Feruloyl Esterase (LaFae) from Lactobacillus acidophilus: Structural Insights and Functional Characterization for Application in Ferulic Acid Production

Ferulic acid and related hydroxycinnamic acids, used as antioxidants and preservatives in the food, cosmetic, pharmaceutical and biotechnology industries, are among the most abundant phenolic compounds present in plant biomass. Identification of novel compounds that can produce ferulic acid and hydroxycinnamic acids, that are safe and can be mass-produced, is critical for the sustainability of these industries. In this study, we aimed to obtain and characterize a feruloyl esterase (LaFae) from Lactobacillus acidophilus. Our results demonstrated that LaFae reacts with ethyl ferulate and can be used to effectively produce ferulic acid from wheat bran, rice bran and corn stalks. In addition, xylanase supplementation was found to enhance LaFae enzymatic hydrolysis, thereby augmenting ferulic acid production. To further investigate the active site configuration of LaFae, crystal structures of unliganded and ethyl ferulate-bound LaFae were determined at 2.3 and 2.19 Å resolutions, respectively. Structural analysis shows that a Phe34 residue, located at the active site entrance, acts as a gatekeeper residue and controls substrate binding. Mutating this Phe34 to Ala produced an approximately 1.6-fold increase in LaFae activity against p-nitrophenyl butyrate. Our results highlight the considerable application potential of LaFae to produce ferulic acid from plant biomass and agricultural by-products.


Introduction
Hydroxycinnamic acids (HCAs), such as ferulic and caffeic acids, are one of the most widely distributed naturally occurring phenolic acids, typically present in the form of esters and found mainly in grains, fruits and vegetables. They function in crosslinking lignin and hemicellulose of the plant cell wall, thereby noticeably enhancing their physical strength and mechanical properties [1,2]. Recently there has been a growing interest in industrial usages of HCAs, as these compounds have been shown to exhibit anti-inflammatory, antioxidant, neuroprotective, anti-aging, anti-carcinogenic and antimicrobial properties [3,4].
Feruloyl esterases (FAEs, E.C. 3.1.1.73), a subfamily of carboxylic acid esterases, are mainly associated with the release and exploitation of HCAs from plant cell walls [5,6]. These enzymes can further facilitate the binding of other enzymes and generate lignincarbohydrate complexes for subsequent biodegradation. FAEs are classified into four subgroups (denoted A, B, C and D) based on their primary sequences and catalytic properties [7,8]. Furthermore, these esterases interact with other cellulases or xylanases, thereby enhancing the biodegradation of wheat bran, corncobs, corn bran and grass lignocellulose materials [9,10]. Accordingly, FAEs appear to have considerable potential with respect to numerous industrial and food-related applications, such as animal feed supplements, biofuel production and bioactive component synthesis. Thus, the discovery of FAEs with improved properties using protein-engineering methods is a noteworthy pursuit. For example, mutants designed and generated based on the simulated structure of FAE from Lactobacillus acidophilus showed an increased ferulic acid production from agricultural waste [11].
Although several fungal FAEs have been characterized, there is currently a growing interest in the potential utility of FAEs derived from lactic acid bacteria (LAB), due to their safety and health-promoting properties. To date, a range of LAB species, including Bifidobacterium animalis [12], Lactobacillus helveticus [13], and Lactobacillus plantarum [14], have been shown to produce FAEs, although only limited information is available regarding their functional properties, structural characteristics and bioavailability. Given that LAB strains are fermentative, easy to culture and safe to use, from a practical perspective, we considered that it would be prudent to investigate FAEs produced by L. acidophilus, which is among the most widely-known probiotic bacteria associated with a diverse range of foods, including dairy products, vegetables, meat and fish [15,16]. In addition, L. acidophilus is one of the most useful bacterial species, particularly with respect to the biotechnological preparations of food products and dairy supplements [17,18]. Given the industrial significance of FAEs and their roles in LAB, in the present study, we sought to identify and characterize a feruloyl esterase (LaFae) from L. acidophilus.
Following the identification of a candidate esterase, we evaluated its biotechnological potential based on its capacity to release ferulic acid from wheat bran, rice bran and corn stalks (CSs) when applied in conjunction with xylanase.

Bioinformatic Analysis of LaFae
Following bioinformatic analyses, a gene encoding an FAE (LaFae, GeneBank I.D.: EEJ76955.1) was identified in L. acidophilus [19]. The phylogenetic relationship between LaFae and other FAEs, including FaeLam from Lactobacillus amylovorus, was examined and is shown in a phylogenetic tree (Supplementary Figure S1A,B) [20]. Most FAEs from LAB are classified as type C. LaFae and its homologs are found not only in L. acidophilus, but have also been identified in other Lactobacillus species, thereby indicating its widespread distribution in this genus. It is also worth noting that gene cluster analyses indicated that genes encoding phosphoglycerate mutase and LaFae are adjacently located within the genome of Lactobacillus species (Supplementary Figure S1C). Considering that LaFae has a sequence similarity of 83.4% identity and 90.89% similarity with L. amylovorus (AOR52353) and shows clustering patterns with other known FAEs, we annotate LaFae as a feruloyl esterase.

