B. longum subsp. longum KABP042 (CECT 7894) was originally isolated from feces of a healthy breast-fed baby and P. pentosaceus KABP041 (CECT 8330) from feces of a healthy child. Both strains belonged to our own collection (Cuñé Castellana et al., 2015). Lacticaseibacillus rhamnosus GG (ATCC 53103), B. longum subsp longum ATCC 15707, Lacticaseibacillus paracasei ATCC 334, Streptococcus pyogenes CECT 191, and Bacillus cereus CECT 5144 were obtained from the American Type Culture Collection (ATCC; Manassas, VA) or the Spanish Type Culture Collection (CECT; Valencia Spain) and used as control strains as indicated. Enteropathogenic Escherichia coli CECT 729 (EPEC), Klebsiella pneumoniae DSM 11678, Salmonella enterica subsp. enterica serovar typhimurium LT2 (ATCC 700720), Staphylococcus epidermidis CIP 81.55T, Staphylococcus aureus CIP 107860 and Enterococcus faecalis ATCC 29212 were obtained from the ATCC, CECT, German Collection of Microorganisms and Cell Cultures (DSM; Braunschweig, Germany) or Institute Pasteur Collection (CIP; Paris, France) and used as indicator strains in the antimicrobial activity assays.

P. pentosaceus and lactobacilli strains were routinely grown in MRS (De Man, Rogosa, and Sharpe) broth or agar (Scharlab, Barcelona, Spain) for 18–48 h at 37°C in chambers with anaerobic sachets (Anaerogen, Thermo Scientific, Barcelona, Spain). Bifidobacteria strains were grown in the same conditions except that MRS was supplemented with 0.1% (w/v) cysteine-HCl (Scharlab) (MRScys). S. pyogenes, enterobacteria, staphylococci and E. faecalis strains were cultured in tryptic soy broth (TSB) or agar (TSA) (Scharlab) for 18–24 h aerobically at 37°C as indicated. B. cereus CECT 5144 was grown in similar conditions on nutrient gelatin agar (per L: peptone 4 g, yeast extract 1 g, gelatin 40 g and agar 15 g).

The antimicrobial susceptibility profile of probiotic strains to a panel of 8 antibiotics required by European Food Safety Authority (EFSA, 2018) was investigated (Table 1). Minimum inhibitory concentrations (MIC) were determined by broth-microdilution method following the standard ISO 10932:2010 IDF 223:210 (2010). B. longum ATCC 15707 and L. paracasei ATCC 334 were used as control strains. Susceptibility of tested strains was determined by comparing MIC values with cut-off values provided by EFSA in their latest Scientific Opinion (EFSA, 2018).

The ability to produce 4 biogenic amines through 5 biosynthetic routes -putrescine from agmatine and from ornithine, histamine from histidine, cadaverine from lysine and tyramine from tyrosine- was studied. Strains were cultured in MRS or MRScys supplemented with 1 mM of the corresponding precursor. Overnight culture supernatants were recovered by centrifugation. Samples were derivatized with diethyl ethoxymethylenemalonate and filtered through 0.2 μM filters and biogenic amines were quantified by UHPLC as described by Redruello et al. (2013) at IPLA-CSIC (Villaviciosa, Spain). Putrescine, histamine, cadaverine and tyramine (Sigma-Aldrich, Madrid, Spain) were used as standards.

Hemolytic activity was determined by streaking the strains on Columbia blood agar plates containing 5% defibrinated sheep blood (Scharlab). After 48 h of incubation at 37°C the plates were observed for hemolytic reaction (Kondrotiene et al., 2020). S. pyogenes CECT 191 was used as positive control. Gelatinase activity was tested by spotting fresh cultures on nutrient gelatin agar plates. Strains were cultured at 37°C for 5 days. Plates were treated with acidic mercuric chloride solution for 10 min to detect opaque halos indicating digestion of gelatin (dela Cruz and Torres, 2012). B. cereus CECT 5144 was used as positive control. To test additional enzymatic activities, API^® ZYM (bioMérieux, Madrid, Spain) experiments were performed by triplicate following manufacturer instructions.

