A Metagenomic and in Silico Functional Prediction of Gut Microbiota Profiles May Concur in Discovering New Cystic Fibrosis Patient-Targeted Probiotics

Cystic fibrosis (CF) is a life-limiting hereditary disorder that results in aberrant mucosa in the lungs and digestive tract, chronic respiratory infections, chronic inflammation, and the need for repeated antibiotic treatments. Probiotics have been demonstrated to improve the quality of life of CF patients. We investigated the distribution of gut microbiota (GM) bacteria to identify new potential probiotics for CF patients on the basis of GM patterns. Fecal samples of 28 CF patients and 31 healthy controls (HC) were collected and analyzed by 16S rRNA-based pyrosequencing analysis of GM, to produce CF-HC paired maps of the distribution of operational taxonomic units (OTUs), and by Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) for Kyoto Encyclopedia of Genes and Genomes (KEGG) biomarker prediction. The maps were scanned to highlight the distribution of bacteria commonly claimed as probiotics, such as bifidobacteria and lactobacilli, and of butyrate-producing colon bacteria, such as Eubacterium spp. and Faecalibacterium prausnitzii. The analyses highlighted 24 OTUs eligible as putative probiotics. Eleven and nine species were prevalently associated with the GM of CF and HC subjects, respectively. Their KEGG prediction provided differential CF and HC pathways, indeed associated with health-promoting biochemical activities in the latter case. GM profiling and KEGG biomarkers concurred in the evaluation of nine bacterial species as novel putative probiotics that could be investigated for the nutritional management of CF patients.


Introduction
Cystic fibrosis (CF) is an autosomal recessive condition occurring among people with European origins, which is caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene. The CFTR mutation leads to the failure or the absence of functional CFTR proteins at the apical membrane of epithelial cells in several body systems [1]. The CFTR protein, in addition to functioning as a chloride channel, can also affect bicarbonate transport. Protein mutations cause the formation of viscous and dehydrated mucus followed by the establishment of aberrant mucosa in The study protocol was approved by the OPBG Ethics Research Committee (protocol No. 534/RA), and was conducted in accordance with the Declaration of Helsinki (as revised in Seoul, Korea, October 2008). A signed informed consent was obtained from the parents of the enrolled subjects. The patients were age-matched with 31 healthy controls (HC) screened by means of a survey of the OPBG Human Microbiome Unit on pediatric gut microbiota programming.
Inclusion criteria for HC were: absence of any inflammatory, infectious, and chronic diseases at the time of the microbiota analysis and no antibiotic and pre-probiotic intake in the previous two months.
For CF patients, the inclusion criteria consisted of being recruited under clinical stability (i.e., absence of infectious exacerbation of pulmonary symptoms) and no pre-probiotic intake in the previous two weeks.

Anamnestic and Laboratory Features
Age, gender, and body mass index (BMI) (for patients over 2 years of age) or Z-score (Weight/Length (W/L) of patients under 2 years of age) were collected for both CF patients and HC, whereas sweat chloride test values, pancreatic status (PS), and antibiotic data for chronic regimen were collected only for CF patients (Table 1).

DNA Extraction and Next Generation Sequencing (NGS) Analysis
Fecal samples (59) were collected from each subject during clinical examination and stored until metagenomics analysis. The genomic DNA was isolated using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany). The V1-V3 region (520 bp) of the 16S ribosomal RNA locus was amplified for pyrosequencing analysis using a 454-Junior Genome Sequencer (Roche 454 Life Sciences, Branford, CT, USA) according to Del Chierico et al. [26]. The nucleotide barcodes, added in forward primers, were composed of 8 unique nucleotides (Roche 454 Life Sciences). The polymerase chain reactions were performed using Hi-Fi Polymerase Chain Reaction (PCR) Taq polymerase (FastStart™ High Fidelity PCR System, dNTPack, Roche Diagnostics, Mannheim, Germany), guaranteeing high specificity, sensitivity, and accuracy during PCR amplification.

