Lactobacillus rossiae strain isolated from sourdough produces putrescine from arginine

This work reports a Lactobacillus rossiae strain (L. rossiae D87) isolated from sourdough that synthesizes putrescine - a biogenic amine that raises food safety and spoilage concerns - from arginine via the ornithine decarboxylase (ODC) pathway. The odc and potE genes were identified and sequenced. These genes respectively encode ornithine decarboxylase (Odc), which participates in the decarboxylation of ornithine to putrescine, and the ornithine/putrescine exchanger (PotE), which exchanges ornithine for putrescine. Transcriptional analysis showed that odc and potE form an operon that is regulated transcriptionally by ornithine in a dose-dependent manner. To explore the possible role of the ODC pathway as an acid stress resistance mechanism for this bacterium, the effect of acidic pHs on its transcriptional regulation and on putrescine biosynthesis was analysed. Acidic pHs induced the transcription of the odc-potE genes and the production of putrescine over that seen at neutral pH. Further, putrescine production via the ODC system improved the survival of L. rossiae D87 by counteracting the acidification of the cytoplasm when the cells were subjected to acidic conditions. These results suggest the ODC pathway of L. rossiae D87 provides a biochemical defence mechanism against acidic environments.


Results
Production of BAs by L. rossiae D87. To check the capacity of L. rossiae D87 to produce histamine, tyramine, cadaverine and putrescine, it was grown in De Man, Rogosa and Sharpe (MRS) supplemented with the corresponding precursors (histidine, tyrosine, lysine, and ornithine or agmatine in the case of putrescine) at a final concentration of 2.5 mM. The strain was not able to decarboxylate histidine, tyrosine or lysine (data not shown). However, it was able to produce putrescine from ornithine, although not from agmatine (data not shown), indicating L. rossiae D87 follows the ODC pathway.
To monitor the production of putrescine from ornithine over the growth curve, L. rossiae D87 was grown in MRS broth and in MRS + Ort (i.e., supplemented with 20 mM ornithine) and the concentrations of arginine, ornithine and putrescine quantified by UHPLC at different times. In the culture supplemented with ornithine, concomitant production of putrescine was observed (Table 1). Surprisingly, the concentration of putrescine was slightly higher than the amount of ornithine added. Indeed, putrescine production was also observed in cultures with no ornithine supplementation. In both cases, the non-expected amount of putrescine detected corresponded to the amount of arginine present in the medium (Table 1). Arginine is not a direct substrate for the production of putrescine. However, many bacteria can follow the arginine deimination (ADI) pathway 31 , in which arginine is deiminated to ornithine with the concomitant production of ATP and NH 4 + . To test whether L. rossiae D87 could deiminate arginine towards ornithine via the ADI pathway and its further decarboxylation to putrescine via the ODC pathway, cultures supplemented with arginine instead of ornithine were monitored for the production of putrescine. As shown in Table 1, most of the arginine present in the medium (23.4 ± 2.8 mM) was converted into putrescine at 48 h (19.9 ± 5.1 mM), indicating that L. rossiae D87 can effectively form putrescine from arginine. Sequence analysis of the L. rossiae D87 ODC cluster. To identify the genes involved in the production of putrescine, a Polymerase Chain Reaction (PCR) strategy using the primers previously described for the detection of the odc gene ODC3/ODC16 32 was followed. A single band of the expected size was obtained (data not shown). The fragment was sequenced and showed similarity to odc genes. To obtain the complete sequence of the putrescine-production-associated genes, the unknown DNA sequences adjacent to the odc fragment gene were examined in both the 5′ and 3′ directions using a combination of reverse PCR and PCR amplification with primers based on the sequence of ODC clusters available in databases. A 4165 bp nucleotide sequence was finally obtained (GenBank accession no. KT020759). Sequence analysis showed the ODC cluster to be composed of two genes with the same orientation: odc and potE (Fig. 1A). The cluster was enclosed between two putative rho-independent terminators (ΔG = −13 kcal upstream of odc and ΔG = −9.8 kcal downstream of potE). A conserved ribosome binding site (RBS) was identified 7 nt upstream of the odc start codon and 8 nt upstream of that of potE. No clear promoter consensus sequence was found upstream of odc or potE.
