Characterization of the Gut Microbiota and Mycobiota in Italian Pediatric Patients With Primary Sclerosing Cholangitis and Ulcerative Colitis

Abstract

Background

Primary sclerosing cholangitis (PSC) is a chronic, fibroinflammatory, cholestatic liver disease of unknown etiopathogenesis, often associated with inflammatory bowel diseases. Recent evidence ascribes, together with immunologic and environmental components, a significant role to the intestinal microbiota or its molecules in the PSC pathogenesis.

Methods

By metagenomic sequencing of 16S rRNA and ITS2 loci, we describe the fecal microbiota and mycobiota of 26 pediatric patients affected by PSC and concomitant ulcerative colitis (PSC-UC), 27 patients without PSC but with UC (UC), and 26 healthy subjects (CTRLs).

Results

Compared with CTRL, the bacterial and fungal gut dysbiosis was evident for both PSC-UC and UC groups; in particular, Streptococcus, Saccharomyces, Sporobolomyces, Tilletiopsis, and Debaryomyces appeared increased in PSC-UC, whereas Klebsiella, Haemophilus, Enterococcus Collinsella, Piptoporus, Candida, and Hyphodontia in UC. In both patient groups, Akkermansia, Bacteroides, Parabacteroides, Oscillospira, Meyerozyma and Malassezia were decreased. Co-occurrence analysis evidenced the lowest number of nodes and edges for fungi networks compared with bacteria. Finally, we identified a specific patient profile, based on liver function tests, bacterial and fungal signatures, that is able to distinguish PSC-UC from UC patients.

Conclusions

We describe the gut microbiota and mycobiota dysbiosis associated to PSC-UC disease. Our results evidenced a gut imbalance, with the reduction of gut commensal microorganisms with stated anti-inflammatory properties (ie, Akkermansia, Bacteroides, Parabacteroides, Oscillospira, Meyerozyma, and Malassezia) and the increase of pathobionts (ie, Streptococcus, Saccharomyces, and Debaryomyces) that could be involved in PSC progression. Altogether, these events may concur in the pathophysiology of PSC in the framework of UC.

Key Messages.

What is already known?

Preliminary observation suggests a pivotal role of microbiota in the pathogenesis of primary sclerosing cholangitis (PSC).

What is new here?

Children with primary sclerosing cholangitis (PSC) associated with ulcerative colitis (UC) have a specific bacterial and fungal dysbiosis state compared with children with UC and healthy subjects.

How can this study help patient care?

The alteration of the micro/mycobiota has a role in the PSC-UC pathogenesis, affecting the production of crucial intestinal molecules (eg, bile acids). Then, interventions on the gut microbial and fungal content could have a protective or therapeutic role in PSC.

Introduction

Primary sclerosing cholangitis (PSC) is a chronic, fibroinflammatory, cholestatic disease involving the intra- or extrahepatic bile ducts, that may lead to biliary cirrhosis and liver failure.¹

Up to 80% of children with PSC have an associated inflammatory bowel disease (IBD), in particular ulcerative colitis (UC).²

The strong clinical association between PSC and IBD led to the pathogenetic situation in which the increase of intestinal permeability, the translocation from the inflamed gut, and the enterohepatic circulation of microbial products (eg, lipopolysaccharide [LPS], lipoteichoic acid, and peptidoglycan) are central pathobiological drivers of PSC.³ Indeed, portal bacteremia, bacterobilia, and traces 16S rRNA in bile have been found in adult patients with PSC.⁴ Moreover, it has been shown that cholangiocytes actively recognize endogenous and exogenous molecules, including pathogen-associated molecular patterns (PAMPs), leading to hepatobiliary inflammation and fibrosis, through an inappropriate innate immune response to intestinal bacterial endotoxins⁵; and there have been reports of a persistent hypersensitivity to LPS and other PAMPs in cultured cholangiocytes from PSC patients.⁵ Lastly, the “gut homing” hypothesis suggests the pathogenesis of PSC is driven by aberrant hepatic expression of gut adhesion molecules and subsequent recruitment of gut-derived T cells to the liver.⁶In this context, an altered microbial gut environment may be the trigger for the PSC onset process. Moreover, the gut microbiota modulation with probiotics and antibiotics like oral vancomycin, metronidazole, and/or minocycline has reported a biochemical improvement in these patients. Over the time, this clinical approach seemed to be promising and has gained wide interest, although, still today, it cannot be recommended.^7–9

