthermore, the L184F mutant and the L184F/L360F double mutant provided lower FAME conversions than LipT6_WT, as predicted by their T_m values in 60% methanol. Unexpectedly, the A187F and L184F/A187F variants displayed similar increases in transesterification activity contrary to their different T_m values. An additive effect in terms of stability was observed mainly when combining A187F and L360F mutations. On the other hand, the L184F mutation had a negative effect in combination with L360F and negligible influence when merged with A187F.

FIG 4.

FAME biosynthesis from waste chicken oil using soluble LipT6 variants. Reaction conditions: oil (2 g), water (20%), methanol-to-oil molar ratio, 5:1 (60% MeOH), and soluble lipase (0.04% based on the oil weight), 1,350 rpm, 45°C. The results represent triplicates (n = 3).

Crystal structure determination.

In an attempt to gain a deeper understanding of the correlation between structure and stability, the crystal structures of all single and double mutants were solved at resolutions of 1.2 to 2.7 Å (Fig. 5A to F, in comparison with LipT6_WT). The crystal parameters and data statistics are summarized in Table S2. Each solved structure was analyzed with two web servers: (i) MOLE 2.0 to reassess the solvent tunnel distribution in the variants (30) and (ii) Arpeggio to calculate and visualize the unique interatomic interactions in LipT6 mutants compared with those in LipT6_WT. The Arpeggio server uses PDB files to calculate all possible intramolecular interactions on the basis of the geometrical and biochemical features of the residues (65). It was selected due to its versatility in identifying a wide range of interactions and its straightforward user interphase in comparison with those of other traditional tools.

FIG 5.

X-ray structures of LipT6 designed mutants superimposed with LipT6_WT (in gold). (A) L184F in green; (B) A187F in pink; (C) L360F in blue; (D) L184F/A187F in cyan; (E) L184F/L360F in brown; (F) A187F/L360F in purple. The α-helix lid in all structures is marked in red, and catalytic S114 is presented in all figures. New interactions induced in the mutants compared with that in LipT6_WT (according to Arpeggio server analysis) are marked in dashed lines as follows: π-π in yellow, CH/π in green, and amide-π in blue.

First, a superposition of all mutant structures with LipT6_WT ensured there were no significant structural changes in the catalytic triad (Ser114, His359, and Asp318), the calcium-binding site (Glu361, Gly287, Pro367, and Asp366), the zinc-binding site (Asp62, His88, Asp239, and His82), and the oxyanion hole-stabilizing backbone residues (Phe17 and Glu115) (48). Thus, all changes in the properties of the variants were associated directly with the induced mutations. The electron densities around the mutated residues in all variants discussed are presented in Fig. S1.

An inspection of the solvent tunnels in the new structures by using the MOLE server confirmed the elimination of tunnel 1, which extends from the outer protein surface toward the hydrophobic pocket (see Fig. S2). As expected, the phenyl rings obstructed the channel by occupying its volume. Moreover, no other newly formed tunnels were identified close to the active site region or at other locations in the crystal structures of the variant.

In all single mutant structures (Fig. 5A to C), the orientation of the phenyl side chains is toward the former occupied region of tunnel 1. Compared with LipT6_WT, the three single mutations did not cause any structural changes in the near environment and overall fold. An analysis with Arpeggio of the new contacts in the mutants revealed new π-π interactions with Phe291 residues in the L184F and A187F variants (Fig. 5A and B). Phe291 is a neighbor to the mutated residues, located on a solvent-accessible loop (a13) and mainly stabilized by CH/π interactions (see Fig. S3). Moreover, Phe184 interacts with Phe17, and several CH/π interactions were formed following mutagenesis. In the L360F variant, a new π-π interaction with the catalytic His359 was generated. This residue is stabilized by several other interactions in LipT6_WT (Fig. S3).

