Introduction

Several studies support the use of probiotics as therapeutic agents against a number of diseases, in particular enteric disorders but also human diseases which are not apparently linked to the microbial gut composition, such as allergies and autoimmune diseases (Taylor et al. 2006; Roessler et al. 2008). The intestinal microbiota plays an important role in human health by providing nutrients that would not be otherwise available to the host, by stimulating the development of the immune system and by creating a barrier for the colonization of pathogens (Palmer et al. 2007; Roncaglia et al. 2011). A balanced composition of the gut microbiota has a positive influence on the health status of the host, both in adults and infants (Guarner 2006; Serikov et al. 2010).

Bifidobacteria are the major components of the microbiota of infants fed exclusively with breast milk and are commensal bacteria of the large intestine of humans and animals (Ventura et al. 2004). They are widely used as probiotics for therapeutic purposes considering their capabilities of colonizing the gastrointestinal tract and their long history of safe use (Sanders et al. 2010). As regards to Bifidobacterium spp. administration to newborns, in early 2002 the Food and Drug Administration has accepted as “safe” the use of Bifidobacterium lactis in formula milks and has given to this microorganisms the “Generally Regarded As Safe (GRAS)” status (Hammerman et al. 2006). Positive effects on the administration of B. lactis Bb12 on the reestablishment of a balanced composition of the gut microbiota were found on preterm, full-term infants and toddlers (Mohan et al. 2006). Among the different species belonging to this genus, Bifidobacterium breve appears to be one of the most used in infants. Li et al. (2004) concluded that the very early administration (at the first days of life) of a B. breve strain to low birth weight infants was useful in promoting the colonization of the bifidobacteria and the formation of a normal intestinal flora. A later study (Wang et al. 2007) evidenced that the administration of a B. breve strain to low birth weight infants reduces the production of butyric acid, which may be helpful in protecting these infants from necroting enterocolitis (Lin 2004). The beneficial effects of B. breve strain Yakult have also been evidenced in immunocompromized pediatric patients on chemotherapy (Wada et al. 2010). These young patients suffered from infectious complications, which were preceded by bowel colonization of pathogenic bacteria followed by translocation through the gut mucosa and systemic dissemination. Following probiotic administration, the use of antibiotics was lower and the gut habitation of anaerobes was enhanced.

Recent results evidenced that probiotics may be also useful for the treatment of minor gastrointestinal problems of newborns such as colics (Indrio et al. 2008; Savino et al. 2010). Infantile colics are characterized by an excessive and inconsolable crying without an identifiable cause in healthy newborns in the first 3 months of life. It affects up to 30 % of infants and it causes considerable stress and concern for parents. The pathogenesis of the conditions remains largely unknown although evidences suggested multiple independent causes, including modification of the gut microbiota with an increased number of gas-forming coliforms in colicky infants with respect to healthy controls (Savino et al. 2009). This finding suggested a role of coliform colonic fermentation in excessive intra-intestinal air production and pain typical of colic infants. The daily administration of Lactobacillus reuteri DSM 17938 in early breastfed infants was found to improve symptoms of infantile colics (Savino et al. 2010) and a recent study evidenced that other Lactobacillus spp. strains possess the ability to in vitro inhibit gas-forming coliforms and therefore have the potential of being used as probiotics for colics treatment (Savino et al. 2011). No studies have been presented up to now on the possibility of using Bifidobacterium spp. strains for this purpose, although, differently from Lactobacillus spp., Bifidobacterium spp. systemic infections upon administration in infants have never been reported (Hammerman et al. 2006).

In this work, a number of bifidobacteria strains were screened for desiderable functional properties for their application as probiotics against enteric disorders and, in particular, colic disease in newborns. The antimicrobial activity against coliforms isolated from colicky newborns and against bacteria most frequently cause of diarrhea in infants was studied, as well as their antibiotic susceptibility, adhesion, and toxicity towards gut epithelial cells and their ability to stimulate the metabolic activity and immune response of epithelial cells. The most interesting strains were also checked for trasmission of antibiotic resistance traits.

Materials and methods

Strains and culture conditions

Forty-six strains of Bifidobacterium spp. were included in this study; the majority of them derived from infant feces (Table 1). Forty-two of them were obtained from the Bologna University Scardovi Collection of Bifidobacteria (BUSCoB), available at the University of Bologna, whereas four were from the American Type Culture Collection (ATCC 15697, ATCC 15707, ATCC 15708, ATCC 27917). Thirty-six of the BUSCoB strains have been previously characterized with phenotypic analyses and by means of the electrophoretic pattern of transaldolase and 6-phosphogluconic dehydrogenase (Scardovi et al. 1979). The remaining six strains (B7710, B7740, B7840, B7947, B7958, B8452) were isolated from preterm newborn feces and characterized as members of the Bifidobacterium genus by means of phenotypic analyses and the fructose 6-phosphate phosphoketolase assay (unpublished results). Bifidobacterium strains were cultivated in tryptone, peptone, yeast extract medium (TPY prepared according to Biavati and Mattarelli 2006) and incubated at 37 °C under anaerobic conditions using an anaerobic atmosphere generation system (Anaerocult A, Merck, Darmstadt, Germany). Four strains have been deposited to the DSMZ culture collection with the following collection numbers: DSM 24706 (B. breve B632), DSM 24707 (B. breve B2274), DSM 24708 (B. breve B7840), DSM 24709 (Bifidobacterium longum subsp. longum B1975).