Characterizations of LaFae
In the present study, we used immobilized Ni 2+ -affinity chromatography to purify LaFae as a His 6 -tagged fusion protein. As shown in Figure 1A, LaFae has a molecular weight of 29 kDa in SDS-PAGE. In addition, strong fluorescence based on activity staining in native-PAGE was detected ( Figure 1B). Furthermore, gel-filtration chromatography indicated that LaFae tends to polymerize into a dimeric conformation ( Figure 1C); the LaFae dimerization in solution was confirmed by the analytical ultracentrifugation (AUC) experimental results. A major peak (95.3%) of dimer appeared at 3.811 sedimentation coefficient position, corresponding to 59.2 kDa ( Figure 1D). The calculated molecular weight of the LaFae monomer was 27.37 kDa based on the amino acid sequence ( Figure 1D). As shown in Figure 1E, LaFae appears to have a strong preference for short or mediumlength substrates, including p-nitrophenyl butyrate (pNB), p-nitrophenyl hexanoate (pNH) and p-nitrophenyl decanoate (pNDe). Similar behavior has been observed in other FAEs, including FaeI from Cellulosilyticum ruminicola [21] and Lp_0796 from L. plantarum [14]. We discovered that the activity of LaFae is optimal within a temperature range of 25 and 37 • C and shows approximately 75% of the optimal activity at 4 • C ( Figure 1F). The LaFae thermostability was determined by incubation for 1 h at temperatures ranging between 25 and 60 • C, which revealed that the activity of LaFae was not substantially altered at temperatures 25 or 37 • C, whereas 70% of the activity disappeared after a 15 min incubation at 50 • C ( Figure 1G). dimerization in solution was confirmed by the analytical ultracentrifugation (AUC) experimental results. A major peak (95.3%) of dimer appeared at 3.811 sedimentation coefficient position, corresponding to 59.2 kDa ( Figure 1D). The calculated molecular weight of the LaFae monomer was 27.37 kDa based on the amino acid sequence ( Figure 1D). As shown in Figure 1E, LaFae appears to have a strong preference for short or medium-length substrates, including p-nitrophenyl butyrate (pNB), p-nitrophenyl hexanoate (pNH) and p-nitrophenyl decanoate (pNDe). Similar behavior has been observed in other FAEs, including FaeI from Cellulosilyticum ruminicola [21] and Lp_0796 from L. plantarum [14]. We discovered that the activity of LaFae is optimal within a temperature range of 25 and 37 °C and shows approximately 75% of the optimal activity at 4 °C ( Figure 1F). The LaFae thermostability was determined by incubation for 1 h at temperatures ranging between 25 and 60 °C, which revealed that the activity of LaFae was not substantially altered at temperatures 25 or 37 °C, whereas 70% of the activity disappeared after a 15 min incubation at 50 °C ( Figure 1G). (E) Substrate specificity analysis of LaFae using different chain lengths of p-nitrophenyl esters. (F) Optimal temperature for LaFae activity. The activity value obtained using pNB as a substrate at 37 °C was considered 100%. (G) Thermal stability of LaFae. Activity at 0 min was considered as 100% for all four temperatures at which the stability was studied. LaFae, L. acidophilus feruloyl esterase; pNB, p-nitrophenyl butyrate. (E) Substrate specificity analysis of LaFae using different chain lengths of p-nitrophenyl esters. (F) Optimal temperature for LaFae activity. The activity value obtained using pNB as a substrate at 37 • C was considered 100%. (G) Thermal stability of LaFae. Activity at 0 min was considered as 100% for all four temperatures at which the stability was studied. LaFae, L. acidophilus feruloyl esterase; pNB, p-nitrophenyl butyrate.
The optimum pH for LaFae activity was found to be 8.0, whereas it retained, approximately, 70% activity at pH 7.0 (Supplementary Figure S2A). Furthermore, we discovered that LaFae retained approximately 65 and 20% of its activity in the presence of 10 and 30% ethanol, respectively, whereas approximately 35% of activity was preserved in the presence of 0.1% SDS (Supplementary Figure S2B). However, only 24% of activity was detected when the enzyme was exposed to 1% of both, Triton X-100 and Tween 20. Interestingly, we observed that the activity of LaFae increased with the increase in NaCl concentration. Specifically, relative activities of 260 and 280% were detected at 1.0 M and 2.0 M NaCl, respectively. Alternatively, only small changes in activity (0-20%) were detected with increasing glycerol concentration (Supplementary Figure S2C).
In addition, we noted a gradual reduction of enzyme activity with the increase in urea concentration. LaFae retained 68% of its original activity when exposed to 0.5 M urea and approximately 25% of its activity remained when the concentration of urea exceeded 2.0 M (Supplementary Figure S2D). Far-UV CD spectra in range 210-220 nm in the presence of urea supported the finding that the secondary structure of LaFae was not completely destroyed in 2.0 M of urea solution (Supplementary Figure S2E). Furthermore, the thermal denaturation of LaFae revealed small changes up to a temperature of 45 • C and the melting temperature was found to be 56 • C (Supplementary Figure S2F).

