Synthesis of γ-Aminobutyric Acid by Lactic Acid Bacteria Isolated from a Variety of Italian Cheeses

Twenty-two Italian cheese varieties were considered in this study. Cheese varieties were manufactured by using milk from different cows (Parmigiano Reggiano, Barricato San Martino, Vento d'Estate, Ubriaco di Raboso, Caciocavallo, Gorgonzola, and Crescenza), buffaloes (Mozzarella), sheep (Pecorino Piemontese, Pecorino Marchigiano, Pecorino Umbro, Pecorino del Reatino, Pecorino Sardo, Pecorino di Filiano, Pecorino del Tarantino, and Pecorino Leccese), sheep and cows in combination (Canestrato Pugliese and Caciotta di Urbino), or goats (Caprino di Valsassina, Caprino di Cavalese, Flor di Capra, and Capritilla) which was used raw (Parmigiano Reggiano, Barricato San Martino, Vento d'Estate, Ubriaco di Raboso, Canestrato Pugliese, Casciotta di Urbino, Pecorino Piemontese, Pecorino Marchigiano, Pecorino del Reatino, Pecorino Sardo, Pecorino di Filiano, Pecorino del Tarantino, and Pecorino Leccese) or pasteurized (all other cheeses). Other main technological traits were as follows: (i) use of primary natural starters (Parmigiano Reggiano, Caciocavallo, Mozzarella, Canestrato Pugliese, Pecorino Sardo, and all goat cheeses) or commercial starters (Gorgonzola and Pecorino Umbro); (ii) use of calf powder or liquid (Parmigiano Reggiano, Gorgonzola, Crescenza, Mozzarella, Canestrato Pugliese, Casciotta di Urbino, and all goat cheeses) or lamb (Caciocavallo) or calf (all other cheeses) paste; (iii) cooking (Parmigiano Reggiano) or stretching (Caciocavallo and Mozzarella); and (iv) very long ripening (18 months; Parmigiano Reggiano), long ripening (8 months; Canestrato Pugliese), medium ripening (2 to 5 months; Barricato San Martino, Vento d'Estate, Ubriaco di Raboso, Caciotta di Urbino, Caciocavallo, and all sheep and goat cheeses), or no ripening to short ripening (0 to 45 days) (Mozzarella, Crescenza, and Gorgonzola).

Cheeses were supplied in triplicate (different batches from the same factory) by local cheese markets and were stored at 4°C for a few hours before analyses. All the analyses were carried out at least in duplicate for each batch of cheese (total of six analyses for each cheese variety).

Moisture and pH were determined as reported by the International Dairy Federation (22, 23). Soluble and total nitrogen (N) were determined by the micro-Kjeldahl method (21).

Thirty grams of cheese was suspended in 90 ml of 50 mM phosphate buffer, pH 7.0, and treated for 10 min with a stomacher (PBI International, Milan, Italy). The suspension was kept at 40°C for 1 h under gentle stirring (150 rpm) and centrifuged at 3,000 × g for 30 min at 4°C. The supernatant was filtered through Whatman no. 2 paper, and the pH of the extract was adjusted to 4.6 using 1 N HCl. The suspension was centrifuged at 10,000 × g for 10 min. Finally, the supernatant was filtered through a Millex-HA 0.22-μm-pore-size filter (Millipore Co., Bedford, MA). Total and individual free amino acids (FAAs) contained in the pH 4.6-soluble nitrogen fraction were analyzed by a Biochrom 30 series amino acid analyzer (Biochrom Ltd., Cambridge Science Park, England) with a sodium cation-exchange column (20 by 0.46 cm [inner diameter]). A mixture of amino acids at known concentrations (Sigma Chemical Co., Milan, Italy) was added with cysteic acid, methionine sulfoxide, methionine sulfone, tryptophan, ornithine, glutamic acid, and GABA and used as standard. Proteins and peptides in the samples were precipitated by addition of 5% (vol/vol) cold solid sulfosalicylic acid, holding the samples at 4°C for 1 h, and centrifuging them at 15,000 × g for 15 min. The supernatant was filtered through a 0.22-μm-pore-size filter and diluted, when necessary, with sodium citrate (0.2 M, pH 2.2) loading buffer. Amino acids were postcolumn derivatized with ninhydrin reagent and detected by absorbance at 440 (proline and hydroxyproline) or 570 (all the other amino acids) nm.