Numerous studies describing bacterial^10–16 and fungal dysbiosis^17,18 of gut microbiota in adult PSC patients have been reported. The number of pediatric studies on the gut microbiota in PSC is still limited, and it has been focused only on gut bacterial description,^19,20 moreover, little is known about the difference between children with PSC and UC and children with only UC.

Herein we present a prospective multicentric cross-sectional study to investigate gut microbial and fungal composition in pediatric patients with PSC associated with UC compared with both UC patients without PSC and heathy subjects to demonstrate a specific intestinal microbiota and mycobiota in these patients.

Materials and Methods

Study Population

We performed a prospective multicentric study conducted over 5 years at the Bambino Gesù Children’s Hospital of Rome, Italy, and the Institute for Maternal and Child Health, Burlo Garofalo, of Trieste, Italy, enrolling pediatric patients aged 2 to 19 years with PSC and concomitant UC (PSC-UC), only UC, and healthy controls (CTRLs). The criteria that allowed the diagnosis of PSC were (1) typical clinical presentation such as pruritus, jaundice, pain in the right upper abdominal quadrant, fatigue, weight loss, and episodes of fever; (2) elevated alkaline phosphatase (ALP) and gamma-glutamyl transferase (GGT) and/or bilirubin, not otherwise explained; (3) presence of characteristic bile duct changes with multifocal strictures and segmental dilatations on endoscopic retrograde cholangiography (ERC) or magnetic resonance cholangiography (MRC); (4) liver histology; and (5) no evidence for secondary sclerosing cholangitis.²¹

Criteria to perform diagnosis of UC were based on clinical history, physical examination, endoscopic appearance, histologic findings, and radiologic studies, according to Porto criteria.²²

The exclusion criteria were (1) use of antibiotics and/or probiotics in the 4 weeks before clinical diagnosis and (2) presence of other concomitant conditions (eg, metabolic, endocrinologic, hematologic diseases).

At the time of stool collection from each patient, the following information was retrieved: gender; age at diagnosis; body mass index (BMI); liver function tests (ie, alanine aminotransferase [ALT], aspartate aminotransferase [AST], total bilirubin [BT], gamma-glutamyl transpeptidase [GGT] and serum albumin); type of IBD, Mayo score, and Montreal classification; medical treatment histories; and presence of cholangitis and/or bile dilatation/stenting and/or cholecystectomy.

The CTRL group was composed of healthy children with normal BMI, gender- and age-matched with patient cohorts, enrolled during an epidemiological survey carried out at the Human Microbiome Unit of Bambino Gesù Children’s Hospital (BBMRI Human Microbiome Biobank, OPBG). Their exclusion criteria were family history of autoimmune or IBDs diseases, gastrointestinal infections, and use of antibiotics and pre/probiotics in the previous 2 months before recruitment.

The study was approved by the OPBG Ethical Committee (PSC protocol: 1191_OPBG_2016; healthy protocol: 1113_OPBG_2016) and was conducted in accordance with the Principles of Good Clinical Practice and the Declaration of Helsinki. Written informed consent was obtained from either parents or a legal representative of children.

Metagenomic Analyses Based on 16S and ITS2-DNA Sequencing

From each subject, a single fecal sample was collected and stored at −80°C until further processing. For bacterial metagenomic analysis, 200 mg of stools were suspended in 500 μL of phosphate-buffered saline (PBS) and treated for mechanical and thermal lysis steps. Then, samples were processed for DNA extraction by QIAamp DNA Stool Mini Kit (Qiagen, Germany) according to the manufacturer’s instructions.