Gastric stress and bile salt tolerance assays were performed as described elsewhere (Kondrotiene et al., 2020) with slight modifications. To study the gastric stress resistance, overnight cultures were used to inoculate (1%) simulated gastric solutions (per L: NaCl 7.3 g, KCl 0.52 g, NaHCO₃ 3.78 g and pepsin 3 g) at pH 2.3 and pH 3 adjusted with 1N HCl. Strains were incubated at 37°C for 30 min in condition at pH 2.3 and for 90 min in condition at pH 3. To determine bile salt tolerance, MRS or MRScys broth containing 0.3% (w/v) bile salts (Sigma-Aldrich) were inoculated at 1% from overnight cultures and incubated for 180 min at 37°C. Routine media inoculated at 1% were used as controls of active cultures. Proliferative bacteria were counted before and after incubation times by serial dilution and counting method on MRS or MRScys agar plates. The plates were incubated for 48 h. Commercial probiotic strain L. rhamnosus GG was used as reference. Results were expressed as mean and standard deviation of log CFU/mL. Determinations were performed in independent triplicates.

The ability of probiotic strains to adhere to the intestinal epithelium was studied in vitro using Caco-2 intestinal epithelial cells. Routinely, cells were grown in DMEM High Glucose with 10% decomplemented fetal calf serum (FCS), 40 μg/mL gentamycin and 50 μg/mL L-glutamine (Euroclone, Milan, Italy) at 37°C in a 5% CO₂ atmosphere. Caco-2 cells were seeded in 24-well plates and cultivated until confluence. Caco-2 cells were washed with HBSS (hank’s balanced salt solution, Sigma-Aldrich) and incubated for 1 h with 875 μL of DMEM + 1% FCS under the same conditions. Bacterial cells from overnight cultures were recovered by centrifugation, washed twice with sterile distilled water, and suspended in sterile DMEM + 1% FCS at 1.5 × 10⁷ CFU/mL. Bacterial suspensions were added to Caco-2 monolayers (MOI 1:5 cells to probiotic) and to wells without Caco-2 cells as controls. After 1 h of incubation at 37°C with 5% CO₂, the medium was removed, Caco-2 cells were washed 3 times with HBSS for 5 min and detached by adding 100 μL of trypsin for 5 min at 37°C, and finally recovered with 900 μL of MRD (Maximum Recovery Diluent, BD Difco). Bacterial suspensions were enumerated by plating serial dilutions on MRS or MRScys. Bacteria in the medium of the control wells were also quantified and bacterial adhesion capacity was calculated as the percentage of log CFU/ml of adhered bacteria relative to log CFU/ml in the controls. The experiments were performed in two independent replicates. L. paracasei ATCC 344 and B. longum ATCC 15707 were used as quality controls with a known adhesion percentage of 64–71 and 47–55, respectively (Sagheddu et al., 2020).

