Raw milk, swabs from cow and goat saliva and vaginal mucosa, ruminal boluses, consumption water and silage were collected from dairy farms (goat and cattle) located in the Universidade Federal de Viçosa, Viçosa, Minas Gerais state, Brazil, with conventional management and production destined to dairy processing. The samples were obtained after agreement of the responsible sector for managing these farms (Animal Science Department, Universidade Federal de Viçosa) and kept refrigerated before the following analyses.
LAB isolation and characterization
All samples were ten-fold diluted with 0.85% NaCl (w/v). Selected dilutions were pour-plated in Man, Rogosa and Sharpe agar (MRS, Oxoid Ltd., Basingstoke, England) and MRS supplemented with 10 mg/L vancomycin (Sigma–Aldrich, St. Louis, MO, USA) for LAB enumeration, according to M Colombo, AEZ Oliveira, AF Carvalho and LA Nero . Representative colonies were selected (10% of the observed count) and tested for Gram stain and catalase reaction. The preliminary LAB characterized isolates (Gram-positive and catalase-negative) were freeze-dried and stored at − 20 °C. Further microbiological analyses were conducted, as described in the sections below.
Gastric pH resistance
Bacterial cells were grown overnight and prepared for the gastric pH resistance test, according to AA Argyri, G Zoumpopoulou, KG Karatzas, E Tsakalidou, GE Nychas, EZ Panagou and CC Tassou . Resistance, assessed in triplicate, was evaluated by viable colony counts on MRS agar after incubation at 37 °C for 0 and 3 h, reflecting the time spent by food in the stomach. The resistance to low pH was performed as described by SD Todorov, DN Furtado, SMI Saad, E Tome and BDGM Franco , with some modifications. The isolates were grown at 37 °C in MRS broth adjusted to pH 2.0, 2.5 and 3.0 with HCl until the cell density reached 3 × 107 CFU/mL. All tests were conducted in sterile flat-bottom 96-well microtiter plates (Thermo Scientific, Waltham, MA, USA). In order to compare the count with the absorbance reading, optical density (OD) measurements were recorded at 650 nm at zero time and after incubation at 37 °C for 3 h (aerobic condition), using a microtiter plate reader (BioTek Instruments, Inc., Winooski, VT, USA). Cultures grown in MRS broth corrected to pH 7.2, served as the control.
After preparing the bacterial inoculum , the resistance to bile salts was assessed, based on SD Todorov, DN Furtado, SMI Saad, E Tome and BDGM Franco , with some modifications. The isolates were grown at 37 °C in MRS broth containing 0.5 and 3% (w/v) bile salts (Sigma), using 96-well microtiter plates, as described above. The OD readings were recorded at zero time and after incubation at 37 °C for 4 h. Cultures grown in MRS broth without bile served as the control.
DNA of 82 selected isolates was extracted using a ZR Fungal/Bacterial DNA kit (Zymo Research, Irvine, CA, USA), and the DNA concentrations determined using NanoDrop (Thermo Scientific). Repetitive-element PCR and gel electrophoresis were performed according to the protocol described by B Dal Bello, K Rantsiou, A Bellio, G Zeppa, R Ambrosoli, T Civera and L Cocolin , using the single primer GTG5 (Additional file 1: Table S1). The electrophorezed gels were stained with Gel Red (Biotium, Inc., Hayward, CA, USA) and the bands were visualized and documented using an ultraviolet transilluminator (LPIX, Loccus Biotecnologia, São Paulo, SP, Brazil). Further differentiation of the LAB strains was achieved by random amplification of polymorphic DNA, as detailed by SD Todorov, M Wachsman, E Tomé, X Dousset, MT Destro, LMT Dicks, BDG de Melo Franco, M Vaz-Velho and D Drider . Taxonomic identification was confirmed by sequencing of PCR-amplified 16S rRNA using the universal pair of primers 8F and 1512R . Sequencing of the amplicons was done at the Center for Human Genome Studies, Institute of Biomedical Sciences, University of São Paulo (São Paulo, SP, Brazil). Obtained sequences were compared to reference sequences in GenBank, using the basic local alignment search tool (BLAST).