The V3-V4 region of the 16S rRNA (460 bp) was amplified by polymerase chain reaction (PCR) using primers reported in the MiSeq rRNA Amplicon Sequencing protocol.²³

For fungal metagenomic analysis, 200 mg of stools were resuspended in 500 μL of lysis solution (50 mM Tris [pH 7.5], 10 mM EDTA, 28 mM 2-Mercaptoethanol, 10 U/mL lyticase; Merck KGaA, Darmstadt, Germany) and incubated at 37°C for 30 minutes at 81 g. After the lysis step, the pellet was resuspended into 500 μL of PBS and treated for mechanical and thermal lysis steps. Afterwards, samples were processed for DNA extraction by QIAamp DNA Stool Mini Kit (Qiagen) according to the manufacturer’s instructions.

The ITS2 region of approximately 350 bp was amplified using the primers ITS2 5’-GTGARTCATCGAATCTTT-3’ and 5’-GATATGCTTAAGTTCAGCGGGT-3’.¹⁷

For each sample, DNA was amplified by PCR as follows: 94°C for 2 minutes, 35 cycles of 15 seconds at 94°C, 52°C for 30 seconds, and 72°C for 45 seconds, followed by 7 minutes at 72°C. The PCR amplicons were purified using AMPure XP Beads (Beckman Coulters, Brea, CA). The purified amplicons were submitted to 15 cycles of PCR, under the previous described conditions, using Illumina adapted ITS2 primers: 5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGARTCATCGAATCTTT-3’ and 5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGATATGCTTAAGTTCAGCGGGT-3’. A second step of PCR amplicons purification was performed with AMPure XP Beads (Beckman Coulters) prior to proceed with PCR indexing in which dual indices and Illumina sequencing adapters were added to each PCR amplicon (Nextera XT Index Kit, Ilumina). Both bacterial and fungal libraries were separately cleaned and quantified using the PicoGreen Staining Kit (Molecular Probes, Paris, France), pooled and sequenced by an Illumina MiSeqTM platform (Illumina). For both 16S rRNA and ITS2 approaches, negative and positive controls were used to monitor and exclude possible external and internal contaminations.

Raw sequences were filtered for quality and read length, filtered for chimera presence, and matched against Greengenes 13.8 database²⁴ for bacteria and against UNITE ITS database²⁵ for fungi (sh_qiime_release_s_04.02.2020) by QIIME software.²⁶

MicrobiomeAnalyst²⁷ was used to calculate α- and β-diversity and related statistical tests (Mann-Whitney U, Kruskal-Wallis, Benjamini-Hochberg tests, co-occurrence network) on taxa relative abundances. Alpha and beta diversity metrics were calculated after normalization by rarefaction (the minimum number of sequences in a sample was 3523 and 2179 sequences for bacteria and fungi, respectively). Cumulative sum scaling (CSS) method was used to normalize the data prior to compare the relative taxa abundances across samples. Alpha diversity was estimated by the Shannon and ChaoI index methods; β-diversity was measured by a Bray-Curtis distance matrix and used to build principal coordinate analysis (PCoA) plots. Raw sequence data are accessible in the NCBI database (accession number PRJNA935155 and PRJNA 280490).

The confounding factors analysis was performed by the Bray Curtis algorithm for age, gender, BMI, and treatments (Figure S1, panels A and B). These analyses confirmed the absence of distinct clusters for age, gender, BMI, and 5-aminosalicylic acid (5-ASA) treatment (PERMANOVA P > .05), excluding these variables as a confounder factors of the microbiota and mycobiota analysis. As for ursodeoxycholic acid (UDCA) medication, it was not possible to separate the effects of this drug from the effects of the disease, inasmuch as 96.2% (25 of 27) of PSC-UC patients were treated with UDCA, and none of the UC patients received this treatment.

