Glutamate Decarboxylase from Lactic Acid Bacteria—A Key Enzyme in GABA Synthesis

Glutamate decarboxylase (l-glutamate-1-carboxylase, GAD; EC 4.1.1.15) is a pyridoxal-5’-phosphate-dependent enzyme that catalyzes the irreversible α-decarboxylation of l-glutamic acid to γ-aminobutyric acid (GABA) and CO2. The enzyme is widely distributed in eukaryotes as well as prokaryotes, where it—together with its reaction product GABA—fulfils very different physiological functions. The occurrence of gad genes encoding GAD has been shown for many microorganisms, and GABA-producing lactic acid bacteria (LAB) have been a focus of research during recent years. A wide range of traditional foods produced by fermentation based on LAB offer the potential of providing new functional food products enriched with GABA that may offer certain health-benefits. Different GAD enzymes and genes from several strains of LAB have been isolated and characterized recently. GABA-producing LAB, the biochemical properties of their GAD enzymes, and possible applications are reviewed here.


Introduction
Lactic acid bacteria (LAB) are Gram-positive, acid-tolerant, non-spore forming bacteria, with a morphology of either cocci or rods that share common physiological and metabolic characteristics. Even though many genera of bacteria produce lactic acid as their primary or secondary metabolic end-product, the term 'lactic acid bacteria' is conventionally reserved for genera in the order Lactobacillales, which includes Lactobacillus, Lactococcus, Leuconostoc, Pediococcus, and Streptococcus, in addition to Carnobacterium, Enterococcus, Oenococcus, Tetragenococcus, Vagococcus, and Weisella. LAB are important for a wide range of fermented foods and are widely used as starter cultures in traditional and industrial food fermentations [1].
Lactic acid formed during the fermentation of carbohydrates as one of the main metabolic products can affect the physiological activities of LAB. Under acidic conditions, several LAB have developed different acid-resistance systems to maintain cell viability. These systems include, for example, the F 0 F 1 -ATPase system or cation/proton antiporter/symporter systems such as K + -ATPase, which contribute to pH homeostasis in the cytosol by the translocation of protons [2]. In addition, glutamate or arginine-dependent systems, which require the presence of glutamate and arginine, respectively, as substrates, contribute to the acid resistance of LAB. The first enzyme in the arginine-dependent system is arginine deiminase, which degrades arginine to citrulline and NH 3 .

Biodiversity of Glutamate to γ-Aminobutyric Acid (GABA)-Producing Lactic Acid Bacteria
LAB are among the most important organisms when it comes to the fermentation of various food raw materials. They efficiently and rapidly convert sugars into lactic acid as their main metabolic product (or one of their main products), and thus contribute to the preservation of these fermented foods. Many of these raw materials or foods contain glutamate in significant amounts, which can be utilized by LAB to increase their tolerance against acidic conditions. Hence, a number of GABAproducing LAB have been isolated from a wide range of fermented foods including cheese, kimchi, paocai, fermented Thai sausage nham, or various fermented Asian fish products [2,13,25,26,30] (Table  1). Table 1. Diversity of glutamate to γ-aminobutyric acid (GABA)-converting lactic acid bacteria (LAB), isolation sources, GABA production, and fermentation conditions. GABA concentrations as found in GABA is primarily produced via different biotechnological approaches using either isolated GAD in a biocatalytic approach or various microbial strains [17], rather than through chemical synthesis due to the corrosive nature of the reactant compound [18]. GABA is currently commercialized as a nutritional supplement, however, interest in GABA-enriched food, in which GABA is formed in situ via fermentation using appropriate microorganisms, has increased lately in parallel to a general interest in functional foods. As GABA is formed as a by-product of food fermentations, LAB, which play an eminent role in the fermentation of a wide range of different products, are of particular importance when talking about GABA-enriched food. Hence, it is not surprising that strains isolated from various fermented food sources had first been shown to have the ability to produce GABA, for example, Lactobacillus namurensis NH2 and Pediococcus pentosaceus NH8 from nham [19], Lactobacillus paracasei Microorganisms 2020, 8,1923 3 of 24 NFRI 7415 from Japanese fermented fish [20], L. paracasei PF6, Lactococcus lactis PU1 and Lactobacillus brevis PM17 from cheese [21], L. brevis CGMCC 1306 from unpasteurized milk [22], L. brevis GABA100 from kimchi [23,24], L. brevis BJ20 from fermented sea tangle [25], Lactobacillus futsaii CS3 from Thai fermented shrimp [26] and L. brevis 119-2 and L. brevis 119-6 from tsuda kabu [12]. Recently, many studies have focused on the identification of novel GABA-producing LAB and investigated the biochemical properties of GAD from different strains in more detail [12,14,15,[27][28][29].
Here, we outline the presence of gad genes in LAB as important and efficient GABA-producing organisms together with a phylogenetic analysis, we summarize the biochemical data available for GAD from different LAB, and finally, we give an outlook on potential applications of GAD in the manufacture of bio-based chemicals.