An examination of the structures of the double mutants, of which two were more stable than LipT6_WT, showed different bond networks depending on the mutation combinations (Fig. 5D to F). The stable L184F/A187F and A187F/L360F mutants had similar residue positions as in the respective single variants. The L184F/A187F mutant (Fig. 5D) displayed the same interactions found in the L184F and A187F mutants, creating two new π-π interactions with Phe291. In addition, Phe184 and Phe187 participated in a new aromatic continuum with Phe291, as was shown in the L184F variant. Likewise, similar to those for the two single mutants, the stable A187F/L360F variant (Fig. 5F) exhibited new π-π interactions between His359 and Phe360 and between Phe291 and Phe187.

In contrast, the methanol-sensitive L184F/L360F variant exhibited a different conformation of Phe184, due to steric hindrance by Phe360 (Fig. 5E). The movement of Phe184 also induced a conformational change of Phe291, now facing “out” in a more solvent-accessible orientation (see Fig. S4). This movement caused the exclusion of π-π or amide-π interactions formerly stabilizing Phe291. In addition, in its “new” orientation, Phe184 was discarded from any π-π interactions, now stabilized by only hydrophobic interactions. The poor methanol stability of the L184F/L360F variant is therefore linked to the aromatic rearrangement in the vicinity of the active site, despite the stabilizing effect of L360F alone due to tunnel 1 obstruction. To further strengthen this hypothesis, the crystal structure of the L184F/A187F/L360F triple mutant was solved (see Fig. S5). The structure displayed the same “flipped” conformation of Phe184 and Phe291 occurring in the L184F/L360F mutant. Despite this, Phe187 managed to interact with Phe291 and Phe360 by π-π interactions, but with no effect on the stability of the variants.

DISCUSSION

Enzyme engineering is one of the major approaches for designing stable biocatalysts for an organic solvent environment (2, 10, 18, 22). The manipulation of the tunnels to obtain stability in organic solvents was first introduced with a primarily randomized design of haloalkane dehalogenase DhaA by Koudelakova et al., which increased its stability in DMSO (33). Most works on tunnel redesign have focused on altering substrate selectivity by influencing substrate access to the active site (26, 31, 52, 53, 66–69).

The present work aimed to stabilize LipT6 in methanol by incorporating aromatic residues into the solvent channels to induce improved hydrophobic packing via π-involving interactions. Some works suggested that such modifications potentially restrict the unnecessary penetration of polar alcohols into the enzyme core (32). Site-directed mutagenesis at selected positions was performed on the basis of the geometric and biochemical properties of the residues. This approach was indeed successful in obtaining new solvent-stable mutants of LipT6 with improved T_ms and biodiesel synthesis yields. As a rational concept, introducing π interactions within lipophilic areas in the enzyme inner tunnels reduces screening efforts compared to those reported with other directed evolution approaches. Dror et al. obtained a 2-fold improvement in FAME synthesis yield with the LipT6 H86Y/A269T double mutant when combining mutations selected from random mutagenesis and structure-guided consensus libraries. The isolation of these mutations required the screening of more than 2,200 colonies (47). A greater improvement in the stability (30-fold higher) in 70% methanol was achieved by Korman et al., who constructed Dieselzyme4 (a Proteus mirabilis lipase variant with 13 mutations, including one introduced disulfide bond), though their overall screening efforts were in an estimated 20,000 colonies (70). On the other hand, a rational approach by Park et al. yielded 7 variants of Candida antarctica lipase B (CaLB) with potentially enriched hydrogen bond networks, while only one mutant possessed 1.5-fold higher stability in 80% methanol (71). Koudelakova et al. discovered stabilizing mutations located in the DhaA access tunnel after screening 5,326 colonies from random mutagenesis libraries for stability in 42% DMSO. The random positive mutations were added to a previously known stable DhaA variant, and a saturation mutagenesis of position Ala171 in the access tunnel was performed. This combined approach resulted in few stable variants possessing superior 2-fold increased stability in 40% DMSO (33). Regarding our screening efforts (16 single variants) and the additive effect accomplished (3-fold improvement in FAME biosynthesis), one can conclude that the introduction of aromatic interactions within solvent tunnels is a promising concept in the quest for solvent-stable enzymes. Moreover, the focus on long and deep tunnels may have even reduced labor and cost efforts.