To identify taxa with significantly different abundances between categories, linear discriminant analysis effect size (LEfSe) analyses was used.²⁸ An α value of 0.05 and an effect size threshold of 2 were used to identify the significant predicted biomarkers.

To predict metagenome functional content from 16S rRNA gene surveys, the PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) tool was applied.^29,30 Furthermore, to find metabolic biomarkers associated with each group, a LEfSe analysis was performed (α = 0.05, logarithmic LDA score threshold of 2.0).²⁸

Pearson’s correlation, Mann-Whitney, and Fisher’s exact tests, and linear discriminant function analysis (LDA), univariate ANOVAs, Fisher’s coefficient, Wilks’ Lambda, Leave-One-Out Cross-Validation tests were computed by SPSS v.21 software (IBM statistics). Principal component analysis (PCA) was computed and graphed by Graph Pad software version 9.5.1 (Prism).

Results

Patients Description

Twenty-six pediatric patients affected by PSC-UC and 27 with UC were enrolled in this study. Clinical, endoscopic, laboratory features and treatment information were exploited to compare PSC-UC and UC groups. The 2 groups were also compared in terms of gender, age, BMI, treatments (UDCA, 5-ASA, azathioprine [AZA], proton-pump inhibitors [PPIs], vitamin D [VD], and steroids), endoscopic features, liver function, and cholestasis tests (ALT, AST, GGT, BT, albumin). The characteristics of the subjects are reported in Table 1.

Table 1.

Patient cohort characteristics.

Patients’ Features		PSC-UC N° (%)	UC N° (%)	* P
No.		26	27
Gender	F	13 (50.0)	14 (51.8)	.556
	M	13 (50.0)	13 (48.2)
Age classes	2-12 years	14 (53.8)	13 (48.2)	.44
	13-19 years	12 (46.2)	14 (51.8)
BMI	Low (>18.5)	11 (42.3)	11 (40.7)	.601
	Medium (18.5 ≤ BMI ≥ 25.0)	14 (53.8)	13 (48.2)
	High (<24.9)	1 (3.8)	3 (11.1)
Treatments
UDCA treatment		25 (96.2)	0 (0.0)	.000
5-ASA treatment		22 (84.6)	24 (88.9)	.478
PPIs treatment		15 (57.7)	11 (40.7)	.169
VD treatment		6 (23.1)	4 (14.8)	.339
Steroids treatment		17 (65.4)	11 (40.7)	.064
AZA treatment		17 (65.4)	6 (22.2)	.002
Endoscopic features
Montreal classification	Ulcerative proctitis	1 (3.8)	4 (14.8)	.342
	Left sided UC (distal UC)	3 (11.5)	4 (14.8)
	Extensive UC (pancolitis)	22 (84.6)	19 (70.4)
Mayo score	0	14 (53.8)	7 (26.0)	.059
	1	1 (3.8)	6 (22.0)
	2	8 (30.8)	7 (26.0)
	3	3 (11.6)	7 (26.0)
Laboratory test		Average ± SD	Average ± SD	** P
ALT (U/l)		72.3 ± 95.2	16.6 ± 11.1	.000
AST (U/l)		63.6 ± 82.3	22.7 ± 12.4	.010
GGT (U/l)		149.2 ± 159.7	10.9 ± 5.8	.000
BT (mg/dL)		0.9 ± 0.8	0.5 ± 0.2	.087
Albumin (g/dL)		4.3 ± 0.4	4.2 ± 0.4	.426

Abbreviations: BMI, Body Mass Index; UDCA, ursodeoxycholic acid; 5-ASA, 5-aminosalicylic acid; PPIs, proton-pump inhibitors; VD, vitamin D; AZA, Azathioprine; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyl transpeptidase; BT, total bilirubin

^*Fisher’s exact test

^**Mann-Whitney test. In bold are reported P values ≤.05.