During this work, an internal fragment of another odc gene from a different L. rossiae strain was deposited in the GenBank database (Accession no. HQ141619). Its nucleotide sequence and that of L. rossiae D87 were 96% identical. The L. rossiae D87 odc gene was also similar to odc from Lactobacillus brevis IOEB 9906 (77%), O. oeni BIFI-83 (74%) and Lactobacillus saerimneri 30a (68%), and encoded a protein of 725 amino acids with a predicted molecular weight of 82 kDa. The deduced Odc protein showed a typical pyridoxal 5′-phosphate binding pocket in its extreme N-terminal, followed by the catalytic centre. potE was similar to other potE genes adjacent to odc in L. brevis IOEB 9906 (71%) and O. oeni BIFI-83 (72%). This gene encodes a protein, PotE, of 437 amino acids with a predicted molecular weight of 46.5 kDa. PotE was found to have 12 transmembrane domains, supporting its role as an ornithine/putrescine membrane antiporter, like that of L. brevis IOEB 9906 22 .
The phylogenetic tree ( Fig. 2) constructed -based on the clustalW alignment of those ODC proteins that have been described involved in the production of putrescine 22 -showed the ODCs of LABs to be closely related (except that of L. saerimneri 30a).
odc and potE are cotranscribed as a monocistronic mRNA. Before starting the transcriptional analysis of factors affecting the expression of the ODC cluster, the transcriptional profile of the ODC cluster was determined. cDNA from total RNA of cultures grown in MRS + Ort was used in Reverse Transcription Polymerase Chain Reaction (RT-PCR) amplifications with a pair of primers (Table 2) designed to amplify the intergenic region between odc and potE. A PCR band of the expected size was obtained (Fig. 1A,B) showing that these genes are cotranscribed.
The expression of odc is induced by ornithine and acidic pH. The influence of ornithine on odc expression was examined by Reverse Transcription quantitative Polymerase Chain Reaction (RT-qPCR) in CDM supplemented with different concentrations of this compound (Fig. 3). Since odc and potE were found to be cotranscribed ( Fig. 1), only the expression of odc was quantified by RT-qPCR. As shown in Fig. 3, the presence of ornithine induced odc expression in a concentration-dependent manner.
The influence of acidic pH on odc expression was also examined. As shown in Fig. 4A, the relative expression of odc gene at pH 4 was about 19 times that recorded at pH 7. This indicates that acidic pH induces odc expression in L. rossiae D87. Accordingly, the production of putrescine increased about 10 times (Fig. 4B).
Effect of putrescine production on bacterial growth and medium acidification. To determine whether the production of putrescine offers any advantage to L. rossiae D87 in terms of growth, the OD 600 and  culture pH in MRS and MRS + Ort were monitored (Fig. 5). No differences in growth were observed, although the acidification of the culture media was less strong in the presence than in the absence of ornithine (pH 4.6 compared to pH 4.2). This indicates that the decarboxylation of ornithine to putrescine might play a role in acid resistance.
Putrescine production improves cell survival under acidic conditions. The viability of L. rossiae D87 under acidic stress was evaluated in resting cells. As shown in Fig. 6, this strain was highly resistant to acid challenge. Little lethality was seen after exposure of the cells to pH 4 for 2 h, although it increased at pH 3 and especially at pH 2. The presence of ornithine significantly improved cell survival at pH 3 (p < 0.05) and particularly at pH 2 (by four orders of magnitude; p < 0.01) (Fig. 6). Thus, putrescine production via ornithine decarboxylation improves the fitness of L. rossiae D87 under acidic conditions. Putrescine production via ornithine decarboxylation counteracts acidification of the cytosol in acidic environments. It has been suggested that the mechanism underlying the resistance to acid conferred by the production of putrescine via the ODC pathway in LAB is exerted by maintaining the pH of the cytosol 28 . Changes in the cytosolic pH of L. rossiae D87 were therefore monitored at external pH values ranging from 8.0 to 5.5 in the absence/presence of ornithine (10 mM). Figure 7 shows that in the absence of ornithine, the internal pH dropped as the external pH decreased. When ornithine was present, a drop in the internal pH was also observed as the external pH is reduced, but the gradient was much shallower than in the absence of ornithine. Between pH out 6 and 5.5, the internal pH values in the presence of ornithine were about 1 pH unit higher than in its absence (pH in 6.7 and 5.8 respectively). This indicates that the production of putrescine via ornithine decarboxylation in   L. rossiae D87 counteracts the acidification of the cytosol produced by exposure to an acidic pH challenge, and that the effect is greater the lower the external pH.