Table 1 List of the 46 Bifidobacterium spp. strains used in this study and evaluation of their antimicrobial activity against four antagonistic strains (E. coli, E. cloacae, K. pneumoniae, and S. enteriditis) expressed as radius (in centimeter) of the inhibition halos obtained on TPY plates in the agar spot test
Full size table

The strains used as antagonistic microorganisms were: Escherichia coli ATCC 11105, Salmonella enteriditis M94 strain and Clostridium difficile M216 strain (both isolated from hospitalized patients and available at BuSCoB), Campylobacter jejuni CIP 70.2T (from the Collection de l’Institut Pasteur, Paris, France) and two gas-forming coliforms isolated from feces of colicky infants, Klebsiella pneumoniae GC23a and Enterobacter cloacae GC23a (Savino et al. 2011). The E. coli, S. enteriditis, K. pneumoniae, and E. cloacae strains were cultivated in nutrient broth (NB; Oxoid, Basingstoke, UK) aerobically at 37 °C. C. difficile M216 strain was grown in Brain Heart Broth (Merck) and incubated under anaerobic condition at 37 °C; C. jejuni CIP 70.2T strain was cultured ad described in Santini et al. (2010).

Genetic typing of the strains

Enterobacterial repetitive intergenic consensus PCR

Total DNA was extracted from 10 mL of overnight pure cultures and purified using Wizard Genomic DNA purification kit (Promega, Madison, WI, USA). Enterobacterial repetitive intergenic consensus PCR (ERIC-PCR) patterns of Bifidobacterium strains were obtained following the procedure described by Ventura et al. (2003). Primers ERIC-1 (5′ATGTAAGCTCCTGGGGATTCAC-3′) and ERIC-2 (5′AAGTAAGTGACTGGGGTGAGCG-3′) were used. The 20 μL reaction mixture contained 10 μL of HotStart Taq Plus Master Mix Kit (Qiagen, West Sussex, UK), 1 μM of each primer, 1.5 mM MgCl2 (Qiagen). PCR reactions were run in a Veriti Thermal Cycler (Applied Biosystem, Foster City, CA, USA). The reference strains used in this study were: Bifidobacterium pseudocatenulatum ATCC 27917T, Bifidobacterium catenulatum ATCC 27539T, B. breve ATCC 15700T, Bifidobacterium bifidum DSM 20456T, B. longum subsp. longum ATCC 15707T, and B. longum subsp. infantis ATCC 15697T.

PCR with genus-specific and specie-specific primers

Bifidobacterium genus-specific PCR was performed on total DNA using 16S rDNA-targeted primers Bif64-f and Bif662-r (Satokari et al. 2001). Species identification was carried out using species-specific PCR primers described by Matsuki et al. (1999).

In vitro inhibition of antagonistic strains

Agar spot test using living cells

To assess the antimicrobial activity of Bifidobacterium spp. strains against selected bacteria (E. coli ATCC 11105, S. enteriditis M94, K. pneumoniae GC23a strain, and E. cloacae GC 6a were used for all the 46 strains, whereas C. jejuni CIP 70.2T and C. difficile M216 only for 16 selected strains) the protocol described by Santini et al. (2010) was employed. Briefly, Bifidobacterium spp. strains were spotted on TPY plates and, after strain growth, the plates were overlaid with a soft agar medium inoculated with each pathogen strain. Inhibition was evaluated after 24 h measuring the radius of the inhibition halo around the Bifidobacterium spot. Each assay was performed in triplicate.

Antimicrobial activity of Bifidobacterium spp. culture supernatants

This assay was performed with the 16 strains showing the most interesting antimicrobial activity in the previously described assay and, as a negative control, with a Bifidobacterium strain not showing any antagonistic activity in the spot agar assay (B7710). Cell-free supernatants were obtained by centrifuging TPY bifidobacteria o.n. cultures (15,000×g, 20 min, 4 °C) followed by filtration through a 0.22 μm pore size cellulose acetate filter. An aliquot of the supernatant was adjusted to pH 7. The antagonist strains used in this assay were: E. coli ATCC 11105, S. enteriditis M94, K. pneumoniae GC23a, and E. cloacae GC23a. The antagonistic strains were grown in NB until absorbance at 600 nm (A600) of 0.9 and used to inoculate 96-well plates. Each well contained: 100 μL of double concentrated NB, 25 or 50 μL of Bifidobacterium spp. cell-free supernatant (both neutralized and non-neutralized), corresponding to a v/v percentage of 12.5 and 25, respectively, and water to 200 μL of total volume. 1 % v/v inoculum of the antagonistic strain was added. Positive controls were prepared by using 50 μL of fresh NB without any supernatants. Plates were incubated aerobically at 37 °C for 22 h; A620 was periodically evaluated in a multiwell plate spectrophotomer (Multiskan, Thermo Electron, Oy, Vaanta, Finland).