Biodiversity of Glutamate to γ-Aminobutyric Acid (GABA)-Producing Lactic Acid Bacteria
LAB are among the most important organisms when it comes to the fermentation of various food raw materials. They efficiently and rapidly convert sugars into lactic acid as their main metabolic product (or one of their main products), and thus contribute to the preservation of these fermented foods. Many of these raw materials or foods contain glutamate in significant amounts, which can be utilized by LAB to increase their tolerance against acidic conditions. Hence, a number of GABA-producing LAB have been isolated from a wide range of fermented foods including cheese, kimchi, paocai, fermented Thai sausage nham, or various fermented Asian fish products [2,13,25,26,30] (Table 1). Table 1. Diversity of glutamate to γ-aminobutyric acid (GABA)-converting lactic acid bacteria (LAB), isolation sources, GABA production, and fermentation conditions. GABA concentrations as found in food products fermented with this strain are given. Lactobacillus spp. are the most predominant species that have been described as GABA-producing organisms including, for example, L. brevis, L. paracasei, L. bulgaricus, L. buchneri, L. plantarum, L. helveticus, or L. futsaii [21,[30][31][32][33]42,43]. Among these, L. brevis, a heterofermentative LAB, is one of the best-studied organisms [43] and is known for forming high levels of GABA under appropriate conditions (Table 1). Traditionally, fermented food samples containing GABA are used to screen for and isolate GABA-producing LAB, and it is not surprising that food samples with high GABA content may result in the isolation of promising strains showing good GABA-forming properties. Furthermore, the adjustment of the pH medium to an acidic condition (pH 4.5-5.5) could improve GABA production since GABA biosynthesis is closely related to the pH. Typical fermented foods used for isolating GABA-producing LAB are kimchi, where in one study, 68 out of 230 LAB isolates showed the ability to convert glutamate to GABA [44]; Thai fermented fish plaa-som [45], or other fermented vegetable (kimchi) [46]; fermented shrimp paste [47]; cheese [16] or milk products as well as various fermented meat or fish products including sausages or traditional fermented Cambodian food, mainly based on fish, where six out of 68 LAB isolates showed a significant GABA-producing ability [1]. These screening/isolation strategies often resulted in the identification of strains capable of efficiently converting glutamate or in the discovery of novel, not-yet-identified producers of GABA, which show promise as starter cultures for various fermented foods enriched in GABA. For example, the novel GABA producer Lactobacillus zymae, which can grow on up to 10% NaCl and is able to utilize D-arabitol as a carbon source, was isolated from kimchi [46]. Recently, Sanchart et al. isolated the novel GABA-forming strain L. futsaii CS3 with probiotic properties from fermented shrimp (Kung-som) [26,47]. This isolate was able to convert 25 mg/mL of monosodium glutamate to GABA with a yield of more than 99% within 72 h. These studies (Table 1) showed that the genera Lactobacillus and Lactococcus are the predominant GABA-producing LAB, but also other genera such as Enterococcus were studied in this respect. A novel GABA-producing Enterococcus avium strain was isolated from Korean traditional fermented anchovy and shrimp (jeotgal) and was shown to produce 18.47 mg/mL GABA within 48 h in a medium containing glutamate as the substrate. A recent study looking at LAB isolated from traditional Japanese fermented fish products (kaburazushi, narezushi, konkazuke, and ishiru) showed that out of 53 randomly picked LAB isolates, 10 showed the ability to transform considerable amounts of glutamate into GABA, and identified Weissella hellenica as a novel GABA producer [41]. Thus, these new genera expand the list of GABA-producing bacteria, which can open up new and different applications in the food industry. This may lead to a wider application and flexibility of starter cultures in the food industry [9]. Production of GABA by different LAB together with fermentation conditions, yields, and productivities has recently been reviewed in detail [15,43,48].

Occurrence and Organization of Glutamic Acid Decarboxylase (GAD) Genes
The conversion of glutamate to γ-aminobutyric acid is catalyzed by glutamate decarboxylase (glutamic acid decarboxylase, GAD, systematic name l-glutamate 1-carboxy-lyase (4-aminobutanoateforming), EC 4.1.1.15), which catalyzes the irreversible α-decarboxylation of glutamate [5,48]. GAD employs pyridoxal-5 -phosphate as its cofactor, and is found in numerous microorganisms such as bacteria [3], fungi [49], and yeasts [50]; furthermore, GAD is found in plants [51], insects, and vertebrates [52]. GAD is an intracellular enzyme that is utilized by LAB to encounter acidic stress by decreasing the proton concentration in the cytoplasm in the presence of l-glutamate ( Figure 2) [2,6,53,54]. This system, the so-called glutamate-dependent acid-resistance system (GDAR), provides protection under the acidic condition, and therefore the ability of LAB to perceive and cope with acid stress is crucial for successful colonization of the gastrointestinal tract (GIT) and survival under acidic environments such as in fermented food. The GDAR system consists of two homologous inducible glutamate decarboxylases, GadA and GadB, and the glutamate/γ-aminobutyrate antiporter GadC [20,48]. The corresponding genes (i.e., gadA, gadB, and gadC) are expressed upon entry into the stationary phase when cells are growing in rich media independently of pH, and are further induced upon hypoosmotic and hyperosmotic stress, or in the log-phase of growth in minimal medium containing glucose at a pH of 5.5 [53,55]. Siragusa et al. demonstrated that three strains with a GDAR system, L. bulgaricus PR1, L. lactis PU1, and L. plantarum C48, were able to survive and synthesize GABA under simulated gastrointestinal conditions [21]. Recently, cell numbers of the GABA-producing strain L. futsaii CS3 were shown to be only decreased by 1.5 log cycles under simulated gastrointestinal conditions, indicating that the GDAR system contributes to resistance to the conditions in the GIT and that GABA-producing LAB thus have the potential as functional probiotic starter cultures [47].  