Samples (20 g) of cheeses were diluted in 180 ml of sodium citrate (2%, wt/vol) solution and homogenized with a Lab-Blender 400 stomacher (PBI International Milan, Italy). Serial dilutions were made in one-quarter-strength Ringer's solution and plated on DeMan, Rogosa, Sharpe (MRS) (lactobacillus) or M17 (coccus) agar (Oxoid Ltd., Basingstoke, Hampshire, England) for viable counts. Mesophilic or thermophilic lactic acid bacteria were enumerated after incubation at 30 or 42°C for 48 to 72 h. At least 10 colonies for each medium and cheese variety, possibly with different morphologies, were isolated from the highest plate dilution. Gram-positive, catalase-negative, nonmotile rod and coccus isolates were cultivated in MRS or M17 broth (Oxoid Ltd.) at 30 or 42°C for 24 h and restreaked into MRS or M17 agar. All the isolates considered for further analyses showed the capacity of acidifying the culture medium. Microbial cultures were stored at −20°C in 10% (vol/vol) glycerol.

Twenty-four-hour-old cells of 440 presumptive lactic acid bacterium strains were harvested by centrifugation (9,000 × g for 15 min at 4°C), washed twice with sterile 0.05 M potassium phosphate buffer (pH 7.0), and resuspended in sterile physiological solution (0.85% NaCl) at an A₆₂₀ of 2.5, which corresponded to a cell density of ca. 8.5 to 9.0 log CFU ml⁻¹. GAD activity was measured as described previously (33). The reaction mixture consisted of 900 μl of 50 mM sodium acetate buffer (pH 4.7) containing 2 mM l-glutamate, 0.1 mM pyridoxal phosphate, 100 μl of cell suspension, and 0.05% (wt/vol) NaN₃ (final concentration). After 24 h at 30°C, the concentration of GABA was determined using a Biochrom 30 series amino acid analyzer (Biochrom Ltd.).

Genomic DNA from each strain was extracted as reported by De Los Reyes-Gavilán et al. (10) from 2-ml samples of overnight cultures grown in MRS or M17 broth. Two primer pairs (Invitrogen Life Technologies, Carlsbad, CA), LacbF/LacbR and LpCoF/LpCoR (7), were used to amplify a 16S rRNA gene fragment of presumptive lactic acid bacteria. Fifty microliters of each PCR mixture contained 200 μM of each 2′-deoxynucleoside 5′-triphosphate (dNTP), 1 μM of both forward and reverse primers, 2 mM MgCl₂, 2 U of Taq DNA polymerase (Invitrogen) in the supplied buffer, and approximately 50 ng of DNA. The expected amplicons of about 1,400 and 1,000 bp (after amplification with primer pairs LacbF/LacbR and LpCoF/LpCoR, respectively) were eluted from the gel and purified by the GFX PCR DNA and Gel Band Purification Kit (Amersham Biosciences, Piscataway, NJ). DNA sequencing reactions were performed by PRIMM srl (San Raffaele Biomedical Science Park, Milan, Italy). Taxonomic strain identification was performed by comparing the sequences of each isolate with those reported in the Basic BLAST database (2).