Antibiotic resistance profiles

Minimal inhibitory concentration

Minimal inhibitory concentration (MIC) for 12 antibiotics was determined with the microdiluition assay in 96 well plates as described in Santini et al. (2010) on the 16 Bifidobacterium strains. Twelve antibiotics were employed, eight of which were suggested in the most recent European Food Safety Authority (EFSA) guidelines (EFSA 2008), i.e., tetracycline, cefuroxime, kanamycin, chloramphenicol, vancomycin, ampycillin, streptomycin, and erythromycin whereas the other four were antibiotics widely used in infant therapy (cefuroxime, amoxicillin, ceftriaxone, and clarithromycin). Antibiotics were used in the concentration range 1–512 μg/mL. Growth or inhibition of the strains was determined by measuring the A620 at regular time intervals for a total incubation of 24 h at 37 °C. Antibiotics were all from Sigma-Aldrich, Milan, Italy.

Screening of resistance genes

The presence of known antibiotic resistance genes was determined by PCR using specific primers: aph (3″)-I, aph (3″)-II, aph (3″)-III coding for kanamycin and neomicine resistance (Ouoba et al. 2008), aadA, aadE, ermA, streptomycin, and erythromycin-resistance genes (Ouoba et al. 2008), tet(M), tet(O), tet(W) coding for tetracycline resistance (Masco et al. 2006) and blaCTX-M-g1, blaCTX-M-g2, ß-lactam resistance (Van Hoek et al. 2008). The amplification conditions are from Ouoba et al. (2008). The annealing temperature varied in the range 45–64 °C, depending on the primer. Lactobacillus casei L9 was used as positive control for aph(3″)-III, aadA, aadE genes whereas Bifidobacterium adolescentis DSM 20087 was the positive control for tet(W) gene. PCR products were separated by electrophoresis on 1.5 % agarose gel.

Plasmid detection

Pure yield plasmid Miniprep System kit (Promega) was used to extract and purify plasmid DNA from the selected 16 Bifidobacterium strains. B. longum B2399, which was known to possess two plasmids, was used as positive control for plasmid DNA extraction. Plasmids were separated after electrophoresis on a 0.7 % agarose gel.

Evaluation of the transferability of the antibiotic resistance traits

Four Bifidobacterium strains (B632, B1975, B2274, B7840) were used as donor strains, whereas Bifidobacterium animalis ATCC 27536, B. longum subsp. suis PCD733B (Santini et al. 2010), three Bifidobacterium strains from this study (B1412, B7840, B632), Lactobacillus plantarum PCS22 (Nissen et al. 2009), and Enterococcus faecium PCD71 (Santini et al. 2010) were used as recipient strains. The transferability of the antibiotic resistance traits was assayed following the protocols of Lampkowska et al. (2008) and Ouoba et al. (2008). A scheme of the experiments is outlined in Table 2. Donor and recipient strains were cultivated to mid-exponential growth phase in liquid medium with appropriate antibiotics, and then mixed in 1:1 ratio in a final volume of 200 μL. The mixture was inoculated into 10 mL of TPY broth, which was incubated anaerobically for 24 h at 37 °C. Cells were then harvested by centrifugation (10 min at 6,000 rpm), resuspended in 1 mL of phosphate-buffered saline (PBS) and plated on donor- and recipient-selective agar plates and on selection plates, in which only recipient strains having acquired the antibiotic resistance can grow (Table 2). The same plates were used to estimate the frequency of spontaneous mutations in the recipient strain. To counter select lactic acid bacteria, the plates were incubated in aerobic conditions.

Table 2 Evaluation of the transferability of the antibiotic resistance traits from B. breve B632, B2274, and B7840 and B. longum B1975 to selected recipient strains
Full size table

In vitro interaction between Bifidobacterium strains and human cells

Growth and maintenance of cell line

The cell lines used were: small intestinal human epithelial cell line H4, derived from human fetal tissue and supplied by Massachusetts General Hospital (Prof. WA Walker), and human blood monocytes/macrophages, referred to as TLT cell line (Cencic and Langerholc 2010). Cells were routinely grown in Dulbecco modified essential medium (DMEM) as described in Nissen et al. (2009). To perform biological assays, the cells were seeded in 96-well plates at the concentration of 1 × 106 cells/mL and incubated for 24 h at 37 °C in 5 % CO2. When cell monolayers were obtained, the 16 selected Bifidobacterium strains, grown in TPY and suspended in DMEM at the concentration of 1 × 108 CFU/mL, were inoculated in each well at the concentration of 107 CFU/mL. In most of the assays described, the well-known probiotic strain Lactobacillus rhamnosus GG (LGG) was used to compare the results obtained. Reagents used were purchased from Sigma-Aldrich.

Cytotoxicity assays

Bifidobacterium strains were inoculated on H4 and TLT cell monolayers as described above and plates were incubated in 5 % CO2 at 37 °C for 90 min. Unbound bacteria were eliminated by washing the cell layers three times with PBS. One hundred microliters of DMEM supplemented with l-glutamine (2 mM) was added to each well, and plates were incubated for 24 h. Cell viability was measured with crystal violet staining, measuring absorbance at 595 nm (A595), and compared with the A595 of non treated cells (i.e., cells not exposed to probiotics).

Adhesion assay

H4 and TLT cell monolayers were washed with PBS and Bifidobacterium strains were applied at a concentration of 9.4 log (CFU/m2). Plates were incubated for 90 min at 37 °C. The monolayers were washed with PBS, then cells with adherent bacteria were harvested with trypsin and the number of bacteria was counted on TPY agar plates. Results of attached bacteria cells were expressed as percentage of adherent bacterial cells compared to initial inoculum.