To study the effect of probiotics on expression of Caco-2 genes, differentiated Caco-2 cells were washed with HBSS and left for 1 h in DMEM. Cells were exposed to single and combined bacterial suspensions (prepared by mixing equal volumes of each strain) for 5 h at 37°C in a 5% CO₂ atmosphere (MOI 1.5:1/cells to probiotic). Caco-2 wells without probiotic addition were used as controls. At the end of the incubation period, total RNA was extracted with Aurum Total RNA Mini Kit (Bio-Rad, Barcelona, Spain) following manufacturer’s instructions and quantified with NanoDrop (Thermo Fisher Scientific). cDNA was obtained from 2 μg of RNA through the IScript Advanced cDNA Syntesis Kit (Bio-Rad). Gene expression was quantified using 100 ng of cDNA by Reverse Transcription Quantitative PCR (RT-qPCR) with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad). RT-qPCR program was set as follows: 50°C for 2 min; 95°C for 7 min; 40 cycles of 95°C for 15 s, 60°C for 1 min and 72°C for 1 min; and 95°C for 10 s. Target genes included zonula ocludens-1 (ZO-1) ocludine (OCLN), claudine (CLDN1) and serotonine trasporter SERT (SLC6A4). β-actin gene was measured as internal control for normalization. Gene-specific primers were obtained from Bio-Rad (PrimePCR SYBR Green Assay) with the following codes: qHsaCID0018062 for ZO-1, qHsaCED0038290 for OCLN, qHsaCID0006097 for CLDN-1, qHsaCID0016255 for SLC6A4 and qHsaCED0036269 for β-actin. Relative mRNA expression was calculated using the 2^ΔΔCt method. RT-qPCR analysis was performed on RNA purified from two independent replicates for each condition.

The ability of probiotic strains to inhibit the growth of Gram-positive and Gram-negative pathogens was studied in liquid cultures by microplate growth inhibition assay using supernatants of probiotic mono-cultures and probiotic + pathogen co-cultures, and in agar cultures by overlay assay.

To obtain mono-culture supernatants of probiotic strains, overnight cultures adjusted to an optical density (OD_{620 nm}) of 0.2 were inoculated (0.1%) into tubes containing 2x concentrated MRS or MRScys and incubated for 16 h at 37°C in anaerobiosis. Then, one volume of 2x TSB medium was added to the probiotic culture and the reconstituted mixture was further incubated for 24 h under the same conditions. Supernatants of pure P. pentosaceus and B. longum cultures were collected by centrifugation, filtered and an aliquot was neutralized with 5N NaOH.

To investigate whether the antimicrobial activity of probiotic liquid cultures could be induced by the presence of particular pathogens (Maldonado-Barragán et al., 2016), probiotic + pathogen co-cultures were performed. To this end, probiotic and pathogen overnight cultures (OD_{620 nm} of 0.2) were inoculated (0.1%) into tubes containing 2x MRS or MRScys (for probiotic strains) or 2x TSB medium (for pathogens) and incubated for 16 h at 37°C in anaerobiosis and aerobiosis, respectively. Pure probiotics and pathogens cultures were then mixed 1:1 and co-cultures were incubated anaerobically for an additional 24 h. Supernatants of mixed probiotic-pathogen cultures were collected and processed as indicated above.

The inhibitory activity of mono-culture and co-culture supernatants against pathogens’ growth was tested in 96-wells plates assays as described by Vijayakumar and Muriana (2015). Overnight cultures of bacterial pathogens were adjusted to an OD_{620 nm} of 0.2 in 2x TSB medium and 100 μL were poured into microplate wells. One hundred μL of crude and neutralized mono-culture and co-culture supernatants were added to the wells containing indicator strains. To investigate the potential synergistic effect of P. pentosaceus and B. longum culture supernatants, 50 μL of each crude or neutralized supernatant (1:1) were added to the wells. As controls, wells containing pathogens cultures were supplemented with 50 μL of 2x MRS and 50 μL of 2x TSB. The OD_{620 nm} of the resulting cultures was monitored for 16 h using a plate reader and growth curves were obtained. The antagonistic activity of the different supernatants was quantified by calculating the Area Under the Curve (AUC) of the treated cultures in relation to that of the control wells and expressed as relative growth inhibition in percentage. The experiment was performed at least by duplicate.

Agar overlay method was performed as described previously (Magnusson and Schnürer, 2001) with modifications adapted from Lee et al. (2011). Overnight cultures of B. longum KABP042 and P. pentosaceus KABP041 were adjusted to an OD_{620 nm} of 0.8. Ten μL of each strain and a mixture (1:1) of both were spot-inoculated on MRScys agar plates. After 24 h of incubation, pathogenic strains (10⁹ CFU approximately) were suspended in soft TSA (0.9% agar) with and without buffering with 0.5 M Tris and poured onto spot-inoculated MRScys agar plates. Plates were further incubated for 48 h and the zones of inhibition were measured. The experiment was performed in two independent occasions.