Enzymatic Properties of LaFae
To evaluate the potential utility of LaFae as a fermentation agent, its activity in the presence of the organic acids, maleic acid, citric acid and lactic acid, as well as sodium metabisulfite was examined (Figure 2A). At 0.05 g/L sodium metabisulfite, LaFae retained approximately 50% of its initial activity, whereas at a 0.25 g/L concentration of the three organic acids, the enzyme showed approximately 60% of its original activity. These observations indicate that LaFae could be effectively used in the fermentation process [22]. Further analysis of LaFae activity in the presence of inorganic salts was performed by exposing the enzyme to different concentrations of CaCl 2 , CoCl 2 , KCl, MgCl 2 and MnCl 2 ( Figure 2B). Results revealed that LaFae was highly activated in the presence of KCl, with approximately 180% of the initial activity being observed in response to treatment with 1.0 M. Moreover, when exposed to 2.0 M KCl, the activity of LaFae remained at 160% of its initial activity. Salt-induced activation was also detected with 0.5 M KCl, MgCl 2 , MnCl 2 and CaCl 2 , respectively, with the enzyme exhibiting >150% of its initial enzyme activity when exposed to 0.5 M CaCl 2 . However, activity was almost completely lost in the presence of ≥1.0 M CoCl 2 . Thus, it seems that that a low ionic strength of cations enhances the activity of LaFae, while high-ionic strength decreases the activity, with the exception of CoCl 2 .
FAEs have a relatively broad substrate specificity in addition to the hydrolysis of HCA esters [14,23,24]. In the present study, we used a colorimetric assay to examine the hydrolytic properties of LaFae with respect to lipids and carbohydrates using a previously described method [25,26]. As shown in Figure 2C, of the two different types of lipids, glyceryl tributyrate GTB) and glyceryl trioleate (GTO), evaluated, substantial LaFae hydrolytic activity was detected only against GTB, as exhibited by the yellow color of the reaction mixture. LaFae activity on carbohydrate esters were examined using three substrates and displayed high activity only toward α-D-glucose pentaacetate ( Figure 2D). In addition, LaFae can synthesize fatty acid esters using soybean oils and alcohols (methanol, ethanol and butanol) as substrates ( Figure 2E). Biosynthesis of methyl, ethyl and butyloleic esters was detected using thin-layer chromatography, indicating that LaFae could be utilized to prepare biodiesels [24,27,28]. Collectively, our biochemical studies indicate that LaFae has promising hydrolytic/synthetic properties that could be harnessed for diverse commercial applications.
Furthermore, we demonstrated that LaFae exhibits a hyperbolic kinetic behavior in the presence of different synthetic substrates. When pNB was used as a substrate, V max and K M values of 0.1 µM s −1 and 1.16 mM were obtained, respectively. Moreover, the catalytic efficiency for pNB (0.55 s −1 mM −1 ) was similar to that obtained for pNA (0.52 s −1 mM −1 ), indicating that LaFae essentially shows an equal affinity toward these two substrates (Supplementary Figure S3). These values are also comparable to those reported for other FAEs, such as Lj0536 [29]. Furthermore, we demonstrated that LaFae exhibits a hyperbolic kinetic behavior in the presence of different synthetic substrates. When pNB was used as a substrate, Vmax and KM values of 0.1 μM s −1 and 1.16 mM were obtained, respectively. Moreover, the catalytic efficiency for pNB (0.55 s −1 mM −1 ) was similar to that obtained for pNA (0.52 s −1 mM −1 ), indicating that LaFae essentially shows an equal affinity toward these two substrates (Supplementary Figure S3). These values are also comparable to those reported for other FAEs, such as Lj0536 [29].

Structural Analysis of LaFae
The unliganded crystal structure of LaFae from L. acidophilus was determined at 2.3 Å resolution ( Figure 3A and Table 1 using the molecular replacement method, which used the crystal structure of cinnamoyl esterase from L. johnsonii (PDB code 3PF8; 70% sequence identity) as a template model. The crystal of unliganded LaFae belongs to the space group of P212121 with two molecules in the asymmetric unit. The crystal structure of LaFae contains nine α helices and eight β strands. The β1 and β2 strands form an antiparallel β-sheet while the β3-β8 strands form parallel β-sheets. Seven α-helices surround the β-sheets and form a core domain that contains an active site with conserved catalytic triad residues (Ser106, Asp197 and His225). The α5-and α6-helices form a cap domain (from Ala133 to Val175) ( Figure 3B). These structural features showed a canonical α/β fold of esterases. Structural homolog search using the DALI server [30] showed that cinnamoyl esterase from L. johnsonii (PDB code 3PF8) had the highest structural similarity with a Z-core of 42.1. In addition, Est1E from Butyrivibrio proteoclasticus (PDB code 2WTM) [31], monoglyceride lipase from Mycobacterium tuberculosis (PDB code 6EIC) [32] and putative hydrolase from Agrobacterium vitis (PDB code 3LLC) showed considerable structural similarities with the LaFae structure (Supplementary Table S1). Effects of (A) organic acids, sodium metabisulfite and (B) inorganic salts on LaFae activity. The LaFae activity in the buffer alone was defined as 100%. All data are presented as mean ± standard deviation (s.d.) (n = 3). The hydrolysis of (C) lipids and (D) carbohydrates was investigated by pH-dependent assay using phenol red. (E) Synthesis of oleic acid esters from oil was investigated by thin-layer chromatography (Soybean oil with 1: methanol, 2: ethanol, 3: butanol). LaFae, L. acidophilus feruloyl esterase; GTB, glyceryl tributyrate; GTO, glyceryl trioleate; α-D-Gluc, α-D-glucose pentaacetate.