Mitochondrial activity assay

The metabolic activity of H4 and TLT cell lines after exposure to Bifidobacterium strains was measured by evaluating their mitochondrial function as described by Ivec et al. (2007). After bacterial exposure to cell monolayers and incubation, a solution of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) in DMEM was added to each well and incubated for 75 min. Solubilization of MTT reduction product (i.e., formazan) was achieved by addition of 0.04 % HCl in isopropanol and formazan was quantified at 650 nm. Results are expressed as: (A650 of treated wells − A650 of untreated wells)/A650 of untreated wells × 100.

Determination of reactive oxygen species; NO, H2O2

NO released by cell monolayers after stimulation of Bifidobacterium strains was determined by measuring the accumulation of nitrate using a modified Griess reagent (Sigma-Aldrich), according to the Griess reaction (Pipenbaher et al. 2009). The release of H2O2 was determined as described by Nissen et al. (2009). Constitutive H2O2 production by bifidobacteria was evaluated by incubating bifidobacteria in DMEM; the amount of H2O2 produced by bifidobacteria was subtracted from the amount produced by the cells.

Dot blot for interleukin 6

Interleukin-6 (IL-6) in supernatants of H4 and TLT cells after probiotc exposure was detected using the dot blot technique as described by Ivec et al. (2007). Briefly, supernatants were blotted onto nitrocellulose membrane (Pierce, Rockford, USA) with a Bio-Rad Dot Blot apparatus (Bio-Rad Laboratoires, Hercules, USA). After primary and the secondary antibodies addition, proteins were visualized with the supersignal West Pico chemioluminescent substrate system (Pierce) and BiomaxMR-1 film (Sigma-Aldrich). Supernatants of monolayers not exposed to bacteria were used as negative control, whereas L. casei Shirota and LGG were used as positive controls.

Experimental design, statistical analysis, and strain selection criteria

For the different trials, the adopted experimental scheme was a fully randomized design. All the tests were performed in triplicate. Data on spot agar tests, cytotoxicity assay, adhesion test, mitochondrial activity test, and reactive oxygen species (ROS; NO, H2O2) production were subjected to one-way analysis of variance (ANOVA) by using the GLM procedure of the SAS statistical package. Means were subjected to Fisher’s test (SAS 1988). When treatments were significant according to Fisher’s test, corresponding means were differentiated by the SNK multiple range test at the 0.05 level of probability.

The correspondence analysis (CA) was applied to the fingerprinting pattern obtained from ERIC-PCR of Bifidobacterium reference strains and investigated strains. CA is a statistical method for visualizing the association between levels of a two-way contingency table (Benzecri 1992). Banding profiles were scored as presence/absence of individual fragments in each investigated strain. The contingency table was analyzed by CA module of Statistica Software (ver. 7.1, StatSoft, Tulsa, OK, USA). Plotting the first two dimensions of the coordinates of cases (ERIC-PCR bands) and variables (strains) gave a global view of the correspondence among reference and investigated strains, and band patterns. The first and second dimensions explained 34 and 28 % of the total variability, respectively.

A first strain selection was based on antimicrobial activity against E. coli, S. enteriditis, K. pneumonia, and E. cloacae allowing the choice of the 16 best performing strains out of the original 46 strains. Among the 16 strains, four bifidobacteria were selected on the basis of a synthetic index, calculated as follows: the outputs of different analyses (spot agar tests, antibiotic resistance or sensitivity assay, cytotoxicity test, adhesion assay, mitochondrial dehydrogenase activity, NO and H2O2 production) were transformed into relative percentages by giving the 100 value to the strain showing the best performance in each test. A correction factor of 0.5 was given to the mitochondrial dehydrogenase activity, NO and H2O2 production tests, in order to give more importance to the other parameters, which are defined in the EFSA guidelines (EFSA 2005). IL-6 production was not considered in this evaluation as it is not a quantitative test. These procedures allowed to select four strains which were checked for the transferability of the antibiotic resistance traits to other gut bacteria and were then deposited to the DSMZ culture collection.

Results

Antimicrobiobial activity with the spot agar test

The results obtained with the 46 Bifidobacterium strains against E. coli, E. cloacae, K. pneumonia, and S. enteriditis evidenced antimicrobial activity to varying degrees (Table 1): three strains (B2531, Re11, B7710) did not show any inhibition halo against all the indicator strains, 27 strains showed inhibition halo’s radius not higher than 0.5 cm, whereas 14 strains (Re12, B632, B1412, B1975, B2021, B2055, B2091, B2101, B2150, B2192, B2195, B2274, B7840, B7958) showed an average inhibition halo’s radius higher than 0.5 cm (Table 1). The elaboration of the results with the ANOVA test allowed to indicate these 14 strains as the most performing; however, we decided to include two more strains (B7947 and B8452) for further studies considering their high antimicrobial activity against E. coli, which is the most abundant coliform in the infant gut, and their potential interest as preterm isolated strains.