Genomic DNA was extracted as reported above. Two primers (Invitrogen) with arbitrarily chosen sequences (M13, 5′-GAGGGTGGCGGTTCT-3′; P4, 5′-CCGCAGCGTT-3′; and P7, 5′-AGCAGCGTGG-3′) (6, 55) were used singly in two series of amplifications. The reaction mixture contained 200 μM of each dNTP, 1 to 2 μM primer, 1.5 to 3 μM MgCl₂, 1.25 U Taq DNA polymerase (Invitrogen), 2.5 μl PCR buffer, 25 ng DNA, and sterile double-distilled water to 25 μl. For amplifications with primer P4, the PCR program comprised 45 cycles of denaturation for 1 min at 94°C, annealing for 1 min at 35°C, and elongation for 2 min at 72°C; the cycles were preceded by denaturation at 94°C for 4 min and followed by elongation at 72°C for 5 min. For primer M13, amplification reactions were performed using one cycle at 94°C for 60 s (denaturing), 42°C for 20 s (annealing), and 72°C for 2 min (elongation). PCR products were separated by electrophoresis (2 h at 130 V) on a 1.5% (wt/vol) agarose gel (Invitrogen), and the DNA was detected by UV transillumination after staining with ethidium bromide (0.5 μg/ml). The molecular weight of the amplified DNA fragments was estimated by comparison with a 1 Kb Plus DNA Ladder (Invitrogen). Gels were acquired using a UNIsave gel documentation system camera, model GAS9200/1/2/3, version 11 (UVItec Limited, Cambridge, United Kingdom). Electrophoretic profiles were compared using Quantity One software (Bio-Rad, Milan, Italy).

Harvested cells of lactic acid bacteria were washed in 50 mM phosphate buffer (pH 7.0), centrifuged at 9,000 × g for 15 min at 4°C, and resuspended in reconstituted skimmed milk (RSM) at a cell density of ca. 6.8 log CFU ml⁻¹. RSM was supplemented with l-monosodium glutamate (20 mM) and incubated at 30°C for 24 h, and the pH was recorded online. Acidification data were modeled according to the Gompertz equation as modified by Zwietering et al. (57), where y is log (ΔpH Δt⁻¹, units of pH h⁻¹), k is the initial level of the dependent variable to be modeled, A (ΔpH) is the difference in pH (units) between the initial value (pH₀) and the value reached after 18 h, μ_max is the maximum acidification rate (ΔpH h⁻¹), λ is the length of the latency phase of acidification expressed in hours, and t is time. After incubation, the pH 4.6-soluble nitrogen fraction was prepared as described elsewhere. Total and individual FAAs contained in the pH 4.6-soluble nitrogen fraction were analyzed using a Biochrom 30 series amino acid analyzer (Biochrom Ltd.).

Total DNAs were obtained as described above. Primers designed from highly conserved regions of GAD, CoreF/CoreR (5′-CCTCGAGAAGCCGATCGCTTAGTTCG-3′ and 5′-TCATATTGACCGGTATAAGTGATGCCC-3′, respectively) (PRIMM), were used to amplify genes for GABA-synthesizing enzymes of selected lactic acid bacteria (34, 40). Fifty microliters of each PCR mixture contained 200 μM of each dNTP, 1 μM of both primers, 2 μM MgCl₂, 2 U of Taq DNA polymerase (Invitrogen) in the supplied buffer, and ca. 50 ng of DNA. PCR amplification was performed using the GeneAmp PCR System 9700 thermal cycler (Applied Biosystems). Amplification conditions were changed according to the primer used. PCR products were separated by electrophoresis on a 1.5% (wt/vol) agarose gel (Gibco BRL, France) and stained with ethidium bromide (0.5 μg ml⁻¹). The amplicons obtained were eluted from the gel and purified by the GFX PCR DNA and Gel Band Purification Kit (Amersham Biosciences). DNA sequencing reactions were performed by PRIMM srl. Sequence comparison was performed by using the Basic BLAST database. Translation of nucleotide sequences analyses was performed by using OMIGA software (Oxford Molecular, Madison, WI) or the ExPASy translation routine at the ExPASy Moleculary Biology Server of the Swiss Institute of Bioinformatics (http://ca.expasy.org/ ). Similarity researches were performed with the advanced BLAST algorithm available at the National Center for Biotechnology Information site (htpp://www.ncbi.nlm.nih.gov/ ). Sequence alignments were conducted with the ClustalW algorithm at the ClustalW server at the European Bioinformatics Institute (http://www.ebi.ac.uk/clustalw/index.html ).

