Genetic and phenotypic strain heterogeneity within a natural population of Oenococcus oeni from Amarone wine

Authors


Correspondence

Sandra Torriani, Dipartimento di Biotecnologie, Università degli Studi di Verona, strada Le Grazie 15, 37134 Verona, Italy.E-mail: sandra.torriani@univr.it

Abstract

Aims

To investigate the Oenococcus oeni population occurring during spontaneous malolactic fermentation (MLF) of Amarone wine, a peculiar and hostile environment for malolactic bacteria.

Methods and Results

Pulsed-field gel electrophoresis (PFGE) analysis showed a high level of genetic heterogeneity within the O. oeni population involved in MLF throughout an industrial vinification of Amarone wine. The 13 strains with distinct PFGE profile displayed different capability to hydrolyse esters and glycosides, as well as great variability to growth under stress parameters, such as high ethanol content (15% v/v), low pH (3·0) and temperature (15°C), and presence of SO2. Moreover, polymorphism in the gene sacB involved in exopolysaccharide production was observed among the strains. The strains showed differences to convert l-malic acid into l-lactic acid in wine.

Conclusions

The occurrence of spontaneous MLF in stressful ecosystems such as Amarone wine is related to the heterogeneity of O. oeni community; biodiversity indexes and strain evolution analyses suggested that its success depends on its initial strain evenness.

Significance and Impact of the Study

Remarkable intraspecies complexity within the O. oeni natural population could explain the great versatility of this species as key of successful adaptation to harsh winemaking conditions.

Introduction

Malolactic fermentation (MLF) is a natural process conducted by lactic acid bacteria (LAB) in wine usually after alcoholic fermentation (Ribéreau-Gayon et al. 2006). Since the importance of this fermentation for winemaking, a large number of investigations on the role of LAB, particularly Oenococcus oeni, have been carried out (Van Vuuren and Dicks 1993; Bartowsky 2005; Bartowsky and Borneman 2011). The oenological benefits of MLF, such as improvement of taste, flavour and microbial stability of wine (Bauer and Dicks 2004; Bartowsky 2005), continue to prompt research for the development of novel O. oeni starter cultures that are tailored for specific wines (Ruiz et al. 2010; Torriani et al. 2011). Indeed, genetic and phenotypic surveys have revealed considerable strain diversity within natural O. oeni populations associated with different types of wine (Cappello et al. 2008; Vigentini et al. 2009; Solieri et al. 2010), and a correlation between such diversity and the peculiarity of certain oenological niches has also been supposed (Larisika et al. 2008; Cañas et al. 2009; Vigentini et al. 2009). Recently, Pramateftaki et al. (2012) reported that native O. oeni isolates can even be winery-specific. Such biodiversity within a certain winery and winemaking area is a valuable source for selecting malolactic starters, and the identification of strain-specific traits is especially important to match individual strains to specific industrial process (Borneman et al. 2010; Torriani et al. 2011). Recent explorations of the genome of different O. oeni strains partly elucidated the molecular mechanisms involved in the fitness and genomic diversity of this bacterium (Bon et al. 2009; Renouf et al. 2009; Borneman et al. 2010).

Amarone wine is one of the most important dry red wines produced from withered grapes (Paronetto and Dellaglio 2011). Owing to the specificity of the winemaking conditions, this raisin wine represents a peculiar ecosystem still poorly investigated. The few studies carried out on the microbial community of Amarone wine underlined high intraspecies heterogeneity both within yeast and bacterial populations (Dellaglio et al. 2003; Malacrinò et al. 2003). MLF is often hampered mainly because of its high ethanol content (>14·5% v/v). An analysis of 100 Amarone wines commercially available revealed that most of them undergone MLF, although 54 of them contained more than 0·30 g l−1 of l-malic acid (E. Tosi, personal communication). The ability of LAB to survive and grow in such harsh conditions implies the presence of fully adapted and specialized populations. Differentiation of several O. oeni strains isolated from Amarone wine produced in different wineries has been reported (Malacrinò et al. 2003); however, information about the level and the meaning of the heterogeneity within a single O. oeni natural population has not been yet obtained.

To this aim, the complexity of a LAB natural population during an Amarone wine industrial vinification was examined. Pulsed-field gel electrophoresis (PFGE) was used to achieve a reliable intraspecies discrimination, and biodiversity indexes were calculated. The heterogeneity of this natural population was investigated by biochemical and physiological characterization of the strains with different PFGE profile through enzymatic activity and growth assays in laboratory tests, and evaluating l-malic acid metabolism in wine.

