Tap water is one of the drivers that establish and assembly the lactic acid bacterium biota during sourdough preparation

This study aimed at assessing the effect of tap water on the: (i) lactic acid bacteria (LAB) population of a traditional and mature sourdough; and (ii) establishment of LAB community during sourdough preparation. Ten tap water, collected from Italian regions characterized by cultural heritage in leavened baked goods, were used as ingredient for propagating or preparing firm (type I) sourdoughs. The same type and batch of flour, recipe, fermentation temperature and time were used for propagation/preparation, being water the only variable parameter. During nine days of propagation of a traditional and mature Apulian sourdough, LAB cell density did not differ, and the LAB species/strain composition hardly changed, regardless of the water. When the different tap water were used for preparing the corresponding sourdoughs, the values of pH became lower than 4.5 after two to four fermentations. The type of water affected the assembly of the LAB biome. As shown by Principal Components Analysis, LAB population in the sourdoughs and chemical and microbiological features of water used for their preparation partly overlapped. Several correlations were found between sourdough microbiota and water features. These data open the way to future researches about the use of various types of water in bakery industry.


Results
The chemical and microbiological characteristics markedly differed among tap water from various Italian regions. Tap  Fixed residue and electrical conductivity, both directly related with concentration of ions, ranged from ca. 186 (tap water B) to 441 (T) mg/L, and from ca. 291 (B) to 606 (U) μS/cm, respectively (Table 1). Hardness varied from ca. 162 (B) to 343 (LA) mg of CaCO 3 /L. Tap water B and TA were characterized by the lowest (P < 0.05) values of electrical conductivity and hardness. Bicarbonate, the most abundant ion in tap water, ranged from 165 (tap water TA) to 400 (LA) mg/L. Calcium was found at the highest (P < 0.05) levels in tap water C, T, U and, especially LA. Samples also differed (P < 0.05) for the levels of nitrate, chloride, sulphate, sodium, potassium and magnesium. Other ions (bromide, nitrite, phosphate, lithium, sulphites and ammonia) were not detectable. Sodium adsorption ratio (SAR), indicating the sodium content referred to calcium and magnesium, was ever lower than 1, except for tap water A and P. Samples T and U showed the lowest (P < 0.05) values of pH (7.3-7.5).
On the contrary, tap water A was the most alkaline (pH 8.1). Heterotrophic psychrophilic microorganisms ranged from 1.0 (tap water LO) to 4.2 (tap water B) log CFU/mL, but were not detected in tap water LA, T and V. Heterotrophic mesophilic microorganisms were found at cell density varying from 0.8 (C) to 3.4 (B) log CFU/mL, but were absent in tap water LA, LO and T. Principal Components Analysis (PCA) was carried out using chemical and microbiological characteristics of tap water as data entries. The two first principal components explained ca. 66% of the total variance ( Fig. 1). PC1 mainly loaded water bicarbonate, hardness and heterotrophic microbes. PC2 mainly distributed samples according to sodium and sulphate levels. PCA grouped in the first quadrant tap water B, C and TA, being characterized by relatively low fixed residue. Tap water A and P, sharing the highest concentrations of fluoride, sodium and value of SAR, fell in the second quadrant. Tap water U and T were grouped in the third quadrant, because they showed the highest and lowest values of electrical conductivity and pH, respectively. The fourth quadrant grouped tap water LA, LO and V, being characterized by relatively high concentration of magnesium.