A genomic survey was conducted by Wu et al. to gain insight on the distribution of the gad operon and genes encoding glutamate decarboxylase in LAB [7]. Most strains of L. brevis (14 strains) as well as some strains of L. reuteri (six strains), L. buchneri (two strains), L. oris (three strains), L. lactis (29 strains), and L. garvieae (five strains) were shown to have an intact gad operon. The majority of these strains were shown to contain either gadA or gadB, whereas gadC is only present in the genomes of certain strains and noticeably lacking in L. plantarum, suggesting that the characteristic of GABA GAD systems and the organization of the gad operons among LAB species are highly variable [56,57]. Numerous studies reported that some LAB species such as Streptococcus thermophilus [5], L. brevis [6,7], or L. lactis [3] have one or two gad genes (i.e., gadA, gadB), together with the antiporter (gadC). Interestingly, E. avium 352 carries three gad genes [58]. Typically, L. brevis contains two GAD-encoding genes, gadA and gadB, which when expressed yield GAD enzymes that share approximately 50% amino acid sequence similarity [6]. In contrast, the gadB gene is absent in strain L. brevis CD0817 [59] and the amino acid sequence identities of GadA and GadC from L. brevis CD0817 against other L. brevis strains are 91% and 90%, respectively. The transcriptional regulator gene gadR plays a crucial role in GABA production and acid resistance in L. brevis. Gong et al. reported that deletion of gadR in L. brevis ATCC 367 resulted in lower expression of both the gadB and gadC gene, a concurrent reduction in GABA synthesis, and an increased sensitivity to acidic conditions [6]. Expression levels of gadR are varied among different LAB strains. The gadR gene was expressed 13-155-fold higher than gadCB in L. brevis NCL912 during the cultivation period [60]. In contrast, expression of gadR in L. brevis CGMCC1306 was observed to be much lower compared to gadCB. The role of GadA and GadB in L. brevis CGMCC1306 was investigated by disruption of the genes gadA, gadB, and gadC, resulting in complete elimination of GABA formation and increased sensitivity to acidic conditions, suggesting that both GAD proteins and the antiporter are essential for GABA production and acid resistance [61].
A genomic survey was conducted by Wu et al. to gain insight on the distribution of the gad operon and genes encoding glutamate decarboxylase in LAB [7]. Most strains of L. brevis (14 strains) as well as some strains of L. reuteri (six strains), L. buchneri (two strains), L. oris (three strains), L. lactis (29 strains), and L. garvieae (five strains) were shown to have an intact gad operon. The majority of these strains were shown to contain either gadA or gadB, whereas gadC is only present in the genomes of certain strains and noticeably lacking in L. plantarum, suggesting that the characteristic of GABA production is strain-dependent. Similar results were obtained by Yunes et al., who showed that L. fermentum (9 strains), L. plantarum (30 strains), and L. brevis (3 strains) typically contain gadB genes. In addition, no antiporter gene was observed next to gadB in L. plantarum 90sk, and the expression of gadB was increased in the early stationary phase and at low pH (3.5-5) [62]. The gadB gene from S. thermophilus encoding 459 amino acids has been investigated. The transposase genes Tn1216 (5 and 3 ) and Tn1546 are located downstream and upstream of hydrolase genes flanking the gadB/gadC operon as a result from horizontal gene transfer. This sequence implies that the order of gadB and gadC in S. thermophilus ST110 is similar to S. thermophilus Y2 [63], but in a different order from that reported for L. lactis [64], L. brevis [60], and L. plantarum [62].
The L. reuteri 100-23 genome was investigated by Su et al. for its gad operon [65]. This genome contains gadB and two genes for the antiporter (gadC1 and gadC2) as well as the glutaminase-encoding gene gls3, indicating that glutamine serves as a substrate for the synthesis of GABA. The organization of the gad operon is in a different order for other species of LAB (L. lactis and L. plantarum) as glutaminase (gls3) is in between the antiporters gadC1 and gadC2, while gadB is accompanied by gadC1 [65]. The full length of gad genes has been cloned and sequenced for several species and strains of LAB. Li et al. cloned gadA from L. brevis NCL912, and the whole gene fragment (4615 bp) including gadR, gadC, gadA, and gts (glutamyl t-RNA synthetase) was successfully amplified. Their work suggested that the high GABA production capacity of L. brevis NCL912 may be linked to the gadA locus, forming a gadCA operon complex that ensures the coordinated expression of GAD and the antiporter [60]. A core fragment of the gad gene from L. brevis OPK3 was cloned and successfully expressed in Escherichia coli. The nucleotide sequence revealed that the open reading frame of the gad gene consisted of 1401 bases encoding 467 amino acid residues. The sequence showed 83%, 71%, and 60% homology to GAD from L. plantarum, L. lactis, and Listeria monocytogenes, respectively [66].
