Selection of cereal-sourced lactic acid bacteria as candidate starters for the baking industry

The quality of sourdough bread mainly depends on metabolic activities of lactic acid bacteria (LAB). The exopolysaccharides (EPS) produced by LAB affect positively the technological and nutritional properties of the bread, while phytases improve the bioavailability of the minerals by reducing its phytate content. In the present study, a pool of 152 cereal-sourced LAB were screened for production of phytases and EPS for potential use as sourdough starter cultures for the baking industry. There was large heterogeneity in the phytase activity observed among the screened isolates, with 95% showing the ability to degrade sodium phytate on plates containing Sourdough Simulation Medium (SSM). The isolates Lactobacillus brevis LD65 and Lactobacillus plantarum PB241 showed the highest enzymatic activity, while the isolates ascribed to Weissella confusa were characterized by low or no phytase activity. Only 18% of the screened LAB produced EPS, which were distinguished as ropy or mucoid phenotypes on SSM supplemented with sucrose. Almost all the EPS producers carried one or more genes (epsD/E and/or epsA) involved in the production of heteropolysaccharides (HePS), whereas the isolates ascribed to Leuconostoc citreum and W. confusa carried genes involved in the production of both HePS and homopolysaccharides (HoPS). Monosaccharide composition analysis of the EPS produced by a selected subset of isolates revealed that all the HePS included glucose, mannose, and galactose, though at different ratios. Furthermore, a few isolates ascribed to L. citreum and W. confusa and carrying the gtf gene produced β-glucans after fermentation in an ad hoc formulated barley flour medium. Based on the overall results collected, a subset of candidate sourdough starter cultures for the baking industry was selected, including Lb. brevis LD66 and L. citreum PB220, which showed high phytase activity and positive EPS production.


Introduction
Cereal-based foods are an essential part of the daily diet worldwide, representing an important source of energy, carbohydrates and fiber, as well as proteins, minerals (e.g., zinc and magnesium) and micronutrients (e.g., vitamins B and E) to a lesser extent [1]. In western countries, the increasing consumer demand for cereal-based foods with improved nutritional and health So far, numerous research papers and literature reviews have described the production of EPS by cereal-sourced LAB. However, only a few of these studies were aimed at analyzing the ability of these microorganisms to produce both phytases and EPS, with a focus on the molecular characterization of the produced EPS and the genes involved in their production. Moreover, there was a paucity of available data on the effect of LAB on the β-glucan content of cereal-based substrates.
Thus, the present study was aimed at exploring a large pool of cereal-sourced LAB for their potential use as sourdough starter cultures for the baking industry, with a specific focus on the following-separate or combined-target applications, namely (i) improvement of digestibility and mineral availability of wholegrain high-phytate bread; (ii) improvement of bread softness and sensory perception; (iii) enhancement of bread health-promoting properties.
To this end, 152 LAB were preliminarily screened for the production of phytases and EPS on the appropriate solid media. The EPS-producing isolates were molecularly screened for the occurrence of key genes involved in the synthesis of EPS. In parallel, the molecular weights and monomer compositions of the EPS purified from the selected pool of isolates were determined. Finally, the effect of fermentation by the selected LAB on the β-glucan content of an ad hoc formulated barley flour medium was evaluated.