Materials and methods

Sampling and bacteria isolation

The eight wine samples analysed in this study were obtained during an industrial Amarone vinification performed in a Valpolicella winery according to the disciplinary regulations of Valpolicella wines (www.consorziovalpolicella.it). A volume of 3000 l of must with grape pomace (traditional red winemaking) was fermented in steel tank, and wine samples were collected during alcoholic fermentation, and MLF until wine was transferred to several barriques, two of which were used for the last sampling.

For the enumeration and isolation of LAB, samples were plated on MRS (Fluka, Seelze, Germany) pH 4·5 supplemented with 10% tomato juice broth (Difco Laboratories, West Molesey, UK) (MRS-tj) and 0·01% actidione (Fluka). After incubation at 28°C for 7 days in jars under anaerobic conditions, using the Anaerocult A kit (Merck, Darmstadt, Germany), a total of 37 colonies were randomly picked up from plates at the highest dilution of wine sample to isolate the dominant cells. Isolates were grown at 28°C for 2–5 days in MRS-tj broth and maintained at −80°C as culture stocks in the same medium supplemented with glycerol (25% w/v).

Chemical analysis of wine

Ethanol content was analysed by NIR spectroscopy using Alcolyzer Wine apparatus (Anton Paar GmbH, Graz, Austria), residual sugars by the Fehling method with automatic titration (Crison, Allela, Spain) and titratable acidity (as tartaric acid) by titration with 1 mol l−1 NaOH. Organic acids were determined using enzymatic kits from La Roche (Basel, Switzerland).

Identification of Oenococcus oeni and PFGE analysis

Cells were observed under optical microscopy, and the identification of O. oeni was carried out by using species-specific PCR method as reported previously (Zapparoli et al. 1998). Genomic DNA preparation, PFGE analysis, ApaI restriction and cluster analysis were performed following the procedures described by Zapparoli et al. (2000). In particular, the clustering analysis of ApaI PFGE patterns was carried out by manually selecting the bands that were clearly visible in the gel excluding the poorly defined ones. This procedure allowed a more precise analysis of PFGE patterns with respect to the use of an automated approach, because the fragments with a pairwise error of >5% were excluded from the analysis. A similarity threshold of 100% was used to identify identical PFGE patterns, that is, patterns with identical numbers and molecular weights of the fragments.

Biodiversity indexes

Simpson diversity index (D) (Simpson 1949) and Shannon–Wiener index (H) (Whittaker 1972) were calculated to have further information about strain composition of O. oeni population in addition to the level of strain richness; these indexes take the relative abundance of different strains in account. The former index was calculated as D = (∑(ni(ni − 1)/(N(− 1)), the latter as H = −∑ni ln N, where ni is number of isolates of the ith strain and N is the total number of isolates.

Enzymatic assays

Esterase and β-glucosidase activities were determined on whole cells of the strains grown in FT80 pH 5·3 (5·0 g l−1 fructose, 5·0 g l−1 glucose, 5·0 g l−1 l-malic acid, 5·0 g l−1 yeast extract, 5·0 g l−1 peptone, 5·0 g l−1 meat extract, 0·5 g l−1 NaH2PO4, 0·05 g l−1 MgCl2, 0·01 g l−1 MnSO4, 0·15 g l−1 KCl, 0·13 g l−1 CaCl2, 1 ml Tween 80), according to Cappello et al. (2010). A total of six different substrates (4-nitrophenyl acetate, 4-nitrophenyl butyrate, 4-nitrophenyl octanoate, 4-nitrophenyl decanoate, 4-nitrophenyl dodecanoate and 4-nitrophenyl octadecanoate) and three substrates (4-nitrophenyl β-d-glucopyranoside, 4-nitrophenyl β-d-mannopyranoside and 4-nitrophenyl N-acetyl-β-d-glucosaminide; Sigma, St. Louis, MO, USA) were used to measure esterase and β-glucosidase activities, respectively.

Esterase and β-glucosidase activities were expressed as μmol or nmol of 4-nitrophenol liberated in 1 min per g dry weight. Quantification of these activities was carried out by standard curves obtained using 4-nitrophenol (Sigma) at increasing concentrations. Enzyme assays were conducted in triplicate.