Statistical Analysis
Reads were analyzed with Qiime 1.8 (Quantitative Insights Into Microbial Ecology, http://qiime. org/1.4.0/) using the default pipeline [27]. After demultiplexing, reads with an average quality score lower than 25, shorter than 300 bp, and with an ambiguous base calling were excluded from the analysis to guarantee a higher level of accuracy in terms of detection of the operational taxonomic units (OTUs). Sequences that passed the quality filter were denoised [28], and singletons were excluded. The denoised sequences were chimera-checked by identify_chimeric_seqs.py.
To characterize the taxonomic structure of the samples, the sequences were organized into OTUs by clustering at a threshold of 97% pairwise identity and by classifying the representative sequences using the Greengenes 13_8 database [29]. The representative sequences were submitted to PyNAST for sequence alignment [30] and to UCLUST for sequence clustering [31].
The OTU Kruskal-Wallis tests were performed by QIIME software (http://qiime.org/1.4.0/) using "group_significance.py" script [32]. The Kruskal-Wallis test was performed on OTU distribution with False Discovery Rate (FDR) correction (p-value ≤ 0.1). To gain more insight into the metagenomics-based function of the microbiome of the CF patients and HC, the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) v1.1.0 tool was used [33], and the resulting function prediction was analyzed using the HUMAnN2 v0.99 program to get KEGG pathways (http://huttenhower.sph.harvard.edu/humann2) [34]. To find possible OTUs and KEGG biomarkers associated with CF and HC, a linear discriminant effect size (LEfSe) analysis was performed [35] with the α value of the statistical test equal to 0.05 and the logarithmic Linear Discriminant Analysis (LDA) score threshold equal to 2.0.

Putative Probiotic Distribution in the GM Profiles
By targeted metagenomics, a total of 316,000 reads was obtained with an average of 5356 reads/sample and an average length of 487 bp. Genus-level comparisons were performed on 24 OTUs belonging to Bifidobacterium spp., Lactobacillus spp., Eubacterium spp., and F. prausnitzii chosen from a total dataset of 165 OTUs, considering their putative probiotic role.
The profiling of targeted metagenomic sequencing pointed out a distribution of 11 bacterial species prevalently associated with the GM of the CF patients ( Figure 1, Panel A), and 9 species prevalently associated with the GM of the HC (Figure 1, Panel B) ( Table 2). The Kruskal-Wallis test identified a statistically significant difference for F. prausnitzii distribution between CF patients and HC, highlighting a higher relative abundance in HC.
On the contrary, the remaining 4 OTUs, namely, Lactobacillus brevis, L. delbrueckii, L. helveticus, and Eubacterium cylindroides were comparably distributed in the GM profiles of the CF patients and HC ( Figure 1, Panel C).  [33], and the resulting function prediction was analyzed using the HUMAnN2 v0.99 program to get KEGG pathways (http://huttenhower.sph.harvard.edu/humann2) [34]. To find possible OTUs and KEGG biomarkers associated with CF and HC, a linear discriminant effect size (LEfSe) analysis was performed [35] with the α value of the statistical test equal to 0.05 and the logarithmic Linear Discriminant Analysis (LDA) score threshold equal to 2.0.

Putative Probiotic Distribution in the GM Profiles
By targeted metagenomics, a total of 316,000 reads was obtained with an average of 5356 reads/sample and an average length of 487 bp. Genus-level comparisons were performed on 24 OTUs belonging to Bifidobacterium spp., Lactobacillus spp., Eubacterium spp., and F. prausnitzii chosen from a total dataset of 165 OTUs, considering their putative probiotic role.
The profiling of targeted metagenomic sequencing pointed out a distribution of 11 bacterial species prevalently associated with the GM of the CF patients ( Figure 1, Panel A), and 9 species prevalently associated with the GM of the HC (Figure 1, Panel B) ( Table 2). The Kruskal-Wallis test identified a statistically significant difference for F. prausnitzii distribution between CF patients and HC, highlighting a higher relative abundance in HC.
On the contrary, the remaining 4 OTUs, namely, Lactobacillus brevis, L. delbrueckii, L. helveticus, and Eubacterium cylindroides were comparably distributed in the GM profiles of the CF patients and HC ( Figure 1, Panel C).

. Metabolic Pathways of Probiotics
To better define the metabolic role of the detected putative probiotic species, a supervised comparison of CF patients' and HC's KEGGs was inferred by LEfSe on the 24 OTU matrix.
The predicted microbial function highlighted differences in metabolic pathways associated with the 24 selected OTUs (Figure 2). In particular, 24 pathways resulted associated with CF and 39 with HC (Table 3).

Metabolic Pathways of Probiotics
To better define the metabolic role of the detected putative probiotic species, a supervised comparison of CF patients' and HC's KEGGs was inferred by LEfSe on the 24 OTU matrix.
The predicted microbial function highlighted differences in metabolic pathways associated with the 24 selected OTUs (Figure 2). In particular, 24 pathways resulted associated with CF and 39 with HC (Table 3).