A phylogenetic tree constructed from available GAD sequences in the NCBI protein database showed that amino acid sequences of GAD are highly conserved within the same species (Figure 3), and that GAD is widely distributed in a number of LAB including L. brevis, L. buchneri, L. delbrueckii subsp. bulgaricus, L. fermentum, L. futsaii, L. paracasei, L. parakefiri, L. paraplantarum, L. plantarum, L. plantarum subsp. argentoratensis, L. reuteri, L. sakei, L. lactis, and S. thermophilus. All of these LAB are commonly found in fermented foods and some of these are commonly used as starter cultures in food industries. In addition, GAD is also found in other lactobacilli including L. acidifarinae, L. aviaries, L. coleohominis, L. farraginis, L. japonicas, L. koreensis, L. nuruki, L. oris, L. rossiae, L. rennini, or L. suebicus ( Figure 3). These organisms have not been studied for their capacity to synthesize GABA nor have their GAD system been studied, and hence they could be of interest with respect to GABA production and GABA-enriched food.

Glutamate Decarboxylase
Glutamate decarboxylase is an intracellular enzyme that is found ubiquitously in eukaryotes and prokaryotes. GAD exhibits different physiological roles, especially in vertebrates and plants, and its presence is highly variable among organisms [52]. GAD is a PLP-dependent enzyme and as such belongs to the PLP-dependent enzyme superfamily. This superfamily comprises seven different folds [67] with GAD from LAB showing the type-I fold of PLP-dependent enzymes [68]. A number of important catalytic reactions including αand β-eliminations, decarboxylation, transamination, racemization, and aldol cleavage are catalyzed by various members of this superfamily of enzymes [69]. GAD activity relies on the binding of its co-factor PLP, and belongs to group II of PLP-dependent decarboxylases [70]. In GAD from L. brevis CGMCC 1306, the active site entrance is located at the re-face of the cofactor PLP. PLP is covalently attached to a lysine (K279) via an imine linkage (Figure 4), referred to as an internal aldimine [68,71]. This lysine is strictly conserved in group II PLP-dependent decarboxylases. The corresponding lysine in E. coli GAD is at position 276, and when mutating this residue, the variant has less flexibility and affinity to both its substrate and the cofactor [72]. In addition to this covalent attachment, PLP is positioned in the active site via a number of H bonds between the phosphate group of PLP and surrounding amino acids, while the pyridine ring of PLP forms hydrophobic interactions with side chains of various amino acids in the active site [68].
its presence is highly variable among organisms [52]. GAD is a PLP-dependent enzyme and as such belongs to the PLP-dependent enzyme superfamily. This superfamily comprises seven different folds [67] with GAD from LAB showing the type-I fold of PLP-dependent enzymes [68]. A number of important catalytic reactions including α-and β-eliminations, decarboxylation, transamination, racemization, and aldol cleavage are catalyzed by various members of this superfamily of enzymes [69]. GAD activity relies on the binding of its co-factor PLP, and belongs to group II of PLP-dependent decarboxylases [70]. In GAD from L. brevis CGMCC 1306, the active site entrance is located at the reface of the cofactor PLP. PLP is covalently attached to a lysine (K279) via an imine linkage (Figure 4), referred to as an internal aldimine [68,71]. This lysine is strictly conserved in group II PLP-dependent decarboxylases. The corresponding lysine in E. coli GAD is at position 276, and when mutating this residue, the variant has less flexibility and affinity to both its substrate and the cofactor [72]. In addition to this covalent attachment, PLP is positioned in the active site via a number of H bonds between the phosphate group of PLP and surrounding amino acids, while the pyridine ring of PLP forms hydrophobic interactions with side chains of various amino acids in the active site [68].
Molecular docking of the substrate glutamate into the active-site of the holo-form of L. brevis GAD showed several noncovalent interactions including hydrogen bonds between the O2, the O3 and the O4 atoms of the substrate L-Glu to various parts of the GAD polypeptide chain. Furthermore, electrostatic interactions between the negatively charged oxygen atom of the α-carboxyl and the γcarboxyl group of L-Glu and the positively charged nitrogen atom of residue R422 as well as H278 and K279 ( Figure 5), respectively, were proposed [68]. The flexible loop residue Tyr308-Glu312 in L. brevis GAD is located near the substrate-binding site ( Figure 4). This loop is important for the catalytic reaction, and the conserved residue Tyr308 plays a crucial role in decarboxylation of L-Glu. Thr 215 and Asp246 are the two catalytic residues in L. brevis GAD ( Figure 5), which are also highly conserved and promote decarboxylation of L-Glu [68,71,73]. Molecular docking of the substrate glutamate into the active-site of the holo-form of L. brevis GAD showed several noncovalent interactions including hydrogen bonds between the O2, the O3 and the O4 atoms of the substrate L-Glu to various parts of the GAD polypeptide chain. Furthermore, electrostatic interactions between the negatively charged oxygen atom of the α-carboxyl and the γ-carboxyl group of L-Glu and the positively charged nitrogen atom of residue R422 as well as H278 and K279 ( Figure 5), respectively, were proposed [68]. The flexible loop residue Tyr308-Glu312 in L. brevis GAD is located near the substrate-binding site ( Figure 4). This loop is important