Lactic acid bacteria isolation
One hundred and fifty-two LAB isolates were subjected to an in vitro pro-technological and functional characterization. Among these, 121 LAB had previously been isolated from: i) soft wheat sourdoughs used for bread-making at 8 local artisan bakeries (93 isolates, labeled as PB) [24]; ii) a soft wheat sourdough produced with stone grinded wholemeal wheat flour (10 isolates, labeled as LM) [25]; iii) a soft wheat sourdough used for production of recurrence sweet leavened cakes at a local industrial bakery (3 isolates, labeled as LD) [26]; and iv) commercial Boza, an unpasteurized fermented cereal-based beverage (15 isolates, labeled as Bz) [27]. The remaining 31 LAB were isolated in the present study from wholemeal wheat (1 isolate, labeled as FG, emmer (23 isolates, labeled as FF), and barley (7 isolates, labeled as FO) flours purchased from a local mill (Luzi S.r.l., Sassoferrato, Italy). All the 152 isolates were maintained at the Culture Collection of the Department of Agriculture, Food and Environmental Science (Università Politecnica delle Marche, Ancona, Italy) at -80˚C in a mixture of glycerol and MRS broth (VWR International, Milan, Italy), at a 1:1 ratio. They were subcultured on MRS agar incubated at 30˚C for 48 h prior to further analysis.
The DNA was extracted, as previously described by Osimani et al. [27], from the 31 newly isolated flour sourced LAB and the 13 still unidentified isolates (labeled as LM and LD, respectively). The extracted DNA was quantified, checked for the purity using a NanoDrop ND 1000 (Thermo Fisher Scientific, Wilmington, DE, USA) and normalized to a concentration of 25 ng μL -1 . Two μL aliquots of the extracts (containing approximately 50 ng of DNA) were amplified by PCR in a My Cycler Thermal Cycler (BioRad Laboratories, Hercules, CA, USA) using the universal eubacterial primer pair 27F and 1495R [28]. The amplification reactions were carried out according to the conditions previously described by Osimani et al. [27]. The PCR products were checked on 1.5% (w v -1 ) agarose gels and sent to Genewiz (Takeley, UK) for purification and sequencing. The obtained raw sequences in FASTA format were checked for chimeras using Uchime software package [29], whereas terminal NNNs and misleading data from the ends of the sequences were trimmed before further analysis. Hence, the processed sequences were compared with 16S rRNA sequences of type strains deposited in the GenBank DNA database (http://www.ncbi.nlm.nih.gov/) using the Basic Local Alignment Search Tool (BLAST) [30]. The sequences of the newly identified isolates were deposited in the GenBank DNA database under accession numbers MT501108-MT501151. A detailed list of the 152 molecularly identified isolates is reported in S1 Table. Phytase activity All the 152 LAB isolates were screened for their ability to degrade phytic acid on slightly modified Sourdough Simulation Medium (SSM) [31] containing: fructose (10 g L -1 ), maltose (10 g L -1 ), glucose (2 g L -1 ), tryptone (10 g L -1 ), yeast extract (12 g L -1 ), L-cysteine HCl (0.5 g L -1 ), MgSO 4 X 7 H 2 O (0.2 g L -1 ), MnSO 4 X H 2 O (0.05 g L -1 ), and Tween 80 (1 mL L -1 ).
Phytase activity was evaluated by the phytase plate assay using a combination of the methods previously described by Anastasio et al. [32], Fossi et al. [33], Manini et al. [34] and Raghavendra and Halami [35], respectively. The isolates were first grown at 37˚C for 48 h in SSM broth supplemented with sodium phytate (1%, w v -1 ) (Merck Life Science S.R.L., Milan, Italy); 3 μL of each broth culture was spotted (in duplicate) with a manual pipette on 90 mm plates containing SSM supplemented with sodium phytate (1%, w v -1 ), CaCl 2 (0.2%, w v -1 ) and agar (18 g L -1 ). For each isolate, three spots were produced on the same plate. The plates, each including six spots (from two isolates), were incubated at 37˚C for 48 h and further examined for the appearance of clearance zones around the spots. To eliminate false positives, after incubation, the colonies were washed off from the agar surface using distilled water. The plates were then flooded with a 2% (w v -1 ) aqueous cobalt chloride solution for 5 min at room temperature and subsequently counterstained with a 6.25% (w v -1 ) ammonium molybdate solution for 5 min at room temperature. The phytase activity was evaluated by measuring the diameter (in mm) of the clear halos surrounding the spots. The isolates were grouped in three arbitrary classes, based on the mean diameter of their clear halos: class 1 (labeled as +++), with a halo diameter ranging from 21 to 27 mm; (ii) class 2 (labeled as ++), with a halo diameter ranging from 11 and 20 mm; (iii) and class 3 (labeled as +), with a halo diameter ranging from 4 to 10 mm.

EPS production
The pool of 152 LAB were further screened for the ability to produce EPS. The LAB were first grown in SSM broth incubated at 30˚C for 48 h. An aliquot of 5 μL of each broth culture was spotted (in duplicate) on 90 mm plates containing SSM agar supplemented with only sucrose (5%, w v -1 ) as a replacement for fructose, maltose, and glucose. For each isolate, three spots were produced on the same plate. The plates, each including six spots (from two isolates), were incubated at 30˚C for 72 h. After the incubation period, the colonies were classified as "mucoid" or "ropy", depending on their appearance and slime production. Mucoid colonies had a slimy and glistening appearance but did not produce long visible filaments when picked with a sterile toothpick. Ropy colonies were grouped into the following three arbitrary classes, based on the mean length of the filaments picked from the colony: class 1 (labeled as +++), with a filament mean length � 11 mm; class 2 (labeled as ++), with a mean length ranging from 6 to 10 mm; and class 3 (labeled as +) with a mean length ranging from 3 to 5 mm.