Putative Probiotic Distribution in the GM Profiles
Bifidobacteria and lactobacilli are recognized as beneficial bacteria for their intrinsic probiotic features [23]. Eubacterium spp. may aid in the digestion, the absorption, or both of food ingredients and minerals, especially under malnutrition conditions usually occurring in CF because of nutrient absorption defects [36]. More generally, F. prausntizii has been recently proposed to provide high butyrate production in the gut [37].
Among the bacteria associated with the CF gut profiles (Table 2), some are actually linked to different pathologic conditions. Indeed, Bifidobacterium dentium was detected in the oral cavity in association with dental caries [38], while Eubacterium dolichum was associated with frailty in the elderly, a condition that represents the biggest problem associated with population aging [39]. Lactobacillus mucosae was detected in the microbiota of short bowel syndrome patients [40]. Other bacterial strains identified for CF patients were associated with metabolic disorders, such as high total cholesterol and low-density lipoprotein levels (Eubacterium biforme) [41], obesity (Lactobacillus reuteri) [42], and nonalcoholic fatty liver disease (NAFLD) (L. zeae and L. vaginalis) [24].
Unlike the aforementioned negative roles of the previous reported bacteria in human health, B. breve is considered a commensal or even a health-promoting microorganism [43] because it improves symptoms in necrotizing enterocolitis [44] and atopic dermatitis [45], as well as those associated with HIV-induced damages [46]. Moreover, B. breve shows antimicrobial activity [47], induces innate immune responses, and has anti-inflammatory effects [48]. Also, L. pentosus was reported to ameliorate colitis in the aged rodent by inhibiting the activation of nuclear factor (NF)-κB, activator protein 1 (AP1), and mitogen-activated protein kinases (MAPKs) [49].
No role in the human GM or putative effects as probiotics have been reported for L. crispatus and L. pontis.
The GM in the HC group seemed to be enriched in species involved in gut integrity and mobility, digestion of specific dietary compounds, and immune system modulation Indeed, in the research of Kanauchi and co-workers [50], E. limosum was presented as an important probiotic candidate for its short-chain fatty acid (SCFA) production, role in maintaining and enhancing mucosal integrity, and anti-inflammatory properties in the intestinal mucosa [50]. Moreover, Bruzzese et al. found that both E. rectale and F. prausnitzii were reduced in the GM of CF children compared to HC, confirming our results [51].
Eubacterium siraeum is able to degrade wheat bran, contributing to the beneficial effects of cereal fiber in human health through their impact on the GM [52].
L. sanfranciscensis is generally considered the most important lactic acid bacterium in the fermentation of rye and wheat sourdoughs [53]. The strain L. sanfranciscensis LBH1068, tested in an induced chronic colitis mouse model, improved mouse health by reducing weight loss, decreasing gut permeability, and modulating cytokine production [54].
In addition, L. fermentum demonstrated intestinal anti-inflammatory effects in the model of sodium dextran sulfate-induced colitis in mice. Among the mechanisms proposed, L. fermentum restored GM composition and modulated the altered immune response by preserving intestinal barrier integrity, decreasing pro-inflammatory cytokine production, and modulating the expression of Th1-, Th17-and T reg -related cytokines [55].
The study of Moya-Pérez and colleagues demonstrated that B. pseudocatenulatum modulated immune cell infiltration and inflammation in the gut in obesity [56].
A reduction in bifidobacteria in CF, especially B. longum, was already reported by Duytschaever et al. [57]. High richness of bifidobacteria species was positively correlated with the maturation of the mucosal immune system [58]. B. longum was found to be an inhibitor of rotavirus, the predominant cause of sporadic diarrhea in infants [59]. A recent study has demonstrated that acetate produced by B. longum acts as an essential cosubstrate for butyrate production and for E. rectale growth [60].
B. bifidum and B. longum possess numerous pathways involved in the catabolism of human milk oligosaccharides (HMO) and may also consume carbohydrates released by other bacteria [61]. B. bifidum and B. longum were described as being more abundant in healthy subjects compared to NAFLD patients, suggesting a protective and beneficial role also in obesity and NAFLD [62].
Finally, each of these microbial species, especially those lacking in patients' GM (Table 2, HC-related species), could be considered suitable for the design of CF patient-tailored probiotics.