Generally, the initial screening results emphasized the dependence on tunnel location and length. Stabilizing mutations (3 of 6 [50%]) were found exclusively in the longest solvent pathway (tunnel 1) leading from the surface to the active site. The overall structure of LipT6 is rigid and compact, similar to that of other thermophilic homologs in the I.5 family (49, 72, 73). Thus, the peripheral tunnels are situated mostly near the hydrophilic surface (Fig. 1) and are less prone to stabilization through core hydrophobic interactions. In addition, tunnel 1 is near the active site; thus, the redesign of such surroundings was expected to have some significant outcomes, as was presented by Biedermannova et al. for haloalkane dehalogenase LinB (51). Furthermore, Phe and Tyr mutations had significantly diverse stabilization features when occupying the same position (Table 3). For example, the L360F variant was 81.2-fold more stable than the wild type, while the L360Y variant was less stable (0.19-fold). These contradicting outcomes highlight the importance and unique nature of this tunnel, near the helix lid, which is expected to have a significant structural rearrangement in the presence of hydrophobic substrates.

Three positions within tunnel 1 were selected for further investigation after mutations of the Phe residues resulted in an increased stability in 60% methanol. L184 and A187 are located on LipT6 helix lid (α9), while L360 is found close to the active site. A reanalysis with MOLE 2.0 of the crystal structures of the variants indicated that the introduced phenyl side chains managed to crowd tunnel 1 (no longer found in mutants) (see Fig. S2 in the supplemental material), as intended, and as described beforehand (32, 33). Conversely, Liskova et al. showed that replacing Phe with Gly in DhaA80 caused a decline in stability due to the disruption of intramolecular hydrophobic packing (32). These findings correlate with our positive results of rationally replacing Leu or Ala with Phe residues. The stability measurements of purified enzymes in 70% methanol validated our screening results, as the L360F variant exhibited the highest residual activity after 6 h followed by stable A187F and L184F mutants. The melting temperature measurements agreed with the initial stability screening results, as the L360F and A187F mutants presented better thermal stability in both buffer and methanol (along with other organic solvents, as presented in Table S1). A similar correlation between T_m and stability in organic solvents was previously obtained for other LipT6 stable variants (47, 48).

The superior stability of the L360F variant was related to its unexpected π-π stacking interaction with catalytic His359 (Fig. 5C). Kannan and Vishveshwara previously reported the existence of one aromatic cluster next to the active site in thermophilic enzymes that was lacking in their mesophilic equivalents (55). An alignment of sequences homologous to LipT6 performed by Dror et al. showed that Leu360 is not an evolutionarily conserved position; however, the formation of a new aromatic cluster explains the stability exhibited by this mutant (47, 55).

Both L184F and A187F interacted separately with Phe291 via new aromatic π-π interactions, while the A187F variant had both improved transesterification activity and thermostability. Since the L184F variant exhibited the lowest improvement in T_m and methanol stability, it can be implied that a minimum stability enhancement threshold is required for improved biodiesel synthesis in comparison to that with LipT6_WT. As previously described by Dror et al., some mutations in LipT6_WT induced methanol stability but at the same time led to decreased FAME synthesis. One example is the neighboring position Gln185, which was mutated to Leu and was found to induce alcohol stability but a lower transesterification yield of soybean oil (47). This phenomenon was attributed to a tighter orientation of the LipT6 helix lid, limiting triglyceride accessibility during biodiesel synthesis. It can be assumed that the L184F mutation had a similar affect as was reported for Q185L. In addition, several studies indicated that mutations of the lid can alter thermostability along with selectivity (74, 75). Khan et al. recently reviewed the influence of mutagenesis of the lid on thermostability and activity, highlighting the importance of this domain in lipases (76). Some studies indicated that a simultaneous change in stability and substrate preferences is caused by an alteration of the residues on the lid (77, 78). Likewise, a decline in activity accompanied with elevated stability was also apparent after introducing bulky residues in the DhaA access tunnel (32, 33).

