Glutamate decarboxylase-dependent acid resistance in orally acquired bacteria: function, distribution and biomedical implications of the gadBC operon


  • Daniela De Biase,

    Corresponding author
    • Istituto Pasteur – Fondazione Cenci Bolognetti, Dipartimento di Scienze e Biotecnologie Medico-Chirurgiche, Sapienza Università di Roma, Latina, Italy
    Search for more papers by this author
  • Eugenia Pennacchietti

    1. Istituto Pasteur – Fondazione Cenci Bolognetti, Dipartimento di Scienze e Biotecnologie Medico-Chirurgiche, Sapienza Università di Roma, Latina, Italy
    Search for more papers by this author

For correspondence. E-mail; Tel. (+39) 0773 1757212; Fax (+39) 0773 1757254.


For successful colonization of the mammalian host, orally acquired bacteria must overcome the extreme acidic stress (pH < 2.5) encountered during transit through the host stomach. The glutamate-dependent acid resistance (GDAR) system is by far the most potent acid resistance system in commensal and pathogenic Escherichia coli, Shigella flexneri, Listeria monocytogenes and Lactococcus lactis. GDAR requires the activity of glutamate decarboxylase (GadB), an intracellular PLP-dependent enzyme which performs a proton-consuming decarboxylation reaction, and of the cognate antiporter (GadC), which performs the glutamatein/γ-aminobutyrateout (GABA) electrogenic antiport. Herein we review recent findings on the structural determinants responsible for pH-dependent intracellular activation of E. coli GadB and GadC. A survey of genomes of bacteria (pathogenic and non-pathogenic), having in common the ability to colonize or to transit through the host gut, shows that the gadB and gadC genes frequently lie next or near each other. This gene arrangement is likely to be important to ensure timely co-regulation of the decarboxylase and the antiporter. Besides the involvement in acid resistance, GABA production and release were found to occur at very high levels in lactic acid bacteria originally isolated from traditionally fermented foods, supporting the evidence that GABA-enriched foods possess health-promoting properties.


In order to colonize the host, orally acquired bacteria must (i) endure transit through the extremely acidic gastric compartment, (ii) resist the detergent-like activity of bile salts, (iii) cope with the decreasing oxygen availability, and (iv) compete with the resident microbiota for gut colonization and, in the case of pathogenic bacteria, to cause infection.

The ability to withstand acid stress, i.e. the combined effect of HCl in gastric juice and of short chain fatty acids produced by the intestinal microbiota, is crucial for successful colonization of the gastrointestinal tract. In this respect stomach acidity (pH ≤ 2.5) plays a major role in killing most of the ingested microorganisms (Giannella et al., 1972). In food-borne pathogens acid resistance is a fundamental pre-requisite to succeed in host colonization. Escherichia coli, Shigella flexneri, Salmonella typhimurium, Listeria monocytogenes and Vibrio cholerae have evolved different strategies to overcome acid stress. A detailed description of the strategies adopted is beyond the scope of this review and is reported elsewhere (Bearson et al., 1997; Foster, 1999; Merrell and Camilli, 2002; Cotter and Hill, 2003). Briefly, each microorganism exhibits a different degree of resistance to acid stress which involves either ATR (acid tolerance response) systems or AR (acid resistance) systems. The two systems differ in that the ATR (log- or stationary-phase) requires pre-exposure of the bacteria to mildly acidic pH before acid challenge at pH ≥ 3.0, whereas AR typically protects stationary-phase cells from an extreme acid stress (pH ≤ 2.5) without pre-adaptation (Foster, 2001; Merrell and Camilli, 2002).

Escherichia coli (pathogenic and non-pathogenic strains) and S. flexneri are noteworthy in their ability to survive in an extremely acidic environment (pH ≤ 2.5) and this correlates with their low infectious dose (< 102 cells; Gorden and Small, 1993; Lin et al., 1995; Lin et al., 1996). The aptitude to withstand extreme acid stress primarily relies on the activation of very potent amino acid-dependent AR systems, which rely either on glutamate or on arginine supplementation to minimal medium at pH ≤ 2.5.

Glutamate-dependent acid resistance in Escherichia coli

Historical background

In his seminal work carried out in the Forties, Ernest Gale provided compelling evidence that in bacteria the α-decarboxylases acting on the L-amino acids lysine, ornithine, arginine, tyrosine, histidine and glutamate are inducible (Gale, 1940). These decarboxylases were expressed in the presence of the corresponding amino acids, during the stationary phase of growth and upon acidification of the growth medium. Gale proposed that the biological function of amino acid decarboxylases resides in a ‘neutralization mechanism evolved by the cell as a method of stabilizing the internal environment against the unfavourable changes in the external environment’ (Gale, 1946). He also showed that the amino acid α-decarboxylases depend on a co-decarboxylase, the vitamin pyridoxine (precursor of pyridoxal 5′-phosphate, PLP). From then on PLP-dependent bacterial amino acid α-decarboxylases were mainly studied at the biochemical level. Glutamate decarboxylase from E. coli was one of the most intensively studied enzyme because it was purifiable in large quantities from naturally occurring overproducing strains (Shukuya and Schwert, 1960a,b,c; O'Leary and Brummund, 1974; Sukhareva, 1986).

The glutamate-dependent acid resistance system: role and regulation

More recently researchers' interests have returned to the biological function of bacterial amino acid decarboxylases, because it became apparent that E. coli (commensal and pathogenic) and S. flexneri possess arginine- and glutamate-dependent AR systems (Lin et al., 1995; 1996). The glutamate-dependent AR (GDAR) system, by far the most potent in these microorganisms, was also reported to be present in L. monocytogenes and in the lactic acid bacterium Lactococcus lactis (Sanders et al., 1998; Cotter et al., 2001a).