EPS extraction
The best EPS-producing LAB (13 isolates, including all the ropy colonies grouped in class 1 and 2 and some randomly selected mucoid colonies) were selected for determination of the monosaccharide composition and molecular weight of the produced EPS. For preliminary EPS extraction, each isolate was grown in SSM broth incubated at 30˚C for 24 h, and the bacterial biomass was harvested by centrifugation at 4000 rpm for 5 min. Once the supernatant was discarded, each cell pellet was resuspended in a sterile physiological solution (0.85% NaCl, w v -1 ), and the concentration of LAB cells was determined spectrophotometrically at 600 nm using a UV-Vis Shimadzu UV-1800 spectrophotometer (Shimadzu Corporation, Kyoto, Japan). Cell suspensions were standardized to a final concentration of approximately 5 x 10 7 cells mL -1 using a sterile physiological solution; an aliquot (500 μL) of each suspension was streaked (in duplicate) on 90 mm plates containing SSM agar supplemented with sucrose (5%, w v -1 ). After incubation at 30˚C for 48 h, the EPS produced by each LAB isolate were washed off from the two plates by sterile deionized water, pooled together and further processed as described by Palomba et al. [36].

EPS monosaccharide composition
Purified EPSs were hydrolysed according to Zhou et al. [37]. Monosaccharides were than derivatized (silylated) according to Lucci et al. [38] and analyzed on a CP-9001 gas chromatograph (Chrompack, Middelburg, The Netherlands) equipped with a column TG-17Sil MS, 30 m length × 0.25 mm i.d., 0.25 μm film thickness (Thermo Fisher Scientific, MA, USA). Helium was used as carrier gas at a linear rate of 35 cm/s. The oven temperature was set at 150˚C for 2 min, then increased at a rate of 5˚C min -1 to the final temperature of 270˚C, which was kept for 10 min. Injector and flame ionization detector (FID) temperatures were set at 270˚C.

EPS molecular weight
The average molecular weight of the purified EPSs was estimated by gel-permeation chromatography. An Agilent 1100 Series quaternary pump (Agilent Technologies, CA, USA) coupled with a Jasco 880 RI detector (Jasco Europe S.r.l., Italy) was used. Separation was carried out using the column PROGEL-TSK GMPWXL, 7.8 x 300 mm (Tosoh Bioscience, Japan), according to the conditions described by Zhou et al. [37]. Commercial dextrans (5, 12, 50, 80, and 410 KDa) from Leuconostoc. mesenteroides (Merck KGaA, Darmstadt, Germany) were used for the calibration of the chromatographic system. A linear regression Log molecular weight vs retention time was used to estimate the dextran-equivalent molecular weights of the purified EPS.

Molecular detection of EPS-related genes
The LAB isolates producing EPS on SSM agar supplemented with sucrose (5%, w v -1 ) were further screened for the presence of the genes gtf (coding for glucansucrase) and lev (coding for levansucrase) and for the genes epsA, epsB, epsD/E plus epsEFG, which are involved in HoPS and HePS production, respectively. The DNA was extracted from these isolates, performed as previously described by Osimani et al. [27], and amplified by PCR using the primer pairs and thermal cycling conditions previously reported by Palomba et al. [36]. Briefly, 3 μL of each DNA extract (containing about 75 ng of bacterial DNA) was amplified in a 50 μL-reaction mixture containing 1 μM of each primer and 25 μL of Green PCR Master Mix Direct-Load 2X (Biotechrabbit, Hennigsdorf, Germany). For each gene under study, amplicons from a few arbitrarily selected PCR-positive isolates were sent to Genewiz for purification and sequencing to confirm the reaction specificity. The obtained sequences in FASTA format were aligned with those deposited in the GenBank database (http://www.ncbi.nlm.nih.gov/) using the Basic Local Alignment Search Tool (BLAST) [30].

Effect of LAB on β-glucan content
The LAB carrying the gtf gene were further assayed in duplicate for their interaction with the β-glucans contained in a sterilized barley flour (BF) medium, prepared by adding 5% barley flour (w v -1 ) to deionized water. Each isolate was inoculated in tubes containing 10 mL of BF medium to reach a final cell concentration of approximately 2 x 10 8 cells mL -1 . The β-glucan content of the liquid medium was determined immediately before inoculation and after 24 h of fermentation at 30˚C using a commercial mixed-linkage beta-glucan kit (Megazyme, Bray, Ireland) according to the assay procedure suggested by the kit manufacturer for liquid samples (including alcohol precipitation). The results were expressed as the mean variance (%) (positive or negative) of the β-glucan content before and after fermentation ± standard deviation.

Statistical analysis
The values of β-glucan content before and after fermentation were subjected to one-way analysis of variance (ANOVA), whereas the significant differences in the EPS monosaccharide composition between different isolates were analyzed using the Tukey-Kramer's Honest Significant Difference test. All statistical analyses were performed using the JMP statistical software, version 11.0.0 (SAS Institute Inc., Cary, NC, USA) with criterion of significance set at P < 0.05.