Structural Analysis of LaFae
The unliganded crystal structure of LaFae from L. acidophilus was determined at 2.3 Å resolution ( Figure 3A and Table 1 using the molecular replacement method, which used the crystal structure of cinnamoyl esterase from L. johnsonii (PDB code 3PF8; 70% sequence identity) as a template model. The crystal of unliganded LaFae belongs to the space group of P2 1 2 1 2 1 with two molecules in the asymmetric unit. The crystal structure of LaFae contains nine α helices and eight β strands. The β1 and β2 strands form an antiparallel β-sheet while the β3-β8 strands form parallel β-sheets. Seven α-helices surround the β-sheets and form a core domain that contains an active site with conserved catalytic triad residues (Ser106, Asp197 and His225). The α5and α6-helices form a cap domain (from Ala133 to Val175) ( Figure 3B). These structural features showed a canonical α/β fold of esterases. Structural homolog search using the DALI server [30] showed that cinnamoyl esterase from L. johnsonii (PDB code 3PF8) had the highest structural similarity with a Z-core of 42.1. In addition, Est1E from Butyrivibrio proteoclasticus (PDB code 2WTM) [31], monoglyceride lipase from Mycobacterium tuberculosis (PDB code 6EIC) [32] and putative hydrolase from Agrobacterium vitis (PDB code 3LLC) showed considerable structural similarities with the LaFae structure (Supplementary Table S1).
Two protomers of LaFae have been identified in the symmetric unit ( Figure 4A) during the crystal structure determination. In the LaFae crystal structure, the α3, α4, α6 and α7 helices, as well as the β1-β2 hairpin, are involved in the dimerization interface. Several hydrogen bonds (Asp9-Leu11, Asp94-Lys86 and Leu118-Arg171) and salt bridges (Arg8-Asp9, Asp74-Lys86, Asp121-Arg171 and Asp121-His153) were observed at the dimerization interface; however, hydrophobic interactions (Ile81, Ala82, Leu118, Phe168, Val175, Leu176, Pro177 and Ile181) were predominant during the dimerization of LaFae ( Figure 4B). Similar composition of reisues and the dimerization was also found in from Lactobacillus johnsonii (UniProtKB code D3YEX6; PDB code 3PF8), Est1E from Butyrivibrio proteoclasticus (UniProtKB code D2YW37; PDB code 2WTM), an α-β serine hydrolase homologue from Coxiella burnetii (UniProtKB code Q83AV9; PDB code 3TRD) and Esterase Pf2001 from Pyrococcus furiosus (UniProtKB code Q8TZJ1; PDB code 5G59). The conservation scores of amino acids are represented using a color spectrum ranging from black (indicating high conservation) to white (indicating low conservation). The catalytic triad residues (Ser106, Asp197 and His225) are indicated with solid black circles and the residue (Phe34) that affects activity is indicated with a black triangle. The cap domain region of LaFae is indicated by a cyan box. LaFae, L. acidophilus feruloyl esterase.
Two protomers of LaFae have been identified in the symmetric unit ( Figure 4A) during the crystal structure determination. In the LaFae crystal structure, the α3, α4, α6 and α7 helices, as well as the β1-β2 hairpin, are involved in the dimerization interface. Several hydrogen bonds (Asp9-Leu11, Asp94-Lys86 and Leu118-Arg171) and salt bridges (Arg8-Asp9, Asp74-Lys86, Asp121-Arg171 and Asp121-His153) were observed at the dimerization interface; however, hydrophobic interactions (Ile81, Ala82, Leu118, Phe168, Val175, Leu176, Pro177 and Ile181) were predominant during the dimerization of LaFae ( Figure  4B). Similar composition of reisues and the dimerization was also found in the cinnamoyl esterase from L. johnsonii (PDB code 3PF8) and Est1E from B. proteoclasticus (Supplementary Table S1). from Lactobacillus johnsonii (UniProtKB code D3YEX6; PDB code 3PF8), Est1E from Butyrivibrio proteoclasticus (UniProtKB code D2YW37; PDB code 2WTM), an α-β serine hydrolase homologue from Coxiella burnetii (UniProtKB code Q83AV9; PDB code 3TRD) and Esterase Pf2001 from Pyrococcus furiosus (UniProtKB code Q8TZJ1; PDB code 5G59). The conservation scores of amino acids are represented using a color spectrum ranging from black (indicating high conservation) to white (indicating low conservation). The catalytic triad residues (Ser106, Asp197 and His225) are indicated with solid black circles and the residue (Phe34) that affects activity is indicated with a black triangle. The cap domain region of LaFae is indicated by a cyan box. LaFae, L. acidophilus feruloyl esterase.  We also resolved the crystal structure of ethyl ferulate-bound LaFae mutant (S106A) at 2.19 Å resolution and analyzed its substrate binding mode. Ethyl ferulate complexed LaFae crystallized in the C121 space group with four molecules in an asymmetric unit. The substrate-binding pocket of LaFae comprised of hydrophobic residues F34 and V200 and polar residues Q107, Q134, D138, T144, Q145, Y169, D197 and H225 ( Figure 6A). A structural superposition of the ligand-free form and ethyl ferulate-bound LaFae yielded a rootmean-square deviation value of 0.261 Å over 209 residues, indicating a highly similar The active site of LaFae is located within the core domain. The substrate binding pocket is comprised of the β3-α1 loop, β7-α8 loop and β8-α9 loop regions ( Figure 5A). In detail, several polar residues (His105, Gln107, Gln134, Asp138, Thr144, Gln145 and Tyr169) and hydrophobic residues (Phe34, Ala132, Leu135, Phe160, Leu165 and Val199) have been shown to form the ligand-binding pocket. Furthermore, the conserved catalytic triad of Ser106, Asp197 and His225 was also observed to be located in this pocket. The residue Ser106 is located at the end of the β5 strand and the residues Asp167 and His225 are located on the β7-α8 loop and the β8-α9 loop, respectively. The surface surrounding the ligandbinding pocket is positively charged, presumably to attract negatively charged substrates, such as acyl group-containing substrates ( Figure 5B). In addition, the cap domain was observed to be located above the ligand-binding site and this domain was hypothesized to have a gatekeeper role for substrate binding and release of the reaction product.  We also resolved the crystal structure of ethyl ferulate-bound LaFae mutant (S106A) at 2.19 Å resolution and analyzed its substrate binding mode. Ethyl ferulate complexed LaFae crystallized in the C121 space group with four molecules in an asymmetric unit. The substrate-binding pocket of LaFae comprised of hydrophobic residues F34 and V200 and polar residues Q107, Q134, D138, T144, Q145, Y169, D197 and H225 ( Figure 6A). A structural superposition of the ligand-free form and ethyl ferulate-bound LaFae yielded a root-mean-square deviation value of 0.261 Å over 209 residues, indicating a highly similar overall structure ( Figure 6B). In a close-up view of the substrate binding site of the complex structure, the electron density map (contoured at 1.0 σ) of ethyl ferulate was observed near the S106A mutation position ( Figure 7A). Notably, the carbonyl group of ethyl ferulate was discovered to interact directly with the NE2 atom of His225 at a 2.5 Å distance and the 4 -hydroxyl group of ethyl ferulate also formed hydrogen bonds with Asp138 at a 2.4 Å distance ( Figure 7A). In the substrate-bound conformation, two anti-parallel β strands formed a sheet within the cap domain (α5-α6 loop region of the apo structure). Additionally, the comparative b-factor analysis showed that the binding of ethyl ferulate caused a reduced structural flexibility on the α5-α6 loop region ( Figure 7B). It is thought that this loop region plays a role in the open/close movement of LaFae when the substrate is bound.