Finally, to verify the proteinaceous nature of the antimicrobial molecules, neutralized mono-culture supernatants were individually treated with proteinase K (Thermo Fisher Scientific) (1 mg/mL) and trypsin (Sigma-Aldrich) (5 mg/mL) at 37°C for 2 h. Enzymes were inactivated by boiling for 3 min. Treated supernatants were tested by microplate growth inhibition assay.

Probiotic strains were grown in MRS or MRScys for 24 h in anaerobic conditions at 37°C. Next day, culture supernatants were filtered, and lactate and acetate concentrations were determined by ion chromatography at Analysis Service Unit facilities of ICTAN-CSIC (Madrid, Spain). L-lactic acid isomer was quantified by enzymatic kit (Biosystems, Barcelona, Spain) according to manufacturer’s instructions. Assays were performed in three different occasions.

Acute oral toxicities of B. longum KABP042 and P. pentosaceus KABP041 were tested following OECD Guideline for Testing Chemicals No 420 (OECD, 2002). In short, ten 7-week-old female Wistar rats (Envigo, Oxon, United Kingdom) were housed in groups of 5 with drinking water and standard diet ad libitum except for the night before the first administration and for the subsequent 3 h. Lyophilized probiotic strains B. longum KABP042 and P. pentosaceus KABP041 produced by AB-Biotics S.A. (Barcelona, Spain) were re-suspended in PBS 2 h prior administration. A sighting test was performed for each probiotic strain in a single animal at a starting dose of 2,000 mg/kg (equivalent to 2 × 10¹¹ CFU/kg). In absence of toxicity, the same animal and an additional group of 4 animals were treated with the same single dose. In total, five rats were treated with B. longum KABP042 and 5 animals with P. pentosaceus KABP041. Clinical observations were made 30 min, 1, 2, and 4 h after dosing and then daily for 14 days. Morbidity and mortality checks were made twice daily. Individual body weights were recorded on the day of dosing and on days 7 and 14. At the end of the observation period the animals were euthanized by cervical dislocation and subjected to gross necropsy. An estimate of the acute oral median lethal dose (LD₅₀) of the treatments was made. Study followed animal welfare requirements of Envigo—Shardlow Ethical policy and United Kingdom’s Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012.

A prospective, observational, multi-site, pilot study was carried out at Azienda Sanitaria Locale Napoli (Naples, Italy) during 2018. Infants enrolled had a diagnosis of FGID (functional constipation and/or infant colic) according to Rome IV criteria for clinical purposes (Benninga et al., 2016). Constipation was diagnosed in infants with two or more of the following symptoms in a month: 2 or fewer defecations per week, excessive stool retention, painful or hard bowel movements, large-diameter stools or presence of a large fecal mass in the rectum. Infant colic was diagnosed in babies up to 5 months old if recurrent and prolonged periods of crying, fusing or irritability occurred without obvious cause and that cannot be prevented or resolved. Other inclusion criteria included: age from 1 to 10 months, ≥ 37 weeks of gestation at birth and ≥ 2500 g of birth weight. Infants with failure to thrive (weight gain < 100 grams/week on average from birth), major medical problems (e.g., immunodeficiencies, developmental problems or genetic abnormalities), severe gastrointestinal diseases, administered with antibiotics for 4 weeks or probiotics for 2 weeks before or during study were excluded.

Sample size was estimated to observe a difference ≥ 1 unit in gastrointestinal symptoms on the Likert scale accepting an alpha risk of 0.05, and a beta risk of 0.2 in a two-sided test. Assuming a standard deviation of 2 and a 10% dropout rate, 40 subjects were needed.