In E. coli GDAR relies on the activity of at least one of the two isoforms of the PLP-dependent, hexameric enzyme glutamate decarboxylase (GadA and GadB) and on the activity of GadC, the functionally associated glutamate/γ-aminobutyrate (GABA) antiporter (Hersh et al., 1996; Waterman and Small, 1996; Castanie-Cornet et al., 1999; De Biase et al., 1999). GDAR is astonishing in that exogenous glutamate in minimal medium is the only component necessary for the system to operate at pH ≤ 2.5 for 2 h or more. In the absence of this amino acid bacteria succumb (Lin et al., 1996; Castanie-Cornet et al., 1999; De Biase et al., 1999). In contrast to S. flexneri, in E. coli the induction of GDAR that occurs in stationary phase does not require growth in a rich medium at acidic pH and/or glutamate supplementation (Lin et al., 1995; Bhagwat and Bhagwat, 2004).

Figure 1A outlines how the system is proposed to work: when the extracellular proton concentration abruptly increases to harmful levels (pH ≤ 2.5), as during transit through the host stomach, the inner membrane becomes leaky to protons (and to chloride ions), intracellular pH drops to 4.2 and import of glutamate (net charge 0) via GadC is triggered (Richard and Foster, 2004). Once inside, glutamate remains protonated on the γ-carboxylate to allow binding to and decarboxylation by GadA/B (Fonda, 1972; Capitani et al., 2003a), which are maximally active in the pH range 4–5 (De Biase et al., 1996; Gut et al., 2006). The α-carboxyl group, leaving as CO2, is thus replaced by a cytoplasmic proton, yielding GABA (net charge +1), which is exported via GadC in exchange for a new glutamate molecule.

Figure 1.

Glutamate-based acid resistance system.

A. Schematic representation of the role played by the major structural components of the system. Glutamate (net charge 0) is taken up by the electrogenic l-glutamate/GABA antiporter GadC, an inner membrane protein. Decarboxylation of glutamate via GadA/B consumes an intracellular H+ at each cycle and contributes to the generation of proton motive force by GABA (net charge +1) export. Biochemical data provided evidence that glutamate binding to GadB occurs only when the γ-carboxylate of glutamate is in the fully protonated form (Fonda, 1972). Crystallographic data support the biochemical data; the crystal structure of GadB clearly shows that the γ-carboxylate of glutamate interacts with the acidic side chain of Asp86 (an active site residue). This interaction is possible only when at least one of the carboxylates (i.e. the γ-carboxylate of glutamate and/or the β-carboxylate of Asp86) is fully protonated (Capitani et al., 2003a).

B. Representation of the E. coli genome map with the location of the gene loci encoding glutamate decarboxylases (red), l-glutamate/GABA antiporter (blue), additional membrane proteins (orange), multidrug exporters (gray), acid stress periplasmic chaperons (magenta), LuxR-like transcriptional regulators (light green), AraC-like transcriptional regulators (dark green) and a small regulatory RNAs (yellow). The bent arrows show the transcripts of the major GDAR genes. The origin of each arrow is where transcripts start.

Functional studies (see next section) on E. coli GadB have unveiled that this enzyme is mostly localized in the cytoplasm at pH > 5.6, whereas below pH 5.6 it colocalizes with the cellular membrane fraction, where it is considered to control the intracellular pH more efficiently (Capitani et al., 2003a). GadA shares with GadB 98.9% and 99.3% sequence identity and similarity, respectively, which suggests that GadA is likely to undergo a similar pH-dependent cellular partitioning. Thus, glutamate decarboxylase-mediated proton consumption and glutamatein/GABAout antiport constitute a simple molecular system that by sensing, consumption and pumping-out of protons, efficiently works to protect E. coli and S. flexneri from the acid stress encountered during transit through the stomach. In vivo studies on E. coli cells also showed that GDAR not only contributes to pH homeostasis but, by transiently accumulating GABA in the cell, it counteracts illicit entry of protons by inversion of the membrane potential, a strategy similar to that adopted by extreme acidophiles (Foster, 2004).

Escherichia coli GadA and GadB isoforms are coded by the corresponding genes (Smith et al., 1992; De Biase et al., 1996), which are 2.1 Mb apart on E. coli K12 chromosome. As outlined in Fig. 1B, the gadA gene is at the 3′ end of a 14 kb genome region, named acid fitness island (AFI), comprising 13 additional genes which contribute at various levels to AR. The genes gadE, gadX, gadW, yhiF (dctR), arrS and gadY code for four specific transcriptional regulators and two small RNAs respectively (Tramonti et al., 2002a; 2006; 2008; Ma et al., 2003; Tucker et al., 2003; Opdyke et al., 2004; Aiso et al., 2011); hdeA and hdeB code for pH-regulated acid stress periplasmic chaperones, which facilitate the refolding of acid-denatured proteins in this cellular compartment (Hong et al., 2005; Kern et al., 2007); slp, yhiD and hdeD code for membrane proteins required for protection from acidic metabolites (lactate, succinate and formate) and for AR at high cell densities (Mates et al., 2007); the mdtE and mdtF genes code for multidrug exporters which are unique amongst the 20 known drug exporters in that their expression is induced in stationary phase and is GadX-dependent (Kobayashi et al., 2006; Nishino et al., 2008). The gadA gene is either independently transcribed or transcribed with gadX (Fig. 1B), to generate a gadAX transcript, whereas the gadB gene is typically co-transcribed with the downstream gadC gene (De Biase et al., 1999; Tramonti et al., 2002a). The AFI genes are mostly organized into operons (Fig. 1B), i.e. slp-yhiF, hdeAB-yhiD, gadE-mtdEF, gadXW and gadAX (Tramonti et al., 2002a; Tucker et al., 2003).