Metabolic Pathways of Probiotics
To evaluate the microbial metabolic and functional KEGG pathways of the chosen putative probiotic species, a supervised comparison of CF patients' and HC's KEGGs was performed by LEfSe (Table 3). Pathways associated with fatty acid biosynthesis, metabolism, and synthesis and degradation of ketone bodies were significantly associated with CF patients, as already described by Fouhy et al. [63]. This increase in fat metabolism probably occurs as a result of a combination of factors, including a reduced intestinal absorption and an altered GM in CF patients. Reduced fat absorption is one reason why most CF patients are traditionally prescribed a high-fat diet to ensure adequate weight maintenance. Moreover, it is possible that the altered GM might contribute to the increase in fat metabolism [63].
On the contrary, primary and secondary bile acid biosynthesis pathways were associated with the GM of HC. It is known that CF patients have a variety of intestinal abnormalities in bile acid metabolism at the intestinal level, including increased fecal bile acid losses, reduced bile acid pool size, and duodenal bile acid concentration [64]. These abnormalities appear to be associated with exocrine pancreatic insufficiency and steatorrhoea. Indeed, improvement of bile abnormalities with amelioration in fat malabsorption was reported after pancreatic enzyme therapy [64].
Pathways involved in xenobiotic metabolism have been significantly observed in the GM pattern of CF patients, including benzoate, fluorobenzoate, dioxin, xylene, aminobenzoate, and ethylbenzene degradation pathways. Bacterial pathways involved in xenobiotic metabolism were also observed by Fouhy et al. in CF patients [63]. The increase in bacteria capable of degrading xenobiotic compounds is probably due to the higher exposure to antibiotics and pharmacological treatments, recurrent in CF patients [65]. An enhanced ability of CF patients' GM to metabolize proteins was highlighted by the increase in amino acid catabolism (e.g., valine, leucine, isoleucine, and lysine degradation) prediction. Indeed, the increase in protein catabolism in CF individuals has been well documented, probably due to the breakdown of both cellular and connective tissue proteins, which is related to the degree of impaired lung function and to the systemic inflammatory response [66]. Moreover, also valine, leucine, isoleucine, lysine, phenylalanine, tyrosine, and tryptophan biosynthesis pathways were linked to putative probiotics in HC. Our results nicely agree with the findings of a study carried out by Palmer et al., in which 11 genes involved in branched-chain and aromatic amino acid catabolism were highly upregulated in CF patients' sputum, while genes involved in the biosynthesis of these amino acids were repressed [67].
Moreover, the flagellar assembly pathway was associated with putative probiotics in CF. Bacterial flagellin is classified as a potent mediator of virulence of Gram-negative bacteria. Recurrent infections caused by Gram-negative strains could be linked to this inferred pathway [68].
Finally, the prediction of lipoic acid metabolism and folate biosynthesis pathways were associated with CF patterns. Consistently, Quinn and colleagues reported the abundance of lipoic acid metabolism in the lung of CF patients [65]. Lipoic acid is an antioxidant and a potent quencher of reactive oxygen species (ROS) [69], and it is used as a metabolic cofactor by Proteobacteria, Gram-positive bacteria, and Pseudomonas aeruginosa [69,70]. Quinn reported also high abundance of folate synthesis in CF patients [65]. Sulfonamides, such as sulfamethoxazole and trimethoprim, are antibiotics commonly used to treat CF infections that target microbial enzymes required for folate biosynthesis [71].
Prolonged exposure to sulfonamides may select microbes with multiple copies of these genes to overcome the drug's effect on folate synthesis [72].

Conclusions
In conclusion, patients with CF usually have an abnormal intestinal microbiota and dysregulated immune mediators resulting from a massive exposure to antibiotics. Probiotics as immunomodulatory and anti-inflammatory substances are considered to improve both the clinical and the biochemical intestinal and pulmonary function in CF patients. The results reported in this study may point out new putative probiotic species on the basis of the GM differential profiles and predicted metabolic pathways of CF patients compared to HC.
On the basis of our data, we speculate that some putative probiotic species, such as B. longum, E. rectale, E. limosum, E. siraeum, L. sanfranciscensis, L. fermentum, B. pseudocatenulatum, B. bifidum, and F. prausnitzii and their produced metabolites may have a protective role against CF disorders. Nevertheless, further in vitro studies and clinical trials should focus on these probiotics to assess whether the administration of selected strains, alone or in combination, may improve the quality of life and the clinical management of CF patients.