A kinetic analysis revealed a general increase in substrate affinity (lower K_m) and a decrease in maximum hydrolytic velocity (lower k_cat) for most variants. As expected, mutations in the vicinity of the active site affected enzyme kinetics, selectivity, or even the mechanism (51, 68, 69). Among the single mutants, the L360F variant displayed the lowest k_cat with similar K_m values, emphasizing the significant interaction with catalytic His359. Lid mutations L184F and A187F also affected LipT6 kinetics, as was described previously by Tang et al. for lid mutations of Penicillium expansum lipase (77). Despite the lower activity rates of the mutants in the hydrolysis reaction, most of them performed better in an organic solvent environment, leading to higher biodiesel yields (Fig. 4).

The merging of stable mutations in all possible double mutant combinations revealed interesting complex correlations with regard to stability in methanol, T_m values, and FAME synthesis. The highest stabilizing effect was observed in the A187F/L360F variant, which displayed the best FAME yield (88% after 24 h) and T_m improvement (T_m increases of +5°C and +7°C compared with those of LipT6_WT in buffer and 60% methanol, respectively). These findings were also confirmed in the purified enzyme stability assay (Fig. 3) and by the melting temperatures in other polar solvents (Table S1). Relatedly, Stepankova et al. previously showed that different organic solvents confer different destabilizing effects on enzymes (79). The results of a structural analysis suggest that new aromatic interactions with both catalytic His359 (by Phe360) and Phe291 (by Phe187) are responsible for the improved performance in the nonaqueous environment. Aromatic interactions were previously found to stabilize xylanase, RNase, and many other protein structures by improving hydrophobic packing and introducing new π interactions (80–83). In addition, these two mutations did not eliminate interactions found in LipT6_WT but enriched the existing network and occupied tunnel 1. Prior work on LipT6 stabilization highlighted the importance of enhancing the hydrogen bond network among surface residues as well as the interactions with water molecules (47, 48). Here, we have discovered the significance of π-π stacking interactions and CH/π interactions within the LipT6 hydrophobic core for improving methanol stability. The kinetic properties of the A187F/L360F double mutant revealed a lower hydrolysis reaction rate and a higher affinity toward pNPL (C₁₂). Despite this, the triglyceride methanolysis rate and yield were 3-fold higher than in the wild type. On the basis of these observations, we conclude that the major factor influencing FAME synthesis in our system is the enzyme stability. This was also shown in other cases (47, 48). Dror et al. showed that the LipT6 H86Y/A269T double mutant had a similar hydrolysis performance to that of LipT6_WT, though its FAME yield from soybean oil was 2-fold higher (47).

The L184F/A187F variant displayed a similar thermostability to that of the A187F/L360F variant, but its transesterification performance was similar to that of the A187F variant. The addition of the L184F mutation did not reduce the already improved transesterification activity of the A187F variant. Only the L184F/A187F double mutant possesses two neighboring lid mutations, which may explain this noncorrelative relationship between alcohol stability and FAME synthesis, as was shown previously (76, 77). The structures of both the L184F and L184F/A187F variants revealed a π-π stacking interaction network involving 16 aromatic side chains (see Fig. S6). The improved thermostability was likely due to this branched aromatic continuum, but no significant transesterification improvement was observed compared with that of the A187F variant (55).