These 16 strains were then assayed against C. jejuni and C. difficile as antagonistic microorganisms. The results obtained (Table 3) evidenced that all Bifidobacterium strains except for B2101 were capable of inhibiting both antagonistic microorganisms. Among them, eight strains (B632, B1412, B1975, B2055, B2192, B2274, B7840, B8452) showed a marked activity against the two pathogens.

Table 3 Antagonistic activity of 16 selected Bifidobacterium strains against C. jejuni LMG8841 and C. difficile M216 expressed as average radius (in centimeter) of the inhibition halos obtained on TPY plates in the agar spot test; mean values followed by different letters (in brackets) are statistically different at P < 0.01 for C. jejuni and P < 0.001 for C. difficile
Full size table

Antimicrobiobial activity of Bifidobacterium culture supernatants against coliforms and S. enteriditis

The culture supernatants, both neutralized (referred to as neutralized culture supernatant; NCS) and non-neutralized (referred to as CS), of the 16 Bifidobacterium strains showing the highest antimicrobial activity (listed in Table 3), plus one strain (B7710) as negative control, were used for evaluating the inhibiting activity towards E. coli ATCC 11105, S. enteriditis M94, K. pneumoniae GC23a, and E. cloacae GC6a. The majority of Bifidobacterium supernatants were capable of exerting their inhibiting activity only when non-neutralized (data not shown), whereas the inhibitory activity of four strains (B632, B1975, B2274, and B7840) was evidenced both with CSs and NCSs. Figure 1 shows details of the experiments performed with B632: the inhibitory activity towards E. coli and S. enteriditis was clearly evident in the early hours of incubation (Fig. 1a, b) with no differences in the use of NC and NCS; the inhibitory activity towards E. cloacae and K. pneumoniae was also present although less marked with respect to the other target strains and, moreover, it was more evident when the non-neutralized supernatants was used. The profiles obtained with B1975, B2274, and B7840 were quite similar to that of B632, showing a higher inhibitory activity of NCS towards E. coli and S. enteriditis (data not shown). No inhibitory activity against all the antagonistic strains was evidenced by the B7710 strain (data not shown).

Fig. 1
figure 1

Effect of culture supernatants (CS) and neutralized culture supernatants (NCS) of E. coli ATCC 11105 (a), S. enteriditis M94 (b), E. cloacae GC6a (c), K. pneumoniae GC23a (d) on the growth of B. breve B632 (filled diamond control with 50 μl NB, filled square 25 μl CS, empty square 50 μl CS, multiplication symbol 25 μl NCS, empty circle 50 μl NCS)

Full size image

Genotypic and phenotypic characterization of the strains

The selected 16 strains were identified and classified at the species level using the ERIC-PCR approach proposed by Ventura et al. (2003). An accurate clustering and identification of the strains was achieved comparing ERIC-PCR banding patterns of the strains used in this work with those retrieved from reference strains. The CA and the scatterplot projections of variables (strains) and cases (ERIC-PCR bands) on the first two dimension evidenced four main clustering groups corresponding to different type strains (Fig. 2). One group was formed by the B. pseudocatenulatum type strain (ATCC 27917T) and the B8452 strain: it was the most divergent cluster due to the exclusive presence of eight DNA fragments. A second main group clustered with the B. longum strains including the B. longum subsp. longum and the B. longum subsp. infantis type strains: six strains clustered close to longum subspecies and were therefore identified as B. longum subsp. longum (B1412, B1975, B2055, B2101, B2192, B7958, Re12). A third cluster grouped with the B. breve type strain (B632, B2021, B2150, B2274, B2195, B7840, B7847). Finally, the B2091 strain clustered close to the B. bifidum type strain.

Fig. 2
figure 2

Relationships established among Bifidobacterium strains by means of CA based on ERIC-PCR band patterns. Numbers correspond to fingerprinting DNA fragments obtained after agarose gel electrophoresis following ERIC-PCR

Full size image

To confirm the results obtained with ERIC-PCR, the strain identification was compared with species-specific standard PCR. 16S-targeted species specific primers allowed to confirm the Bifidobacterium identification at the species level, except for the B. pseudocatenulatum strain which was only inserted in the “catenulatum group” with this technique.

Antibiotic resistance profiles

MIC and screening of resistance genes

The resistance or sensitivity of the selected 16 strains to 12 antibiotics and the relative MIC values obtained are shown in Table 4. All the strains were found to be sensitive to chloramphenicol, erythromycin, vancomycin (apart from B2091), and gentamycin according to most recent EFSA guidelines (EFSA 2008). Moreover, most of the strains were sensitive to tetracycline except a few strains (B2055, B2150, B2195, B2274, B7840, and B7958). All the strains were resistant to ampicillin and the majority of them to kanamycin (except B1412). Nine strains out of 16 were resistant to streptomycin. Regarding cefuroxime, ceftriaxone, and clarithromycin, whose breakpoints are not present in the mentioned EFSA guidelines, the majority of the strains presented low MIC values so they can be considered sensitive to them. All the strains but one (B632) presented a high MIC value for amoxicillin.