Forty-three patients were screened and 36 who satisfied inclusion and exclusion criteria were enrolled in the study. Infants received a formula containing B. longum KABP042 and P. pentosaceus KABP041 (1:1) in sunflower oil which was produced by AB-Biotics S.A. A daily dose of 10 drops (2 × 10⁹ CFU of total bacteria) was administered orally for 14 days, preferably with the first meal of the day. Vials were kept at room temperature during the study. The trial was conducted according to the Helsinki Declaration and Good Clinical Practices. Study protocol was approved by the local ethics committee of Azienda Sanitaria Locale Napoli 3 Sud (Italy) and registered as NCT04944628. Signed informed consent was obtained from the caregivers of all participating infants.

The co-primary outcomes of the study were to assess the safety and tolerability of the probiotic product, and the efficacy measured as the difference in total FGIDs severity (summatory of excessive crying score as indicator of colic severity and constipation score) at the end of study relative to baseline. Secondary outcomes included investigating the differences in excessive crying and constipation severity within infant colic and constipation subpopulations, respectively, and change in parental anxiety after the 14 days of treatment.

Demographic characteristics, delivery mode, type of feeding, and past and current medications were obtained from medical records and anamnesis at baseline visit. Excessive crying and constipation severity were individually scored by the pediatricians using 5-point Likert scales (0: no symptoms, to 4: extremely severe symptoms) at baseline and at the end of the treatment. Parents’ anxiety was recorded through the validated Generalized Anxiety Disorder 7-item (GAD-7) scale (Spitzer et al., 2006) by each caregiver before and after treatment period. Any adverse events experienced by the participant were daily recorded by caregivers in the patient’s data-collection diary.

Statistical analysis was performed using GraphPad Prism Software 5.0 (GraphPad Software Inc., La Jolla, CA, United States). Data are showed as mean and SD. Data from in vitro experiments was analyzed by one-way ANOVA followed by Bonferroni post hoc test. Clinical data was assessed for normal distribution by Shapiro-Wilk normality tests. Non-parametric continuous variables were compared with the Wilcoxon signed rank test for paired data. Statistical tests were two-sided, at 5% level of significance. The primary efficacy outcome (total FGID score in all recruited subjects) was assessed by Wilcoxon matched-pairs test both on an intent-to-treat (ITT) basis and a per protocol (PP) basis, to ensure the robustness of the findings. For ITT, missing data were imputed using last observation carried forward (LOCF). Secondary outcomes were assessed on a per protocol (PP) basis. Comparisons of individual variables against baseline were assessed by Wilcoxon matched-pairs test, while comparisons between population subgroups was performed by two-way ANOVA followed by Bonferroni post hoc test.

The safety of B. longum KABP042 and P. pentosaceus KABP041 was assessed by combining in silico, in vitro and in vivo analyses. Genome sequences for both strains were successfully obtained (Supplementary Table 2) and subjected to in silico analyses.

VirulenceFinder analysis did not detect true virulence factors for any of the strains. PathogenFinder returned probabilities of being a human pathogen of 0.172 and 0.223 (1 would correspond to 100% probability of being a pathogen) for P. pentosaceus KABP041 and B. longum KABP042, respectively, with no matches with pathogenic gene families.

Antibiotic susceptibility experiments revealed that P. pentosaceus KABP041 was phenotypically resistant to kanamycin, streptomycin, and tetracycline (Table 1). However, genomic analyses using the two maintained databases CARD and ResFinder did not detect any AMR gene, indicating that such resistances are likely driven by intrinsic mechanisms and therefore are not transferable. B. longum strain KABP042 displayed phenotypic resistance to erythromycin and clindamycin (Table 1). Genomic inspection revealed that B. longum KABP042 harbors the resistance gene erm(49), which confers resistance to both antibiotics and cannot be mobilized (Martínez et al., 2018). Hence, according to EFSA criteria (EFSA, 2018, 2021), P. pentosaceus KABP041 and B. longum KABP042 do not pose any concern in relation to potential AMR dissemination.