Growth assays

The effects of various conditions on O. oeni growth were examined in 6-ml test tubes containing 5 ml FT80 medium, which were inoculated as described below and incubated at 28°C. Cells growing in FT80 pH 5·3 were considered as reference tubes. The ability of the isolates to grow at 22 and 15°C was also evaluated. To test the effect of ethanol, the same medium was supplemented with increasing concentrations of ethanol (9, 12 and 15%, v/v). The effect of pH was evaluated in FT80 modified with the addition of casein acid hydrolysate (Amicase; Sigma) instead of meat and yeast extracts in order to avoid the formation of a precipitate at low pH, and the original pH was adjusted to 3·5 and 3·0 with HCl. To assay the SO2 sensitivity, cells were inoculated in FT80 pH 3·5 containing 15 and 30 mg l−1 SO2, adding aliquots of Na2S2O5 (Fluka) solution (1 g l−1) sterilized by filtration (0·22 μm). Control tubes were used to measure the fraction of free SO2 in the medium during the growth test by iodine titration (Ripper method). The cell inoculation was standardized at an equivalent of 0·05 optical density read at 600 nm (OD600) from a stationary-phase FT80 pH 5·3 broth culture. The increase in cell number was monitored by measuring the optical density of the cultures at 600 nm (OD600), and a change of 1 unit of OD600 (1 UOD) was equivalent to 0·4 g of dry matter l−1 (Cavin et al. 1989). The strain growth rate was expressed as average biomass produced per day (mg day−1). All tests were carried out in triplicate.

Microvinification trials

l-Malic acid metabolism was examined by microvinification trials using Valpolicella wine, produced from fresh grapes of the same varieties (Corvina and Rondinella) used for Amarone wine production, with ethanol 11·5% (v/v), pH 3·37, titratable acidity 7·7 g l−1 as tartaric acid and l-malic acid 2·4 g l−1. Cells grown in FT80 pH 3·5 were inoculated at about 106 CFU ml−1 in 200 ml of wine, sterilized by filtration (0·22 μm) and incubated at 22°C. Microvinification trials were carried out in triplicate.

Amplification of genes related to exopolysaccharide production

To detect the gene dsrD coding for dextransucrase, involved in the production of glucose-based homopolysaccharides, the degenerated primers dxt-F1 (5′-AACGAYRTIGAYAAYTCIAAYCC-3′) and dxt-R1 (5′-CKRTCIGTRAAIGCRTAICCRTT-3′) were designed on the basis of the alignment of the dsrD gene (GenBank ID: AY017384.1) from the strain Leuconostoc mesenteroides Lcc4 (Neubauer et al. 2003) and its homologous sequence in O. oeni PSU-1 (Gene ID: 4416854).

A primer pair was designed also for the detection of the sacB gene coding for a putative fructansucrase through a specific and random amplification PCR (Specific and random amplification, SARA-PCR; Knijff et al. 2001). The primers ftf-F1 (5′-GATGTTTGGG ATTCGTGG-3′) and ftf-R1 (5′-GCCGAACCTGACCAT TGTT-3) were designed in the positions corresponding to the conserved amino acid residues A and B, respectively, of the catalytic core (Van Hijum et al. 2006) of the sacB gene of O. oeni ATCC BAA-1163 (EMBL-CDS: EAV39810.1). Both the PCR amplifications were carried out in a 20-μl reaction volume, using 1× GoTaq Flexi Buffer (Promega, Madison, WI), 1·5 mmol l−1 MgCl2, 100 μmol l−1 each deoxynucleotide triphosphates, 1 μmol l−1 each primer, 0·5 U of GoTaq Polymerase (Promega) and approximately 10 ng of the genomic DNA. The thermal programme comprised initial denaturation at 94°C for 5 min; 30 cycles at 94°C for 30 s, 54°C for 40 s, 72°C for 90 s, in the case of the primer pairs dxt-F1/dxt-R1, and 32 cycles at 94°C for 30 s, 56°C for 30 s, 72°C for 40 s for the primer pairs ftf-F1/ftf-R1. The samples were held at 72°C for 5 min to complete the extension of the products. Oenococcus oeni PSU-1 and O. oeni ATCC BAA-1163 were used as reference strains and positive controls in both the PCR assays.

For sequencing, PCR products were purified using the NucleoSpin Gel and PCR Clean-up (Macherey-Nagel, Düren, Germany) and processed by BMR-Genomics (Padova, Italy). Sequence alignments were conducted at NCBI (www.ncbi.nlm.nih.gov).