The comparison between the 2 groups showed no difference in anthropometric (age, gender, BMI) and endoscopy features (Montreal/Mayo scores). A large majority of patients with PSC were treated with UDCA and AZA (P < .05) compared with patients affected by only UC.

We found a significant statistical increase in ALT, AST, and GGT values in PSC-UC; however, no differences for albumin and BT values were reported between the 2 groups.

Bacterial Gut Dysbiosis in Patients Affected by PSC-UC and UC

A total of 4 777 097 sequencing reads were obtained by 16S rRNA-based metagenomic analysis of fecal samples, with a mean value of 60 469.58 ± 46 894.18 sequences per sample. Alpha diversity revealed a decreased and similar richness in both PSC-UC and UC groups compared with CTRL (Figure 1A and B). Beta diversity, based on Bray Curtis metric, revealed high similarity amongst CTRL samples, whereas PSC-UC and UC samples were sparse and not well clustered depending on disease group (Figure 1C).

Figure 1.

Biodiversity of the bacterial gut microbiota. The bacterial subset of gut microbiota was analyzed in PSC-UC, UC, and CTRL groups. Alpha diversity was estimated by Shannon (A) and Chao1 (B) indexes. Median, percentiles and minimum and maximum values of indexes were plotted; Kruskal-Wallis test was used for comparisons. Beta diversity was assessed by principal coordinate analysis of Bray-Curtis distance (C). The fraction of diversity captured by the coordinate is reported as a percentage. The PERMANOVA test was used for group comparisons. Linear Discriminant Analysis Effect Size (LEfSe) was calculated at phylum (D) and genus (E) levels. The P values were FDR adjusted (P < .05) and LDA score was >2.

The LEfSe algorithm revealed an increase of Verrucomicrobia and Bacteroidetes in CTRLs, and of proteobacteria in UC patients (Figure 1D). At the genus level, Klebsiella, Haemophilus, Enterococcus and Collinsella appeared increased in UC and Streptococcus in PSC-UC, whereas Akkermansia, Bacteroides, Dialister Parabacteroides, and Oscillospira resulted decreased in both groups compared with CTRLs (Figure 1E).

Fungal Gut Dysbiosis in Patients Affected by PSC-UC and UC

A total of 3 323 260 sequencing reads were obtained by the ITS2-based metagenomic analysis of fecal samples, with a mean value of 43 498.31 ± 34,035.92 sequences per sample. No statistical significance was observed in the comparison of Shannon and ChaoI indices between groups of samples (Figure 2A, B). Interestingly, the mycobiota β-diversity showed the highest similarity between PSC-UC and UC feces compared with CTRL samples, which appeared sparser (Figure 2C).

Figure 2.

Biodiversity of the fungal gut mycobiota. The fungal subset of gut mycobiota was analyzed in PSC-UC, UC, and CTRL. Alpha diversity was estimated by Shannon (A) and Chao1 (B) indexes. Median, percentiles and minimum and maximum values of indexes were plotted; Kruskal-Wallis test was used for comparisons. Beta diversity was assessed by principal coordinate analysis of Bray-Curtis distance (C). The fraction of diversity captured by the coordinate is reported as a percentage. The PERMANOVA test was used for comparisons. Linear Discriminant Analysis Effect Size (LEfSe) was calculated at phylum (D) and genus (E) levels. The P values were FDR adjusted (P < .05) and LDA score was >2.

At the phylum level, we identified a statistically significant increase of Ascomycota for PSC-UC and of Basidiomycota for CTRLs (Figure 2D). At the genus level, Saccharomyces, Sporobolomyces, Tilletiopsis, and Debaryomyces appeared increased in PSC-UC; Piptoporus, Candida, and Hypodontia were increased in the UC subset and Meyerozyma and Malassezia were decreased in both patient groups (Figure 2E).