Expression of the GDAR structural genes, which is maximal in the stationary-phase of growth, was reported to occur also in the exponential-phase of growth under acidic conditions and, in general, under many stress conditions, namely starvation, acidic pH, hyper- and hypo-osmolar stress and anaerobiosis (Castanie-Cornet et al., 1999; De Biase et al., 1999; Weber et al., 2005; Hayes et al., 2006). The log phase induction is under the control (direct or indirect) of the two-component systems EvgS/EvgA, RcsC/RcsB, PhoQ/PhoP (Castanie-Cornet et al., 2007; Itou et al., 2009; Burton et al., 2010). Both log- and stationary-phase inductions are the outcome of a complex interplay between specific regulators, such as GadE, RcsB, GadX and GadW (Tramonti et al., 2002a; 2008; Ma et al., 2003; Castanie-Cornet et al., 2007), and global regulators such as H-NS (the histone-like nucleoid-structuring protein) and RpoS (the stationary phase sigma factor of the RNA polymerase) (Hommais et al., 2001; Weber et al., 2005; Oshima et al., 2006). The molecular details have not yet been fully elucidated.

Functional and structural insights into E. coli GadB

The crystal structure of E. coli GadB, the glutamate decarboxylase isoform coexpressed with gadC (Fig. 1B), was reported in 2003 (Capitani et al., 2003a). This was the first structure of a glutamate decarboxylase to be released, followed shortly after by the release of the structure of E. coli GadA in complex with glutarate, a substrate analogue (Dutyshev et al., 2005). More recently, the structures of the two isoforms of human GAD (GAD65 and GAD67) and of the Arabidopsis thaliana Gad1 isoform were disclosed (Fenalti et al., 2007; Gut et al., 2009). Human GADs are active as dimers at neutral pH and share little sequence identity (≤ 12%) with E. coli GadB (Fenalti et al., 2007). Conversely, A. thaliana Gad1 is more similar to E. coli GadB with which it shares 39% sequence identity as well as a hexameric assembly (Gut et al., 2009). Regardless of the level of sequence identity, human, A. thaliana and E. coli GADs all possess the PLP-fold typical of the largest of the PLP-dependent enzyme families (Momany et al., 1995a). However, each GAD is unique in terms of distinctive structural features, such as mobile loops (human GAD), a calmodulin binding domain (plant Gad1) or regions undergoing pH-dependent conformational changes (E. coli GadB), relevant for activity in the corresponding organism.

pH-Dependent conformational changes

The GadB hexamer is notable in that it is a true structural hexamer (Capitani et al., 2003a) unlike prokaryotic ornithine, lysine and arginine decarboxylases, which are dodecamers (the former) and decamers (the latter two), respectively, more prone to dissociate into the minimal functional unit, the dimer (Momany et al., 1995b; Andrell et al., 2009; Kanjee et al., 2011). The GadB crystal structure, solved both at acidic pH (4.6) and at neutral pH (7.6), shows that the GadB hexamer can be more appropriately described as a trimer of dimers, in which monomers from each dimer belong to different layers (Fig. 2). The entangled structure is held together by the N-terminal domains (residues 1–57) of each subunit and is primarily, though not exclusively, evident in the structure at neutral pH (Fig. 2A, left side). Via the narrow central channel, the residues of the N-terminal domain act like a ‘hook’, i.e. they depart perpendicularly from the subunit surface to reach the opposite layer where they make contacts both with the counterpart in the functional dimer and with a subunit belonging to a neighbouring dimer.

Figure 2.

Side views of the E. coli GadB hexamer at neutral (left) and acidic (right) pH.

A. The two-layered GadB hexamer is shown in colours. Each dimer contributes to both layers (dimers are in red-magenta, light green-dark green, blue-cyan).

B. Major conformational changes in GadB. The hexamer is in gray, but the 1–14 N-terminal region (red), the 452–466 C-terminal tail (green) and the 300–313 β-hairpin (magenta). PLP is shown in filled space (yellow). Figure prepared with PyMol.

The comparison of the structures obtained at pH 4.6 (Fig. 2A, right side) and 7.6 provided evidence that the overall structure of the hexamer remains unaffected by the pH change. However, the pH changes caused three major structural reorganizations; one in the first 15 residues of the N-terminal domain (residues 1–57), a second in the last 15 residues of the C-terminal small domain (residues 347–466) and the third in residues 300–313, which form a β-hairpin (Fig. 2B, highlighted in magenta) in the large PLP-binding domain (residues 58–346).

The crystal structures at different pHs and the analysis of a N-terminal truncated mutant (GadBΔ1–14) suggested the sequence of events that lead to the conversion of GadB from the inactive into the active form and vice versa (Capitani et al., 2003a; Gut et al., 2006; Pennacchietti et al., 2009). In its inactive form GadB has (i) the N-terminal residues 1–14 of each subunit mainly involved in dimerization and hexamerization; (ii) the C-terminal residues 452–466 ordered and protruding into the active site (with residues His465 and Thr466), like a plug, thus occupying the binding site of the physiological substrate (i.e. glutamate); (iii) the β-hairpin 300–313 contacting the C-terminal tail of the other subunit in the dimer as to hold it in place (Fig. 2B, left side). Therefore, at neutral pH the enzyme is in a compact conformation with access to the active site precluded by steric hindrance of some structural elements.

At acidic pH the N-terminal regions, rich in acidic residues, become protonated: this event abolishes negative charges repulsion of their side chains and allows the formation of α-helices, which give rise to two triple helical bundles, one on each top and bottom side of the hexamer (Fig. 2A and B, right side). These bundles are parallel to the threefold axis of the hexamer and have a hydrophobic core. Functional studies on the GadBΔ1–14 deletion mutant showed that the formation of the triple helical bundles is instrumental both to the recruitment of GadB to the cytosolic side of the inner membrane, where its proton-consuming activity will exert a more beneficial effect to the cell, and to the formation of the binding sites for chloride ions, which act as positive allosteric modulators of GadB (Iyer et al., 2002; Capitani et al., 2003a; Gut et al., 2006). The exit of the C-terminal tail from the active site is accompanied by a movement of the β-hairpin 300–313 towards the centre of the active site funnel and also by increased accessibility of the active site to solvent and substrate (Capitani et al., 2003a; Gut et al., 2006) (Fig. 2B). Together, the changes occurring in GadB when it is shifted to acidic pH fully account for its intracellular buffering role (by proton uptake) and for its activation (Capitani et al., 2003a).