Statistical treatment of the data

Data of enzymatic activities and growth parameters were represented by Box plot that indicates the degree of dispersion and skewness among the strains. Data showed in Table 2 were analysed statistically by t-test, while biochemical data of all strains (Supporting Information) were used for clustering analysis by paired group algorithm and Euclidean similarity measure.

Results

Genetic strain characterization and biodiversity index

All the 37 LAB isolated during the Amarone vinification were identified by species-specific PCR as O. oeni (data not shown).

Pulsed-field gel electrophoresis analysis of genomic DNA from these isolates, restricted with ApaI enzyme, produced a total of 13 different patterns consisting of 13–18 DNA fragments in the range of 30–160 kb (Fig. 1). As shown in the dendrogram, the percentage of similarity between these profiles varied from 63 to 90% (Fig. 2). Profile E was the most frequent, as it was detected in 15 isolates, while profiles A and G were found in 7 and 3 isolates, respectively. Two isolates showed profiles F and H, while the remaining profiles (B, C, D, I, J, K, L and M) were represented by only one isolate. Because the isolates that displayed different profiles (owing to the presence/absence of one or more bands) can be considered different strains by convention, a representative of each PFGE profile was taken and designed as strain A-M.

Figure 1.

Different pulsed-field gel electrophoresis profiles (A-M) obtained from the ApaI restriction of genomic DNA extracted from the 37 Oenococcus oeni isolates during an industrial Amarone vinification; mk, molecular marker.

Figure 2.

Dendrogram derived from the 13 different pulsed-field gel electrophoresis profiles and numbers of isolates that displayed identical profile (100% of similarity).

Table 1 describes the dynamic of O. oeni population during the Amarone vinification. Strains with profiles E, F, K and M were isolates from wine samples collected during the exponential growth phase of the LAB, when l-malic acid consumption rate was highest (wine samples 6 and 7 in Table 1). However, profile E was also recognized in bacterial isolates at the beginning of vinification (wine sample 2).

Table 1. Sampling, chemical and microbiological data of Amarone wine vinification
Wine sampleTime* (d)Ethanol (% v/v)l-malic acid (g l−1)Bacteria (CFU ml−1)Pulsed-field gel electrophoresis profiles
  1. * From grape crushing.

  2. † Wine pressed.

  3. ‡ Wine transferred to barrel.

1102·721·75<102A, B
22311·051·752 × 102A, C, D, E, G
33014·071·723 × 102A, I
43915·581·704 × 102A, E, G, L
5551·655 × 102E, H, J
6781·378 × 103E, F, M
7900·129 × 105E, K
81600·057 × 105E, F

The remarkable intraspecies diversity found within the O. oeni population was confirmed by Simpson and Shannon–Weiner indexes, 0·803 and 1·981, respectively. The former (1–D) ranges from 0 to 1 (1–1/number of strains) approaching 1 as the number of strains increases, the latter (H) from 0 to log(1/number of strains) (generally between 1·5 and 3·5), while Shannon–Wiener equitability (E) ranges between 0 and 1 with 1 begin complete evenness.

Biochemical characterization of strains

Thirteen strains with different PFGE profiles were selected and assayed for esterase and β-glucosidase activities (Supporting information). Oenococcus oeni strains were found to have higher esterase activity towards short-chain esters such as acetate (C2) and butyrate (C4) than long-chain esters such as octanoate (C8), decanoate (C10) and dodecanoate (C12). No activity was observed towards octadecanoate (C18) (data not shown). Strains were also capable of hydrolysing glucosides, as shown by β-glucosidase activity measured towards three glucosides. Great variability in both activities and substrate specificity among the strains was observed as highlighted by Box plots of esterase and glucosidase data (Fig. 3). Table 2 shows the different enzymatic activities of three representative strains (A and E as the most frequent during AF and MLF, respectively, L as one of eight strains with the lowest frequency during vinification).

Figure 3.

Box plots of esterase (a) and β-glucosidase (b) activities (μmol min−1 g dry weight−1) measured in whole cells of 13 Oenococcus oeni strains using different substrates. The line across the box is the median; bar indicates maximum and minimum values.

Table 2. Enzymatic and technological characterization of three Oenococcus oeni strains isolated during an industrial Amarone wine vinification. Data are mean ± standard deviation
 Strain
AEL
  1. A different letter means that the values differed significantly (P < 0·05).