Analysis of culture supernatants of B. longum KABP042 and P. pentosaceus KABP041 strains grown in the presence of five biogenic amines precursors demonstrated that the strains did not produce such detrimental compounds. In agreement with these results, genes encoding for the key enzymes responsible for the biosynthesis of putrescine (via agmatine and ornithine), histamine, tyramine, and cadaverine were not found in their genomes. None of the studied strains presented hemolytic or gelatinase activities (data not shown). API ZYM experiments revealed that the two strains display different and complementary lipidic, aminoacidic and carbohydrate enzymatic activities (Supplementary Table 3).

To investigate the potential toxicity of the strains in vivo, the acute oral toxicity protocol OECD 420 was performed for the two strains using the highest dose of 2,000 mg/kg (equivalent to 2 × 10¹¹ CFU/kg). No signs of systemic toxicity were observed during the study and no macroscopic findings were recorded at necropsy in any animal for any treatment. Moreover, all rats showed a normal body weight gain throughout the experiment (Supplementary Table 4), confirming that strains B. longum KABP042 and P. pentosaceus KABP041 were safe and non-toxic even if consumed at very high doses.

Tolerance of B. longum KABP042 and P. pentosaceus KABP041 to the gastric conditions was assessed by exposing the strains to a simulated fast-gastric passage (pH 2.3 for 30 min) and a slow postprandial digestion (pH 3 for 90 min). The two studied strains and the well-known probiotic strain L. rhamnosus GG showed a loss < 1 log CFU/mL in both challenges (Table 2). As expected, the 3 strains had a lower survival in the more acidic condition. B. longum KABP042 showed a slightly lower resistance to pH 3 condition than control L. rhamnosus GG (p < 0.05) but tolerated better pH 2.3 challenge.

Furthermore, B. longum KABP042 and P. pentosaceus KABP041 were found to be highly tolerant to bile salts with a loss < 0.5 log CFU/mL and showing no significant differences with control strain (Table 2). In line with these results, one copy of the bsh gene -encoding for a Bile Salt Hydrolase enzyme- was identified in each genome, corroborating that both strains are adapted to the gastrointestinal tract.

Adhesion experiments indicated that both strains efficiently adhere to human intestinal epithelial cells Caco-2. P. pentosaceus KABP041 showed 73.4% adherence and B. longum KABP042 70.8% (Figure 1A). Both probiotic strains adhered slightly greater than the moderately adherent control strain L. paracasei ATCC 334 (p < 0.05 for P. pentosaceus KABP041) and significantly more (p < 0.001 for both strains) than slightly adherent control B. longum ATCC 15707 (Sagheddu et al., 2020).

In silico analyses revealed that all strains are equipped with an array of predicted adhesion proteins and domains (Figures 1B,C). Genomic analysis results for B. longum ATCC 15707, B. longum KABP042 and P. pentosaceus KABP041 mostly mirrored in vitro results, as number of adhesion proteins and domains correlated with their adhesion capacity. Of note, P. pentosaceus KABP041 encode for two mucus-binding proteins containing 14 Mub_B2 or MucBP_2 domains, 1 Pep-Gly associated domain, 1 collagen associated domain and 1 fibronectin associated domain, whereas B. longum KABP042 encode for a larger number of pilin-associated proteins and domains. These results suggest that both strains may occupy a different niche in the GIT tract.