Tap water hardly affected the microbiological features during propagation of a traditional and mature Apulian sourdough. Before propagation, the pH value of the traditional and mature sourdough, produced by P water, was ca. 4.0. Cell density of presumptive LAB and yeasts were ca. 9.0 (both on modified de Man-Rogosa-Sharpe (mMRS) and Sour Dough Bacteria (SDB) agar media) and 7.0 log CFU/g, respectively. The sourdough showed a 400 ml volume increase in 6 h of fermentation (data not shown). It was propagated at laboratory for nine days under the same conditions but using different tap water collected from the other 9 Italian regions. In addition, a control sourdough (P sourdough) was propagated using the P water. After sourdough propagation, the values of pH ranged from 3.96 to 4.03, being not significantly different (P > 0.05) from the value before propagation. Similarly, no significant differences (P > 0.05) were detectable for the number of LAB, which varied from 8.9 to 9.2 CFU/g. In addition, the values of pH and LAB cell density were not significantly (P > 0.05) different among the ten propagated sourdoughs. The yeast cell density decreased (P < 0.05) by approximately one log cycle (5.8-6.2 log CFU/g) and, consequently, the leavening capacity decreased (P < 0.05) to 300 ml, regardless of the tap water used during sourdough propagation (data not shown).
LAB were isolated from the sourdough before propagation and from the 10 sourdoughs after propagation. As shown by Randomly Amplified Polymorphic DNA (RAPD)-PCR analysis and identification of bacterial isolates, two strains of Lactobacillus sanfranciscensis and two strains of Lactobacillus plantarum dominated the traditional and mature sourdough before propagation ( Supplementary Fig. S1a). Both species were detected in all the sourdoughs after propagation, regardless of the water used. In detail, at least three out of the four lactobacilli strains identified in the sourdough before propagation were also detectable in all the propagated sourdoughs. Besides these strains, the sourdoughs propagated with A or P water harbored another strain of L. plantarum (s3) and one strain (s1) of Pediococcus pentosaceus ( Supplementary Fig. S1a,c). L. plantarum s3 was also found in the sourdoughs propagated with B ( Supplementary Fig. S1b) or U water ( Supplementary Fig. S1d). The sourdoughs propagated with C, LA Tap water affected hardness, gumminess and chewiness of breads made using a traditional and mature sourdough. After nine days of propagation with different water, each traditional and mature sourdough was used as starter (in combination with baker's yeast) for bread-making. No significant (P > 0.05) differences were found among the breads for elasticity (89-100%), cohesiveness (0.29-0.38), specific volume (2.52-2.82 cm 3 /g) and percentage of gas cell area (black pixel area ranging from 54 to 57%) ( Table 2). Hardness ranged from ca. 15.5 (bread started with traditional sourdough propagated with tap water P and LO) to 28 N (C), with significant (P < 0.05) differences among most of breads. The breads started with sourdough propagated using tap water C, LA and U showed higher (P < 0.05) gumminess (43.81 N on average) than the breads obtained with sourdough propagated using tap water LO and P (26.08 N on average). The same three breads were characterized by higher (P < 0.05) values of chewiness (average value: 40.33 N) than the breads started with traditional sourdough propagated using tap water A, LO, P and TA (27.39 N). Correlations between tap water characteristics and bread hardness, gumminess and chewiness were calculated. Positive (P < 0.05) correlations (r > 0.70) were found between concentration of bicarbonate in water and the above-mentioned bread textural features. Weaker correlations (r > 0.50) were found between calcium and bread hardness and chewiness, as well as between potassium in water and bread hardness. All the three textural features of bread were negatively correlated (r < −0.63) with chloride in water. In addition, a weaker (r = −0.52) negative correlation was found between sulphate and bread hardness.
Tap water affected the establishment and assembly of the LAB biome during sourdough preparation.
During preparation of the 10 sourdoughs, the values of pH already significantly (P < 0.05) differed after the first 8 h of fermentation ( Supplementary Fig. S2). Indeed, fermented doughs obtained using tap water T, TA and U showed the lowest values of pH (5.63-5.67), whereas those obtained using tap water A, LA, LO and P showed the highest values (5.92-5.94). The differences of pH among the fermented doughs diminished after the third fermentation (4.43-4.52), but the sourdoughs produced using water T and U were characterized by the lowest values. After the fourth fermentation, all the sourdoughs showed values of pH lower than 4.5. Afterwards, final pH continued to decrease slightly and then reached constant values after the sixth (sourdoughs produced using tap water A, B, C, P, T, U and V) and eighth (LA, LO and TA) fermentation. After the last fermentation, pH values of the sourdoughs ranged from 4.15 to 4.22.