Table 4 MIC of various antibiotics of the selected strains
Full size table

The screening of the resistance genes via PCR amplification of known genes in the 16 strains of bifidobacteria allowed to detect the tet (W) amplicon only in two (B2274 and B7840) of the six tetracyclin resistance strains, whereas none of them was positive to tet(M) and tet (O). Only three strains (B1975, B2192, and B7947) out of the 15 resistant to kanamycin were positive to aph(3″)-III amplification, whereas aph(3″)-I and aph(3″)-II were not amplified in any strain. With regard to the β-lactam-resistance determinants, almost all the tested strains carried blaCTX-M-g1 apart from B2021, B2101, B2150, B2274, and B7958. No strains were found to be positive to the amplification of the aadA and aadE streptomycin-resistance genes. The presence of plasmids was detected only in B. longum subsp. longum B2192 strains, which possessed two plasmids (data not shown).

In vitro interaction between Bifidobacterium strains and human cells

Cytotoxicity and adhesion

Cytotoxicy assays showed that a number of strains (B1412, B2021, B2101, B2150, B2192, B7947, B7958, and Re12) at the bacterial concentration of 107 CFU/mL after 90 min incubation exherted a cytotoxic effect to the H4 monolayers higher than the control strain LGG (p < 0.05). Referring to TLT monolayers, only B7958 strain was more citotoxic than LGG (p < 0.05; data not shown). A number of strains showed a low reduction of viability of TLT cells when compared to untreated cells, although not statistically significant; however, it has to be considered that a direct contact between the content of the intestinal lumen with macrophages is not an in vivo real condition. On the contrary, three breve strains (B632, B2274, and B7840), B. longum B2055, and B. pseudocatenulatum B8452 showed positive effects on both cell monolayers, in particular B632 and B2274 seem to increase the viability of cells after the exposure (data not shown).

All strains showed a good ability to adhere to polarized human epithelial H4 cells and TLT macrophages (data not shown). B. breve B632, B. pseudocatenulatum B8452, and B. longum B2192 showed a higher attachment to H4 cells with respect to the control LGG strain whereas the majority of Bifidobacterium strains presented an adhesion capability comparable to LGG or slightly higher.

Stimulation of cell activity: mitochondrial activity, production of ROS and of interleukin

The results of the mitochondrial activity enhancement with the MTT assay are shown in Fig. 3. The mitochondrial dehydrogenase activity of H4 cell lines increased after exposure to B. breve B632 and B2195 strains at the concentration of 1 × 107 CFU/mL. B632 was the only strain able to strongly stimulate the activity of mitochondrial dehydrogenase of macrophages, whereas only a slight enhancement was obtained with B2021and B2274, although not statistically significant. Therefore, the percentage of stimulation obtained for most of the Bifidobacterium strains, such as for the reference strain LGG was, generally, negligible apart from B632. However, the values obtained for all the strains were not as negative as those obtained with the potential pathogens S. enteriditis and E. coli.

Fig. 3
figure 3

Effect of 16 Bifidobacterium spp. strains on the mitochondrial dehydrogenase activity of H4 and TLT cell monolayers. Results are expressed as the average of three independent experiments (±SD). LGG strain is used as a reference strain for the evaluation of the probiotic effect on the cells, E. coli and S. enteriditis are used to evidence the effects of potential pathogen microorganisms on the cells

Full size image

Among the 16 Bifidobacterium strains applied at the concentration of 107 CFU/mL on H4 cell line, only B2274 induced an increase of NO production statistically higher than LGG strain (Table 5). Except for B632, B2091, and B7840 strains, the remaining Bifidobacterium strains exhibited a lower stimulation effect on NO production than LGG strain. As concerns the stimulation of NO production on TLT cell line, the strongest induction was observed for B1412 strain (approximately five times higher with respect to LGG strain). A moderate increase of NO production, comparable with that observed for LGG strain, was reported for B2091, B2274, B7840, B7947, and B7958 strains. Twelve out of 16 Bifidobacterium strains stimulated H2O2 production on H4 cell lines, while all Bifidobacterium induced an increase of hydrogen peroxide on TLT cell lines (Table 5). On H4 cell line, the B1412, B2021, B2055, B2150, and B2195 strains induced an increase of H2O2 production statistically higher than LGG strain. In contrast, only one strain (B7947) was more efficacious in stimulating H2O2 production of TLT cell line than LGG. E. coli and S. enteriditis, used as potential entheropatogens, induced the strongest stimulation of ROS production (i.e., nitric oxide and hydrogen peroxide) in both H4 and TLT cell lines (Table 5).

Table 5 ROS production (nitric oxide, hydrogen peroxide) by different intestinal cell lines (H4, TLT) as a function of the stimulation from different bacterial strains
Full size table

Dot blot was performed to determine the presence of pro-inflammatory cytochine IL-6 in cell-free culture supernatants after exposure of cells to the bacteria for 24 h. A notable production of IL-6 was achieved with H4 and TLT cells using all bacteria except for B. longum subsp. longum B1412. The highest IL-6 production was noted for B632 and B2055 (Fig. 4).