Mutational Analyses of LaFae
To investigate the roles of the aforementioned amino acids, site-directed mutagenesis was performed to generate six mutants (F34A, S106A, D138A, Q145A and I154A). Following expression and purification, the enzymatic properties of these mutants were examined using p-nitrophenyl esters of different lengths from C2 to C16. In line with our expectations, we observed that the activity of the S106A mutant was almost lost, whereas there were notable reductions in the activities of D138A and I154A mutants ( Figure 8C-E). In contrast, compared to wild-type LaFae, F34A exhibited noticeably enhanced activities ( Figure 8A) of an approximate 1.6-fold increase for pNB, whereas three other mutants (D138A and I154A) retained only about 30% of the wild-type activity for the same substrate. Furthermore, Q145A retained approximately 100% of the activity exhibited by wild-type LaFae for p-nitrophenyl hexanoate as a substrate ( Figure 8B). With respect to the activity of the F34A mutant, we hypothesize that the bulky nature of the Phe side chain enhances binding to short-chain substrates. Within the LaFae structure, Phe34 forms hydrophobic interactions with the Phe72, Met75, Phe160, Leu165 and Tyr169 residues and conceivably stabilizes and controls the movement of the cap domain via hydrophobic interactions. Thus, the F34A mutation may confer a greater cap domain flexibility, as well as a broader substrate binding space. In summary, these mutants could regulate the ferulic acid activity of LaFae, thereby highlighting the important functional roles of these amino acids in catalysis.

Mutational Analyses of LaFae
To investigate the roles of the aforementioned amino acids, site-directed mutagenesis was performed to generate six mutants (F34A, S106A, D138A, Q145A and I154A). Following expression and purification, the enzymatic properties of these mutants were examined using p-nitrophenyl esters of different lengths from C2 to C16. In line with our expectations, we observed that the activity of the S106A mutant was almost lost, whereas there were notable reductions in the activities of D138A and I154A mutants ( Figure 8C-E). In contrast, compared to wild-type LaFae, F34A exhibited noticeably enhanced activities ( Figure 8A) of an approximate 1.6-fold increase for pNB, whereas three other mutants (D138A and I154A) retained only about 30% of the wild-type activity for the same substrate. Furthermore, Q145A retained approximately 100% of the activity exhibited by wild-type LaFae for p-nitrophenyl hexanoate as a substrate ( Figure 8B). With respect to the activity of the F34A mutant, we hypothesize that the bulky nature of the Phe side chain enhances binding to short-chain substrates. Within the LaFae structure, Phe34 forms hydrophobic interactions with the Phe72, Met75, Phe160, Leu165 and Tyr169 residues and conceivably stabilizes and controls the movement of the cap domain via hydrophobic interactions. Thus, the F34A mutation may confer a greater cap domain flexibility, as well as a broader substrate binding space. In summary, these mutants could regulate the ferulic acid activity of LaFae, thereby highlighting the important functional roles of these amino acids in catalysis. Relative activity of wild-type LaFae against pNB were defined as 100%. All the results from mutants (black) were compared to the wild-type LaFae (gray). All data are presented as mean ± standard deviation (s.d.) (n = 3). LaFae, L. acidophilus feruloyl esterase; pNB, p-nitrophenyl butyrate.

Ferulic Acid Production Using LaFae
The ability of LaFae to release ferulic acids from plant biomass was investigated using de-starched wheat bran (DSWB), de-starched rice bran (DSRB) and corn stalks (CSs). Initially, we conducted a plate assay in which LaFae was used to hydrolyze ethyl ferulate on an agar plate. The appearance of halo rings indicated the presence of FAE activity in the different enzymes screened. Although we screened several enzymes derived from LAB Relative activity of wild-type LaFae against pNB were defined as 100%. All the results from mutants (black) were compared to the wild-type LaFae (gray). All data are presented as mean ± standard deviation (s.d.) (n = 3). LaFae, L. acidophilus feruloyl esterase; pNB, p-nitrophenyl butyrate.