Detection of enzymatic activity
The enzymatic activity of each of the selected isolates was established, according to the API ZYM Kit (bioMérieux, Marcy-l’Étoile, France) manufacturer’s manual. The following enzymes were tested: alkaline phosphatase, esterase, esterase/lipase, lipase, leucine arylamidase, valine arylamidase, cysteine arylamidase, trypsin, α-chymotrypsin, acid phosphatase, naphthol-AS-BI-phosphohydrolase, α-galactosidase, β-galactosidase, β-glucuronidase, α-glucosidase, β-glucosidase, N-acetyl-β-glucosaminidase, α-mannosidase and α-fucosidase.
Resistance to simulated gastric and intestinal conditions
The tolerance of the selected strains to gastric and intestinal conditions was evaluated through an in vitro model described by KMO Santos, ADS Vieira, FCA Buriti, JCF Nascimento, MES Melo, LM Bruno, MF Borges, CRC Rocha, ACS Lopes and BDGM Franco . The assay was performed three times for each strain, and the enumeration was done in duplicate. The survival rate (SR) of strains after gastric and enteric simulation were calculated using the equation: SR (%) = [log CFU N/log CFU N0] × 100 , where N0 and N are the populations before and after the assay, respectively. The mean counts of log populations were compared by analysis of variance (ANOVA) and Tukey’s test (p < 0.05) using XLSTAT 2016.01.26192 (Addinsoft, New York, NY, USA).
Aggregation and co-aggregation properties
Aggregation abilities of the 15 selected LAB were tested using the method proposed by SD Todorov, DN Furtado, SMI Saad, E Tome and BDGM Franco  and Y Zhang, L Zhang, M Du, H Yi, C Guo, Y Tuo, X Han, J Li, L Zhang and L Yang . Auto-aggregation was determined using the following equation: % auto-aggregation = [(OD0 – OD60) / OD0] × 100. OD0 and OD60 refer to the initial OD and the OD determined at 60 min, respectively.
For evaluation of co-aggregation, the 15 selected strains were grown in 10 mL of MRS and Listeria monocytogenes Scott A, Enterococcus faecalis ATCC 19443 and Lactobacillus sakei ATCC 15521 in brain heart infusion (Oxoid) and MRS (Oxoid), respectively, at 37 °C . Co-aggregation was calculated using the following equation: % co-aggregation = [(OD0− OD60) / OD60] × 100. OD0 refers to the initial OD, taken immediately after the relevant strains were paired. OD60 refers to the OD of the supernatant at 60 min. Experiments were conducted in triplicate on two separate occasions.
The bacterial adhesion to hydrocarbons was tested as described by RJ Doyle and M Rosenberg , using 15 selected LAB strains. The percentage hydrophobicity was calculated as follows: % hydrophobicity = [(OD580 reading 1 – OD580 reading 2) / OD580 reading 1] × 100. Experiments were conducted in triplicate.
Finally, DNA obtained from the selected strains was analyzed by PCR for the presence of genes (Additional file 1: Table S1) related to the adhesion characteristics. The target genes included EF2380, EF2662, prgB, EF1249 , map, mub and EFTu .
Bile salt deconjugation
The selected strains were evaluated by their ability in deconjugate bile salts, as described by KMO Santos, ADS Vieira, FCA Buriti, JCF Nascimento, MES Melo, LM Bruno, MF Borges, CRC Rocha, ACS Lopes and BDGM Franco , using sodium salts of taurocholic acid (TC), taurodeoxycholic acid (TDC), glycocholic acid (GC) and glycodeoxycholic acid (GDC) (all from Sigma–Aldrich), in two repetitions and in duplicate.
The assay described by KMO Santos, ADS Vieira, FCA Buriti, JCF Nascimento, MES Melo, LM Bruno, MF Borges, CRC Rocha, ACS Lopes and BDGM Franco  was considered to assess the β-galactosidase activity of the selected strains, using sterile filter paper discs impregnated with o-nitrophenyl-β-D-galactopyranose (ONPG discs, Fluka, Buchs, Switzerland), in two repetitions and in duplicate.
The ability of LAB strains to metabolize lactose was tested by the strains cultivation in modified MRS, with 2% lactose as the single carbon source, at 37 °C for 24 h. Cultures obtained under the same conditions but on MRS with 2% glucose as the carbon source were used as the controls. The growth of the strains was estimated by viable cell counts, after plating 10-fold serial dilutions on MRS agar medium . The mean counts of log populations were compared by ANOVA (p < 0.05) using XLSTAT 2016.01.26192 (Addinsoft).