As shown by crystallographic ‘snapshots’ and kinetic analyses, during the conversion of GadB from the active into the inactive form, the unfolding of the α-helices on both sides of the hexamer is not synchronized and affects the rate at which the C-terminal ends regain access into the active site (Gut et al., 2006). Notably, truncation of the N-terminal region is responsible for slowing down the closure of the active site (Gut et al., 2006), whereas a GadBΔHT-deletion mutant (missing the last two residues in the sequence, i.e. His465 and Thr466) and a His465Ala site-specific mutant are unable to undergo efficient closure of the active site, as shown by spectroscopic analyses (Pennacchietti et al., 2009).

A signature for bacterial GAD

Unlike the aminotransferases, the PLP-dependent amino acid decarboxylases seem to be evolved along multiple lineages (Sandmeier et al., 1994). Bacterial GAD belongs to group II, which includes the decarboxylases acting on glutamate, histidine, tyrosine and aromatic-L-amino-acid. According to the original multiple sequence alignment (Sandmeier et al., 1994) and to the PFAM multiple sequence alignment of 5264 sequences ( from the conserved domain of PLP-dependent decarboxylase (PF00282, Pyridoxal_deC), 8 residues appear to be highly conserved in the decarboxylases of the evolutionary related group II (Sandmeier et al., 1994; Momany et al., 1995a). These residues are the PLP-binding lysine (Lys276 in GadB; Tramonti et al., 2002b); the aspartic acid residue interacting with the pyridine N of the coenzyme (Asp243 in GadB); an alanine residue in hydrophobic contact with the pyridine ring of PLP (Ala245 in GadB); a histidine residue (His241 in GadB), buried in the protein core and hydrogen bonded with a serine residue (Ser269 in GadB), with which it plays a critical structural role in stabilizing the interaction between two β-strands and a long α-helix in the large domain (Capitani et al., 2003b); a threonine residue (Thr212 in GadB) tilted towards the hydroxyl group of PLP in the inactive (closed) conformation and which might affect intramolecular proton transfer, as suggested in Dopa decarboxylase (Lin and Gao, 2010); two glycine residues (Gly120 and Gly210 in GadB), the role of which has not been investigated yet, but is likely to be purely structural.

A Clustal X (version 2.1) alignment generated with 20 bacterial GADs (Fig. S1), mostly from Gram-positive and Gram-negative enteric bacteria (see Fig. 5), shows that 84 residues (18 % of the E. coli GadB sequence) are strictly conserved. Several of these residues are known to occupy critical positions, i.e. in or near the active site of E. coli GadB or at sites where the conformational changes occur. In addition to the eight residues listed above, which are shared with group II decarboxylases, the concurrent presence of the residues listed in bold in Table 1 (numbering refers to E. coli GadB) may be regarded as the signature for bacterial GAD. The latter residues play different roles: Asp86* (*from the neighbouring subunit in the functional dimer), with its side-chain tilted towards the active site in the open conformation, interacts with the γ-carboxylate of the substrate glutamate; Glu89* in the low-pH structure is not far away from Asp86*, from which in the neutral-pH structure it points in opposite direction (Capitani et al., 2003a); Gln163 with its side-chain alkyl portion acts as stacking residue for the PLP ring; Thr62 side chain hydroxyl and Phe63 amide nitrogen provide additional hydrogen bonds to the γ-carboxylate of the substrate; Phe63 with its side-chain ring plays also an important role in preventing the binding of the α-carboxylate of glutamate to Arg422 (corresponding in sequence and spatial position to the arginine side-chain which binds the α-carboxylate of the amino acid substrate in PLP-dependent transaminases); His275, preceding the PLP-binding residue Lys276, is critical for keeping the cofactor in place by making an hydrogen bond with an oxygen of the PLP phosphate group via its side-chain τ nitrogen (Tramonti et al., 1998; Capitani et al., 2003a); Tyr305* and Leu306*, at the tip of the β-hairpin 300–313, interact with residues 461–463 at the C-terminus in the inactive form; Tyr305* may also be involved in protonating the substrate during the decarboxylation reaction, similarly to Tyr332 in Dopa decarboxylase (Bertoldi et al., 2002); His465 is responsible for GadB auto-inhibition via formation of an aldamine, an entirely novel species originating from the reaction of His465 side-chain τ nitrogen with the Lys276-PLP Schiff base (Gut et al., 2006; Pennacchietti et al., 2009).

Table 1. Amino acid residues composing the bacterial GAD signature
  1. aNumbering refers to E. coli GadB.
  2. bResidues in plain text are the residues strictly conserved in the decarboxylases of the PFAM PF00281.
  3. The asterisk ‘*’ indicates residues contributed by the neighbouring subunit in the functional dimer.
Thr62 (side chain OH)Binding of substrate γ-carboxylate

Phe63 (amide N)

Phe63 (side chain)

Binding of substrate γ-carboxylate

Prevents Arg422 binding to substrate α-carboxylate

Asp86* (β-carboxylate)Binding of substrate γ-carboxylate
Glu89* (side chain)Significant change in orientation upon pH-shift
Gly120bStructural role
Ser126 (amide N)Hydrogen bond with PLP phosphate oxygen
Ser127 (side chain OH)Hydrogen bond with PLP phosphate oxygen
Gln163 (alkyl chain)PLP stacking
Gly210Probable structural role
Thr212 (side chain OH)Interacting with PLP OH (intramolecular proton transfer?)
His241 (side chain π N)Hydrogen bond with Ser269
Asp243 (β-carboxylate)Interacting with the pyridine N of PLP
Ala245 (side chain)Hydrophobic contact with the pyridine ring of PLP
Ser269 (side chain OH)Hydrogen bond with His241
His275 (side chain τ N)Hydrogen bond with PLP phosphate oxygen
Lys276 (ε-amino group)Forming the Schiff-base with the C4' of PLP
Tyr305* (side chain OH)Interacts with residues 461–463 of the C-terminus (reprotonation?)
Leu306*Interacts with residues 461–463 of the C-terminus
Gly307*In the β-turn of β-hairpin 300–313
His465 (side chain τ N)Covalent bond with PLP-Lys276 Schiff base