At the beginning of the first fermentation, presumptive LAB ranged from 2.4 to 2.6 (estimated on mMRS) log CFU/g and from 3.4 to 3.6 (estimated on SDB) log CFU/g, with no significant difference (P > 0.05) among the doughs (Supplementary Table 1). Presumptive LAB increased of ca. one log cycle after the first fermentation. At the end of the last fermentation (day 9), cell density of LAB estimated on SDB varied from 9.0 (sourdough back-slopped with tap water LO) to 9.6 (tap water TA) log CFU/g. Under our experimental conditions, only two significant differences (P < 0.05) were found at the end of the last fermentation: LAB cell density in sourdough LO vs. sourdough TA, estimated on both SDB and mMRS. Within the same mature sourdough, cell density estimated on mMRS did not differ (P > 0.05) from that found using SDB. Cell density of yeasts was determined in the sourdoughs and ranged from 5.7 to 6.0 log CFU/g, with no significant difference (P > 0.05) (Supplementary Table 1).
After the first fermentation, 75 presumptive LAB isolates were bio-typed by RAPD-PCR analysis and the representative strains were identified by partial sequencing of 16S rRNA and pheS genes. All the fermented doughs harbored a LAB consortium composed of Lactococcus lactis, Leuconostoc citreum and Weissella confusa (Supplementary Table 2). In addition, doughs produced by using tap water B, T and TA also harbored Leuconostoc pseudomesenteroides. The highest number of strains was found for W. confusa. Some of them varied depending on the tap water used.   Table 2. Physical characteristics of breads made by traditional and mature sourdough propagated for nine days using tap water collected in ten Italian regions. a The name of each sourdough bread derived from the tap water used during the propagation of traditional sourdough. b Black pixel area was the percentage of the image area filled up with black pixel (gas cells). Mean values (±S.D.) in the same column with a common letter were not significantly different at a P value of 0.05.
SCIenTIFIC RepoRTS | (2019) 9:570 | DOI:10.1038/s41598-018-36786-2 One hundred and twelve Gram-positive, catalase-negative, non-motile, cocci or rods acidifying isolates from the ten mature sourdoughs (after 9 fermentations) were subjected to RAPD-PCR analysis resulting in 31 strains ( Supplementary Fig. S3). The following species were identified: Lactobacillus plantarum (12 strains), Pediococcus pentosaceus (12), Ln. citreum (5), Lactobacillus curvatus (1) and Pediococcus acidilactici (1). L. plantarum was identified in all the sourdoughs and was the only LAB species found in the sourdough prepared using tap water C (Table 3). P. pentosaceus was found as co-dominant in seven sourdoughs, prepared with tap water B, LA, LO, P, T, TA and U. Besides L. plantarum and P. pentosaceus, the sourdoughs obtained using tap water B and LO also harbored L. curvatus. These three species were flanked by Ln. citreum in the sourdough obtained using tap water U. Ln. citreum was also found (together with L. plantarum and P. pentosaceus) in the sourdoughs prepared with tap water LA and TA. Besides harboring the previously mentioned four species, the sourdough prepared with tap water T also hosted P. acidilactici. In the sourdoughs prepared with tap water A and V, three and four, respectively, strains of L. plantarum were flanked by one strain of L. curvatus. Within L. plantarum, some strains were detected in two or more sourdoughs (Table 3). For instance, L. plantarum s4 and s6 were found in the sourdoughs obtained with tap water B, C and V, s5 was found in the C, LA, LO and V sourdoughs, and s9 and s10 were found in the A and P sourdoughs. Similar results were found for strains of P. pentosaceus.