RT-qPCR experiments on selected Caco-2 genes [i.e., tight junction (TJ) proteins zonula occludens-1, occludin, claudin-1, and the serotonin transporter SERT] were performed to investigate probiotic-host interaction. The presence of B. longum KABP042 or P. pentosaceus KABP041 did not affect ZO-1 gene expression significantly (Figure 2A). However, exposure of Caco-2 to a mixture (1:1) of both strains resulted in a low but significant upregulation of ZO-1 compared to control (p < 0.05) and single strain conditions (p < 0.05 vs. B. longum strain). OCLN expression showed a similar trend although differences did not reach statistical significance (Figure 2B). Similar to ZO-1, CLDN-1 had a low but still significant increased expression in response to the two-strains combination compared to B. longum KABP042 (p < 0.05) or P. pentosaceus KABP041 (p < 0.01) alone (Figure 2C). Although differences were not significant, an interesting trend was observed for the expression of the gene encoding serotonin transporter SERT. SLC6A4 transcript was slightly modulated by the presence of the probiotic strains with the mixture of the two strains having a stronger upregulating effect (Figure 2D). These results indicated that B. longum KABP042 and P. pentosaceus KABP041 synergistically modulate Caco-2 genes, with significant effects in genes involved in TJ formation.

Antagonistic activity of supernatants of B. longum KABP042 and P. pentosaceus KABP041 axenic cultures were investigated with and without neutralization, alone or in combination (1:1) against a panel of bacterial pathogens (Figure 3A). All crude B. longum KABP042 and P. pentosaceus KABP041 culture supernatants and their combination drastically inhibited the planktonic growth of bacterial pathogens (p < 0.0001). Neutralization of individual culture supernatants resulted in the loss of antagonistic activity against E. coli, E. faecalis, S. epidermidis and S. aureus strains, but still retained inhibitory effect against K. pneumoniae, and S. typhimurium (p < 0.0001). Interestingly, neutralized mixed supernatants (B. longum KABP042 plus P. pentosaceus KABP041) maintained antagonistic effect against all pathogens, reaching statistical significance against growth of K. pneumoniae and S. typhimurium. These results indicated that the main antagonistic activity of B. longum KABP042 and P. pentosaceus KABP041 supernatants is due to the accumulation of organic acids, but also suggested the presence of additional inhibitory compounds that may act synergistically when combined.

To investigate whether the production of antimicrobial compounds is induced by the presence of pathogenic bacteria, co-cultures of B. longum KABP042 and P. pentosaceus KABP041 with the aforementioned pathogens were performed and their supernatant activity was evaluated (Figure 3B). The results obtained were very similar to those of the mono-culture supernatants, indicating that potential antimicrobials produced by B. longum KABP042 and P. pentosaceus KABP041 are not induced by the presence of most pathogens investigated. Of note, and in line with mono-culture results, B. longum KABP042 and P. pentosaceus KABP041 seems to markedly produce antimicrobial compounds against K. pneumoniae and S. typhimurium when grown in the presence of such pathogens. Such activity appears to be independent of organic acid accumulation, suggesting that production of antimicrobials by B. longum KABP042 and P. pentosaceus KABP041 is targeted to and induced upon sensing these two particular pathogens.

Finally, the soft agar overlay antagonistic assay was conducted to evaluate the antimicrobial activity of probiotic strains during sessile growth (Figure 3C). As in liquid culture supernatants, both probiotic strains were able to inhibit growth of all pathogens when grown on agar media. Inhibition of pathogens’ growth was mainly mediated by the production of organic acids as well, since neutralization of the soft agar inoculated with the pathogen reduced the zone of inhibition produced by B. longum KABP042 and P. pentosaceus KABP041 strains. Nonetheless, for S. aureus, S. epidermidis, and S. typhimurium, the antagonistic activity of both probiotic strains was retained to some extent even under buffered conditions, indicating that antimicrobial compounds are also produced by B. longum KABP042 and P. pentosaceus KABP041 during sessile growth. Moreover, B. longum KABP042 and P. pentosaceus KABP041 displayed synergistic antagonism independent of organic acid production against E. coli and S. typhimurium (p < 0.001) when grown in combination on agar media.

Overall, antagonism experiments indicated that the main inhibitory activity of B. longum KABP042 and P. pentosaceus KABP041 is mediated by the production of organic acids, but neutralization experiments strongly suggested that both strains can produce additional antimicrobial compounds that display an accumulative effect when combined.