Molecular identification of LAB
The results of the 16S rRNA sequencing of the 31 LAB isolated in this study from wheat, emmer, and wholegrain barley flours (labeled as FG, FF and FO, respectively) and of those previously isolated from sourdoughs (labeled as LD and LM) are reported in Table 1. The 16S rRNA gene sequencing is widely used for the identification and classification of bacteria. This method relies on the comparison of the 16S rRNA gene sequence of an unknown isolate against the deposited sequences of bacterial type or reference strains. Recently, a 98.65% cutoff has been set for the identification of species [39]. Though 16S rRNA gene sequencing is recognized as one of the most efficient molecular techniques for the establishment of taxonomic relationships between prokaryota, it is not suitable for the discrimination of some closely related species, as those belonging to the Lactobacillus casei, Lactobacillus plantarum, Lactobacillus buchneri, and Lactobacillus sakei groups, which are characterized by a high-level similarity (� 99%) in the 16S rRNA gene sequences. This implies that advanced methods, including whole genome sequencing or multilocus sequence typing (MLST), can be applied to differentiate these microorganisms [40].
The species Lb. curvatus and P. pentosaceus have previously been detected in emmer flour [41,42]. The seven LAB isolated from wholegrain barley flour were identified as P. pentosaceus (5 isolates) and Lb. plantarum (2 isolates). Both these species have previously been isolated from barley sourdoughs [43], where the latter was found to dominate after two-month refreshments [44,45]. The sole LAB isolate obtained from wheat flour was ascribed to Enterococcus casseliflavus. As reported by Corsetti et al. [46], this species is frequently isolated from wheat grains, bran, and non-conventional flours. Finally, the 13 sourdough-sourced isolates were ascribed to Lactobacillus brevis (4 isolates) and Lactobacillus paralimentarius (9 isolates), with both species being recognized as typical sourdough LAB [47]. The remaining 108 isolates

Phytase activity
The results of the phytase activity screening carried out on SSM agar (supplemented with sodium phytate and CaCl 2 ) used to simulate the typical sourdough ecological environment are shown in Table 2. More details are reported in the S1 where the 40 LAB assayed were able to hydrolyze hexacalcium phytate, but only two of these isolates (P. pentosaceus CFR R38 and CFR R35) were able to degrade sodium phytate. However, when CaCl 2 was added to modified MRS agar supplemented with phytate salt (calcium or sodium), all the assayed LAB hydrolyzed sodium phytate, thus indicating a possible involvement of calcium ions in this enzymatic activity.
In regard to the diameter of the clear haloes, mean values between 4 and 27 mm were recorded, thus revealing the high heterogeneity in the phytase activity of the 152 LAB, even within the same species. The isolates classified as class 1 (+++) and hence producing the largest clear halos (with a mean diameter of 21-27 mm) belonged to the following species: Lb. brevis, Lb. buchneri, Lb. curvatus, Lb. fermentum, Lb. plantarum, Lb. sanfranciscensis, and P. pentosaceus. Among these isolates, Lb. plantarum PB241 and Lb. brevis LD65 showed the highest phytase activity, with a mean halo diameter of 27 mm. Of note, most of the isolates belonging to the latter two species showed a high or intermediate phytase activity, with only 2 out of 50 Lb. plantarum isolates being classified as weak phytase producers (class 3, +). Similarly, in a recent study by Sümengen et al. [50], Lb. brevis isolated from Hatay boiled cheese produced high levels of both extracellular and intracellular phytases. To the authors' knowledge, to date, most of the phytase activity assays carried out on LAB have used solid media other than SSM agar, thus making the comparison of our data with those available in the literature cumbersome. In this regard, Anastasio et al. [32] reported clear halos ranging from 3 to 13 mm on modified Chalmers agar supplemented with 1% hexacalcium phytate, with a L. plantarum strain isolated from pizza dough producing a 13 mm-diameter clear halo. The study performed by Manini et al. [34] reported even smaller halo diameters (1-3 mm) on the same medium, except for the Lb. brevis, Lb. plantarum and L. citreum strains producing clear halos with diameters larger than 3 mm. In the present study, the majority of P. pentosaceus isolates showed high (+++) or intermediate (++) phytase activity, with the two isolates FF86 and FO40 collected from emmer and barley flour, respectively, both producing 21 mm clear halos. The latter finding agrees well with what was reported by Cizeikiene et al. [51], where the two strains P. pentosaceus KTU05-8 and KTU05-9 were recommended as suitable starters for the preparation of wholemeal wheat sourdough, due to their high phytase activity. More recently, Rizzello et al. [52] reported that in fava beans fermented by the strain P. pentosaceus VTTT E-153483, the phytic acid content was reduced to more than half the initial concentration. Among the five Lb. sanfranciscensis isolates assayed in the present study (PB213, PB219, PB221, PB223 and PB276), only the first showed no phytase activity on SSM agar, whereas the remaining 4 produced clear halos with diameters between 14 mm and 21 mm. Again, this finding agrees well with those reported in a study performed by De Angelis et al. [49], where Lb. sanfranciscensis strains showed a high phytase activity on sodium phytate. As far as the isolates ascribed to W. confusa are concerned, most of them had low (+) or no phytase activity (-). For this species, a very different picture emerged from the study of Mandhania et al. [53], where all the strains assayed produced phytases. Table 3 reports the results of the LAB screening for EPS production on SSM agar supplemented with 5% sucrose. The visual inspection of colonies grown on a solid medium is one of the most frequently used methods for the screening of EPS-producing LAB, being particularly suitable for the detection of high EPS producers [15]. EPS production mainly depends on the composition of the growth medium in terms of carbon and nitrogen sources, as well as on other parameters, including the pH of the growth medium, incubation temperature and time. Palomba et al. [36] clearly demonstrated that a growth medium supplemented with sucrose is more suitable for screening EPS-producing LAB compared to the same growth medium supplemented with other carbohydrate sources, such as maltose, glucose, galactose, lactose, and fructose. As shown in Table 3, only 28 out of the assayed 152 LAB (18%) were able to produce EPS. This finding is consistent with the results reported in the study carried out by Palomba et al. [36], where 34 out of 177 LAB (19%) isolated from doughs and sourdoughs were able to produce EPS on modified Chalmers agar supplemented with sucrose.