DNA extracted from the doughs was used as template for 16S metagenetic analysis in order to describe the diversity within the Firmicutes phylum (Fig. 2). Doughs before fermentation harbored 15 major Operational Taxonomic Units (OTUs), which were ever detected at values of relative abundance that did not differ (P > 0.05) among the doughs. W. confusa was the OTU found at the highest relative abundance (ca. 61%), followed by Lc. lactis (13.8%), L. curvatus (9.68%) and Lactococcus garvieae (4.41%). After the first fermentation (day 1), L. curvatus, W. confusa and L. plantarum were the dominant OTUs. Compared to doughs after 1 day of fermentation, the relative abundance of Firmicutes found in the mature sourdoughs (day 9) changed. Overall, 11 major OTUs were attributed to LAB (Table 4, Fig. 2). Six species (L. curvatus, Lactobacillus fermentum, L. plantarum, P. pentosaceus, Ln. citreum and W. confusa) were detected in all the sourdoughs, but their relative abundance varied among sourdoughs. When tap water A and P were used for preparing the respective sourdoughs, L. plantarum was the most abundant OTU (relative abundance of ca. 49%), followed by L. curvatus (ca. 33%) and Lactobacillus brevis (ca. 16%). L. plantarum and L. curvatus were also the two most abundant OTUs in the sourdoughs prepared with tap water B, C and V. In these sourdoughs, W. confusa was also detected at relative abundance ranging from ca. 5 to 19%. L. curvatus and L. fermentum were the dominant OTUs in the sourdoughs produced with tap water LA and LO. The sourdoughs produced with tap water T and U were characterized by L. curvatus at the highest relative abundance (ca. 81%) and by other species (L. plantarum, P. pentosaceus, Ln. citreum, W. confusa and Lc. lactis) whose abundance varied between ca. 1 and 8%. When tap water TA was used for preparing sourdough, L. curvatus was the most abundant OTU (relative abundance of ca. 65%), followed by L. plantarum (ca. 13%), P. pentosaceus (ca. 10%) and W. confusa (ca. 6%).
When the 10 newly prepared sourdoughs were used as starters for bread-making, the resulting breads approximately showed the same differences as those found for breads obtained with traditional sourdough propagated using different tap water (Supplementary Table 3).
Correlations between sourdough microbiota and chemical composition of tap water. PCA based on the chemical characteristics of water and on the microbiological characteristics (LAB cell density, percentage of strains per species, relative abundance of OTU) of the mature sourdoughs showed that two first principal components explained ca. 54% of the total variance (Fig. 3). Sourdoughs produced with tap water C, V and TA were grouped in the second quadrant because of the highest values of relative abundance of W. confusa, Lc. lactis and Ln. citreum. Sourdoughs prepared with tap water LA, LO, T and U fell in the third quadrant. They were characterized by the highest relative abundance of L. curvatus and L. fermentum and presence of strains allotted to P. pentosaceus. Sourdoughs A and P, characterized by the highest relative abundance of L. brevis, fell in the fourth quadrant. Positive correlations (P < 0.05) were found between concentration of potassium in tap water and number of L. plantarum strains (r = 0.86) and relative abundance of W. confusa (r = 0.80). On the contrary, potassium and L. curvatus were negatively correlated (r = −0.63). Magnesium concentration in tap water was correlated positively with L. fermentum (r = 0.69). Relative abundances of Ln. citreum, W. confusa and Lc. lactis were negatively correlated with magnesium and, especially, chloride (r ≤ −0.60). Positive correlations (P < 0.05) were found between sodium and relative abundance of L. brevis (r = 0.78) and L. plantarum (r = 0.25). Sodium was negatively correlated with L. fermentum (r = −0.69). Concentration of sulphate in water was positively correlated with number of P. pentosaceus (r = 0.82) and Ln. citreum (r = 0.69) strains. Bicarbonate and relative abundance of L. fermentum were also correlated (r = 0.75).