  2. * Growth rate (mg biomass day−1) determined in FT80 medium.

  3. † l-Malic acid consumption rate (g l−1 day−1) measured in wine.

Esterase activity (μmol min−1 g dry weight−1)
4-N-Acetate10·3 ± 1·1a38·3 ± 3·4b9·8 ± 0·5a
4-N-Butyrate17·2 ± 0·4a38·5 ± 1·9b14·2 ± 0·7c
4-N-Octanoate7·9 ± 0·2a8·9 ± 0·3b6·2 ± 0·2c
4-N-Decanoate2·3 ± 0·0a3·6 ± 0·2b1·9 ± 0·3c
4-N-Dodecanoate0·15 ± 0·00a0·07 ± 0·00b0·07 ± 0·00b
β-Glucosidase activity (nmol min−1 g dry weight−1)
4-N-β-d-Glucopyranoside681·4 ± 72·3a1054·9 ± 20·4b960·0 ± 58·3b
4-N-N-ac.-β-d-Glucosaminide4·0 ± 0·3a1·5 ± 0·1b76·1 ± 1·9c
4-N-β-d-Mannopyranoside24·8 ± 0·5a4·2 ± 0·9b42·3 ± 1·4c
Temperature (°C)
28*185·5 ± 25·5a85·8 ± 6·4b67·7 ± 11·9b
22*111·3 ± 2·0a9·3 ± 2·0b63·9 ± 7·3c
15*30·1 ± 0·6a3·9 ± 1·3b22·4 ± 5·0a
Ethanol (% v/v)
9*41·2 ± 6·3a6·5 ± 0·3b8·1 ± 0·5c
12*0·7 ± 0·0a0·0 ± 0·0b3·7 ± 0·2c
15*0·0 ± 0·0a0·0 ± 0·0a1·5 ± 0·0b
pH
3·5*88·0 ± 0·0a36·0 ± 12·7b54·0 ± 0·0b
3·0*8·2 ± 0·0a9·4 ± 2·5a14·4 ± 0·1b
SO2 (mg l−1)
15*19·0 ± 0·0a0·0 ± 0·0b28·5 ± 2·1c
l-Malic acid cons. rate0·29 ± 0·00a0·16 ± 0·00b0·60 ± 0·01c

Physiological characterization of strains

Growth assays showed different behaviours among strains under stressful parameters (Supporting information). With respect to the reference condition (FT80 pH 5·3, 28°C), growth rates significantly decreased in relationship to the decrease in incubation temperature and pH, and increase in ethanol and SO2 content (Fig. 4). All strains displayed negligible growth rate at 30 mg l−1 of SO2 (data not shown). In this latter test, the amount of molecular SO2, calculated by measuring free SO2, remained at about 0·5 mg l−1 (a lethal level for LAB) for about 3 days from the inoculation time. Different l-malic acid metabolism kinetics was also observed among strains (Supporting information). No correlation was found between stress resistance and dominance in MLF monitored in Amarone wine vinification. In fact, growth rates of the strain E that dominated MLF in Amarone wine decreased strongly at 22 and 15°C and failed to grow at 9–15% v/v ethanol; in addition, l-malic acid consumption rate measured in wine was lower when compared to other strains that did not contribute to MLF in Amarone wine, as strains A and L (Table 2).

Figure 4.

Box plots of daily growth rate (mg biomass day−1) of 13 Oenococcus oeni strains measured in FT80 medium at different conditions. The line across the box is the median; bar indicates maximum and minimum values.

No correlation was also found between genetic and biochemical data of all strains as shown by the dendrogram obtained by analysing enzymatic activities, growth assays and l-malic acid consumption (Fig. 5).

Figure 5.

Dendrogram derived from the data of enzymatic activities, growth and l-malic acid consumption rate of 13 Oenococcus oeni strains.

Amplification of genes related to exopolysaccharide production

PCR amplification of the gene dsrD coding for dextransucrase generated the expected fragment (1501 bp) with the reference strains PSU-1 and ATCC BAA-1163 and with all strains isolated from the Amarone wine (data not shown). The identity of the 1501-bp amplification product was confirmed by sequencing. Conversely, the SARA-PCR assay with the new primers ftf-F1 and ftf-R1 designed on the gene sacB coding for a putative fructansucrase generated different profiles. Indeed, the expected band of 350 bp was obtained only with the reference strain ATCC BAA-1163 and with the strain I. A PCR product of approximately 600 bp long was observed with the strain PSU-1 and other seven strains (A, B, D, F, J, K and L), while the presence of both bands was observed in the profile of the remaining strains (C, E, G, H and M) (Fig. 6).