Notably, the residue corresponding to E. coli GadB His465 is not present in the GadB homologue from Lactobacillus brevis ATCC367 (LVIS_0079). The absence of a His residue near the C-terminus in L. brevis LVIS_0079 may be explained by the different structural organization of this enzyme. Hiraga et al. (2008) showed that GAD from L. brevis IFO12005 is dimeric in the inactive form and tetrameric in the active. Treatment of GAD with high concentrations of ammonium sulphate and subsequent dilution with sodium glutamate were essential for tetramerization and activation. GAD was cloned from different strains of L. brevis, i.e. OPK-3 and IFO12005 (Park and Oh, 2007a; Hiraga et al., 2008). The nucleotide sequence of L. brevis OPK-3 did not match with that of L. brevis ATCC367 and L. brevis IFO12005 (Hiraga et al., 2008). GAD from L. paracasei, which shares with L. brevis the property of being a high GABA-producing strain (Komatsuzaki et al., 2008), was also reported to be dimeric.

It should be noted that the protein products of L. brevis LVIS_2213 and of Edwardsiella tarda ETAE_0786 (Fig. 5) were excluded from the sequence alignment (Fig. S1) because of the very low level of sequence identity (< 12% with respect to all the analysed sequences), which places them on distinct branches of the dendrogram (Fig. 3).

Figure 3.

Clustal X-generated dendrogram of bacterial glutamate decarboxylases. The amino acid sequences used to generate the alignment (Fig. S1) and the dendrogram are from the 18 representative species reported in Fig. 5. Bacterial Phyla are shown on the right. The bootstrap values calculated with Clustal X (version 2.1) using the N-J method are placed on the tree branches.

Notably, the residues listed in Table 1, but His465, are found in plant Gad1 too (Gut et al., 2009). The absence of His465 may well be explained by the different levels of regulation of plant Gad (pH-dependent and/or Ca2+/calmodulin-dependent), allowing this enzyme to be active also at pH 7.5 as part of an abiotic stress response (Gut et al., 2009).

Functional and structural insights into E. coli GadC

The recently published crystallographic structure of E. coli GadC adds an important piece of information to our knowledge on the biochemical basis underlying the pH-dependence in the activity of the major structural components of the GDAR system (Ma et al., 2012). Unlike GadB, the structure of GadC was solved at a single pH value, pH 8.6. It is therefore not surprising that the GadC structure displays all the features of the inactive conformation (inward-open) of this protein, which in vivo and in vitro (i.e. in reconstituted proteoliposomes) studies show to be active only at pH ≤ 5.5 (Richard and Foster, 2004; Ma et al., 2012). The activity profile of GadC as a function of pH remarkably resembles that of GadB (Pennacchietti et al., 2009; Ma et al., 2012). Despite the different function, GadB and GadC not only share an almost overlapping activity profile, but also adopt similar mechanisms to control their activity in the cell.

Like the GadB C-terminal tail, which plugs the active site, the GadC C-terminal fragment (involving residues 477–511; named the C-plug) is structurally arranged so to block the path to the putative substrate-binding site (Fig. 4). In GadB His465 is the residue which primarily contributes to an efficient closure of the active site and its mutation causes a shift in the midpoint of the activity profile as a function of pH by 0.4 pH units towards alkaline (Pennacchietti et al., 2009). Similarly, in GadC the deletion of the C-plug, in which several His and Arg residues contribute to the stability of the C-plug itself and to its interaction with other residues in the protein, cause a shift in the midpoint of the activity profile by 0.5 pH units towards alkaline (Ma et al., 2012). Thus, the presence and structural integrity of the C-plug are crucial to ensure that GadC activation occurs only at pH ≤ 5.5. It is tempting to speculate that, as in case of GadB, the pKs of the His residues in the C-plug play a major role in controlling the C-plug displacement following intracellular acidification.

Figure 4.

Side view of E. coli GadC at alkaline pH. The overall structure is shown in gray, but the C-plug (residues 477–511), which is in green as the C-terminal tail of GadB (Fig. 2B, left side). The relative orientation in the inner membrane (blue lines) is shown. Figure prepared with PyMol.

The crystal structure of GadC provides also insights into the mechanism of substrate transport (Ma et al., 2012). The availability of the structures of E. coli AdiC, the arginine/agmatine antiporter of the arginine-dependent acid resistance system, in the free (outward-open) and Arg-bound forms (Fang et al., 2009; Gao et al., 2009; 2010; Kowalczyk et al., 2011) helped to identify the gating residues in GadC (Ma et al., 2012). The residues Tyr96, Tyr214, Glu218, Trp308, Tyr378 and Tyr382 are all located within or in close proximity to the putative transport path. The six mutants generated by missense mutation (into Ala) show a significant decrease (> 90%) in substrate transport, thus suggesting a conserved transport mechanism between AdiC and GadC, which are otherwise significantly different in terms of substrate specificity and pH-dependent activity profile.

Noteworthy is the finding that using the proteoliposome-based transport assay GadC, besides glutamate, efficiently transports the amino acid glutamine (Ma et al., 2012). This finding agrees with a previous report showing that in vivo glutamine is transported by GadC and can be as efficient as glutamate in protecting from acid stress (Waterman and Small, 2003). In this respect it is worth to recall that in E. coli ybaS, the gene coding for glutaminase (the enzyme converting Gln into Glu with concomitant release of ammonia) was reported to be upregulated under the same stress conditions in which gadBC and the AFI genes were also upregulated (Tucker et al., 2002; 2003; Weber et al., 2005; Hayes et al., 2006; Shepherd et al., 2010).