for the catalytic reaction, and the conserved residue Tyr308 plays a crucial role in decarboxylation of L-Glu. Thr 215 and Asp246 are the two catalytic residues in L. brevis GAD ( Figure 5), which are also highly conserved and promote decarboxylation of L-Glu [68,71,73]. During catalysis, a transamination reaction occurs, and PLP, which is covalently attached to a Lys in the active site of GAD in its resting state, now becomes covalently bonded to the substrate glutamate, forming a Schiff base or what is referred to as an external aldimine. This Schiff base can then be transformed to a quinonoid intermediate [67,74]. In a small fraction of catalytic cycles, when glutamate is decarboxylated, a subsequent alternative transamination of the quinonoid intermediate of the reaction can occur, and succinic semialdehyde (SSA) and pyridoxamine-5′-phosphate (PMP) are formed. The latter will immediately be released from the enzyme, resulting in inactive apoGAD ( Figure 6), which can be regenerated to the active GAD-PLP complex when free pyridoxal-5′phosphate is present, thus completing a cycle of inactivation and activation. However, when free PLP is not present, GAD will be inactivated as a function of time and substrate concentration [62,[67][68][69][74][75][76][77].   During catalysis, a transamination reaction occurs, and PLP, which is covalently attached to a Lys in the active site of GAD in its resting state, now becomes covalently bonded to the substrate glutamate, forming a Schiff base or what is referred to as an external aldimine. This Schiff base can then be transformed to a quinonoid intermediate [67,74]. In a small fraction of catalytic cycles, when glutamate is decarboxylated, a subsequent alternative transamination of the quinonoid intermediate of the reaction can occur, and succinic semialdehyde (SSA) and pyridoxamine-5 -phosphate (PMP) are formed. The latter will immediately be released from the enzyme, resulting in inactive apoGAD (Figure 6), which can be regenerated to the active GAD-PLP complex when free pyridoxal-5 -phosphate is present, thus completing a cycle of inactivation and activation. However, when free PLP is not present, GAD will be inactivated as a function of time and substrate concentration [62,[67][68][69][74][75][76][77].
with carbons being green. All images were made using the PyMOL Molecular Graphics System, v. 2.3.0. for Linux. Figure 5. Active site of glutamate decarboxylase from L. brevis (PDB code 5GP4). The conserved catalytic residues T215 and D246 are shown in sticks, colored by atom type, with carbons shown in green, nitrogen in blue, and oxygen in red. The prosthetic group PLP is represented as sticks colored by atom type with carbons in magenta. The residues R422, H278, and K279, proposed to be involved in electrostatic interactions with the substrate glutamate [68], are represented as sticks colored by atom type. The rest of the chain is shown as a transparent orange cartoon. K279 is also involved in forming the imine linkage to PLP. The image was made using the PyMOL Molecular Graphics System, v. 2.3.0. for Linux.
During catalysis, a transamination reaction occurs, and PLP, which is covalently attached to a Lys in the active site of GAD in its resting state, now becomes covalently bonded to the substrate glutamate, forming a Schiff base or what is referred to as an external aldimine. This Schiff base can then be transformed to a quinonoid intermediate [67,74]. In a small fraction of catalytic cycles, when glutamate is decarboxylated, a subsequent alternative transamination of the quinonoid intermediate of the reaction can occur, and succinic semialdehyde (SSA) and pyridoxamine-5′-phosphate (PMP) are formed. The latter will immediately be released from the enzyme, resulting in inactive apoGAD ( Figure 6), which can be regenerated to the active GAD-PLP complex when free pyridoxal-5′phosphate is present, thus completing a cycle of inactivation and activation. However, when free PLP is not present, GAD will be inactivated as a function of time and substrate concentration [62,[67][68][69][74][75][76][77].

Biochemical Insights into Glutamate Decarboxylase from Lactic Acid Bacteria
GAD from LAB typically consists of identical subunits with molecular masses ranging from 54 to 62 kDa and is formed in its mature holo-form, even when produced heterologously. The oligomerization, typically resulting in the formation of a homodimer, is crucial for activity of the Lactobacillus spp. enzymes. Some ambiguity about the active form of GAD isolated from different isolates of L. brevis and its quaternary structure exists in the scientific literature. Hiraga et al. reported that treatment with high concentrations of ammonium sulfate resulted in an active tetrameric form with the enzyme from L. brevis IFO12005 GAD [78]. The presence of ammonium sulfate apparently stabilized GAD from this source as the purified enzyme was found to be rather unstable, and the dimeric form showed no activity. Moreover, the presence of ammonium sulfate apparently did not affect the overall conformation but had effects on the active site of the protein. Studies by Yu et al. showed that GAD from L. brevis CGMCC 1306 is active as a monomer, while GAD from other LAB are generally active as dimers [71]. Subsequent structural studies on this enzyme revealed, however, that GAD from L. brevis CGMCC 1306 is active as a dimer (Figure 7), even though elucidation of the crystal structure resulted in a distorted asymmetric trimer. The authors concluded that this observed trimer only resulted from the crystallographic packing and not the biological form [68].