Figure 6.

SARA-PCR profiles from DNA of Oenococcus oeni strains PSU-1 (PS), ATCC BAA-1163 (AT) and A-M, obtained with primers ftf-F1 and ftf-R1 designed on the gene sacB; mk, molecular marker.

The identity of the SARA-PCR products was assessed by sequencing the fragments obtained from representative strains, that is, ATCC BAA 1163 and I (which generated the 350-bp fragment), PSU-1 and J (600-bp fragment) and H and M (both fragments). The sequence of the 350-bp PCR product corresponds to ORF OENOO_45011 (sacB) from strain ATCC BAA-1163. The 600-bp PCR product corresponds to the predicted ORF OEOE_0229 in the strain PSU-1 (pseudogene), which shares the 95% of the sequence with the predicted ORF OENOO_52035 in the strain ATCC BAA 1163 (l-arabinose isomerase). This sequence divergence could explain the absence of the 600-bp PCR product for the strain ATCC BAA-1163.

Discussion

All 37 LAB isolates, including those coming from wine samples collected at the beginning of fermentation, were exclusively assigned to the species O. oeni. This result is supported by the peculiarity of Amarone wine vinification where low temperatures, pH generally below 3·5 and, especially, high alcohol content are selective parameters for LAB species (Ribéreau-Gayon et al. 2006). The strain heterogeneity of O. oeni population revealed by PFGE typing in this study confirmed that this species is characterized by high intraspecies diversity, as previously demonstrated (Sato et al. 2001; Cappello et al. 2008, 2010; Vigentini et al. 2009; Solieri et al. 2010). The clustering analysis of ApaI PFGE patterns, carried out using the semiautomated procedure of manually selecting only visible and clean bands, could also underestimate the strain discrimination level because some of the DNA fragments were overlapped (Fig. 1) and were not considered for comparison.

Bridier et al. (2010) reported high Simpson indexes (>0·9) in O. oeni populations of different origin analysed by multilocus sequence typing scheme, while Lopez et al. (2008) obtained an index of 0·885 analysing O. oeni strains isolated from Spanish wines by PFGE. However, it is not possible to compare directly these indexes as they were generated by different strain typing tools and experimental protocols (sampling, number of isolates, type of number of vinifications, etc.). Such limitations reduced greatly the cases, like the study of Pramateftaki et al. (2006) and Ruiz et al. (2008), to compare. Values of Simpson and Shannon–Weiner indexes measured in this study were similar to those calculated from the data of Ruiz et al. (2008), which discriminated the isolates using PFGE with ApaI, and were higher than those calculated from PFGE with NotI reported by Pramateftaki et al. (2006). In the present study, the high number of wine samplings carried out during the entire vinification permitted a better estimation of strain evolution than these two previous investigations. Moreover, the value of Simpson diversity index calculated in Amarone wine population was included in the range reported recently by Gonzalez-Arenzana et al. (2011) that analysed the PFGE patterns of several O. oeni isolates from Spanish wines.

The biodiversity reduction during MLF is likely a consequence of selective pressure favouring strains better adapted to the wine conditions, as previously suggested (Reguant and Bordons 2003; Reguant et al. 2005). Nevertheless, based on this assumption, it would be difficult to explain the conservation of strain heterogeneity of O. oeni populations. Indeed, although the strain E dominated the MLF, the presence of several other strains during vinification shows a significant degree of complexity in the natural population structure, as indicated by biodiversity indexes. According to Wittebolle et al. (2009), who investigated the biodiversity–stability relationship of denitrifying bacterial community, the degree of evenness is a key feature in the case of selective stress. This could also be true for wine ecosystems, especially for those characterized by hostile conditions for bacteria, such as Amarone wine. In fact, the strain evenness was higher before than during MLF because 10 of 13 strains were not detected in the last three wine samples (Table 1).