Discussion
In this study ten tap water, collected from as many Italian regions characterized by cultural heritage in leavened baked goods, were used for propagating or preparing firm (type I) sourdoughs. All the tap water used in this study could be classified as either hard (120-300 mg of CaCO 3 /L; water A, B, C, P, TA, V) or very hard (more than 300 mg/L; the remaining ones). Although some researchers claimed that hardness of water should range from 50 to 200 mg/L 25 , in Italy, one of the major countries in the world in terms of bread cultural heritage, water evidently show, on average, hardness values above that range. When trying to assess whether tap water would influence sourdough, we found two fairly opposed effects, depending on the sourdough protocol of use (propagation vs. preparation). When different tap water were used during nine days of laboratory propagation of a traditional and mature Apulian sourdough, LAB cell density did not differ, regardless of the type of tap water used. Regarding the composition of LAB population of the propagated sourdoughs, around 60-80% of all the strains were identical to those detected before propagation. Previously, it had been shown that five out of seven traditional sourdoughs, propagated for 20 days under laboratory conditions, harbored from 50 to 100% of the LAB strains detected before starting the propagation 18 . However, in the current study two strains belonging to L. plantarum and P. pentosaceus, which had not been detected in the sourdough before propagation, were found in eight propagated sourdoughs. The retrieval of these additional LAB could be explained by temporary fluctuations, although we cannot exclude that these LAB could have the chance to replace the lactobacilli strains that dominate the traditional sourdough 6 . The results found in the current study showed that the type of tap water used during propagation of a mature and stable artisan sourdough had modest effect on the LAB population and, consequently, on the acidifying capacity. During laboratory propagation, cell density of yeasts decreased. This could be due to the place of sourdough propagation, being laboratories less subjected than bakeries to environmental contamination with baker's yeast 18 . In order to scout the effect of tap water on sourdough bread, the propagated sourdoughs were used for making breads, in combination with a low concentration of baker's yeast. The breads differed for some textural features. Overall, bread texture results from interactions between gluten matrix (whose behavior differs depending on flour), type of leavening agent used, fermentation conditions and chemical composition of tap water 26 . In the experimental design of this study, the tap water used for sourdough propagation, preparation and bread-making was the only variable. Thus, the differences among breads in terms of gumminess, chewiness and hardness might be attributed to tap water. These textural attributes positively correlated with the concentration of bicarbonate, but negatively with chloride. Other weaker positive correlations were between bread hardness and chewiness, and concentration of calcium. Besides, bread hardness also positively correlated with the concentration of potassium. Previous studies showed that pH and hardness of tap water affect both rheological behavior of dough 21 and texture of bread 20,22 . Water hardness indicates its content of calcium and magnesium 26 . In detail, temporary hardness refers to calcium and magnesium bicarbonates, which decrease upon boiling, whereas permanent hardness, not affected by boiling, refers to calcium and magnesium sulphates 26 . In the current study, relatively weak positive correlations were found between water hardness and bread textural features. This could probably be due to the fact that values of hardness determined for the ten tap water express calcium, magnesium, bicarbonate and sulphate (total hardness), but magnesium in water varied at much lesser extent than calcium and bicarbonate.  When tap water were used for preparing the corresponding sourdoughs, the values of pH became lower than 4.5 after two to four fermentations (back-slopping), depending on the tap water. These values of pH limit the growth of many microorganisms that normally contaminate cereal-based doughs, such as some members of the Enterobacteriaceae family. Given that the same batch of flour was used, the reasons for different pH values found during early fermentations should be searched among: (i) different pH and buffering capacity of water; and (ii) specific effect of minerals on the fermentative metabolism of LAB 27 . Tap water T and U had the lowest pH and relatively high total concentration of cations, which overall favor acidification. Regarding the effect of minerals on LAB, no clear relationships were found at the early fermentation steps. However, some minerals (e.g., sulphate, sodium, potassium and magnesium) are normally contained in the media routinely used for LAB cultivation 28 . Although the concentration of most of minerals (especially potassium and magnesium) in flour is much higher than in tap water 29 , some of them could not be available to microorganisms, because sequestered by flour components. For instance, phytic acid in cereals chelates magnesium and other divalent cations 26 . It may be hypothesized that tap water used in dough preparation, along with flour, could provide LAB with minerals necessary to their growth. Yet, such a hypothesis would fall before the results obtained when different tap water were used during propagation of a traditional sourdough. In order to solve this issue, we hypothesize that minerals provided by water (besides those by flour) could drive LAB population just during preparation of sourdough, whose early fermentations are due to bacteria present at low cell density. Vice versa, the influence of minerals from tap water would be negligible during propagation of a mature sourdough, inhabited by high microbial numbers.