Interestingly, the L184F/L360F double mutant, comprising two single stabilizing mutations, showed decreased methanol stability, lower T_m (in all solvents tested), and the lowest k_cat, K_m, and biodiesel yield. Unlike the effect of L184F on A187F (in the L184F/A187F variant), in combination with L360F, a dramatic decrease in stability occurred. An inspection of the crystal structure of the L184F/L360F double mutant revealed an aromatic cluster rearrangement involving Phe184, Phe291, and Phe360 (Fig. 5 and S3). The steric collision of static Phe360 (found in the same orientation in all mutants) and Phe184 caused a conformational change in the latter, leading to movement of Phe291 as well. Subsequently, fewer intramolecular interactions were possible. It was previously shown that helix-stabilizing residues (as Phe291) impact protein stability when comparing thermophilic and mesophilic homologs (84). In addition, an analysis of the L184F/A187F/L360F triple mutant affirmed that A187F did not restore stability to the L184F/L360F variant or the favored ring conformation. The fact that the stabilizing mutation L360F did not improve the stability of the L184F or L184F/A187F variants demonstrated the significant impact of Phe291 conformation on LipT6 stability and organic synthesis capability.

Overall, this new systematic approach of rational tunnel engineering by incorporating aromatic residues to facilitate π-involving interactions, with a focus on deep and long solvent channels, can be considered for stabilizing other enzymes in organic solvents.

MATERIALS AND METHODS

Chemicals.

Methanol, glycerol, NaCl, and Triton X-100 were purchased from Bio-Labs (Jerusalem, Israel). 2-Propanol, ethanol, and sodium citrate were purchased from J.T. Baker (Deventer, The Netherlands). Sodium formate, DMSO, and sodium acetate were purchased from Merck (Darmstadt, Germany). Ethyl acetate was purchased from Gadot (Haifa, Israel) while acetonitrile and CaCl₂ were from Spectrum Chemical MFG (Gardena, CA, USA). Trizma base, pNPL, polyethylene glycol (PEG) 400, PEG 3350, heptadecanoic acid methyl ester, and kanamycin were purchased from Sigma-Aldrich (Rehovot, Israel). Waste chicken oil was kindly donated by Miloubar (Miloubar Mixture Institute ACS, Miluot, Israel). All materials used were of the highest purity available.

Bacterial strains, plasmids, and enzymes.

Recombinant Geobacillus stearothermophilus T6 lipase (EMBL, AF429311.1) fused to a His tag was expressed in Escherichia coli BL21 cells (DE3; Novagen, Darmstadt, Germany) as previously described (47, 48, 50).

Solvent tunnel analysis using MOLE 2.0.

To detect tunnels in LipT6_WT (PDB 4X6U), the crystal structure was analyzed with the MOLE 2.0 online generator (http://old.mole.upol.cz/) using default parameters (30). The analysis revealed 9 tunnels, and a superposition display (enzyme structure and coordinates of the tunnels) was utilized to define the closest residues to these channels (4 to 5 Å direct distance) using Pymol (85). After eliminating essential residues (catalytic, metal binding, and multiple hydrogen bond donor) and inspecting the MOLE 2.0 job review, 10 positions were selected for site-directed mutagenesis (Table 1), which were mutated into at least one bulky residue (F, Y, or W). The mutants were generated and subsequently evaluated for activity, stability, and structural characterization.

Site-directed mutagenesis of LipT6.