Correlations Amongst Microbes, Fungi, and Liver Function Test Features

Network analysis showed more abundant bacterial (Figure 3A) connections compared with fungal ones (Figure 3B). In particular, all bacterial correlations appeared positive, with Atopobium, SMB53, Eggerthella, Actinomyces, Blautia, and Lachnospira at the top of the most interconnected nodes. In fungi, the number of the edges was limited, and the nodes appeared interconnected by a low number of edges. A negative correlation linked Saccharomyces with an unidentified fungus.

Figure 3.

Correlations amongst microbes, fungi, and liver function features. Network analysis of intestinal microbiota (A) and mycobiota (B) represented by Pearson’s correlation between PSC-UC, UC, and CTRL groups. The nodes refer to genera; the edges represent the correlation values (filtered for adjusted P < .05) between genera (blue lines, negative correlations; red lines, positive correlations). Nodes are colored according to their relative abundance in each group. Spearman’s correlation amongst bacteria, fungi, and liver function features (C). Red blocks: positive Spearman’s coefficient with P ≤ .005; blue blocks: negative Spearman’s coefficient with P ≤ .005.

Spearman’s correlation test revealed positive correlations: amongst AST, ALT, and Clostridiales, Barnesiellaceae, Odoribacter, Coprobacillus, Methanobrevibacter, and Sporobolomyces; between BMI and Erwinia, Fusobacterium, Serratia, and Acinetobacter; between GGT and Barnesiellaceae, Odoribacter, Methanobrevibacter, Erysipelotrichaceae, Sporobolomyces, Tremellomycetes; between BT and Debaryomyces and Candida.

Finally, negative correlations were found among the Montreal score and Clostridiaceae, Eubacterium, Prevotella and Zygosaccharomyces; and among ASA-5 and Erwinia, Methanobrevibacter, Fusobacterium, Serratia, and Sporobolomyces (Figure 3C).

Microbial, Fungal, and Liver Function Test Features Strictly Classify PSC-UC and UC Patients

We applied the LDA and leave-one-out cross-validation analysis on bacterial, fungal genera, and liver function test features to evaluate their capability to classify patients on the bases of the disease and to identify features able to describe the differences between PSC-UC and UC groups (Table 2). Linear discriminant function analysis creates a function capable of classifying phenomena, considering a series of discriminant variables and a grouping probability. By this approach, the 86.8% of samples were correctly classified in UC or PSC-UC. When the cross-validation test was applied, the capability of the model to classify samples increased to 92.5%. Moreover, we identified the minimum number of features contributing to the correct patients’ classification, that is GGT, BT, Bacteroides, Collinsella, Dorea, Saccharomyces, Zygosaccharomyces, and Entylomatales. The unsupervised principal component analysis (PCA) confirmed the LDA results, clearly assigning Collinsella and Dorea to UC and Bacteroides and Saccharomyces to PSC-UC patients (Figure 4).

Table 2.

Linear discriminant analysis (LDA) applied on microbes, fungi, and liver function features of the 2 patient groups.

	Original Group	Prediction of Grouping
		UC	PSC UC
Original groups	UC	85.2%	14.8%
	PSC-UC	11.5%	88.5%
Cross-validated	UC	100%	0%
	PSC-UC	15.4%	84.6%

Figure 4.

Principal component analysis (PCA) plot. Red dots: PSC-UC samples; blue dots: UC samples; black dots: loadings.