Ferulic Acid Production Using LaFae
The ability of LaFae to release ferulic acids from plant biomass was investigated using de-starched wheat bran (DSWB), de-starched rice bran (DSRB) and corn stalks (CSs). Initially, we conducted a plate assay in which LaFae was used to hydrolyze ethyl ferulate on an agar plate. The appearance of halo rings indicated the presence of FAE activity in the different enzymes screened. Although we screened several enzymes derived from LAB for FAE activity, only LaFae was observed to form a large clear zone of hydrolysis on the opaque plates containing ethyl ferulate in medium ( Figure 9A). Similar hydrolysis was obtained for the F34A and Q145A mutants, whereas the I154A mutant produced a notably smaller hydrolysis ring zone ( Figure 9B). Furthermore, HPLC analysis showed that the retention time of LaFae reaction mixture with ethyl ferulate corresponds with ferulic acid standard ( Figure 9C). opaque plates containing ethyl ferulate in medium ( Figure 9A). Similar hydrolysis was obtained for the F34A and Q145A mutants, whereas the I154A mutant produced a notably smaller hydrolysis ring zone ( Figure 9B). Furthermore, HPLC analysis showed that the retention time of LaFae reaction mixture with ethyl ferulate corresponds with ferulic acid standard ( Figure 9C). To evaluate the amounts of ferulic acid from agricultural waste, we initially determined the contents of ferulic acid in DSWB, DSRB and CSs following alkaline hydrolysis with sodium hydroxide, which yielded 3.5, 2.9 and 2.7 mg g −1 , respectively (Supplementary Figure S4). Following incubation with LaFae alone, only relatively small amounts of ferulic acid were released from DSWB (6.3% of that released via chemical extraction), DSRB (3.1%) and CSs (2.9%). In previous studies, it was shown that a combination of xylanase and FAE can enhance the release of ferulic acid [10,33,34]. Therefore, we examined the efficacy of the combination of LaFae and xylanase, derived from the fungus Aspergillus niger. We found that the amount of ferulic acid produced in response to the combined treatment with LaFae and xylanase was approximately 6-10 times higher than that of LaFae alone. Specifically, 37.7%, 34.8% and 42.2% higher amounts of ferulic acid were produced from wheat bran, rice bran and CSs, respectively.

Chemicals and Reagents
All DNA modification enzymes were purchased from Takara Korea Biomedicals (Seoul, Republic of Korea). Nucleic acid preparation kits were obtained from GE Healthcare (Seoul, Republic of Korea) and Qiagen (Seoul, Republic of Korea. Other chemicals were obtained from Sigma-Aldrich (Seoul, Republic of Korea).

Preparation of Wild-Type and Mutants of LaFae
L. acidophilus NCFM (KCTC 3145) was cultured in deMan Rogosa Sharpe medium. The LaFae gene was obtained using polymerase chain reaction (PCR) from L. acidophilus chromosomal DNA. The final product was inserted into a pET-21a vector using BamHI and XhoI restriction enzymes and subsequently introduced into Escherichia coli BL21 (DE3) for heterologous expression of His6-tagged LaFae. Site-directed mutagenesis was per- To evaluate the amounts of ferulic acid from agricultural waste, we initially determined the contents of ferulic acid in DSWB, DSRB and CSs following alkaline hydrolysis with sodium hydroxide, which yielded 3.5, 2.9 and 2.7 mg g −1 , respectively (Supplementary Figure S4). Following incubation with LaFae alone, only relatively small amounts of ferulic acid were released from DSWB (6.3% of that released via chemical extraction), DSRB (3.1%) and CSs (2.9%). In previous studies, it was shown that a combination of xylanase and FAE can enhance the release of ferulic acid [10,33,34]. Therefore, we examined the efficacy of the combination of LaFae and xylanase, derived from the fungus Aspergillus niger. We found that the amount of ferulic acid produced in response to the combined treatment with LaFae and xylanase was approximately 6-10 times higher than that of LaFae alone. Specifically, 37.7%, 34.8% and 42.2% higher amounts of ferulic acid were produced from wheat bran, rice bran and CSs, respectively.

Chemicals and Reagents
All DNA modification enzymes were purchased from Takara Korea Biomedicals (Seoul, Republic of Korea). Nucleic acid preparation kits were obtained from GE Healthcare (Seoul, Republic of Korea) and Qiagen (Seoul, Republic of Korea). Other chemicals were obtained from Sigma-Aldrich (Seoul, Republic of Korea).