EPS production
In the present study, 13 and 15 isolates out of the 28 EPS-producing LAB produced ropy and mucoid colonies, respectively, with the first group including isolates ascribed to Lb. brevis, Lb. curvatus/graminis, Lb. fermentum, and P. pentosaceus. Among these ropy isolates, the colonies formed by Lb. brevis LD66 (sourced from sourdough) and by P. pentosaceus FO27 (sourced from wholegrain barley flour) produced the longest filaments (mean length � 11 mm) when  [55], who found that most of the LAB isolates from wheat bran sourdough ascribed to P. pentosaceus were EPS producers.
As expected, three out of the five Lb. brevis isolates assayed in the present study produced EPS on SSM agar. In fact, among LAB, Lb. brevis is widely recognized as one of the major EPS producers [9]. As early as 1995, Figueroa et al. [56] reported that this species produces ropy colonies on glucose-or sucrose-containing media.

PB327
Weissella confusa sourdough M --+ -+ -1.08 ± 1.08 The remaining two species producing exclusively ropy colonies were Lb. curvatus/graminis and Lb. fermentum. Regarding the latter species, two (Bz2 and Bz5) out of the four isolates sourced from Boza were able to form colonies producing 5-to 8-mm filaments when picked. To date, EPS production by Lb. fermentum has been widely documented [9,57]. Concerning Lb. curvatus/graminis, only one (FF7) out of the 13 isolates assayed (sourced from wholegrain emmer flour) produced EPS. Minervini et al. [58] have previously reported a highly efficient synthesis of EPS by the strain Lb. curvatus DPPMA10 during its growth on hydrolyzed wheat flour agar.

PB330
Regarding L. pseudomesenteroides, two out of three isolates tested in this study synthesized EPS, with the isolates PB288 and PB170 being characterized by a ropy and mucoid EPS phenotype, respectively. To date, numerous studies have established the EPS-producing behavior of L. pseudomesenteroides, as shown for some strains isolated from Turkish sourdoughs [59] or soybean paste [60]. Moreover, Du et al. [61] purified and characterized dextran, an exopolysaccharide synthesized by the strain L. pseudomesenteroides DRP-5, with a moderate antioxidant activity in in vitro experiments. Among the other properties of dextrans produced by L. pseudomesenteroides, the polymer produced by the strain YB-2 showed a high water solubility index, a water retention capacity and inhibitory effects against pathogenic microorganisms such as Escherichia coli and Staphylococcus aureus [62].
In the present study, all the isolates belonging to Lb. paralimentarius, Lb. plantarum, L. citreum, and W. confusa produced mucoid colonies. According to De Vuyst and De Vin [63], this EPS phenotype is associated with the production of extracellular HoPS. In more detail, only one (PB127) out of the twenty Lb. paralimentarius sourdough isolates analyzed were able to produce EPS. This finding agrees well with those reported by Dertli et al. [59], who found a very limited capacity of producing EPS by some sourdough Lb. paralimentarius isolates.
For L. citreum, the two assayed sourdough isolates showed a mucoid EPS phenotype, analogous to the strain L. citreum L3C1E7 isolated from Pico cheese by Domingos-Lopes et al. [64].
Regarding W. confusa, 10 out of the 18 assayed isolates were EPS producers. In 2015, EPS produced by W. confusa were analyzed by Tinzl-Malang et al. [65], who explored the use of members of this species to improve the quality of bakery products, specifically bread, both in terms of shelf life and texture. In a more recent study [66], the strain W. confusa QS813, isolated from a type I sourdough, synthesized high molecular weight dextrans with a positive effect on hydration, water distribution and microstructure of gluten during the freeze-thaw process.
Surprisingly, only one (PB181) out of the fifty Lb. plantarum isolates screened in this study produced EPS on SSM agar. This finding was quite unexpected, given the numerous available reports on the EPS produced by Lb. plantarum, which are known to have antitumor, antioxidant and antibiofilm activities. The EPS synthesized by Lb. plantarum have also been exploited by the food industry as emulsifiers, gelling agents and stabilizers, as well as texture improvers in fermented foods [67].

