Metagenetic Analysis for Microbial Characterization of Focaccia Doughs Obtained by Using Two Different Starters: Traditional Baker’s Yeast and a Selected Leuconostoc citreum Strain

Lactic acid bacteria (LAB) decisively influence the technological, nutritional, organoleptic and preservation properties of bakery products. Therefore, their use has long been considered an excellent strategy to improve the characteristics of those goods. The aim of this study was the evaluation of microbial diversity in different doughs used for the production of a typical Apulian flatbread, named focaccia. Leavening of the analyzed doughs was obtained with baker’s yeast or by applying an innovative “yeast-free” protocol based on a liquid sourdough obtained by using Leuconostoc citreum strain C2.27 as a starter. The microbial populations of the doughs were studied by both a culture-dependent approach and metagenetic analyses. The flours used for dough preparation were also subjected to the same analyses. The metagenetic analyses were performed by sequencing the V5–V6 hypervariable regions of the 16S rRNA gene and the V9 hypervariable region of the 18S rRNA gene. The results indicate that these hypervariable regions were suitable for studying the microbiota of doughs, highlighting a significant difference between the microbial community of focaccia dough with baker’s yeast and that of the dough inoculated with the bacterial starter. In particular, the dough made with baker’s yeast contained a microbiota with a high abundance of Proteobacteria (82% of the bacterial population), known to be negatively correlated with the biochemical properties of the doughs, while the Proteobacteria in dough produced with the L. citreum starter were about 43.5% lower than those in flour and dough prepared using baker’s yeast. Moreover, the results show that the L. citreum C2.27 starter was able to dominate the microbial environment and also reveal the absence of the genus Saccharomyces in the dough used for the production of the “yeast-free” focaccia. This result is particularly important because it highlights the suitability of the starter strain for obtaining an innovative “yeast-free” product.


Introduction
Traditionally, dough leavening of bakery products has been obtained using sourdough, but with the development of industrial baking at the beginning of the 20th century, baker's yeast (Saccharomyces cerevisiae) almost completely replaced sourdough as a more rapid leavening agent, which also occurred in several traditional productions [1]. In fact, baker's yeast is a ready-to-use starter able to produce high amounts of CO 2 and flavor compounds [2,3]. The typical Apulian focaccia flatbread has recently been included in the Italian list of unprotected typical/local products (PAT, Prodotti Agroalimentari Tradizionali) [4]. Its production is characterized by a fermentation process using baker's yeast on a blend of soft and durum wheat flours and water, extra-virgin olive oil and salt [5]. The nutritional values of a commercial Apulian focaccia (100 g) are approximately the following: Focaccia dough (DY ca. 172) with liquid sourdough (DS) was prepared by mixing durum wheat semolina (27% w/w, Divella, Rutigliano, ITA), soft wheat flour type "00" (27% w/w; Casillo, Corato, ITA), tap water (18.36% v/w), extra-virgin olive oil (2.1% v/w; Agridè, Bitonto, ITA), malt barley flour (0.54% w/w; Antico Molino Rosso, Buttapietra, ITA) and S (25% v/w). DS was compared with focaccia dough (DB) prepared using baker's yeast (2% w/w, corresponding to a final yeast density of ca. 8 log cfu/g) and without liquid sourdough. Three independent tests were carried out. Dough portions of 180 g were placed in round non-stick pans and fermented at 30 °C for 6 h for DS or 1 h for DB. After fermentation, the microbial communities were analyzed, together with the pH drop (ΔpH, pH units) and total titratable acidity (TTA, ml di NaOH 0.1 N/10 g) for S, DS and DB [30].