Preparation of Wild-Type and Mutants of LaFae
L. acidophilus NCFM (KCTC 3145) was cultured in deMan Rogosa Sharpe medium. The LaFae gene was obtained using polymerase chain reaction (PCR) from L. acidophilus chromosomal DNA. The final product was inserted into a pET-21a vector using BamHI and XhoI restriction enzymes and subsequently introduced into Escherichia coli BL21 (DE3) for heterologous expression of His 6 -tagged LaFae. Site-directed mutagenesis was performed and all mutants obtained (F34A, S106A, D138A, Q145A and I154A) were confirmed through DNA sequencing. Primers used to generate the clone and mutants are listed in Supplementary Table S2.
Recombinant E. coli BL21 (DE3) were cultured in LB medium containing ampicillin (100 µg/mL) to an OD 600nm of 0.6 at 37 • C. Subsequently, expression of the recombinant LaFae was induced by addition of 1 mM of isopropyl-β-D-1-thiogalactosideand incubation for 4 h, after which cells were harvested through centrifugation at 2000 rpm for 15 min. Cell disruption was carried out by sonication in the lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 30 mM Imidazole). The supernatant was separated by centrifugation at 16,000 rpm for 40 min and loaded onto a HisTrap column (GE Healthcare, Amersham, UK). The recombinant LaFae protein was subsequently eluted with the elution buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 300 mM Imidazole). Subsequent purification and buffer exchange were conducted through size-exclusion chromatography using the protein storage buffer (20 mM Tris-HCl, pH 8.5) and the purity of the protein was assessed using SDS-PAGE.

Oligomerization of LaFae
Oligomerization of LaFae in solution was investigated by gel-filtration chromatography and analytical ultracentrifugation (AUC). For gel-filtration chromatography, the protein standard mix, including cytochrome C (12.4 kDa), carbonic anhydrase (29 kDa), albumin (66 kDa), alcohol dehydrogenase (150 kDa) and β-amylase (200 kDa) was eluted in 20 mM Tris-HCl (pH 8.0) using HiPrep S-200R column and the standard curve was generated. The LaFae was eluted using the same method described above and the molecular weight in the solution was calculated. The AUC was performed in 20 mM Tris-HCl (pH 8.0) buffer using a ProteomeLab XL-A (Beckman Coulter, Carlsbad, CA, USA). The LaFae was centrifuged at 14,225× g for 10 min. The sedimentation profile monitored at 280 nm was analyzed using the SEDFIT program. Coomassie Brilliant Blue R-250 and 4-methylumbelliferone acetate were used for activity staining during the protein electrophoresis analysis [35]. Briefly, the overlay activity assay was conducted using native-PAGE analysis. The resulting gel was then washed three times using a storage buffer. After the washing steps, the gel was incubated with 4-methylumbelliferone acetate, allowing the detection of a fluorescent signal under UV illumination.

Chemical Stability Assay
The organic solvents, ethanol, isopropyl alcohol, sodium dodecyl sulfate (SDS), Triton-X100 and Tween-20 were used to examine the chemical stability of LaFae. Additionally, the enzymatic activities of LaFae in 0-2 M of NaCl, 5-40% of glycerol and 0-3 M of urea solution were investigated to assess its industrial application potential. The LaFae activity was estimated after incubation for 1 h in the respective solutions and the activity of LaFae in the storage buffer alone was considered as 100%. Additionally, the effects of various types of acids and salts on LaFae was estimated after 1 h incubation in 0, 0.05, 0.1, 0.25, 0.5 g/L of three types of acid (maleic acid, citric acid, lactic acid) and 0, 0.1, 0.5, 1.0, 2.0 M of five types of salt (CaCl 2 , CoCl 2 , KCl, MgCl 2 , MnCl 2 ) using an Epoch 2 Microplate Spectrophotometer (BioTek, Winooski, VT, USA). The activity of LaFae using p-nitrophenyl butyrate as substrate in buffer alone was set as 100% of relative activity and all experiments were performed in triplicate. The mean ± standard deviation (s.d.) (n = 3) were calculated and presented in figures.

Thermal Stability and Circular Dichroism Analysis
The effect of temperature on the stability of LaFae was examined by incubating the enzyme at temperatures between 25 and 60 • C and measuring the residual activity for 1 h, at 15 min intervals. While measurements were performed in triplicate, the p-nitrophenyl butyrate was used as substrate and the activity at 25 • C was set as 100%. Far-ultraviolet (UV) circular dichroism (CD) analysis was performed to monitor the structural denaturation profile with increasing temperature. Data collection was carried out in 1 mm pathlength cell with a 0.5 nm bandwidth and 1 s response time. An average of three accumulations was used to generate the final spectra. The spectra were obtained at 190-240 nm (Jasco, Tokyo, Japan) to evaluate the secondary structure of LaFae, and the thermal unfolding profile was monitored from 25 to 80 • C at 222 nm in a thermostatic cell holder.

Hydrolytic Activity Assay
To examine the hydrolytic activity of LaFae on lipids and carbohydrate substrates, a pH-indicator-based colorimetric assay was performed on a microplate [25,27]. Glyceryl tributyrate and trioleate were used as lipid-type substrates and α-D-glucose pentaacetate, cellulose acetate and N-acetyl-glucosamine were used as acetylated carbohydrate substrates. In the phenol-red containing solution of each substrate, 100 µg LaFae enzyme was added to start a hydrolytic reaction. As a result of the pH shift assay of LaFae, the hydrolysis reaction changed the color of the phenol-red containing mixture to yellow. For the synthesis of oleyl esters, three types of alcohols (methanol, ethanol and 1-butanol) were incubated with soybean oil and 100 µg LaFae by continuous shaking at 37 • C, and reaction products were analyzed by thin-layer chromatography.