Molecular detection of EPS-related genes
Twenty-eight LAB showing a ropy or mucoid phenotype on SSM agar supplemented with sucrose (5%, w v -1 ) were further assayed for the occurrence of key genes involved in the synthesis of both HoPS (e.g., gtf and lev) and HePS (e.g., epsA, epsB, epsD/E and epsEFG). The results of the molecular screening are reported in Table 3. The sequencing of the amplicons from randomly selected PCR-positive isolates confirmed the specificity of all the amplification protocols used.
Twenty-two out of the 28 tested isolates harbored at least one of the EPS-related genes of interest. Some LAB belonging to L. citreum and W. confusa harbored both HoPS-and HePS-related genes, as previously reported by Van der Meulen et al. [68] and Palomba et al. [36]. In the present study, the glycosyltransferase genes epsD/E were found with the highest frequency (18 out of 22 screened isolates). Among the two EPS-producing isolates ascribed to Lb. fermentum, only one (Bz2) carried epsD/E. The occurrence of the genes involved in EPS synthesis in Lb. fermentum has recently been documented by Harris et al. [69], who sequenced the whole genome of the strain Lb. fermentum Lf2, highlighting 3 gene clusters potentially involved in the EPS production process. In regards to P. pentosaceus, three (FF71, FF78 and FO8) out of the six EPS-producing isolates assayed harbored epsD/E, while the remaining three isolates, including FO27, one of the best EPS producers, were negative for all the tested genes. This could be explained by a possible variation in one or both the primer binding sites of the EPS genes, which in turn might have hindered the successful amplification of these target genes. An alternative explanation relies on the involvement of other genes in EPS synthesis, as suggested by a recent investigation [70] on the EPS-biosynthetic gene cluster of a P. pentosaceus strain (LP28) isolated from Euphoria longana, where 12 open reading frames (ORFs) containing a priming enzyme, five glycosyltransferases, and a putative polysaccharide pyruvyl transferase were identified.
All the Lb. brevis EPS-producing isolates on SSM agar harbored epsD/E and epsA, thus confirming the results previously reported by Dertli et al. [59], whereas the strain Lb. brevis E-25 harbored epsD/E, epsA, epsEFG and gtf. The epsD/E gene was also detected in the isolates belonging to Lb. paralimentarius, Lb. plantarum, and L. pseudomesenteroides, in agreement with previous findings on the same three species [59]. Regarding the sole EPS-producing Lb. curvatus/graminis isolate (FF7) assayed, it was found to carry the epsA gene. HePS production by the species Lb. curvatus was first demonstrated by Van der Meulen et al. [68] with a preliminary screening of 174 LAB for EPS production, followed by PCR detection of the EPS-related genes, including epsA. A subsequent study [71] showed the presence of glycosyltransferasesrelated genes in the strain Lb. curvatus MTW 1.624, which was used to improve the quality of gluten-free baked products. In the same years, Palomba et al.
Both the EPS-producing L. citreum isolates assayed in this study harbored epsD/E and gtf. Coda et al. [41] have recently analyzed the performance of a L. citreum strain (FDR241) during the propagation of a type I wheat sourdough, revealing the presence of at least 5 different dextran sucrases after transcriptional analysis of the genes involved in EPS synthesis.
In regards to W. confusa, whose members produced mucoid colonies on SSM agar, one isolate (PB325) was negative in all PCR amplifications, whereas the remaining nine isolates harbored at least one of the three genes (epsA, epsD/E and gtf), thus showing the possibility for this species for producing both HoPS and HePS. The isolates PB323, PB333 and PB333 were positive for all these three genes, whereas PB330, PB331 and PB332 carried only the gtf gene. As early as 2009, Malik et al. [72] analyzed the gtf gene of two W. confusa strains (MBF8-1 and MBF8-2), highlighting a high sequence similarity with the same gene in Lactobacillus reuteri and Lb fermentum. None of the 28 EPS-producing LAB screened in the present study were positive for the genes epsEFG, epsB or lev under the applied PCR conditions, thus not excluding the possibility of false negative results.
Finally, negative PCR results were obtained from the isolates Lb. fermentum Bz2, L. pseudomesenteroides PB288, P. pentosaceus FF95, FO27 and FO30 and W. confusa PB325. This latter evidence might be tentatively ascribed to a low gene stability leading to false negatives, as previously suggested [73].
As far as the BLAST analysis of the EPS gene sequences is concerned, the following considerations can be made. As a general trend, most sequences were closely related to deposited sequences of LAB, as in the case of the gtf gene sequence from the isolate L. citreum PB220 showing a 89% similarity with a deposited sequence of the glucansucrase gene from the strain L. citreum 2.8 (DQ873511) or the epsD/E gene sequence from the same isolate showing a 97% similarity with a deposited sequence of the putative glycosylphosphotransferase gene from the strain Vagococcus lutrae MIS7 (KC841159). In addition, the epsD/E gene sequence from the isolate Lb. plantarum PB181 showed a 96% similarity with the deposited sequences of the glycosyltransferase gene from the strains Lb. curvatus FS9 (KC841158) and P. pentosaceus CRL 777 (DQ873506), whereas the epsD/E gene sequence from the isolate Leuconostoc pseudomesenteroides PB170 showed a 88% similarity with the deposited sequence of the epsE gene from the strain Streptococcus thermophilus BN1 (FN396421.1). Finally, epsA amplicons from the isolate Lb. curvatus/graminis FF7 obtained using primers targeting the epsA gene of S. thermophilus [74] showed a 100% similarity with a deposited sequence of the epsA gene of S. thermophilus (AF053346).