Biochemical Insights into Glutamate Decarboxylase from Lactic Acid Bacteria
GAD from LAB typically consists of identical subunits with molecular masses ranging from 54 to 62 kDa and is formed in its mature holo-form, even when produced heterologously. The oligomerization, typically resulting in the formation of a homodimer, is crucial for activity of the Lactobacillus spp. enzymes. Some ambiguity about the active form of GAD isolated from different isolates of L. brevis and its quaternary structure exists in the scientific literature. Hiraga et al. reported that treatment with high concentrations of ammonium sulfate resulted in an active tetrameric form with the enzyme from L. brevis IFO12005 GAD [78]. The presence of ammonium sulfate apparently stabilized GAD from this source as the purified enzyme was found to be rather unstable, and the dimeric form showed no activity. Moreover, the presence of ammonium sulfate apparently did not affect the overall conformation but had effects on the active site of the protein. Studies by Yu et al. showed that GAD from L. brevis CGMCC 1306 is active as a monomer, while GAD from other LAB are generally active as dimers [71]. Subsequent structural studies on this enzyme revealed, however, that GAD from L. brevis CGMCC 1306 is active as a dimer (Figure 7), even though elucidation of the crystal structure resulted in a distorted asymmetric trimer. The authors concluded that this observed trimer only resulted from the crystallographic packing and not the biological form [68]. As above-mentioned, a number of LAB carry two GAD-encoding genes, gadA and gadB. Frequently, studies have focused on the purification and characterization of GadB (e.g., from L. plantarum [79], L. sakei [80], L. brevis [78], Enterococcus raffinosus [75], and L. paracasei [18]), since the expression levels of recombinant GadB are typically higher than those for GadA [55]. A recent study by Wu et al. showed that the gadA transcript was highly upregulated (55-fold) in strain L. brevis NPS-QW-145 at the stationary phase of growth [7]. Subsequently, both GadA and GadB were recombinantly produced and characterized. GadA showed a pH profile of activity near the neutral region, with the optimal activity found in the range of pH 5.5-6.6, in contrast to GadB, which is more active under acidic conditions (3.0-5.5). Presence of both of these two enzymes, GadA and GadB, in the L. brevis genome will give the organism a significant advantage to produce GABA over a broad range of pH (3.0-6.0), and thus to more efficient maintenance of pH homeostasis. These findings As above-mentioned, a number of LAB carry two GAD-encoding genes, gadA and gadB. Frequently, studies have focused on the purification and characterization of GadB (e.g., from L. plantarum [79], L. sakei [80], L. brevis [78], Enterococcus raffinosus [75], and L. paracasei [18]), since the expression levels of recombinant GadB are typically higher than those for GadA [55]. A recent study by Wu et al. showed that the gadA transcript was highly upregulated (55-fold) in strain L. brevis NPS-QW-145 at the stationary phase of growth [7]. Subsequently, both GadA and GadB were recombinantly produced and characterized. GadA showed a pH profile of activity near the neutral region, with the optimal activity found in the range of pH 5.5-6.6, in contrast to GadB, which is more active under acidic conditions (3.0-5.5). Presence of both of these two enzymes, GadA and GadB, in the L. brevis genome will give the organism a significant advantage to produce GABA over a broad range of pH (3.0-6.0), and thus to more efficient maintenance of pH homeostasis. These findings suggest that extending the activity of GadA to the near-neutral pH region offers a novel genetic diversity of gad genes from LABs [7].
A number of GAD have been expressed and characterized from a variety of LABs. In general, the N-and C-terminal regions of GAD from different sources show significant differences, and this might affect recombinant GABA production. As shown in a sequence alignment (Figure 8), the sequence HVD(A/S)A(S/F)GG was highly conserved among LAB GAD, and a lysine residue (Lys279 in L. brevis GAD) played a crucial role in the PLP binding site. Table 2 summarizes the biochemical properties of GAD from different strains [18,42,81,82]. Typically, the pH optima of GAD are found between 4.0 and 5.0. GAD from L. zymae, E. avium M5, S. salivarius subsp. thermophilus Y2, and L. paracasei NFRI 7415 have an optimum activity of above 40 • C, which does not coincide with the optimal temperature for growth of these strains [46,72,82,83]. Different ions can affect the stability and activity of GAD from different sources ( Table 2). GAD from E. avium M5 is activated in the presence of CaCl 2 and MnCl 2 but the activity is decreased by CuSO 4 and AgNO 3 [82]; comparable results were also obtained for GAD from other LAB sources, L. zymae [46] and L. sakei A156 [80].
Since GAD is mainly active under acidic conditions, several engineering approaches have been employed to broaden its activity, especially at the near-neutral pH region. To this end, Shi et al. applied both directed evolution and site-directed mutagenesis at the β-hairpin region and C-terminal end of L. brevis GAD [84]. By using a plate-based screening assay employing a pH indicator as assay principle, they could identify several variants and positions that improved activity at pH 6.0. Furthermore, they selected three residues (Tyr308, Glu312, Thr315) in the β-hairpin region for site-directed mutagenesis based on homology modeling, since these residues exhibit different interactions with surrounding amino acids in the model at different pH values. By combining various positive mutations, they could increase the catalytic efficiency of GAD from L. brevis 13.1-and 43.2-fold at pH 4.6 and 6.0, respectively, when compared to the wild-type enzyme [84]. The role of the C-terminus for the pH dependence of catalysis of L. plantarum GAD was investigated by Shin et al. employing mutagenesis [79]. Deletions of three and eleven residues in the C-terminal region Ile454-Thr468 of this enzyme increased activity in the pH range of 5 to 7, with the ∆11 variant showing significantly better results, increasing the catalytic efficiency of the variant at pH 5.0 and 7.0 by a factor of 1.26 and 28.5, respectively. The authors concluded that the C-terminal region is involved in decreasing the activity of L. plantarum GAD at higher pH values by closing up the catalytic site as a result of pH-induced conformational changes [79]. In a similar way, a C-terminally truncated variant of L. brevis GAD, in which the terminal 14 amino acids had been removed by site-directed mutagenesis, showed improved activity at higher, around neutral pH values [85]. These studies point to the importance of the C-terminus of GAD for improved accessibility of the active site and increased activity, especially at higher pH values, and thus the C-terminal loop is an essential target for enzyme engineering for GABA production at fluctuated pH conditions [79,85].    [48,55]. Furthermore, the residues SINA/V/TSGHKYGM/LVYPGI/V/LGWI/VV/LW/R/K/V are part of the PLP-binding domain [26].   [48,55]. Furthermore, the residues SINA/V/TSGHKYGM/LVYPGI/V/ LGWI/VV/LW/R/K/V are part of the PLP-binding domain [26].