Crystallization and Data Collection
The wild-type apo and S106A-ethyl ferulate complex of LaFae were crystallized by the sitting-drop vapor-diffusion method at 296 K using a mosquito crystallization robot (TTP LabTech, Melbourn, UK). Commercially available crystallization screening solution kits were used, including MCSG I-IV, MIDAS, Morpheus (Molecular Dimensions, Rotherham, UK), Index and SaltRx (Hampton Research, Aliso Viejo, CA, USA). Each crystallization drop was set with 300 nL protein solution and 300 nL reservoir solution and equilibrated against an 80 µL reservoir solution at 296 K. The optimized crystals of wild-type apo LaFae were observed using 0.1 M Bis-Tris:HCl at pH 6.5 and 20% (w/v) PEG MME 5000 (MCSG I #F1) after 1-2 d. For co-crystallization of the S106A mutant with substrate, ethyl ferulate was dissolved in ethanol. The molar ratio of protein to ligand was 1:10 at a final ligand concentration of 3.5 mM. The complex crystals of S106A-ethyl ferulate were observed after 4 d in optimized conditions of 0.02 M magnesium chloride, 0.1 M HEPES:NaOH, pH 7.5 and 22% (w/v) polyacrylic acid 5100 (MCSG II #F5), using hanging-drop vapordiffusion method at 296 K. The single crystal of wild-type LaFae was mounted without cryoprotectant and the S106A-ethyl ferulate complex crystal was mounted after being gently soaked in 20% glycerol mixed in a crystallization reservoir solution as cryoprotectant. X-ray diffraction data were collected from crystals flash-frozen at 193 K under liquid nitrogen at BL-5C beam line of the Pohang Accelerator Laboratory (PAL, Pohang, Republic of Korea) with an oscillation of 1 • per image. A total of 200 and 360 diffraction images of the wild-type and the S106A-ethyl ferulate complex crystals were collected, respectively. Each data set was indexed, integrated and scaled using the HKL-2000 software [36]. The detailed X-ray diffraction data collection statistics are shown in Table 1.

Structure Determination and Refinement
The crystal structures of wild-type apo and S106A-ethyl ferulate complex of LaFae were determined at 2.3 Å and 2.19 Å resolution, respectively. The phase problem was solved through the molecular replacement method using the MOLREP program from the CCP4i suite [37]. The crystal structure of cinnamoyl esterase from Lactobacillus johnsonii (PDB code 3PF8), which shares 70% sequence identity with LaFae, was used as a search model to define the apo structure of LaFae. The refined apo structure was used as a template for the structural analysis of S106A-ethyl ferulate complex. During the manual model building of the S106A-ethyl ferulate complex structure, all ligands shown on each chain were modeled with an occupancy of 1.0. The initial models of both apo and complex structures were iteratively rebuilt and refined using Coot [38], REFMAC [39] and phenix.refine programs [40]. The final structures exhibited R cryst and R free values of 0.227 and 0.280 in apo and 0.186 and 0.213 in complex structures, respectively. Validation of both final models was carried out using MolPorbity [41]. Structure refinement statistics are listed in Table 1. All structural figures were generated using the PyMOL molecular-graphics system [42]. The atomic coordinate and structure factor of wild-type apo and S106A-ethyl ferulate complex of LaFae have been deposited in the Protein Data Bank under accession codes 7XRH and 7XRI, respectively.

Determination of LaFae Feruloyl Esterase Activity
To verify FAE activity, the plate assay was performed at 37 • C, as previously described [43]. The circular zones were generated as a result of ethyl ferulate hydrolysis. Previously characterized enzymes were used to examine feruloyl esterase activity [44]. LaFae activity was estimated using ethyl ferulate (1 mM) as a substrate; 10 µL of 1 mg/mL LaFae was added to 300 µL of reaction mixture. After incubation, the mixture was analyzed using an Eclipse Plus RR-C18 column with a 1220 Infinity II LC system (Agilent, Santa Clara, CA, USA). The samples were separated in a mobile phase of acetonitrile: water (80:20) at 30 • C [45].
To produce ferulic acid from agricultural biomass by LaFae, wheat bran, rice bran and corn stalks were firstly de-starched by incubation with 1 M NaOH at 70 • C. The extraction of ethyl ferulate was performed in an 800 µL reaction mixture containing 20 mg of destarched wheat bran, rice bran and corn stalks and 5 mg LaFae in 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl [46,47]. To investigate the synergism between LaFae and xylanase, we analyzed the amounts of ferulic acid produced after incubation of substrates with LaFae and A. niger xylanase (Megazyme, Bray, Ireland). The amounts of released ferulic acid were measured using HPLC, as described above. All experiments were performed in triplicate and each error rate was calculated and presented in figures.

Conclusions
Herein, we provide a comprehensive characterization of a universal feruloyl acid esterase (LaFae) from LAB, and its biotechnological applications in ferulic acid production from plant biomass. Several residues surrounding the active site were mutated to identify the catalytically important residues for substrate binding and specificity of LaFae and their activities relative to the wild-type were measured. Interestingly, the F34A mutant exhibited relatively increased catalytic activities against pNB and pNH. Since it is thought that the Phe34 plays a role of physically supporting and connecting the cap domain through hydrophobic interactions, the F34A mutation may provide a more flexible cap domain movement, as well as a broader substrate binding space. Furthermore, the conformational change of the cap domain region in substrate-free, as well as the ethyl ferulate bound structure, presented open and closed movements of LaFae when harboring ethyl ferulate as a substrate. Additionally, we showed that LaFae from the probiotic bacterium L. acidophilus can be directly used to enhance the yields of high-value hydroxycinnamates from natural resources. Thus, we believe that LaFae will have potential applications in foods, cosmetics, animal feed and biodiesel industries.