Simulated gastric and intestinal fluids were created (13, 43). Stationary-phase-grown cells of lactic acid bacteria were harvested at 9,000 × g for 15 min at 4°C, washed with physiologic solution (0.85% NaCl), and resuspended in 50 ml of simulated gastric juice (ca. 10 log CFU ml⁻¹) which contained NaCl (125 mM liter⁻¹), KCl (7 mM liter⁻¹), NaHCO₃ (45 mM liter⁻¹), and pepsin (3 g liter⁻¹) (Sigma) (56). The final pH was adjusted to 2.0, 3.0, and 8.0. The value pH 8.0 was used to investigate the influence of the components of the simulated gastric juice apart from the effect of low pH (13). The effect of nitrite on lactic acid bacteria was also investigated by adding N (1 mg liter⁻¹), l-ascorbic acid (5 mg liter⁻¹), iodide (1 mg liter⁻¹), or thiocyanate (30 mg liter⁻¹) (54). The suspension was incubated at 37°C with gentle agitation at 100 revolutions min⁻¹ (29, 54). Aliquots of this suspension were taken at 0, 90, and 180 min, and viable counts were determined. The effect of gastric digestion was also determined by suspending cells in RSM (25 mg ml⁻¹) before inoculation of simulated gastric juice at pH 2.0. The final pH after the addition of RSM was ca. 3.0. This condition was assayed to simulate the effect of the food matrix during gastric transit (56). After 180 min of gastric digestion, cells were harvested and resuspended in simulated intestinal fluid which contained 0.1% (wt/vol) pancreatin and 0.15% (wt/vol) oxgall bile salt (Sigma) at pH 8.0. The suspension was incubated at 37°C under gentle agitation at 100 revolutions min⁻¹, and aliquots were taken at 0, 90, and 180 min (13).

RSM (25 mg/ml), containing ca. 9 log CFU ml⁻¹ of lactic acid bacteria, was subjected to pepsin and pancreatin digestion as described above (43). After digestion, the suspension was further incubated for 24 h at 37°C, at 100 revolutions min⁻¹, to mimic the synthesis of GABA by bacteria colonizing the GI tract. Digested samples were recovered after 2, 4, 16, and 24 h of incubation (8) and subjected to free fatty acid analysis. All treatments were carried out also using RSM without bacterial inoculum as the control.

Experimental data were subjected to analysis of variance, and pairwise comparison of treatment means was achieved using Tukey's procedure at P < 0.05 using a statistical software program, Statistica for Windows (Statistica 6.0 per Windows 1998). Multivariate statistical analysis and principal component analysis were performed using Statistica 6.0 and Statistical software for MS Excel per Windows 1998.

The sequences in this study were deposited in GenBank, National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov ). The accession numbers were EF174472, EF174473, EF174474, and EF174475 for the GAD gene fragments of Lactobacillus paracasei PF6, Lactobacillus delbrueckii subsp. bulgaricus PR1, Lactobacillus plantarum FC48, and Lactococcus lactis PU1, respectively.

Twenty-two Italian cheese varieties having features that may influence directly or indirectly the synthesis of GABA were considered in this study. The values of pH, moisture, and pH 4.6-soluble N were in the range of 4.68 to 6.70, 30 to 60%, and 7.0 to 34%, respectively. The concentrations of total FAAs and glutamate, contained in the pH 4.6-soluble N extracts of the 22 Italian cheese varieties, markedly varied from 19.17 to 231,800.0 mg kg⁻¹ and from 0.0 to 7,190.0 mg kg⁻¹, respectively. The concentration of GABA varied from 0.260 to 391 mg kg⁻¹. The lowest concentrations were found for Vento d'Estate (0.260 mg kg⁻¹), Mozzarella (0.32 mg kg⁻¹), and Crescenza (0.77 mg kg⁻¹) cheeses. The majority of the cheeses (Parmigiano Reggiano, Barricato San Martino, Ubriaco di Raboso, Caciocavallo, Gorgonzola, Canestrato Pugliese, Casciotta di Urbino, Pecorino del Tarantino, Pecorino Piemontese, Flor di Capra, Caprino di Cavalese, Caprino di Valsassina, and Capritilla) had concentrations of GABA which ranged from 4 to 100 mg kg⁻¹. The highest values of GABA were found in Pecorino Marchigiano (289 mg kg⁻¹), Pecorino del Reatino (290 mg kg⁻¹), Pecorino Leccese (290 mg kg⁻¹), Pecorino Umbro (330 mg kg⁻¹), and, especially, Pecorino di Filiano (391 mg kg⁻¹). The values of pH for the cheeses containing the highest concentrations of GABA were 4.68 to 5.70. Overall, no statistical correlation was found between the concentration of GABA and the levels of l-glutamate or other free fatty acids. As determined by multivariate statistical analysis, the use of sheep milk and cheese ripening for 1.5 to 5 months seemed to be related to the highest concentration of GABA. The other technological traits or cheese characteristics did not show a statistical correlation.