The biochemical and physiological strain characterization carried out in this study allows to ascertain the presence of high heterogeneity within a single O. oeni natural population also at a phenotypic level. This result confirms, at the same time, the lack of correlation between phenotypic traits and genotypic clustering, as previously reported (Guerrini et al. 2003; Vigentini et al. 2009; Cappello et al. 2010). The variability of enzymatic activity and growth performances was particularly high. Wide range of glucosidase activity among different O. oeni strains was already noticed (Grimaldi et al. 2005; Gagnè et al. 2010), while great variability on esterase activity among O. oeni strains such as described in this study was never reported previously (Matthews et al. 2007; Cappello et al. 2010). According to Matthews et al. (2007), O. oeni strains were found to have higher activity towards short-chain than long-chain esters. High diversity among Amarone wine strains on the ability to hydrolyse esters was particularly evident by comparing the data on O. oeni isolated from Malvasia Nera wine (Cappello et al. 2010). These strain-dependent activities, especially esterase, could have an important role in the modification of Amarone wine aroma depending on the diversity of the bacterial population present during the vinification.

An interesting point of discussion is offered by the results obtained from genetic assays of genes coding for dextransucrases and fructansucrases involved in the production of homopolysaccharides, a group of exopolysaccharides constituted by only one monomer, that is, glucose in dextrans and fructose in fructans (levans) (Van Hijum et al. 2006). Indeed, the differences shown by a comparative genomics analysis between two reference strains PSU-1 and ATCC BAA1163 (both positive for dsrD and only the latter positive for sacB) stimulated the investigation on the diffusion of these genes also among other O. oeni strains. In our case, the variability among the strains was observed by the amplification of the gene sacB coding for a putative fructansucrase. The exopolysaccharide production even at low level can be advantageous for strains, because it could enhance the cell survival in a stressful environment, such as Amarone wine (Ciezack et al. 2010). Therefore, it would be interesting to enlighten the biological and technological meaning of the presence and expression of such genes in the strains isolated from wine, where sucrose, the only source of both dextransucrases and fructansucrases, is not usually present. Moreover, this study evidenced the suitability of SARA-PCR to reveal the genomic variability within O. oeni populations. A similar approach was adopted also by Holt and Cote (1998), who designed primers targeting dextransucrase gene to be used in a randomly amplified polymorphic DNA (RAPD) PCR protocol for the differentiation on dextran-producing Leuconostoc spp. strains.

This study underlines that in some cases there is no correlation between prevalence during Amarone wine vinification and performance assayed by laboratory tests. This fact appears evident analysing the phenotypic divergence between the strains A and E, those are the strains more frequently isolated from wine samples collected in the early stages (A) and in the last stages (E) of the MLF. The former displayed better growth rate in ethanol and l-malic acid consumption rate (Table 2), when theoretically it would be expected the contrary. Similarly, strain L that showed to be resistant to ethanol was only isolated in a wine sample taken before MLF (Table 1). Therefore, the results of physiological tests could be not always reliable to screen strains for the selection of potential malolactic starters. The incongruence between MLF performance in own natural and that in ‘artificial’ environments demonstrates the need for further characterization of strains before the identification of malolactic strain specific for Amarone wine. The contribution of viable-but-nonculturable (VBNC) subpopulations of malolactic bacteria to MLF has been previously described (Quirós et al. 2009). The occurrence of VBNC cells in this Amarone vinification cannot be excluded, although the observed population level, monitored by plate counts, and MLF kinetics were in agreement with previous observations carried out on other spontaneous vinifications of Amarone wine (our unpublished data) and also during MLF induced by inoculation with a commercial starter (Zapparoli et al. 2009). Improved knowledge on MLF of this wine could be obtained by using protocols more suitable for the recovery of cells in the VBNC status, especially when the growth conditions for bacteria are more stressful (i.e. during ageing). It is also of foremost importance for the development of reliable rapid detection methods, such as culture-independent methods that use DNA-binding dyes capable of discriminating between dead and viable cells at low concentrations (Andorrà et al. 2010). Cell viability assays should also address the possibility to study microbial populations at strain level to acquire deeper information on the diversity and frequency of strains colonizing a peculiar ecological niche like Amarone wine.

In conclusion, this study provided some evidences on the importance of intraspecies biodiversity of malolactic bacterial populations in wine ecosystems. Although the mechanisms involved in the strain evolution and diversification may be similar to other wines, for Amarone wine they could be amplified or different, because of its peculiar winemaking conditions and physical–chemical properties which makes difficult the survival and growth of cells. Finally, further investigations should be focused on the interactions between different strains that colonize the same wine in order to understand the functional and ecological significance of their presence during vinification.

Ancillary