In line with this, the biosynthesis of organic acids was confirmed in individual culture supernatants. B. longum KABP042 produced 36.1 ± 1.1 mM lactate (100% L-lactate isomer) and 64.9 ± 3.3 mM acetate (ratio lactate 1: acetate 1.8) while P. pentosaceus KABP041 accumulated 107 ± 1.7 mM lactate (77% L-lactate) and no acetate. Moreover, the genomes of B. longum KABP042 and P. pentosaceus KABP041 were screened for the presence of bacteriocins using BAGEL4 tool. While no bacteriocin genes were detected in the genome of P. pentosaceus KABP041 through this analysis, the gene encoding for the bacteriocin Lantibiotic B (Bisin) was identified in the genome of B. longum KABP042 and further BLAST analyses revealed the presence of the accessory genes lanR2, lanK, lanA, lanR1, and lanI (data not shown). These results evidence that, beside accumulation of organic acids, the probiotic strains could at least produce the bacteriocin Lantibiotic B, which displays broad antimicrobial spectrum (Lee et al., 2011). To confirm the proteinaceous nature of the antimicrobials, neutralized mono-culture supernatants were treated with proteases and retested against the indicator strain K. pneumoniae DSM 11678. The results strongly suggested that both probiotic strains produce antimicrobials of proteinaceous nature as treated supernatants significantly lost inhibitory activity (Supplementary Figure 1).

Thirty-six infants aged between 1 and 40 weeks, meeting inclusion and exclusion criteria, were enrolled in the pilot study. Main demographic and clinical features at baseline are shown in Table 3. No infant changed feeding type during study period. Two participants discontinued the study, one of them because of experiencing intestinal spams and facial redness. Withdrawal cause of the other participant was unknown. These were the only adverse events reported which were mild in intensity and rapidly mitigated with no medication needed. No serious adverse events were observed, indicating the safety and tolerability of the probiotic mixture.

ITT analysis revealed statistically significant differences in total FGID severity score between baseline (3.65 ± 1.48) and end of the treatment (1.69 ± 1.70, p < 0.0001; Figure 4A). PP analysis confirmed these significant differences (n = 32, 3.58 ± 1.49 vs. 1.38 ± 1.45, p < 0.0001). Then, we investigated changes in excessive crying and constipation scores in correspondingly colicky and constipated patients’ subpopulations by PP analysis. A significant reduction of excessive crying score (2.29 ± 0.64 at baseline vs. 0.81 ± 0.80 at day 14, p < 0.0001) and constipation severity score (2.18 ± 0.79 at baseline vs. 1.00 ± 1.00 at day 14, p < 0.0001) was observed (Figures 4B,C). These outcomes indicate that probiotic treatment may reduce FGID severity in infants suffering colic and constipation symptoms.

There is evidence of the role of parental anxiety in the incidence of FGIDs such as infant colic (Thomas et al., 2016). We assessed parent’s anxiety before and after infant’s probiotic administration. Parents presented a reduced anxiety state at the end of the study. GAD-7 score was significantly lower at day 14 in comparison with baseline (7.66 ± 5.77 vs. 2.40 ± 2.21, p < 0.0001; Figure 4D).

We further evaluated whether feeding type, delivery mode and administration of previous or concomitant medications affected FGID amelioration under probiotic therapy. We analyzed changes in total FGID symptoms in subgroups of breastfed, formula-fed, mixed-fed, vaginally delivered, C-section delivered, previously medicated and unmedicated, concomitant treated and untreated babies. Significance of the primary endpoint (total FGID severity) was retained in all the subgroups studied (Supplementary Table 5). Analysis of score changes in subgroups within colic and constipation subpopulations also found significant reduction of excessive crying score and constipation severity score after probiotic treatment for all variables analyzed (p < 0.0001 for time and p > 0.05 for interaction effects).