Molecular weight and monosaccharide composition of purified EPS
The monosaccharide composition and molecular weight of EPS greatly influence the functionality of these biopolymers when used as food ingredients [58,75]. The results of the determination of the molecular weights and monosaccharide compositions of the EPS produced by the selected 13 LAB are reported in Table 4.
The molecular weights of the EPS were between 1.1x10 4 Da for those produced by Lb. brevis LD66 and 9.0x10 4 Da for those produced by P. pentosaceus FF78. These values match with the molecular weight of the HePS, which are generally between 10 4 to 10 6 Da [76]. EPS with a molecular weight � 10 4 Da are considered as low molecular weight EPS [14]. Wang et al. [77] have previously reported a positive correlation between the low molecular weight of EPS and their antioxidant activity. By contrast, high molecular weight EPS are known to have a stronger antitumor activity with respect to the low molecular weight polymers [78].
The monosaccharide composition analysis revealed that these EPS were almost stably composed of glucose, mannose, and galactose, except for those produced by P. pentosaceus FO30, which did not include mannose. The overall results collected indicated that all the LAB assayed were HePS producers.

PLOS ONE
The EPS produced by both the Lb. brevis isolates assayed in the present study were characterized by a prevalence of galactose (60.4%-67.7%), followed by glucose (26.6%-31.9%) and mannose (5.7%-7.7%). A different EPS composition was found by Dertli et al. [79] in the EPS produced by the sourdough-sourced strain Lb. brevis E25, with glucose as the sole detected monosaccharide.
Similar to what was found for Lb. brevis, even the isolates ascribed to W. confusa showed a high homogeneity in the monosaccharide composition, with a prevalence of glucose (54.5%-59.9%), followed by galactose (39.2%-43.8%) and mannose (0.9%-1.7%). Malang et al. [80] have previously reported the production of HePS by the strains W. confusa 8CS-2, 11GU-1 and 11GT-2 isolated from sour milk. However, W. confusa is generally recognized as a HoPSproducing species, with the ability to release large amounts of high molecular weight dextrans. The latter biopolymers are exploitable by the baking industry for the improvement of the quality and texture of breads, including gluten-free breads [81]. A larger variability in the EPS monosaccharide composition was seen among the isolates belonging to P. pentosaceus. This finding agrees well with those reported by Imran et al. [82] on the high variability in the monosaccharide composition of the EPS produced by different P. pentosaceus strains. In the present study, the EPS produced by the isolates FF78 and FO30 showed the highest content in galactose (73.1%) and glucose (53.2%), respectively. In addition, three out of the four P. pentosaceus isolates assayed produced EPS containing from 2.4% (isolate FF78) to 6.7% (isolate FF71) mannose. Again, these results agreed with those of a research study [70] carried out on the EPS extracted from fruit-sourced isolates ascribed to P. pentosaceus, which were mainly composed of glucose, galactose, and mannose. In the present study, the isolate Lb. fermentum Bz2 produced EPS with the highest amount of mannose (23.1%) when compared with all the other tested isolates, followed by Lb. paralimentarius PB127 and L. pseudomesenteroides PB170, whose EPS included 17.4 and 16.1% mannose, respectively. Mannose is a minor component of the EPSs produced by LAB, whereas in those produced by yeasts, it represents a major component. To date, different authors have reported the occurrence of a positive correlation between the high mannose content of EPS and their anticancer activity [83].
The EPS produced by the isolate Lb. fermentum Bz2 were also characterized by the highest glucose content (68.7%) when compared to the EPS produced by all the other LAB assayed, and by the lowest amount of galactose (8.2%). Interestingly, a similar monosaccharide composition has recently been reported for a Lb. fermentum strain (YL-11) isolated from fermented milk [84].
In addition to mannose, the EPS produced by the isolate L. pseudomesenteroides PB170 also included glucose (41.3%) and galactose (42.6%). Similarly, the EPS purified from the isolate L. citreum PB3 were also characterized by glucose (41.3%), galactose (44.9%) and mannose (2.1%), but at a different relative ratio. As reported above, both these Leuconostoc isolates carried HePS-related genes, with the isolate ascribed to L. citreum also being positive for the gtf gene involved in HoPS production. To date, very few studies have described the ability of Leuconostocs to produce HePS. According to Ziadi et al. [85], members of the genus Leuconostoc produce HePS that are mainly composed of glucose and mannose and, to a lower extent, of rhamnose and arabinose. This was not expected since Leuconostocs are acknowledged as primary producers of dextran as well as other HoPS, such as alternan and levan [85,86].