The number of presumptive mesophilic and thermophilic lactobacilli in cheeses varied from 6.90 to 8.51 log CFU g⁻¹ and from ≤3.0 to 6.28 log CFU g⁻¹, respectively. Lactococci and streptococci were ≤3.0 to 7.8 log CFU g⁻¹ and ≤3.0 to 5.43 log CFU g⁻¹, respectively. As expected, the highest numbers of thermophilic lactobacilli and streptococci were found in cheeses where primary starters were used. For each cheese, 20 gram-positive, catalase-negative, nonmotile, and acidifying isolates (total of 440) were randomly taken from the plates containing the highest dilutions and screened for GAD activity. Only 61 isolates synthesized GABA under in vitro conditions at 30°C for 24 h, pH 4.7. All these isolates were further identified by partial sequencing of the 16S rRNA gene (Table 1). Twelve species were found: Lactobacillus plantarum (17 isolates), Leuconostoc mesenteroides (two), Weissella cibaria (one), Lactobacillus paracasei (16), Lactobacillus brevis (three), Lactobacillus casei (five), Streptococcus thermophilus (six), Leuconostoc pseudomesenteroides (two), Lactococcus lactis (two), Lactobacillus delbrueckii subsp. bulgaricus (two), Enterococcus durans (one), and Lactobacillus rhamnosus (two). Two isolates of unidentified Lactobacillus species were also found. Except for Gorgonzola, Pecorino cheeses especially contained the highest numbers of GABA-producing strains (four to six). Apart from the cheese source, 17 and 16 isolates corresponded to the species L. plantarum and L. paracasei, respectively. Overall, 53 isolates were mesophilic lactic acid bacteria (45 mesophilic lactobacilli) and only eight were thermophilic lactic acid bacteria (six S. thermophilus and two L. delbrueckii subsp. bulgaricus strains). To exclude clonal relatedness, three primers, M13, P4, and P7, with arbitrarily chosen sequences, were used to characterize the 61 GABA-producing isolates by RAPD-PCR analysis. The reproducibility of RAPD fingerprints was assessed by comparing the PCR products obtained from three separate cultures of the same strain. The size of the amplified fragments ranged from 100 to 1,000 bp, and the number of fragments varied from one to seven per primer per isolate. The RAPD profiles generated by the above primers were highly discriminative and reproducible with consistent fragment patterns. The polymorphic bands in total distinguished the isolates from each other with at least a 2.5% dissimilarity level (data not shown). The capacity for synthesizing GABA of the 61 isolates was further assayed in RSM. After 24 h of fermentation at 30°C, all isolates reached the cell density of 8.5 to 9.0 log CFU g⁻¹. The kinetics of acidification was characterized by a median value of ΔpH of 2.28. The majority of the strains caused ΔpHs which ranged from 2.00 to 2.71. The median value for μ_max was 0.03 ΔpH min⁻¹. The median value for the concentration of GABA was 1.71 mg kg⁻¹, and the 25th and 75th percentiles of the data ranged from 1.0 to 3.67 mg kg⁻¹, respectively. L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, L. lactis PU1, L. plantarum C48, and L. brevis PM17, which showed the highest synthesis of GABA (99.9, 63.0, 36.0, 16.0, and 15.0 mg kg⁻¹, respectively), were located out of the error bars of the box plot (data not shown). Overall, the concentration of total FAAs in the fermented RSM showed increases which ranged from 120 to 390.6 mg kg⁻¹. The concentrations of GABA, glutamic acid, and total FAAs of each fermented RSM were subjected to principal component analysis using a covariance matrix (Fig. 1). The first two principal components explained ca. 83.26% of the total variance. PC1 showed the distribution of the samples according to the concentration of FAAs, while PC2 differentiated samples based on the concentration of GABA. The fermented RSMs started with L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, L. lactis PU1, L. plantarum C48, or L. brevis PM17 were characterized by the highest synthesis of GABA and were located in different zones of the plane delimited by the two principal components.