The proximity of the gadB and gadC genes: a need for a coordinated action?

Several reports have provided compelling evidence that a functional gadBC operon is important for stationary phase AR in E. coli, S. flexneri, L. monocytogenes and L. lactis (Hersh et al., 1996; Sanders et al., 1998; Small and Waterman, 1998; Castanie-Cornet et al., 1999; De Biase et al., 1999; Cotter et al., 2001a). The occurrence of the gadBC (or gadCB) operon, though, seems not to be restricted to the above bacteria, i.e. in several Gram-positive and Gram-negative bacteria the gadB and gadC genes lie next or near each other (Fig. 5). Such gene arrangement is probably important to ensure co-regulation of the gadB and gadC genes. The bacteria listed in Fig. 5 might exploit a functional gadBC to pass through the gut. In most of these microorganisms the role and the regulation of the gadBC system has just being started to be investigated.

Figure 5.

Schematic representation of the genetic loci which code for E. coli homologues of GadB (dark grey arrows) and GadC (light grey arrows) in different species of bacteria, mostly enteric. Dashed grey arrows are for putative GadC homologues. The arrow orientation indicates the direction of transcription with respect to the positive DNA strand (right direction). Genes in between gadB and gadC are shown as white-filled arrows. The gene identifiers are shown above each arrow. The asterisk indicates species possessing more than one gene coding for glutamate decarboxylase.

For example, in the genome sequence of Brucella microti, a recently described, fast-growing Brucella species (isolated from common vole, red fox and soil) the gadB and gadC genes are intact. On the contrary in the classical species of Brucella (i.e. B. ovis, B. abortus, B. melitensis, B. canis, B. suis, B. neotomae) gadB and/or gadC are inactivated by stop codons and/or frameshift mutations (Audic et al., 2009). Brucella is a genus of Gram-negative, facultative, intracellular bacteria that are highly pathogenic for a variety of mammals, including humans, who become infected through contact with contaminated animal products (non-pasteurized milk and dairy products), through the air, or by direct contact with infected animals or carcasses. Recently, the WHO cited brucellosis as the world's most widespread zoonosis.

The integrity of the gadBC operon was proposed to help B. microti to survive in acidic soils, but also to play a role in intracellular survival, as within host macrophages (Audic et al., 2009). Notably, B. microti is more resistant than B. suis to acid pH in vitro: B. microti survival after 7 h of incubation in minimal medium at pH 4.5 was 25% compared with 0.9% of that of B. suis (Jimenez de Bagues et al., 2010).

The role of gadBC in B. microti acid resistance was recently investigated in light of a possible involvement in acid survival within the host (Occhialini et al., 2012). B. microti was shown to possess a GDAR, i.e. single and double mutants of gadB and gadC of B. microti failed to survive under extreme acid stress (pH 2.5). In addition, B. microti gadBC functionally complemented for the acid sensitive phenotype of the E. coli MG1655 isogenic derivatives ΔgadAB and ΔgadC. Notably, the gadBC system, dispensable for survival and virulence of B. microti during macrophage infection, was shown to contribute to survival of the microorganism in the spleen and liver of mice following oral infection.

PCR on gadAB genes and GAD activity assays provided evidence that, in addition to E. coli, also Escherichia albertii (originally identified as Hafniae alvei) and Citrobacter braakii exhibit a glutamate decarboxylase-dependent acid resistance (Park and Diez-Gonzalez, 2004). So far the presence of a gadBC operon can only be confirmed for E. albertii (Fig. 5), for which the complete genome sequence is available. The enterobacteria Enterobacter cloacae, Morganella morganii, Providencia spp. and Proteus mirabilis were all reported to have no detectable gadAB genes and GAD activity, but to exhibit GDAR (Park and Diez-Gonzalez, 2004). This phenotypic trait was proposed to be markedly different from that of E. coli based on the finding that Morganella, Providencia and Proteus consumed much less glutamate than E. coli (Park and Diez-Gonzalez, 2004). At least in the case of P. mirabilis this conclusion might be reconsidered because an intact gadBC operon is present in its genome (Fig. 5).

Interestingly, in the genome of several bacterial species more than one gene coding for glutamate decarboxylase is present (Fig. 5, species indicated with an asterisk). This might be due to the metabolic need to possess at least one glutamate decarboxylase active in the cell. However, in some microorganisms the duplication (or more) is not limited to the decarboxylase gene, but extends to the antiporter, so gadBC (or gadCB) is present at distinct genetic loci. Besides L. monocytogenes, the best characterized organism, L. brevis and E. tarda appear to belong to this group.

In the food-borne pathogen L. monocytogenes the gadD1T1 (lmo0447–lmo0448 in Fig. 5) operon is primarily responsible for growth under moderately acidic conditions, whereas the gadT2D2 (lmo2362–lmo2363 in Fig. 5) operon ensures survival under extremely acidic conditions (Cotter et al., 2001a; 2001b; 2005). The former operon is absent from most serotype 4 strains and this finding was linked to their reduced ability to grow at low pH (Cotter et al., 2005). This same operon is also absent in L. innocua, a non-pathogenic species belonging to the same genus, which lacks a 10-kb virulence locus, a cluster of genes responsible for pathogenicity in L. monocytogenes (Buchrieser et al., 2003). Karatzas et al. (2010) provided evidence that in L. monocytogenes strain 10403S (serotype 1/2a) GABA accumulates intracellularly at very high levels (> 80 mM) in response to an acid challenge. In addition, in a chemically defined medium, the intracellular accumulation of GABA occurs without glutamate supplementation and can be uncoupled from its efflux. This finding suggests that the first response to extracellular acidification consists in intracellular accumulation of GABA (produced from an existing intracellular glutamate pool), which provides protection against cytoplasmic acidification (Karatzas et al., 2010). In the same work, the GadT2 transporter is proposed to be the only transporter mediating Glu-GABA exchange under extremely acidic conditions in L. monocytogenes. A recent report shows that intracellular accumulation of GABA commonly occurs in various food and clinical isolates of L. monocytogenes (Karatzas et al., 2012). The work carried out in L. monocytogenes sheds light on the biological implications of two different gadBC operons, and its results might apply also to other bacteria.