Improvement of GAD Activities and GABA Production
GABA biosynthesis can be achieved by using whole cell reactions, recombinant bacteria, and purified GAD (Table 3). gad genes from various sources of LAB have been overexpressed in different hosts including E. coli [86], L. sakei [87], L. plantarum [88], Corynebacterium glutamicum [89], and Bacillus subtilis [90]. Utilization of whole cells for the biocatalytic conversion of glutamate to GABA has some drawbacks including the conversion of GABA to succinic semialdehyde by the enzyme GABA transaminase (GABA-T), which is often found in bacteria and might decrease GABA yields during cultivation. To prolong and thereby increase GABA production, continuous cultivation [91], fed-batch fermentation [92] as well as immobilized cell technology [93][94][95] have been employed. All of these approaches effectively increased GABA productivity by improving cell viability resulting in extended periods of cultivation.
GABA biosynthesis and production could be enhanced by optimizing fermentation conditions, with attention given to different factors including the carbon source, concentration of added glutamate, pH regulation, incubation temperature, nitrogen sources, cofactor, and feeding time [34,94]. A study by Lim et al. showed that under optimized conditions, L. brevis HYE1 produced 18.8 mM of GABA. Monosodium glutamate (MSG) or l-glutamate are the main substrate for the production of GABA using either appropriate GAD-containing cells or pure GAD [27]. LAB with GAD activity may furthermore require the supplementation of PLP to the medium to enhance GABA production. The addition of 0.5% MSG increased GABA production by E. faecium JK29, which reached 14.9 mM after 48 h of cultivation [38]. A concentration of 6% MSG and the addition of 0.02 mM PLP were found to be optimal conditions for L. brevis K203 for GABA production [42]. This strategy of increasing glutamate supplementation could not be used for all strains though; when l-glutamate was added at concentrations of 10 to 20 g/L to the growth medium of S. thermophilus, GABA production could not be enhanced. It was suggested that this strain is not able to tolerate high glutamate concentrations [36]. High glutamate concentrations increase the osmotic pressure in the cells, and this stress can disturb the bacterial metabolism [39]. Fermentation time and temperature are also key factors for GABA production. Villegas et al. investigated GABA formation by L. brevis CRL 1942, and found that 48 h of fermentation at 30 • C employing 270 mM of MSG resulted in a maximum GABA production of 255 mM in MRS medium, indicating that the GABA production occurs in a time-dependent manner [96].
Metabolic pathway engineering has been performed to achieve enhanced GABA production. The key points here are the direct modulation of GABA metabolic pathways. A whole-cell biocatalyst based on E. coli cells expressing the gadB gene from L. lactis was used as the starting point of this engineering approach. An engineered strain was constructed by (i) introducing mutations into this GadB to shift its decarboxylation activity toward a neutral pH; (ii) by modifying the glutamate/GABA antiporter GadC to facilitate transport at neutral pH; (iii) by enhancing the expression of soluble GadB through overexpression of the GroESL molecular chaperones; and (iv) by inhibiting the degradation of GABA through inactivation of gadA and gadB from the E. coli genome. This engineered strain achieved a productivity of 44.04 g/L of GABA per h with an almost quantitative conversion of 3 M glutamate [97].
Several mutational approaches such as directed evolution and site-specific mutagenesis are considered as powerful tools for optimizing or improving enzyme properties. Several researchers have applied these approaches to improve GAD activity [84,[97][98][99][100][101] and were applied in whole-cell biocatalysts. In order to improve GAD activity over an expanded pH range, recombinant C. glutamicum cells were obtained by expressing L. brevis Lb85 GadB variants. These variants were constructed by combining directed evolution and site-specific mutagenesis of GadB to improve activity at higher pH values (see above), since C. glutamicum grows best around neutral pH [84]. C. glutamicum is an industrial producer of glutamate, and by introducing these GadB variants into this organism, GABA could be produced without the need of exogenous glutamate on a simple glucose-based medium, with yields of up to 7.13 g/L [84].
Insufficient thermostability is often a common problem associated with industrial enzymes, and most GAD show low stability even at moderate temperatures. A rational strategy for improving thermostability is to identify critical regions or amino acid residues by sequence alignments. Alternatively, structural information indicating flexible regions can be used, and subsequently, these regions are strengthened [102]. Identification of the consensus sequences can also improve the thermostability of proteins [103]. Recently, Zhang et al. developed a parallel strategy to engineer L. brevis CGMCC 1306 GAD. They compared the sequence and structure of this mesophilic GAD with homologous thermophilic enzymes to identify amino acid residues that might affect stability. Two mutant enzymes were obtained and showed higher thermostability with their half-inactivation temperature 2.3 • C and 1.4 • C higher than that of the wild-type enzyme. Furthermore, the activity of the variants was 1.67-fold increased during incubation at 60 • C for 20 min. They suggested that this approach can be an efficient tool to improve the thermostability of GAD [102].