Microbiological Analyses of Doughs
After fermentation, liquid sourdough (S) was immediately subjected to decimal dilutions and plating, while 20 g aliquots of focaccia doughs (DS and DB) or of a mixture (1:1 ratio) of durum wheat semolina and soft wheat flour type "00" (F) were each homogenized with 180 mL of sterile NaCl solution (0.85%, w/v) in a Stomacher (Seward, London, UK) for 2 min. After serial dilutions, the microbial suspensions were plated on an mMRS agar (Oxoid, UK) [38] supplemented with 100 mg/l of cycloheximide (EMD Millipore Corp., Billerica, MA, USA) for the determination of lactic acid bacteria (LAB), and on a Sabouraud Dextrose Agar (SDA, Oxoid, UK) supplemented with 200 mg/l chlo- ramphenicol (Sigma, Milan, Italy) for the enumeration of yeasts and molds. Moreover, an aliquot of each microbial suspension was heat treated for 20 min at 90 • C, plated on a Plate Count Agar (PCA, Difco, Franklin Lakes, NJ, USA) and incubated for 24 h at 30 • C for spore-forming bacteria counts. A total of 20% of the colonies from the countable mMRS agar and SDA plates (incubated at 30 • C for 48 h and at 25 • C for 72 h, respectively) were randomly taken, purified and stored at −80 • C.

Characterization and Identification of LAB and Yeasts
Bacterial DNA was extracted from overnight cultures grown in mMRS broth (Oxoid, UK) at 30 • C, using a Clonsaver Card Kit (Whatman, Maidstone, UK) and analyzed by Repetitive Extragenic Palindromic-PCR (REP-PCR) [39]. The identification of L. citreum C2.27 was based on the comparison of its strain-specific REP-PCR profile with that of each LAB isolate from liquid sourdough and doughs. Bacterial isolates showing an REP-PCR profile different from that of the starter strain were identified by the sequencing of the almost complete 16S rRNA gene as previously described [24,39], using an ABI Prism 3730 × l DNA Analyzer (Thermo Fisher Scientific, Waltham, MA, USA). The species Lactiplantibacillus paraplantarum (basonym, Lactobacillus paraplantarum) [40] and Lacticaseibacillus paracasei (basonym Lactobacillus paracasei) were also identified by multiplex-PCR methods as described by Torriani et al. [41] and Ventura et al. [42], respectively.
The DNA of yeast isolates was extracted from 1.5 mL cultures grown in YEPG (Yeast Extract 1% w/v, Peptone 1% w/v and Dextrose 2% w/v) at 25 • C for 24 h, using the Wizard Genomic DNA Purification kit (Promega Corporation, Madison, WI, USA), and amplified by the oligonucleotide (GTG) 5 [43]. The isolates were identified by amplification and sequencing of the D1/D2 domain of the 26S rDNA using the primers NL1 and NL4 [44]. Bacterial and yeast strains were assigned to the species on the basis of the highest scores of alignment and percentage of identity (>99%) between their 16S rRNA/26S rRNA gene sequences and those of type strains in the NCBI Nucleotide database [24].

Culture-Independent Community Identifications
Ninety milliliters of saline solution was added to 10 g of flour mixture (F) or doughs DS and DB and homogenized for 3 min. Homogenates were centrifuged (1000× g for 5 min at 4 • C) and the supernatants were recovered and centrifuged (5000× g for 15 min at 4 • C). Each pellet was suspended in 0.5 mL of saline solution and the suspension was subjected to DNA extraction. Total genomic DNA was extracted using the FastDNA Pro Soil-Direct kit (MPBiomedicals, Santa Ana, CA, USA) coupled to the Retsch MM301 instrument (Retsch Gmbh, Germany), according to the manufacturer's instructions. Quality and quantity of DNA extracts were estimated using NanoDrop ND1000 (NanoDrop Technologies, Inc., Wilmington, NC, USA) and by 1% (w/v) agarose gel electrophoresis in TAE buffer. DNA extraction was carried out in triplicate on each sample.