Based on culture-dependent analysis, the newly prepared sourdoughs harbored from one to five species of LAB. L. plantarum and P. pentosaceus were the most commonly isolated species, whereas L. curvatus, Ln. citreum and P. acidilactici were occasionally isolated. L. plantarum and P. pentosaceus have been detected in 43% and 14%, respectively, of more than 500 sourdoughs 5 . Upon culture-independent analysis, the sourdoughs were distinguished based on relative abundance of eight LAB OTUs, among which L. plantarum and L. curvatus were the most important. Compared to the results from culture-independent analysis, L. curvatus was isolated at lower frequency, whereas L. plantarum was more frequently isolated. We hypothesize that L. curvatus would grow in SDB medium at lower extent than L. plantarum. Indeed, the medium chosen for culture-dependent analysis may affect the results describing sourdough bacterial diversity 30 . As shown by PCA, LAB population in the sourdoughs and chemical and microbiological features of tap water used for their preparation partly overlapped. This occurred for tap water A and P and the corresponding sourdoughs. Water LA, LO, T and U, sharing high values of hardness and electrical conductivity (PCA in Fig. 1), gave sourdoughs which harbored L. curvatus at high relative abundance and strains of P. pentosaceus. In some cases, correlations found between relative abundance of LAB OTUs and chemical features of tap water would suggest that some minerals, such as magnesium, potassium, sodium and sulphate, drive the establishment and assembly of the sourdough LAB biota. Microbial cells need magnesium for several biochemical processes (e.g., biosynthesis of nucleic acids and proteins). Bacteria may show different ability to accumulate magnesium in the cytoplasm. It has been reported that L. brevis accumulates higher concentration of this mineral than L. plantarum 31 . This different ability was reflected in higher magnesium concentration in bread started with Saccharomyces cerevisiae and L. brevis 32 . Compared to sodium chloride, magnesium sulphate caused faster growth of Lactobacillus sakei in MRS 33 . Although milk contains sufficient magnesium for microbial growth, an exogenous surplus (ca. 16-50% of the normal supply in milk) of this mineral stimulated growth of two streptococci 34 . Potassium is another macro-nutrient for microbial cells, because it is involved in many enzyme activities 27 . In this study, we found that potassium was positively correlated with one member of the Leuconostocaceae family. Potassium (along with manganese and phosphate) is essential for growth of Leuconostoc mesenteroides. Cell density of LAB was higher in cheeses wherein potassium chloride partially replaced sodium chloride 35 . Similarly, potassium chloride showed lower inhibitory effect than sodium chloride  Table 4. Relative abundance (%) * of Firmicutes OTU found in the mature sourdoughs prepared using tap water collected in ten Italian regions, at the end of the last fermentation. *Values are the mean (±S.D.) of two analytical replicates performed using DNA extracted from a pool composed of sourdoughs obtained in two parallel and independent series of fermentations. Values in the same row with a common letter were not significantly different at a P value of 0.05. a The name of each sourdough derived from the tap water used during the preparation of sourdough.  against Lactobacillus casei 36,37 . Conversely, it has been recently reported that cheeses wherein sodium was partially replaced with potassium contained lower cell number of Lc. lactis and L. casei 38 . Overall, these studies suggest that the same mineral may have different effects depending on the bacterial species. This study showed that the type of tap water drives the assembly of the LAB biota in newly prepared sourdoughs, but hardly influences the microbiota of stable sourdough. Although we found different counts of heterotrophic microorganisms in the tap water used in this study, the role of water microbiota during sourdough preparation could perhaps be neglected, because of lower level of microbial contamination, compared to that of doughs during fermentation. On the contrary, we found that water mineral concentrations could be an additional driver, along with flour and fermentation parameters, for sourdough bacterial biota establishment.