Discussion
Putrescine is one of the most commonly encountered BAs in fermented food products [33][34][35] . Several LAB that produce putrescine from agmatine have been identified and characterized 12,16 ; however, far fewer ODC + strains have been described. This work reports L. rossiae D87, isolated from sourdough, to be ODC + , increasing the spectrum of species able to produce putrescine from ornithine. The two possible substrates for the biosynthesis of putrescine -agmatine and ornithine -come from the same amino acid, arginine. While agmatine is the product of arginine decarboxylation, ornithine is the product of its deimination. Putrescine is thus synthesized from arginine by two different routes, each requiring two steps 17 . However, it is important to note that these two steps are not usually performed by the same bacterial species and, therefore, cooperation between different species is usually required for putrescine to accumulate in foods. Remarkably, L. rossiae D87 appears able to perform both of the steps required to produce putrescine from arginine via the ODC pathway -the deimination of arginine to ornithine and its further decarboxylation to putrescine. As far as we know, Weisella halotolerans W22 is the only other microorganism equipped to undertake both these steps 36 . L. rossiae D87 is the first putrescine producer isolated to have been from sourdough, a food in which arginine is available.  L. rossiae has been first isolated from sourdough by Corsetti et al. in 2005 37 , and it is now well known that it is widely distributed and present in this fermented food 38 . However, the capability of L. rossiae sourdough strains to produce BAs has not been previously shown, only one strain isolated from a wine starter has been described as histamine producer 39 . In this work, we describe for first time a strain of L. rossiae with capability to produce the BA putrescine. Moreover, the ability of L. rossiae D87 to undertake both steps from arginine to putrescine constitutes a property that naturally increases the risk of putrescine accumulation.
This strain could, therefore, easily produce putrescine during the fermentation process. Although the presence of BAs -including putrescine -has not been extensively studied in sourdough, some authors have reported the presence of BAs in sourdoughs made from different flour types 35,40 . The potential presence in sourdough of BA-producing microorganisms other than L. rossiae D87 described cannot, therefore, be ruled out.
Since the first description of an ornithine decarboxylation system in the LAB L. saerimneri 30a 41 , few new data on the ornithine decarboxylases of this bacterial group have been collected 16 . The ODC of L. saerimneri 30a appears to be atypical of those seen in LAB. As shown in Fig. 2, it is more similar to that of microorganisms such as Staphylococci than to those of other LAB, including L. rossiae D87. In addition, L. saerimneri 30a has a unique genomic organization in which odc does not have an adjacent specific transporter gene. In contrast, it has a three-component decarboxylase system that, in addition to the odc gene, has a lysine decarboxylase gene (aadc) and a promiscuous amino acid-amine transporter gene (aat) 42 . Phylogenetic tree analysis grouped the L. rossiae D87 Odc with the other LAB Odc proteins, suggesting they have a common origin. In O. oeni and L. brevis, this cluster was acquired through horizontal gene transfer 28,43 , but further work is needed to determine if this is also the case in L. rossiae D87.  Transcriptional studies were performed as a first step in the physiological characterization of the ODC pathway in L. rossiae D87. odc and potE were found to be cotranscribed in a single mRNA (Fig. 1A,B). As far as we know, there are no other data available on transcriptional studies of the ODC system in LAB. In Gram-negative bacteria, however, these genes are also cotranscribed and form an operon 44,45 . No previous studies have been performed on the factors that regulate the expression of the ODC system genes in LAB. The present results show that substrate availability and an acidic pH enhances odc expression. Both conditions play a determining role in the expression of other amino acid decarboxylases in LAB, such as those involved in the decarboxylation of tyrosine 29,46 and histidine 47,48 . No putative regulatory gene has been found close to the ODC system that might be a responsible for the control gene expression. The expression of the ornithine decarboxylase gene (speF) in E. coli is similarly induced by acidic pH and ornithine 45 , and no regulatory gene has been identified either. It has been suggested that RNase III might be involved in the upregulation of the operon 45 .
Putrescine production by ornithine decarboxylation had no effect on L. rossiae D87 growth; the same OD was reached in MRS and MRS supplemented with ornithine (Fig. 5). Amino acid decarboxylation has been proposed as a means of obtaining metabolic energy through an electrogenic amino acid and/or amine antiporter system leading to the generation of proton motive force 25,26 . However, physiological studies on other decarboxylation systems involved in BAs production in enterococci or streptococci indicate them not to have a major impact on cell growth 29,47,49 . Instead, the production of putrescine by ornithine decarboxylation seems to play a role in acid resistance. As shown in Fig. 5, cell cultures grown in the absence of ornithine reach a more acidic pH than those supplemented with ornithine. In addition, the expression of odc in L. rossiae D87 is enhanced at acidic pH, with a concomitantly greater production of putrescine observed (Fig. 4A,B). This might be related to its role in protection against acid stress. The increased survival of L. rossiae D87 at pH 3 and pH 2 in the presence of ornithine supports this idea (Fig. 6). It has been shown that the mechanism by which amino acid decarboxylation protects cells against acidic environments is mediated by the control of the internal pH 28,29 , and the present data appear to confirm this. Ornithine decarboxylation reduced the acidification of the intracellular pH at different external pHs (Fig. 7), more effectively so as the external pH fell.