Library Preparation and Sequencing
The total DNA extracted from the flour mixture and dough samples was used as template for 16S and 18S metagenetic analyses. All DNA samples were equalized at a final concentration of 10 ng/µL for NGS library preparation. Library preparation was performed using a bidirectional fusion primer set to specifically amplify the target regions. This was accomplished by combining primers targeting the regions of interest to the Ion Torrent sequencing adapters Ion A and truncated P1 (trP1). Adapter A included unique Ion Xpress Barcode sequences for sample multiplexing. Bidirectional sequencing was achieved by swapping adapter sequences A and trP1. Therefore, the V5-V6 hypervariable region of the16S rRNA gene was amplified with primers 785F (GGATTA-GATACCCTGGTA) and 1100R (GGGTTGCGCTCGTTG). For 18S libraries, the V9 hypervariable region was amplified with primers 1380F (CCCTGCCHTTTGTACACAC) and 1510R (CCTTCYGCAGGTTCACCTAC). The PCR reactions were carried out in triplicates using Platinum SuperFi DNA Polymerase (Invitrogen, USA) with the following conditions: Amplified libraries were verified on 2% agarose gel, and PCR products were purified using Agencourt AMPure XP magnetic beads (Agencourt Bioscience, Beverly, MA, USA). Purified libraries were quantified with Qubit dsDNA HS Assay Kit (Invitrogen, Carlsbad, CA, USA) and pooled at a final concentration of 100 pM. Libraries were sequenced on an Ion S5 Sequencing System (Thermofisher, Waltham, MA, USA). Following the sequencing run, a FASTQ file was generated for each sample and demultiplexed by using the built-in software of Torrent Suite.

Statistics
Analysis of variance (ANOVA) combined with the Tukey-Kramer method as a post hoc test was applied for microbiological analysis. Significant differences (p < 0.05) among focaccia dough samples are marked with different letters. Rank abundance and alpha and beta diversity indexes were estimated using Microbiome Analyst [47]. Moreover, principal coordinate analysis (PCoA) was calculated using the Bray-Curtis index. Hypothesis testing was conducted by the analysis of molecular variance (AMOVA) test (p < 0.05) [48]. Statistical comparison among taxonomic categories was performed using STAMP software [49]. Differences between groups were analyzed using Welch's t-test (p < 0.05) [50], and the Benjamini-Hochberg false discovery rate (FDR) method (q-value < 0.05) was used [51].

Data Availability
The sequence data are available at NCBI SRA under BioProject ID: PRJNA668040.

Culture-Dependent Microbiological Analyses
Durum wheat semolina and soft wheat flour type "00" in a 1:1 ratio (F) and, at the end of fermentation, S, DS and DB doughs were subjected to microbiological analyses to assess the cell load of LAB, yeasts, molds and spore-forming bacteria ( Table 1). The use of sourdough influenced the pH values of DS, which, after fermentation, showed pH 4.2 and ∆pH 1.3, while DB presented a pH value (5.38) that was significantly higher (p < 0.05). Moreover, S showed pH and TTA values lower than DS. This apparently contrasting result is due to the higher value of DY in S than in DS. In fact, it is known that a low DY value, as in DS, increases the buffering capacity of the flour, lowering the rate of acidification, even if organic acids are present at higher levels [52].
S and DS showed cell densities of LAB (9 log cfu/g) that were significantly higher than DB (4 log cfu/g). LAB isolates from liquid sourdough and doughs (20% of the colonies from countable plates) were characterized by REP-PCR. The analysis of their electrophoretic profiles revealed the presence of only the starter strain in S and DS, while four LAB strains were detected in DB, belonging to Leuconostoc mesenteroides, Furfurilactobacillus rossiae (basonym Lactobacillus rossiae), Lp. paraplantarum and Lc. paracasei. These species are typically contained in sourdough [17] and are among the species found in doughs made with different flours from the same geographical area [24]. In addition to the absence of LAB strains isolated from DB, the only strain isolated directly from the flours, belonging to Loigolactobacillus coryniformis (basonym Lactobacillus coryniformis), was also absent in DS. It is probable that this strain was no longer found in the doughs because it was dominated during fermentation by the starter strain or by the other strains, the latter evidently present in the flours in quantities lower than the detection limit.
The presence of spore-forming bacteria in the doughs was also evaluated. Bacterial counts were very low (<1 log cfu/g) and therefore below the quantities which could cause product alteration (ropy bread) (≥2 log cfu/g) or constitute a risk for consumer health (>5 log cfu/g) [53].
Yeasts were absent in S and DS, while molds were present in very low quantities. On the contrary, DB contained a high density of yeasts (8.25 log cfu/g), while molds were not detected. The REP-PCR analysis of the yeasts showed the presence of only one strain belonging to S. cerevisiae. Finally, only one yeast strain belonging to Cryptococcus victoriae was isolated from the flours but was not subsequently found in the doughs, probably due to the characteristics of the ecosystem and the presence of the starters. Overall, the results show the capacity of the L. citreum C2.27 strain to dominate the microbiota in DS, and, at the same time, the ability of S. cerevisiae to dominate the mycobiota in DB.