As stated earlier in this review, in the L. brevis ATCC367 and E. tarda genomes some of the genes annotated as glutamate decarboxylase are evolutionarily quite distant (Fig. 3), thus suggesting that they are not performing the expected reaction in the cell. It is possible that if a gadBC duplication has occurred in these microorganisms this might have happened early in evolution and given rise to novel activities, still awaiting to be identified. Indeed, in the L. brevis strains OPK-3 and IFO 12005 only one gadB gene was cloned (Park and Oh, 2007a; Hiraga et al., 2008). In L. brevis strain IFO 12005 there is a gene similar to gadC upstream of the gadB gene (LVIS_0078 in L. brevis ATCC367 reference strain, Fig. 5). This gene shares 42.5% identity with L. lactis gadC. It was therefore proposed that IFO 12005 has an acid tolerance mechanism similar to that of L. lactis (Hiraga et al., 2008).

The Search Tool for the Retrieval of Interacting Genes (STRING;; Szklarczyk et al. 2011), which provides complete coverage and ease of access to experimental and predicted interaction (direct or indirect) information, assigns the highest possible confidence score (> 0.9) to gadC when searching for gadB in the genomes of Yersinia enterocolitica, E. tarda and Lawsonia intracellularis. In addition, the proximity of the gadB and gadC genes suggests that also in the above organisms gadB and gadC are in an operon arrangement and functionally linked.

To our knowledge, there are no reports on GDAR in Y. enterocolitica, which was shown to possess an efficient urease system enhancing its survival in the stomach and in other acidic environments (de Koning-Ward and Robins-Browne, 1997; Bhagat and Virdi, 2009).

In the E. tarda virulent strain PPD130/91, transposon insertion in the gadB gene (which is missing in all known avirulent strains) yielded a mutant strain unable to survive and to cause infection inside the host fish (Srinivasa Rao et al., 2003). The presence of the gadBC operon in the emerging pathogen E. tarda, the only species of the genus pathogenic for humans (Schlenker and Surawicz, 2009), suggests that these genes might play a role in protecting from the stomach acidity. Indeed ingestion of E. tarda-contaminated water or improperly cooked fish is responsible for Salmonella-like gastroenteritis (Leung et al., 2012). The involvement of gadBC in E. tarda extreme acid survival, however, still awaits to be experimentally proved.

Salinispora tropica (Gram-positive) and Bordetella avium (Gram-negative) are two non-enteric microorganisms in which the gadB and gadC genes lie next to each other in the chromosome. In both microorganisms the gadB and gadC genes are divergently oriented. S. tropica belongs to the order Actinomycetales and has been proven to be a rich source of secondary metabolites, including salinosporamide A, which is currently in clinical trials for the treatment of cancer (Gulder and Moore, 2010). It is tempting to speculate that in this marine microorganism gadB and gadC play a role in the response to osmotic stress (De Biase et al., 1999), typically encountered by aquatic microorganisms.

Notably, in B. avium the gadB and gadC genes are spaced by the gene coding for the acid shock chaperone HdeB (Fig. 5). The physical association of these three genes, which are known to be involved in acid resistance in E. coli, suggests that in B. avium they perform a physiological function linked to protection from acid stress. Whether such genes are important to B. avium lifestyle is currently unknown. This specific issue deserves to be investigated as it might shed light either on a different route of transmission (oral?) of this microorganism or on a novel function played by the gad genes, besides protecting from acid stress.

Similarly to B. avium, the gadB and gadC genes are not adjacent also in other bacteria, where they are rather spaced by one or two functionally linked genes. In Bacteroides thetaiotaomicron and Clostridium perfringens (Fig. 5) the intervening genes code for proteins that participate either to glutamate biosynthesis (B. thetaiotaomicrom BT_2571 is a glutaminase) or to GABA export and GDAR (C. perfringens CPE2059 is the homologue of E. coli yhiM, which encodes a membrane protein involved in GABA export and GDAR; D. De Biase, pers. comm.; Nguyen and Sparks-Thissen, 2012). C. perfringens, a major cause of food poisoning in developed countries, is a representative member of Clostridia, which from the evolutionary perspective are considered the most ancient bacteria. GAD from C. perfringens was purified to homogeneity and partially characterized (Cozzani et al., 1970). Intracellular GAD levels increase in the late logarithmic phase of growth and are positively influenced by reduction of the pH of the culture medium (Cozzani et al., 1975). It is therefore likely that in C. perfringens the gadBC system participates in protection from gastric acidity and/or in other acidic environments (Tennant et al., 2008).

Based on the annotated genomes (source: NCBI genome database, complete and in progress), B. dentium, B. adolescentis L2-32 and B. angulatum DSM 20098 are the only species of the genus Bifidobacterium, which possess the gadB and gadC genes. Moreover, the two genes are adjacent on the chromosome and may form an operon. To date approximately 40 species belonging to the Bifidobacterium genus have been isolated from humans and increasing interest is being paid to bifidobacteria as probiotics because they may exert a beneficial effect on human health (see next section). However, in addition to health-promoting taxa, the genus Bifidobacterium also includes B. dentium, an opportunistic pathogen of the oral cavity, associated with dental caries (Mantzourani et al., 2009a,b). A recent analysis of the B. dentium Bd1 genome has provided evidence that B. dentium strains share a high level of genome conservation and that a limited number of horizontal gene transfer acquisition events were sufficient to differentiate this opportunistic cariogen from the commensal species belonging to the gut microbiota (Ventura et al., 2009). Notably, when B. dentium is exposed to acidic conditions, the expression of gadB and gadC increases by 90- and 51-folds respectively, consistently with their well-known role in acid resistance (Ventura et al., 2009). Indeed, B. dentium was reported to exhibit a significant ability to survive a 3 h exposure at pH 4.0 and that at an external pH of 4.0 it maintains the internal pH at 5.25 (Nakajo et al., 2010). In the same study it was proposed that, because no energy source was available for the H+-ATPase (to pump out protons) during the cell viability test, B. dentium must employ another mechanism to protects itself from acidification. The expression under acidic conditions of B. dentium Bd1 gadBC genes may contribute to the fitness and competitiveness within the oral niche of this opportunistic pathogen.