In conclusion, L. rossiae D87 is a putrescine-producing microorganism found in sourdough that appears to produce this BA as a defence against the acidic pH of this fermented food. In addition to their negative effects on the organoleptic characteristics of the final product, putrescine-producers may thus pose a potential health risk to consumers, suggesting such foods should be checked for their putrescine content.

Methods
Bacterial strains and culture conditions. L. rossiae D87, isolated from sourdough was grown at 30 °C without aeration in De Man, Rogosa and Sharpe (MRS) broth (Oxoid, UK).
To test this organism's capacity to produce BAs it was grown in the same medium supplemented with either histidine, tyrosine, lysine, ornithine or agmatine (final concentration 2.5 mM) as previously described 50 . Where indicated, ornithine or arginine (Sigma-Aldrich, Spain) was added to the medium at different concentrations. When needed, the final pH of the medium was adjusted to pH 2, 3, 4 or 5 with HCl. When necessary, viable cells were counted by pour-plating 10-fold serial dilutions of cell cultures on MRS-agar plates. To study the influence of ornithine on the production of putrescine and the transcriptional activity of odc, the strain was grown at 30 °C in the chemically defined medium (CDM) described by Looijesteijn and Hugenholtz 51 .
Nucleic acid extraction. Total DNA of L. rossiae D87 grown in MRS broth was extracted using the Genomic DNA Purification Kit (Sigma-Aldrich) according to the manufacturer's recommendations. RNA was extracted from 2 ml of cell cultures at mid-exponential growth, following centrifugation in a refrigerated benchtop microcentrifuge (Eppendorf, Germany) running at maximum speed. Total RNA was extracted using TRI reagent (Sigma-Aldrich) as previously described 46 . To eliminate any contaminating DNA, 2 μg of total RNA samples were treated with 2U of DNAse I (Fermentas, Lithuania) for 2 h at 37 °C. The absence of contaminating DNA was checked via reverse transcription quantitative Polymerase Chain Reaction (RT-qPCR) using primers that amplify the 16S rRNA gene, ROSQ16F and ROSQ16R (Table 2). Total RNA concentration was determined in an Epoch Microplate Spectrophotometer (BioTek, USA). DNA manipulation. L. rossiae D87 was identified at the species level by partial amplification and sequencing of the 16S rRNA gene. PCR was performed using total DNA as a template (1 ng), the universal primers pA and pH 52 , and 5PRIME Taq DNA polymerase (5 PRIME GmbH, Germany), following the manufacture's instructions, in an iCycler thermocycler (Bio-Rad, Spain). The PCR programme consisted of 5 min at 95 °C followed by 35 cycles of amplification (30 s at 95 °C, 30 s at 55 °C, and 90 s at 68 °C) and a final extension step of 10 min at 68 °C.
The amplified fragments were then purified using the ATP ™ Gel/PCR Extraction Kit (ATP TM Biotech Inc., Taiwan) and sequenced at Macrogen (The Netherlands). The resulting sequences were assembled and deposited in the GenBank database (accession number MG149553) and compared with the eubacterial 16S rRNA gene sequences available in that database, and those in the EMBL database, using BLAST software 53 .
The primer pair ODC3 and ODC16 was used to PCR-amplify a 1.49 kb fragment within odc from L. rossiae D87 following the method described by Marcobal,et al. 32 . PCR products were purified and sequenced as described above, and used together with the ODC cluster sequences available in public genome databases to design primers (Table 2) for progressive "genome walking" sequencing analysis combined with reverse PCR. For this, total DNA from L. rossiae D87 was digested with EcoRI or HindIII, diluted, and ligated with T4 DNA ligase (Fermentas) following the manufacturer's instructions. Ligated fragments were used as templates, directly and 10-fold diluted, for inverse PCR with the primers ODC5 and ODC6C ( Table 2). The various PCR fragments generated with standard and reverse PCR were purified and sequenced at Macrogen, and assembled using Vector NTI Advance software, v.9.1 (Invitrogen, USA). Sequence analysis was performed using the University of Wisconsin Genetics Computer Group software package. BLAST and BLASTP programs were used to determine the similarities of the deduced amino acid sequences to those present in databases. Multiple alignment was performed using ClustalW2 at EBI (http://www.ebi.ac.uk/Tools/msa/clustalw2/) and phylogenetic trees obtained using TreeView v.1.6.6. The transmembrane elements of the predicted proteins were examined using the TMHMM tool (http:// www.cbs.dtu.dk/services/TMHMM/; Danish Technical University, Denmark). The nucleotide sequence of the ODC cluster reported in this paper is available at GenBank (http://www.ncbi.nlm.nih.gov) under accession number KT020759.