Metabolic Pathway Predictions

A total of 22 metabolic pathways differentially expressed between PSC-UC, UC, and CTRL were resulted by applying the PICRUSt algorithm on gut microbiota composition (Figure 5). In particular, we showed the upregulation in CTRL of pathways belonging to (1) the biosynthesis of unsaturated fatty acids (FAs), ubiquinone and other terpenoid, lipopolysaccharide’s proteins, valine, leucine, isoleucine, phenylalanine, tyrosine, and tryptophan; (2) the metabolism of sphingolipid, glyoxylate and dicarboxylate; (3) the degradation other glycans; (4) bacterial motility proteins, (5) protein folding and associated processing, (6) carbon fixation pathways in prokaryotes, (7) oxidative phosphorylation, and (8) the 2 component system. In PSC-UC, the following pathways were upregulated: (1) biosynthesis of glycosphingolipid (globo series); (2) metabolism of carbohydrates, purine and pyrimidine, and FAs; and (3) homologous recombination. In UC, the metabolism of FAs and fructose and mannose was increased (Figure 5).

Figure 5.

Metabolic biomarkers associated with PSC-UC, UC, and CTRL. A linear discriminant effect size (LeFse) analysis have been performed (α = 0.05, logarithmic LDA score threshold = 2.0).

Discussion

Our study focuses on analyzing clinical features and gut microbiota and mycobiota characteristics in UC pediatric patients with or without PSC compared with healthy subjects (ie, controls) to find microbial and fungal signatures for PSC-UC disease and to explore whether the possible differences may clarify the etiopathogenesis of PSC.

By analyzing the gut microbiota, we observed a reduction of bacterial richness in both PSC-UC and UC groups, highlighting a dysbiotic status of the gut, possibly supporting a correlation between microbial diversity and bowel health. The gut dysbiotic status of UC patients relied in the increase of Klebsiella, Haemophilus, Enterococcus, and Collinsella taxa. The rising of Klebsiella and Haemophilus in UC patients was already described comparing active vs remission stage of the disease.³¹ Moreover, the abundance of Enterococcus in the gut microbiota was associated with colon inflammation in mice models.³² The role of Enterococcus in inducing colitis is probably linked to its capability to produce bile acids (BAs) and to generate reactive oxygen species (ROS).³²

Collinsella is an essential intestinal bacterium that produces UDCA and other secondary BAs.³³ Ursodeoxycholic acid is a treatment for PSC and other cholestatic disorders³⁴ and exerts anti-inflammatory properties by suppressing pro-inflammatory cytokines like Tumor necrosis factor (TNF)-α, interleukin (IL)-1β, IL-2, IL-4, and IL-6.³⁵ To explain the presence of Collinsella in UC, we could speculate 2 hypotheses: Collinsella in UC produces UDCA in the gut, protecting patients against PSC; or, the iatrogenic UDCA, administered to PSC-UC patients, could have an inhibitory effect on the microorganism growth.

Interestingly, in PSC-UC patients, the unique overexpressed microbial biomarker was Streptococcus. Pereira et al demonstrated the association of Streptococcus in bile samples and PSC progression—but not with the onset of the disease.³⁶

Lastly, in our results Akkermansia, Bacteroides, and Parabacteroides emerged as the biomarkers of CTRLs, being reduced in both PSC-UC and UC patients. The anti-inflammatory properties of Akkermansia and its role in maintaining the gut eubiotic status by the stimulation of mucin production are overall well established.³⁷ Moreover, Bacteroides and Parabacteroides are intestinal commensals highly abundant in the healthy human gut with anti-inflammatory properties.³⁸

Additionally, Oscillospira was increased in CTRLs. Notably, the growth of Oscillospira is probably influenced by high bile levels and may contribute to the formation of secondary BAs that have anti-inflammatory properties.³⁹ Finally, Dialister appeared reduced in all patients; its inverse relationship with plasma levels of the pro-inflammatory cytokine IL-6 has been described.⁴⁰ These findings reinforce the evidence of the lacking of some bacterial commensals in PSC-UC and in UC that are involved in anti-inflammatory activities and in BAs’ metabolism against biliary injury.