GABA production by the decarboxylase-antiporter system in health and disease

In the previous section, the occurrence of the gadBC operon in different species of bacteria was mainly dealt with as a trait improving fitness in pathogenic bacteria. The gadBC system, however, is present also in non-pathogenic bacteria, in which protection from acid stress is just one of the functions assigned to the GAD system (Siragusa et al., 2007; Li and Cao, 2010; Su et al., 2011).

Indeed, in lactic acid bacteria (LAB) the ability to produce and release GABA has been linked to additional physiological functions, besides being required to survive the passage through the gastrointestinal tract en route to the intestine where these bacteria exert a beneficial effect (Siragusa et al., 2007). Sanders et al. (1998) and Higuchi et al. (1997) suggested an alternative fascinating hypothesis for the function of the gadBC operon: intracellular glutamate decarboxylase-mediated proton consumption combined with the electrogenic antiport carried out by GadC (Glu0in/GABA+1out at acidic pH) generates a proton motive force, which can be coupled with energy production. Indeed, three successive cycles of decarboxylation-antiport would be sufficient to establish an electrogenic gradient to be used to promote the synthesis of one ATP molecule (Higuchi et al., 1997; Sanders et al., 1998). This strategy can be advantageous not only to anaerobes such as LAB, but also to facultative anaerobes such as E. coli and Shigella that colonize the gut environment. An hypothesis which is reinforced by the recent finding that in E. coli the expression of the gadBC operon increases under conditions of respiratory stress (Shepherd et al., 2010). GABA synthesis and antiport may therefore contribute to the establishment of the proton motive force required for ATP production under O2-limiting conditions.

Several reports have shown that LAB possess the remarkable ability to produce and release GABA in the growth medium (for a review see Li and Cao, 2010). GADs from different LAB strains were characterized at the biochemical level and the corresponding genes were cloned and expressed (Park and Oh, 2007a; Hiraga et al., 2008; Komatsuzaki et al., 2008; Fan et al., 2012). LAB possess special physiological activities and are generally regarded as safe. All the GABA-producing LAB strains were isolated from traditionally fermented foods such as kimchi, cheese, sourdough (Li and Cao, 2010). In particular glutamate decarboxylation plays an important role in the fermentation of protein rich foods, which physiologically contain high levels of free and protein-derived glutamate or glutamine (Su et al., 2011).

The interest in GABA-enriched foods, such as yogurt, cheese, green tea, rice germ (Abe et al., 1995; Oh et al., 2003; Hayakawa et al., 2004; Zhang et al., 2006; Park and Oh, 2007b) and their commercial potential as functional foods (Gobbetti et al., 2010), resides in the functions attributed to GABA, which does not act only as neurotransmitter, hypotensive, tranquillizer and diuretic (Wong et al., 2003). In fact, GABA was reported to be a strong secretagogue of insulin (Adeghate and Ponery, 2002), to promote the healing process of cutaneous wounds (Han et al., 2007) and to enhance survival of dermal fibroblasts exposed to oxidative stress (Ito et al., 2007). Notably GABA-enriched food was shown to lower blood pressure in spontaneously hypertensive and normotensive Wistar-Kyoto rats as well as in humans (Inoue et al., 2003; Hayakawa et al., 2004) and to exert a positive effect in chronic alcohol related symptoms (Oh et al., 2003). More recently, a study showed that grape must fermented with Lactobacillus plantarum DSM 19463 yields GAD-derived GABA at millimolar levels that induce the expression of human genes involved in skin protection, i.e. β-defensin-2, hyaluronan synthase and filaggrin (Di Cagno et al., 2010). This intriguing finding could lead to new cosmetic formulations that promote the antimicrobial barrier of the skin.

Conclusions and perspectives

The gadBC system (or any other gene arrangement in which glutamate decarboxylase is expressed together with the Glu/GABA antiporter) is interesting for several reasons.

In many pathogenic bacteria the presence and expression of the gadBC genes has been strongly linked to ability to survive under extremely acidic conditions, such as those encountered during the transit through the stomach of human host and in the phagosome. On the other hand, the ability of non-pathogenic bacteria to produce and release GABA in the growth medium during fermentation does not seem to respond exclusively to the need to survive in the acidic environment produced by the fermentation products.

Glutamate decarboxylase is a widespread enzyme and its function, structure and biochemical properties have attracted a lot of interest because GABA, its product, a non-proteinaceous amino acid, plays different important roles in living organisms. For this reason there is a growing interest in the use of LAB, also in the form of immobilized cells, as cell factories for GABA not only to develop foods with health-promoting properties but also to support the industrial production of GABA. Interestingly, agricultural surplus (Di Cagno et al., 2010) and waste streams from bioethanol production (Lammens et al., 2011) are now being considered as low cost sources of glutamate for biotechnological conversion into GABA, by using either live cells (Di Cagno et al., 2010; Li and Cao, 2010) or purified glutamate decarboxylase (Lammens et al., 2009).


We are grateful to Dr G. Capitani and Professor Robert A. John for useful suggestions and for critically reading the manuscript. E.P. is recipient of a bursary from the Istituto Pasteur-Fondazione Cenci Bolognetti. This work is dedicated to Professor Robert John, Professor Donatella Barra and Professor Francesco Bossa on the occasion of their seventieth birthday.