Bacterial and Fungal Microbiota in Flour and Doughs
After quality filtering, a total of 3,201,028 and 7,194,310 reads were generated and 3,055,906 (95.5%) and 6,648,063 (92.4%) were assigned to a taxon for 16S and 18S, respectively.
The abundance-based coverage estimators (ACE) and the alpha diversity index Chao1 indicated that the richness of the samples was DB > F > DS (Figure 2A). The inoculum of L. citreum reduced the abundance of OTUs, if compared with DB and F samples. The loss of OTUs also reduced both richness (Chao1 and ACE) and diversity (Shannon and Simpson) indexes in DS samples, indicating that the use of L. citreum as a bacterial starter had a strong effect on microbial community diversity. Considering the same population indexes, even the mycobiota diversity was strongly influenced by the presence of S. cerevisiae in the DB sample ( Figure 2B). Table 1. Acidification (∆pH, pH units), total titratable acidity (TTA, ml di NaOH 0.1 N/10 g) and microbiological characteristics of durum wheat semolina and soft wheat flour type "00" in a 1:1 ratio (F), liquid sourdough (S), doughs started with the liquid sourdough (DS) and dough made with baker's yeast (DB).  The PCoA analysis showed that the microbial communities in DB, DS and F samples were distinct from each other. The sample triplicates resemble each other very closely ( Figure 3). The results were also validated with analysis of molecular variance (AMOVA). This confirmed that the inoculum of L. citreum or S. cerevisiae significantly affected microbial consortia in DS and DB samples, respectively. The PCoA analysis showed that the microbial communities in DB, DS and F samples were distinct from each other. The sample triplicates resemble each other very closely ( Figure 3). The results were also validated with analysis of molecular variance (AMOVA). This confirmed that the inoculum of L. citreum or S. cerevisiae significantly affected microbial consortia in DS and DB samples, respectively.
The relative abundances of identified genera were further compared between sample groups, as shown in Figures 5 and 6. In addition to the dominance of the genus Leuconostoc (relative abundance 39.6% higher than in DB), DS showed a reduction in abundance of 18.5% for Stenotrophomonas, 6.3% for Pseudomonas, 4.8% for Acinetobacter and 1.7% for Paenibacillus, compared to DB. Furthermore, within LAB, Lactobacillus was significantly lower and Pediococcus was significantly higher in DS than in DB ( Figure 5). In terms of relative abundance, flour (F) had significantly higher percentages of genera Pseudomonas (up to 39%), Paenibacillus, Enterobacter and Xanthomonas than both doughs (Figure 6), while Sphingomonas and Lactobacillus were higher in F than in DS. Novosphingobium, Rosenbergiella and Staphylococcus genera were present only in F.
The metagenetic analysis of mycobiota revealed that the Saccharomyces genus dominated the mycobiota in DB and was not detected in other samples (Figures 7 and 8).