Impact of LAB fermentation on the β-glucan content of a barley flour medium
The production of EPS is frequently associated with LAB metabolism, though the synthesis of a specific group of HoPS, referred to as β-glucans, is much less frequently associated with these microorganisms. To date, only a few LAB strains, ascribed to the species Pediococcus parvulus, Oenococcus oeni, and Lactobacillus diolivorans, have been reported to produce β-glucan in wine and cider, being responsible for the undesired ropiness of these alcoholic beverages [87]. However, β-glucans are principally known for their health-promoting properties, such as the ability of significantly reducing serum cholesterol levels and in turn, heart disease risk. Furthermore, β-glucans have immunomodulatory, anti-osteoporotic, anticytotoxic, antitumorigenic and antimutagenic effects, in addition to prebiotic traits that are associated with their ability to improve the growth, metabolism and beneficial activities of probiotics [12]. The βglucans produced by LAB have already been used to produce fermented cereal-based products and yogurt to increase the functional and technological characteristics of these food products [88,89].
In the present study, out of the 28 EPS-producing isolates assayed, 11 LAB belonging to L. citreum (isolates PB3 and PB220) and W. confusa (isolates PB315, PB323, PB324, PB327, PB330, PB331, PB332, PB333 and PB337) carried the gtf gene, which is involved in β-glucan production. Therefore, these isolates were further subjected to the evaluation of their effect on the β-glucan content of an ad hoc formulated BF medium. Our results revealed that 6 out of the 11 LAB individually inoculated into the BF medium were able to significantly increase the concentration of β-glucan after 24 h of fermentation. Among these, W. confusa PB332 and L. citreum PB220 performed the best, leading to a β-glucan increase as high as 7.32% and 6.86%, respectively, when compared with the non-inoculated controls (Table 3).

Conclusions
Overall, the screening of the 152 LAB isolated from cereal-based substrates revealed a widespread capacity of these isolates (95%) for degrading phytic acid, though a very high variability was seen among the isolates, even within the same species. Lb. brevis LD65 and Lb. plantarum PB241 showed the highest phytase activity, while the isolates assigned to W. confusa could be distinguished for their low or no phytase activity. EPS production was limited to 18% of the tested LAB, with most of them carrying one or more genes involved in the production of HePS (epsD/E and/or epsA) and L. citreum and W. confusa isolates carrying also the genes involved in the production of HoPS. The monosaccharide composition analysis revealed that all the tested isolates were HePS producers, with glucose, mannose, and galactose being the most abundant monomers of the latter EPS. Interestingly, a few W. confusa and L. citreum isolates carrying the gtf gene were able to produce valuable amounts of β-glucans. Based on the overall results collected, a selected pool of isolates (including Lb. brevis LD66 and L. citreum PB220) showing both a high phytase activity and an appreciable EPS production, was identified for the formulation of sourdough starters with a potential in the manufacturing of breads with improved digestibility and mineral availability and/or improved softness and sensory traits. In addition, a few cultures with a potential for the manufacturing of fiber-rich functional bread, with an increased β-glucan content, were also found.
Further research efforts are needed to test the performance of these selected LAB in doughs prepared with both refined and phytate-rich flours from different cereal grains (e.g., wheat, emmer, and barley) for an in vivo assessment of their technological and functional potentials. Similarly, the finding of the capacity of several LAB isolates for enhancing β-glucan content in a barley flour medium is worth further investigation under in vivo conditions. Supporting information S1 Table. Complete list of the lactic acid bacteria (LAB) isolates and results of their screening for the phytase activity and the exopolysaccharides (EPS) production. a The isolates were identified by comparing their 16S rRNA gene sequences with those deposited in the Gen-Bank DNA database (http://www.ncbi.nlm.nih.gov/) using the Basic Local Alignment Search Tool (BLAST); � halo diameters (mm); �� R, ropy colonies (in brackets: the lengths, in mm, of the filaments picked from the colonies); M, mucoid colonies; -, no EPS production on SSM agar plates added with sucrose (5%, w v -1 ). (DOCX)