Rational mutagenesis of the pET9a-LipT6_WT plasmid was performed using the QuikChange protocol for site-directed mutagenesis. The reaction mixture was composed of 5 μl Taq polymerase buffer, 2 μl DNA template (50 ng/μl), 1.5 μl of each primer solution (1 μg/μl), 2 μl deoxynucleoside triphosphates (dNTPs; 20 mM A/T/C/G, 1:1:1:1), 0.5 μl Taq polymerase (EurX, Gdańsk, Poland), and 37.5 μl distilled water (dH₂O). The different primers are listed in Table 2. Taq polymerase was added and the reaction mixtures were incubated in a thermocycler (Labcycler; SensoQuest, Göttingen, Germany). The PCR program had an initial denaturation step for 1 min at 95°C, and then 20 cycles of 30 s at 95°C, 45 s at 65°C, and 6 min at 68°C, followed by a final elongation step for 7 min at 68°C. PCR products were run on an agarose gel (1% [wt/wt]) to validate single-band products, and the template plasmid was digested for 18 h at 37°C with Dpn1 (New England BioLabs, Ipswich, MA, USA). The resulting plasmid was used for transformation and selection on LB agar plates containing 25 μg/ml kanamycin. The plasmids from positive colonies were extracted using a plasmid miniprep kit (Qiagen, Hilden, Germany) and sequenced for verification (HyLabs, Rehovot, Israel).

Soluble lipase activity assay.

The soluble lipase hydrolytic activity on pNPL was determined using a colorimetric assay as described previously (47, 48). This method was also used to determine the specific activity of purified enzyme samples in buffer and in solvent solutions.

Stability screen in 60% methanol.

The screening for methanol-stable mutants was performed as described previously with a few modifications (47). The TB medium inoculation volume was 35 ml, and after centrifugation, the cells were resuspended in 9 ml buffer to increase the sensitivity of the method.

Purification of LipT6 variants.

LipT6 variants were purified with AKTA Prime Plus (GE Healthcare Bio-Sciences AB, Sweden) according to previously described procedures (47, 48).

T_m determination of purified LipT6 variants.

Denaturation temperatures were determined with a nanoDSF device according to a previously described procedure (50).

Determination of kinetic parameters.

K_m and k_cat values for purified LipT6 variants were determined using the pNPL hydrolysis colorimetric assay in 96-well plates as previously described (47, 48). The results were analyzed using SigmaPlot software.

Stability validation of LipT6 variants in 70% methanol.

The stability of purified LipT6 variants in 70% methanol was determined by measuring the residual hydrolytic activity after incubating for several hours, as described previously (47, 48).

Enzymatic transesterification of waste chicken oil by soluble lipase.

The transesterification reactions were carried out in triplicates according to the work of Dror et al. (48), with a few modifications. Briefly, 14-ml closed glass vials were filled with 2 g waste chicken oil, and methanol (5:1 alcohol-to-oil molar ratio) and 400 μl of lipase buffer (2 mg/ml enzyme solution, 0.04% enzyme, and 20% water content based on oil weight) were added.

Gas chromatography analysis of FAME.

Gas chromatography analysis of FAME formation was carried out according to the work of Dror et al. (48).

LipT6 variant crystallization, data collection, and structure determination.

The crystallization of LipT6 variants was performed as described previously (48), with a few modifications. The hanging drop contained 2 μl protein solution (0.5 to 2 mg/ml) and 2 μl crystallization condition (0.2 M sodium citrate and 25% PEG 3350 or 0.2 M sodium formate and 20% PEG 3350). The cryoprotectant solution was crystallization condition enriched with 25% PEG 400. The X-ray diffraction data of LipT6 variants were collected at the European Synchrotron Radiation Facility (ESRF), Grenoble, France, with beamlines as described in Table S2 in the supplemental material. The diffraction data were indexed, integrated, and reduced with either MOSFLM (86), Scala (87), autoPROC (88), or EDNA (89). All structures were solved by molecular replacement using Phaser (90) and the coordinates of the LipT6_WT structure (PDB 4X6U). Refinement was performed using PHENIX (91). Manual model building, real-space refinement, and structure validations were performed using Coot (92). The crystal parameters, beamlines used by the ESRF, and data statistics are summarized in Table S2.

Calculation and visualization of interatomic interactions in LipT6 variants.

The interaction repertoire in LipT6_WT and other variants was determined by using the PDB file of each mutant and the Arpeggio web server (65). The default settings were used to calculate and analyze each structure, including the graphical presentations with Pymol (85) as shown in Fig. 5.