Discussion
The use of selected starter cultures for dough fermentation is useful to obtain products with determined characteristics and to standardize the production processes [12,13,54]. Starters are known to have a strong impact on the microbial populations of doughs, consequently affecting product features. Therefore, the present work used the

Discussion
The use of selected starter cultures for dough fermentation is useful to obtain products with determined characteristics and to standardize the production processes [12,13,54]. Starters are known to have a strong impact on the microbial populations of doughs, consequently affecting product features. Therefore, the present work used the

Discussion
The use of selected starter cultures for dough fermentation is useful to obtain products with determined characteristics and to standardize the production processes [12,13,54]. Starters are known to have a strong impact on the microbial populations of doughs, consequently affecting product features. Therefore, the present work used the metagenetic method and the culture-dependent approach to analyze the biodiversity of doughs started with sourdough (type II) or baker's yeast. It is well known that during traditional sourdough (type I) fermentation, the bacterial evolution is characterized by a decrease in Gram-negative and an increase in Gram-positive biota [55]. Ercolini et al. [29] observed flours with great diversity, mainly contaminated by genera (Acinetobacter, Pantoea, Pseudomonas, Comamonas, Enterobacter, Erwinia and Sphingomonas) belonging to the phylum Proteobacteria, although this population was almost completely inhibited after 1 day of sourdough propagation. On the contrary, although the phylum Firmicutes (mainly represented by LAB) was present at low relative abundances in the flours, it already became dominant after the first fermentation [29,55]. These microbial dynamics from flour to mature sourdough are well known and are in accordance with the microbial ecology of other fermented foods [16]. Flours analyzed in the present study were also found to be strongly contaminated by Proteobacteria (ca. 82%), and Pseudomonas dominated (49%) this microbiota. Contaminating bacteria originate from grain milling or are epiphytic and endophytic populations of wheat. In this study, sourdough type II (25%) or baker's yeast (2%) was used as a single starter for dough leavening. The metagenetic analysis highlighted that, after fermentation, the Proteobacteria in DS were about 43.5% lower than in F and DB. At the same time, the flour population was replaced mainly by the starter strain. In fact, the metagenetic analysis showed that the Leuconostoc genus constituted 78.6% of LAB in DS. Moreover, as supported by the identification of cultivable LAB, L. citreum C2.27 was the only LAB strain isolated from DS.
On the contrary, Proteobacteria was the dominant phylum in F and DB samples. This phylum represents metabolically active populations, and the abundance of species belonging to the phylum Proteobacteria was mostly negatively correlated with dough and sourdough qualities [29,56]. It can be hypothesized that in dough with baker's yeast, the fermentation conditions and the autochthonous microbiota of wheat flour, characterized, in particular, by the low LAB density, did not cause the reduction in Proteobacteria, which constituted, as in F, 82% of the bacterial population. In fact, the activity of the acid-producing bacterial starter during fermentation caused a rapid increase in the acidity of DS. Such an increase, together with other ecological parameters [16], may influence the microbial succession favoring LAB. On the other hand, LAB showed an increase within Firmicutes also in DB; in particular, Lactococcus, Lactobacillus, Pediococcus and Leuconostoc increased compared with F, but this was not sufficient to decrease the pH.
As described in other studies, Staphylococcus, Streptococcus and Paenibacillus had low ability to survive and colonize the doughs, especially DS. In fact, these genera are not generally found in doughs [17,31]. Regarding the presence of the genus Leuconostoc in F and DB detected by the metagenetic approach, it should be noted that this is probably due to the species L. mesenteroides isolated from DB together with Fb. rossiae, Lp. paraplantarum and Lc. paracasei. Metagenetic analysis also showed the presence of Lactococcus in the doughs, although it was not isolated from DS, probably due to the high starter load, or from DB, due to competition with other genera. In this regard, it should be considered that differences can be noted between metagenetic analysis and plating, especially for those LAB species that are generally difficult to cultivate [35]. Therefore, the flour plays a key role in establishing microbial consortia, but only species and/or strains adapted to the sourdough environment in relation to the nutrient availability and physic-chemical parameters of the process will grow and dominate in this ecosystem [17]. In type II sourdough, fermentation occurs after the inoculation of a starter culture. This starter culture can dominate and inhibit the growth of autochthonous dough microbiota because it is added in a high density [13]. However, the starter strain should be well adapted to the cereal environment in order to compete with the endogenous microbiota and be suitable for the process [57]. L. citreum C2.27, which was selected for its good leavening capacity, was isolated from durum wheat semolina and was suitable for the production process of yeast-free bread in the bakery [24]. In fact, it was the only strain isolated from liquid sourdough and DS, and the present study also confirmed its dominance by culture-independent analysis. This result is fundamental to guarantee the technological characteristics of the product made without using baker's yeast, resulting, at the same time, in positive chemical-physical and nutritional characteristics of the dough and final product, as determined in a previous work [27].
The results show that yeasts were present in very low quantities in DS and F. They are normally associated with flours up to 3.3 log cfu/g [57], and the genera were those typically found in flours and doughs [17,24,58]. Ascomycota was the only phylum present in doughs, but not in the flours evaluated. The mycobiota changed from flour to doughs; in fact, the genera Hyphopichia and Cryptococcus were found only in F, and the latter was also the only genus isolated directly from F. Considering the mycobiota of flours, molds were dominant in F, but, overall, in very low quantities. Therefore, the metagenetic approach made it possible to obtain more information on the mycobiota composition even if present in small numbers.
The absence of Saccharomyces in F and DS was interesting. S. cerevisiae is the species of yeast most frequently isolated in sourdoughs from central and southern Italy [52]. However, several studies have shown different compositions of yeast species between artisanal bakery and spontaneous laboratory sourdoughs and, in particular, have hypothesized the presence of S. cerevisiae in bakery sourdoughs due to contamination of the bakery environment with commercial baker's yeast [58,59]. Conversely, as observed in the present study, yeasts detected in sourdough and doughs made in the laboratory could only come from flours [24,60].
Finally, the results for DS, which was dominated by L. citreum C2.27, also show a reduction in the fungal diversity, probably related to the antifungal properties of the starter strain [36], while DB, with a S. cerevisiae-dominated mycobiota, presented a greater bacterial diversity.