L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, L. lactis PU1, and L. plantarum C48, corresponding to the highest GABA-producing strains, were screened for GAD genes. Primers designed from highly conserved region of GAD genes gave a PCR product of approximately 540 bp for all the strains. The predicted amino acid sequences of L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, L. lactis PU1, and L. plantarum C48 showed 98, 100, 86, and 95% identity to gadB of L. plantarum WCFS1 (accession number NP_786643.1), respectively. Alignments of the internal GAD gene fragment of L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, L. lactis PU1, and L. plantarum C48 with the eight most similar sequences of the GAD gene are shown in Fig. 2.

L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, L. lactis PU1, and L. plantarum C48 were incubated at 37°C in simulated gastric fluid at pH 2.0, 3.0, and 8.0. L. lactis PU1 was the only strain which had a very poor survival in all the conditions assayed (data not shown). Therefore, it was excluded from further characterization. After 180 min of incubation in simulated gastric juice at pH 3.0, all the other strains showed decreases lower than 1 log cycle with respect to their initial cell density (10 to 9.6 log CFU ml⁻¹). At pH 8.0 no decrease of survival was found for all strains (Fig. 3A to C). After 180 min at pH 2.0, all strains decreased to 6.0 to 5.0 log CFU ml⁻¹. When RSM (25 mg/ml) was added to the juice at pH 2.0, lactobacilli were resistant to the simulated gastric juice. Probably, this was due to the increased value of pH (3.0) caused by addition of RSM or to the direct protective effect on microbial cells by the food matrix. After 180 min of gastric digestion, cells were exposed to simulated intestinal fluid for a subsequent 180 min at pH 8.0 (Fig. 3A to C). Cell survival depended on the pH of gastric digestion. The decrease for cells previously treated at pH 8.0, 3.0, or 2.0 with the addition of RSM was always lower than 1 log cycle. When the gastric digestion was at an initial pH of 2.0, the survival of simulated intestinal fluid was approximately 3 and 4 log CFU ml⁻¹ for L. paracasei PF6 and L. plantarum C48, respectively. Nonculturable cells of L. delbrueckii subsp. bulgaricus PR1 were found in 1 ml of sample.

The hypothesis to be investigated concerned the capacity of selected strains to synthesize GABA during gut transit and without the addition of l-glutamate. RSM containing 9 log CFU ml⁻¹ of L. paracasei PF6, L. delbrueckii subsp. bulgaricus PR1, or L. plantarum C48 was subjected to sequential hydrolysis by pepsin (pH 2.0) and pancreatin (pH 8.0) as described above. After digestion, the suspensions were further incubated for 24 h at 37°C under stirring conditions, to mimic the bacterial GI transit. As previously shown (Fig. 3A to C), RSM protected the cells and the numbers of the three strains did not vary with respect to the initial cell density. Nevertheless, all strains synthesized GABA at a lower concentration than that found in fermented RSM (30°C for 24 h, initial pH 6.6) (Fig. 4). Among the three strains, L. paracasei PF6 showed the highest synthesis of GABA, reaching ca. 20 mg kg⁻¹ after 24 h of incubation.