Fig. 4
figure 4

Dot blot of IL-6 detection. The experiment was performed with 16 Bifidobacterium spp. strains. LGG and L. casei Shirota were used as positive controls; negative controls do not have any applied Bifidobacterium strain (H4 or TLT untreated cells). 1a B632/H4, 1b B1412/H4, 1c B1975/H4, 1d B2021/H4, 1e B2055/H4, 1f 2101/H4. 2a B2150/H4, 2b B2192/H4, 2c B2195/H4, 2d B2274/H4, 2e B7840/H4, 2f 7958/H4, 3a B8452/H4, 3b Re12/H4, 3c B2091/H4, 3d B7947/H4, 3e LGG/H4, 3f LGG /H4, 4a L.casei Shirota/H4, 4b L.casei Shirota/H4, 4c neg control/H4, 4d neg control/H4, 4e neg. control/TLT, 4f neg control/TLT, 5ª B1412/TLT, 5b B2091/TLT, 5c B1975/TLT, 5d B2021/TLT, 5e B2055/TLT, 5f 2101/TLT, 6a B2150/TLT, 6b B2192/TLT, 6c B2195/TLT, 6d B2274/TLT, 6e B7840/TLT, 6f B632/TLT, 7a B7947/TLT, 7b B7958/TLT, 7c B8452/TLT, 7d Re12/TLT, 7e LGG/TLT, 7f L. casei Shirota/TLT

Full size image

Selection of the best probiotic strains with the use of a synthetic index

The outputs of all the analyses described above were transformed into relative percentages as described in the “Materials and methods” section. The matrix thus completed allowed to calculate a synthetic index (see Supplementary material, Table S1). The strains with the highest synthetic index were selected, i.e., B632, B2274, and B7840. In addition, B1975 strain was also chosen for further studies because of its considerable synthetic index and its high antimicrobial activity against coliforms and potential pathogens.

Transferability of antibiotic resistance traits of selected strains

The capability of B632, B1975, B2274, and B7840 of transferring the antibiotic resistance traits to Bifidobacterium spp. strains and lactic acid bacteria (L. plantarum PCS22, L. casei L9, and E. faecium PCD71) was studied according to the scheme proposed in Table 2. No recipient strains could receive the antibiotic resistance trait from all the donors and, in addition, no spontaneous mutants of the four donor strains were detected (data not shown).

Discussion

Probiotics are increasingly being used for the treatment of diseases and minor gastrointestinal problems in infants. A recent study has evidenced positive effects on infant colics after treatment of newborns with a L. reuteri strain (Savino et al. 2010), whereas no studies have been performed up to now regarding the use of bifidobacteria for this purpose. This work was therefore aimed at the characterization of Bifidobacterium spp. strains possessing in vitro capabilities of inhibiting the growth of pathogens typical of the infant gastrointestinal tract without exerting toxic activities on the gut epithelium and harmful effects to the host.

The majority of Bifidobacterium spp. strains used in this work derives from infant feces (Scardovi et al. 1979), i.e., from the source which constitutes the target population of the potential probiotic (Arboleya et al. 2011). Preterm isolates were also included considering the high stressing environment of the preterm infant gut, which shows a higher prevalence of C. difficile compared with term infants (Penders et al. 2006). Sixteen strains out of the 46 assayed in this study were capable of effectively contrasting the growth of pathogens which are the main cause of infectious diarrhea of bacterial origin in infants, such as E. coli, S. enteriditis, C. difficile, and C. jejuni (Rowland 2008; Van Niel et al. 2002). Moreover, the same Bifidobacterium strains showed marked antimicrobial activity against gas producing coliforms isolated from stools of colicky infants. Considering that gas forming coliform concentration is higher in colicky infants with respect to healthy controls (Savino et al. 2009, 2011), the results obtained are interesting in the perspective of developing a probiotic based therapy for colic treatment in newborns. The number of Bifidobacterium strains showing antimicrobial activity was lower by using NCS. However, this experiment pointed out that at least in some strains, such as B. breve B632, the inhibitory activity may not result only from the production of acidic metabolites, but also from the action of other cell excreted metabolites such as bacteriocins. This result represents an interesting starting point for further studies aimed at the characterization of inhibitory molecules in this strain.

A clear taxonomic identification is necessary for the use of a probiotic strain in humans (Arboleya et al. 2011). The genotypic characterization approach used in this work allowed to cluster the majority of the16 strains into two species, i.e., B. breve and B. longum subsp. longum, whereas only two strains were clustered within the B. pseudocatenulatum and B. bifidum species. The results of this analysis confirm that B. pseudocatenulatum and B. catenulatum, which are indistinguishable by standard PCR, can be easily and quickly distinguished via the ERIC-PCR approach (Ventura et al. 2004). The strain B1412, which has been previously identified as B. longum subsp. infantis, has now been included in the longum subspecies.