Propagation of a traditional and mature sourdough with different tap water. An Apulian, durum
wheat semolina 20-years-old, type I sourdough (dough yield, DY, of 160), dominated by Lactobacillus sanfranciscensis s1, L. sanfranciscensis s2, Lactobacillus plantarum s1, L. plantarum s2 and Saccharomyces cerevisiae s1, was continuously propagated for nine days, using the same batch of durum wheat semolina (Grandi Molini Italiani, brand "Ca' Bianca", Rovigo, Italy) but ten different tap water. One of these water, namely that collected in Puglia (P), was the same used for conventional propagation of the Apulian traditional and mature sourdough considered in this study. The gross composition of semolina (alias "flour") was the following: moisture, 12.9%; protein, 11.1%; carbohydrates, 71.0%; fiber, 3.5%, fat, 1.5%. Propagation was performed for nine days because we reasonably expected to find the (eventual) effect of water on sourdough in such a lapse of time. The following parameters were adopted in the propagation (back-slopping): DY, 160; inoculum ratio, 20%, fermentation temperature, 30 °C; fermentation time, 6 h; resting time (at 4 °C), 18 h. Before starting propagation and after the last fermentation (day 9), pH, volume increase (in ml), cell densities of LAB and yeasts were monitored in the sourdoughs.
Sourdough preparation. The ten tap water were used for preparing as many sourdoughs, according to the scheme in Fig. 4. Before fermentation, 10 g of dough were sampled and stored at −20 °C for extraction of total DNA. After fermentation (30 °C, 8 h), 10 g of dough were sampled, whereas the remaining dough was stored at 4 °C for 16 h (resting). Upon the 1 st back-slopping (inoculum ratio, 20%; DY, 160), dough was fermented (30 °C, 6 h) and then stored at 4 °C for 18 h. After this second fermentation, other seven fermentation steps followed, keeping the same inoculum ratio, DY and conditions of fermentation. At the end of the last (ninth) fermentation, 10 g of dough were sampled. Based on the available literature, a number of back-slopping steps ranging from 5 to 10 is usually needed to obtain mature sourdough 5,46 . Serial fermentations were performed in duplicate. pH was measured at the beginning and end of each fermentation. Presumptive LAB were enumerated in all the doughs at the beginning and end of the first fermentation and at the end of the ninth fermentation. Yeasts were enumerated at the end of the ninth fermentation.
Acidification and microbial cell densities in doughs and sourdoughs. The values of pH were determined by direct insertion of a Foodtrode (Hamilton, Bonaduz, Switzerland) electrode. Before plate count, the two biological replicates for type of dough produced by using different water were pooled. Each pool was homogenized, serially diluted with sterile peptone water and plate counted. LAB were enumerated using modified de Man-Rogosa-Sharpe (mMRS) (Oxoid, Basingstoke, Hampshire, UK) and Sour Dough Bacteria (SDB) agar media, both supplemented with cycloheximide (0.1 g/L). Yeasts were enumerated using Sabouraud Dextrose agar (SDA) supplemented with chloramphenicol (0.1 g/L) 47 .