The metabolic inferred analysis showed the reduction of valine, leucine, and isoleucine biosynthesis in PSC-UC and UC patients. This result is consistent with the study of Kummen et al who observed a reduction of the microbial metabolism of the branched chain amino acids valine, leucine, and isoleucine in adult patients affected by PSC-UC and IBD.¹⁵ Moreover, in our results, purine and pyrimidine metabolism was increased in PSC-UC patients. Interestingly, Xiong and colleagues identified pyrimidine and purine microbial metabolites as important mediators of “gut-liver” axis in promoting liver cirrhosis development.⁴¹ This evidence reinforces the key role of gut bacterial metabolites as intermediaries between gut microbes and PSC disease processes. In this study, the alpha diversity indices were lower for fungi compared with bacteria, indicating a lower richness and evenness of fungal communities in respect to bacteria. This can be confirmed by the co-occurrence analyses that revealed the highest number of nodes and edges of bacteria networks compared with the poorest of fungi; this could be due to the lower number of fungal species in the gut niche compared with the bacteria or even to the simplicity of the fungal ecosystem. This evidence was also reported by Maas and colleagues who showed a lower fungal diversity compared with bacterial diversity and lower Spearman correlation values for a fungal ecosystem compared with a bacterial one, confirming our results.⁴²

Amongst the fungi increased in UC cohort, Candida is an intestinal commensal, which has also been identified as being responsible for the worsening of intestinal inflammation in mouse models of colitis.⁴³ Additionally, in the mouse model of dextran sulfate sodium (DSS)-induced colitis, a positive correlation between the increase of colitis susceptibility and the intestinal fungal load has been described; conversely, an attenuation of colitis was indeed obtained by antifungal treatment.⁴⁴

In our PSC-UC cohort, we detected the increase of Saccharomyces, Sporobolomyces, Tilletiopsis, and Debaryomyces. Our evidence was opposite to that of the Lemmoine et al study, in which the decrease of Saccharomyces was found in a PSC-UC adult cohort.¹⁷ Nevertheless, Muratori et al reported high levels of anti-Saccharomyces cerevisiae antibodies (ASCAs) in PSC patients, independently from the disease stage or the concomitant presence of IBD.⁴⁵

Debaryomyces has been found in inflamed mucosal tissues of patients affected by CD.⁴⁶ Altogether these results showed a pro-inflammatory mycobiota profile in PSC patients. This was also confirmed by the reduction of Meyerozyma and Malassezia in both patient groups. In fact, Meyerozyma was even proposed as a probiotic for its ability to produce antioxidant, anti-inflammatory, antimutagenic, and antitumor substances (eg, phenolic compounds),⁴⁷ and Malassezia was shown to correlate with health subject fecal samples.⁴⁸

Lastly, the application of the LDA classification model on the entire patients’ set, comprising liver function tests, bacterial and fungal features, indicated GGT, BT, Bacteroides, Collinsella, Dorea, Saccharomyces, Zygosaccharomyces, and Entylomatales as variables able to distinguish PSC-UC from UC patients. This result confirms a complex PSC-UC phenotype, not entirely explained by the gut microbiome and mycobiome composition, but rather influenced by all of them.

This is the first study reporting a description of the gut mycobiota in a pediatric PSC cohort. The strength of our study is based on the dimension and homogeneity of the patient cohorts investigated (eg, gender, age, body weight, treatments) from 2 different geographical origins. Even if this were a homogeneous pediatric cohort, a larger sample size, including naïve patients and patients affected by only PSC should be required to exclude the effect of treatments and to validate the gut microbiota and mycobiota profiles associated with the disease. In the end, metabolomic experiments could help in determining the functional profile of host and of the gut microbiota in PSC-UC.

In conclusion, we have shown the presence of a gut microbiota and mycobiota imbalance in PSC patients in the context of an underlying UC presence. The gut dysbiosis may lead to an altered production of secondary BAs and to a pro-inflammatory activity exerted by some pathobionts, all contributing to the etiopathogenesis of pediatric PSC.

Funding

This work was supported also by the Italian Ministry of Health with “Current Research funds.”

Conflicts of Interest

None of the authors have any conflicts of interest to report.