According to the most recent EFSA guidelines (EFSA 2008), the spread of resistance to antimicrobials in bacteria requires the examination of the sensitivity/resistance to a number of antibiotics for potential probiotic strains as well as the risks of the resistance traits to be transferred to other bacteria. Except for a number of antibiotics for which the majority of the assayed Bifidobacterium strains are resistant, such as ampicillin, kanamycin, and amoxicillin or sensitive, such as chloramphenicol, erythromycin, and vancomycin, there is a great variability among strains also belonging to the same species, as already evidenced in the literature (Masco et al. 2006; Ammor et al. 2008). Intrinsic resistance to aminoglycosides such as streptomycin and kanamycin is commonly present in bifidobacteria (D’Aimmo et al. 2007); however, information on streptomycin-resistance genes is limited for Bifidobacterium strains (Kiwaki and Sato 2009). Aminoglycoside-resistance genes, including aadE which was evidenced in a B. longum strain (Ouoba et al. 2008), were not found in the genome of the assayed strains as well as the kanamycin-resistance genes aph (Ouoba et al. 2008). Conversely, all the strains were sensitive to the aminoglycoside gentamycin in agreement with the data present in the literature on bifidobacteria (Ammor et al. 2008). The MICs for tetracycline obtained for most of the tested strains suggested the presences of tetracycline-resistance genes. Tet genes, coding for ribosomal protection protein, are involved in resistance to tetracycline and tet(M) and tet(W) have been exclusively found in bifidobacteria (Aires et al. 2007). However, only two of the assayed strains, B. breve B2274 and B7840, presented the tet(W) amplicon. Bifidobacteria are usually susceptible to β-lactams, such as ampicillin and amoxicillin (Ammor et al. 2008; Matto et al. 2007), whereas the majority of the strains considered in this analysis are resistant. Consequently, resistance to some β-lactams can be considered an acquired resistance and therefore has the potential for lateral spread (EFSA 2008). There is very little information on the mechanisms responsible for horizontal gene transfer in anaerobic gut bacteria like bifidobacteria; however, the most widespread is the conjugation of plasmids carrying the antibiotic resistance genes. All the 16 Bifidobacterium spp. strains potentially considered interesting for the aims of this study did not carry any plasmids, although plasmids have been identified in several bifidobacteria species and strains (Ventura et al. 2008). However, other genetic mechanisms can influence the likelihood of genetic transfer (Burrus and Waldor 2003), such as transposons, which can carry resistance genes and can move from chromosome to plasmids and vice versa, thereby increasing the mobility of these genes. Therefore, the transferability of the antibiotic resistance traits to Bifidobacterium spp. strains and lactic acid bacteria was assayed in the four strains which were selected as the most interesting ones for the aim of this study (B. breve B632, B2274, B7840 and B. longum subsp. longum B1975) and the results allowed to conclude that there was no transfer of the antibiotic resistances neither to the bifidobacteria nor to the lactic acid bacteria assayed.

Finally, adhesion and cytotoxic effects to human cells of the 16 putative probiotic strains was evaluated using nontumorigenic cell lines, which have already been used as a reliable in vitro method for the selection of lactic acid bacteria with potential probiotic properties (Maragkoudakis et al. 2010; Nissen et al. 2009), but have never been tested with Bifidobacterium spp. strains. It is well assessed that the phenotype of tumorigenic cell lines traditionally used for this purpose distinguishes them profoundly from the normal gut epithelium (Tremblay and Slutsky 2007). The ability to adhere to the intestinal epithelium is one of the most important features as it allows to persist in the colon preventing the elimination by peristalsis and the adhesion of pathogenic bacteria. All the tested bacteria showed a good adhesion to both cell types, epithelial cells, and macrophages. Anyway adhesion cannot singly determine the biological activity of these putative probiotic strains. It is a combination of different factors which determines epithelial integrity, viability, and immunoresponse. Treatments with B. breve B632, B2274, and B7840, B. longum B2055, B. pseudocatenulatum B8452 manifested no cytotoxicity over H4 and TLT cell lines at the concentration of 107 CFU/mL. In addition, B. breve B632 and B2274 at the same concentration were able to increase the metabolic activity of cell mitochondria. These results indicate that these strains are not harmful when exposed to a healthy intestine. Most of the tested strains increased the production of ROS in small intestinal epithelial cells and in macrophages. The ability of probiotic bacteria to induce NO secretion from intestinal epithelium may offer a significant contribution to prevent the enteric pathogens from infecting the host. The ability to stimulate NO production in eukaryotic cells is not a common ability of the genera Lactobacillus and Bifidocbacterium, but rather of individual strains (Pipenbaher et al. 2009). Furthermore, most of the bacterial strains tested induced H2O2 release in both types of cells. Moderate production of H2O2 and NO induced by probiotics could have a beneficial effect in maintaining a balance and increasing resistance to infections. However, it should be noted that high concentration of H2O2 and NO, as displayed by potential enteropathogens such as E. coli and S. enteriditis (Table 5), can cause tissue injury, disseminated intravascular coagulation, and shock (Park et al. 1999). Last but not least, there is extensive evidence that cytokines play pivotal roles in host defence, inflammatory response, and autoimmune disease (Park et al. 1999). Therefore, IL-6 production is likely to be a good indicator of a degree of macrophages and endothelial cells activation. In the present work, exposure of H4 and TLT cells to Bifidobacterium and Lactobacillus strains resulted in marked increase of IL-6 production.

In conclusion, the large array of aspects examined in this study has allowed the identification of four Bifidobacterium strains, B. breve B632 (DSM 24706), B2274 (DSM 24707), B7840 (DSM 24708), and B. longum subsp. longum B1975 (DSM 24709), as potential probiotics for the treatment of enteric disorders in newborns such as infantile colics or as preventive agents for infantile diarrhea of bacterial origin. They all possess strong antimicrobial activity against coliforms and other pathogenic bacteria, do not possess transmissible antibiotic resistance traits and are not cytotoxic for the gut epithelium. Studies are currently being performed in order to develop suitable ways of administering the selected probiotic strains to newborns with the aim of planning a validation clinical trial.