Isolation, identification and biotyping of LAB in sourdoughs. Ten colonies from SDB agar plates inoculated with the intermediate dilution of each single dough (obtained on the first day of sourdough production) and sourdough (traditional sourdough before and after propagation and sourdoughs prepared in laboratory) were randomly picked up to isolate LAB. Isolation of pure bacterial cultures was obtained upon two repeated streaks, using SDB broth. DNA was extracted from Gram-positive, catalase-negative cocci and rods able to acidify the medium, using the Qiagen (Courtaboeuf, France) DNeasy Blood and Tissue Kit, following the manufacturer's instructions 48 . LAB were biotyped through RAPD-PCR using P4, P7 and M13 primers, applying the same conditions previously described 17,49 . The PCR products (3 µl) were separated by electrophoresis at 100 V for 120 min on 1% (w/v) agarose gel containing (0.005%, v/v) GelRed ® Nucleic Acid Gel Stain (Biotium, Inc., Fremont, CA), for detecting the DNA by UV trans-illumination. The molecular sizes of the amplified DNA fragments were Extraction of total DNA from sourdoughs and Illumina MiSeq analysis. 10 g of doughs and related sourdoughs were homogenized with 90 ml of sterile saline solution. Pellet was treated with FastPrep and DNA was extracted using the FastDNA Spin Kit for Soil (MP Biomedicals, Illkrich, France), following the manufacturer's instructions 54 . Concentration and purity of DNA was assessed spectrophotometrically using the Nanodrop ND-1000 (Thermo Fisher Scientific Inc.). DNA was used as template in PCR using primers Firm350F/Firm814R, targeting the region V3-V4 of Firmicutes 55 . PCR were run by RTL Genomics (Lubbock, TX), using internal protocols. PCR products were sequenced at RTL Genomics, using the Illumina 2 × 300 bp paired-end MiSeq platform. Processing of sequenced reads was carried out using, in sequence, the PEAR Illumina paired-end read merger 56 , UCHIME software 57 , USEARCH global alignment algorithm 58 , and UPARSE OTU selection algorithm 59 , as previously described 8 . Operational Taxonomic Units (OTUs) were identified using a database of high quality sequences derived from the NCBI.
Bread-making. Both traditional and mature, and the newly prepared sourdoughs were used as starters in bread-making. In this latter case, the sourdoughs were considered to be mature after nine fermentations. In detail, 60 g of sourdough (10% inoculum rate) was used as ingredient in bread dough together with 337.5 g of durum wheat semolina (the same type and batch used during sourdough propagation/preparation), 202.5 g of tap water (differing depending on the sourdough used) and 1.5 g (corresponding to 0.25% inoculum rate) of baker's yeast (Zeus IBA s.r.l. Firenze, Italy). Two loaves per sourdough bread were shaped, each weighing ca. 250 g. After fermentation (30 °C, 3 h) and baking (220 °C, 25 min), bread loaves were cooled at room temperature for 18 h.
Analyses of breads. Specific volume (cm 3 /g) was determined using the rapeseed displacement method 60 .
Breads were subjected to Texture Profile Analysis (TPA) by using the Universal Testing Machine 3344 (Instron, Norwood, USA), equipped with a cylindrical (diameter, 3.6 cm) probe and a 1 kN load cell. Bread loaves were compressed (30%, two cycles) and the following parameters were registered: hardness, defined as the peak force (expressed in N) during the first compression cycle; elasticity, defined as the elastic recovery (expressed as percentage) that occurs when the compressive force is removed; gumminess (expressed in N); chewiness (expressed in N); cohesiveness, defined as the non-dimensional ratio of the positive force area during the second compression cycle to the positive force area recorded during the first compression cycle. Bread loaves not used for TPA were sliced and the relative images were scanned full-scale using an Image Scanner (Amersham Pharmacia Biotech, Uppsala, Sweden), at 300 dots per inch, and analyzed in grey scale (0-255). Image analysis was performed using the UTHSCSA ImageTool program (Version 3.0, University of Texas Health Science Centre, San Antonio, Texas, freely downloadable at, http://en.bio-soft.net/draw/ImageTool.html) 61 .
Statistical analyses. Two parallel and independent series of fermentations were run for type of water.
One-way ANOVA was applied to the data from two biological replicates (analyzed in triplicate for getting the results about pH and volume increase of sourdough, and physical features of bread) and Tukey's procedure (at P = 0.05) was used to perform pair comparison of treatment means. In addition, the following data were subjected to PCA, using Statistica 7.0 for Windows, and cluster analysis, using PermutMatrix 62 : (i) chemical and microbiological characteristics of the water; (ii) characteristics of sourdough bacterial community (assessed through culture-dependent and -independent methods). Spearman correlations between chemical characteristics of the water and OTUs found in the newly prepared sourdoughs were computed using Statistica v. 7.0.

Data Availability Statement
The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.