The use of purified GAD seems to be economically more feasible than whole-cell biocatalysis when aiming at producing pure GABA due to simplified downstream purification of this compound from less complex reaction mixtures. A number of immobilization techniques have been applied for re-use of the biocatalyst such as immobilization of GadB in calcium alginate beads that are then employed in a bioreactor [104], a GAD/cellulose-binding domain fusion protein immobilized onto cellulose [105], and GAD immobilized to metal affinity gels [106]. The performance of immobilized GAD in a fed-batch reactor was evaluated, which showed high productivity of GABA as the substrate concentration in the medium was kept constant by feeding solid glutamate. Moreover, no significant decrease in enzyme activities was observed during the reaction when the inactivation reaction of PLP to succinic semialdehyde and pyridoxamine-5 -phosphate during catalysis was avoided by the addition of a small amount of α-ketoglutaric acid to the reactor, which regenerated PLP [101]. Sang-Jae Lee et al. performed immobilization of L. plantarum GAD using silica beads and showed high stability under acidic and alkaline conditions with improved thermostability [105]. In addition, the immobilized GAD converted 100% of glutamate to GABA [106]. These results suggest that immobilization gives advantageous results for industrial application when using (partially) purified GAD for GABA production from glutamate.

The Role of Glutamate Decarboxylase in the Manufacturing of Bio-Based Industrial Chemicals
Agricultural waste and waste streams from biofuel production are now being considered as a low-cost source of glutamate for biotechnological conversion into GABA and production of bio-based chemicals [107]. These protein-rich materials are mainly bioethanol by-product streams including dried distiller's grains with solubles (DDGS) from maize and wheat, or vinasse from sugarcane or sugar beet, but also plant leaves, oil, or biodiesel by-products and slaughterhouse waste. In the future, algae could also provide an additional source for biodiesel and thus become a natural low-cost source of glutamic acid.
The protein-rich fraction of plants can be further split into more-and less-nutritious fractions, for example, by hydrolyzing the proteins and separating the essential (nutritious) amino acids from the non-essential (less nutritious) ones. Non-essential amino acids such as glutamic acid and aspartic acid, which have no significant value in animal feed, can be utilized for preparing functionalized chemicals. Recently, a by-product from the tuna canning industry, tuna condensate, was shown to be a useful material for the production of GABA. Tuna condensate contains significant amounts of glutamine, but relatively little glutamate. Glutamine was first converted to glutamate by a glutaminase from Candida rugosa, and in a second step, L. futsaii GAD converted glutamate to GABA. Both steps were catalyzed by immobilized whole cells [108]. Recently, it was shown that supplementation of arginine to media containing glutamate could enhance GABA production, and that the simultaneous addition of arginine, malate, and glutamate enabled GABA production already during exponential growth at relatively high pH (6.5) [109].
The structure of glutamic acid resembles many industrial intermediates, so it can be transformed into a variety of chemicals using a relatively limited number of steps. Decarboxylation of glutamic acid to GABA, enzymatically performed by GAD, is an important reaction of the pathway from glutamic acid to a range of molecules. GABA is, for example, an intermediate for the synthesis of pyrrolidones. Such an approach can be used to produce N-methyl-2-pyrrolidone (NMP), which is used as an industrial solvent. Combining the enzymatic decarboxylation of glutamate performed by GAD with the one-pot cyclization of GABA to 2-pyrrolidone and subsequent methylation will thus yield NMP [110]. Another interesting material synthesized by ring-opening polymerization of 2-pyrrolidone is Nylon 4 [111], a four-carbon polyamide suitable for application as an engineering plastic due to its superior thermal and mechanical properties [112]. Contrary to other nylon polymers, Nylon 4 is heat-resistant, biodegradable, biocompatible, and compostable [112].

Future Trends and Conclusions
The demand for functional foods is increasing and marked by the awareness of consumers in maintaining health and prevention of degenerative diseases. Therefore, exploration of bioactive compounds such as GABA are important. The GAD system plays a crucial role in GABA biosynthesis. A number of studies on cloning, expression, and characterization of both gadA and gadB and the encoded enzymes GadA and GadB has led to deciphering the role of the gad genes in the GABA metabolic pathway and its importance for LAB. Since the production of GABA is dependent on the biochemical properties of GAD, more study on the biochemical properties of GAD are important, especially for those enzymes derived from LAB isolated from food fermentation processes, as this will facilitate the optimization of the fermentation process and support the selection of suitable starter cultures for these processes that will bring more GABA-enriched food to the consumer. Recent structural information of GAD from LAB will facilitate enzyme-engineering approaches to improve GAD toward enhanced thermostability or improved activity over a broad range of pH. However, structural information is currently only limited to GAD from L. brevis, and thus structural studies on GAD from other GABA-producing LAB are needed in order to understand their catalytic and structural properties in more depth. The elucidation of molecular mechanisms and roles of GABA production, knowledge of the regulatory aspects of GABA production, and profound comprehension of GABA-producing cell physiology will offer the basis and tools to increase GABA yields at genetic and metabolic levels.