Conclusions
This study showed how the use of a microbial starter deeply affects the composition of the dough microbiota, which is directly responsible for the quality of the product. To our knowledge, this work is the first that analyzes, also using the metagenetic method, the bacterial and fungal microbiota of doughs leavened with baker's yeast or type II sourdough (started with a selected bacterial strain), also making it possible to verify the absence of S. cerevisiae in the latter used to obtain an innovative "yeast-free" product.
Metagenetic analyses indicated that the V5-V6 hypervariable regions of the 16S rRNA gene and the V9 hypervariable region of the 18S rRNA gene were suitable for studying the microbiota of doughs, providing a comprehensive overview of the microbial community. The culture-independent approach allowed gaining deeper and wider knowledge of the starter impact on the microbial populations of the doughs, even if the association with the culture-dependent analysis is still very useful to obtain more precise information at the species level on certain microbial categories. This study highlighted that, differently from doughs produced with the L. citreum starter, the dough made with baker's yeast contained a microbiota with a high abundance of Proteobacteria (82% of the bacterial population), which were negatively correlated with the biochemical properties of the doughs [27,56]. Furthermore, the analyses showed the ability of the L. citreum C2.27 starter to dominate the microbiota, also inhibiting the growth of S. cerevisiae. This result is particularly important because L. citreum C2.27 has been adopted for its leavening abilities in a biotechnological protocol for the production of "yeast-free" bakery products.