Reverse transcription PCR (RT-PCR) analysis. cDNA was synthesized from 2 μg of total RNA extracted from L. rossiae D87 cells grown in MRS, and MRS supplemented with ornithine (10 mM), using the iScript ™ cDNA Synthesis Kit (Bio-Rad, UK) following the manufacturer's recommendations. PCR was performed as described above but using 2 μl of the cDNA preparation as a template and 0.4 μM of each primer, i.e., QODCD7F and PotQLRR (Table 2). Amplifications were performed over 35 cycles (94 °C for 30 s, 50 °C for 25 s, and 68 °C for 1 min), and the products analyzed in 0.8% agarose gels in TAE buffer. The absence of contaminating DNA was checked by PCR performed under the conditions described above, using total RNA as a template.
RT-qPCR analysis. RT-qPCR was performed using the SYBR Green PCR Master Mix Kit (Applied Biosystems, UK) employing the primer pair QODCD7F and QODCD7R for the amplification of odc, and primer pair ROSQ16F and ROSQ16R for the amplification of 16S rRNA ( Table 2). All reactions were performed in 20 μl volumes, including 1 μl of cDNA (obtained as described above) as a template, 900 nM of each primer, and 10 μl of SYBR Green PCR Master Mix (which contains ROX as a passive reference dye) (Applied Biosystems). Amplification and detection were performed using an ABI Prism Fast 7500 sequence detection system (Applied Biosystems) under standard running conditions. The cycle thresholds (Ct) used for comparison were those automatically assigned by the thermocycler software (7500 Software v2.0.4, Applied Biosystems). Relative gene expression was calculated using the ΔΔCt comparative method, employing 16S rRNA as a housekeeping gene 54 . Analysis of BAs production by UHPLC. The production of BAs, including putrescine, in L. rossiae D87 supernatant cultures -prepared as previously described 47 -under each set of experimental conditions (MRS or CDM supplemented with different substrate, histidine, tyrosine, lysine, agmatine, arginine or ornithine concentrations) was examined by UHPLC. Cultures were centrifuged at 2250 g for 10 min, and 1 ml of the resulting supernatant filtered through a 0.2 μm Supor membrane (Pall, UK) and analysed for the presence of putrescine. Briefly, supernatants (100 μl) were derivatized using diethyl ethoxymethylenemalonate (Sigma-Aldrich) and injected onto a Waters Acquity UPLC TM BEH C 18 1.7 μm column in an H-Class Acquity UPLC system (Waters, USA) coupled to a photodiode array detector at 280 nm and controlled by Empower 2 software (Waters). The gradient and detection conditions were those previously described 55 . Resistance to acidic pH. To test the resistance of L. rossiae D87 to acid stress, mid-exponential cultures of L. rossiae D87 grown in MRS were harvested by centrifugation and resuspended at a final concentration of 10 8 cfu ml −1 in phosphate-buffered saline (PBS) adjusted to pH 2, pH 3, pH 4 or pH 5, and in the same buffers but supplemented with 20 mM ornithine, for 2 h. After this time, cell survival was evaluated by counting.
Measurement of intracellular pH. Cytosolic pH measurements were performed using carboxyfluorescein succinimidyl ester (cFSE, Sigma-Aldrich) (an internally conjugated fluorescence pH probe) following a previously described protocol 29 with minor modifications. Briefly, instead of tyramine, 10 mM ornithine was added before the quantification of fluorescence. The fluorescence intensities of 200 μL cells were measured for 15 min (intervals of 0.5 s) in a Cary Eclipse fluorescence spectrophotometer (Varian Inc., USA) (excitation wavelengths 440 nm and 460 nm, emission wavelength 525 nm) 56 . Background fluorescence levels were assessed by measuring non-fluorescent control cells; these values were subtracted from the fluorescence results. The cytosolic pH values were determined from the ratio of the fluorescence signal at 440/490 nm taken from a calibration curve constructed using buffers at pH 4.5-8.0, after equilibrating the internal (pH in ) and external (pH out ) pH with 0.1% triton 57 . The value given for each condition is the average of measurements for three independent replicates. Statistical analysis. Means ± standard deviations were calculated for at least three independent replicates and compared using the Student t test. Significance was set at p < 0.05.