Gut mucosa-associated microbiota signatures in healthy individuals and patients at different stages of liver disease: a pilot study

Patients with HCV-related liver disease (CHC, LC, or HCC) naïve to antiviral treatment who underwent screening colonoscopy for colorectal cancer were prospectively enrolled at the Department of Gastroenterology and Hepatology of the University Federico II of Naples, Italy, between January 2014 and December 2018. Chronic HCV infection was confirmed based on anti-HCV antibodies and HCV RNA levels >12 IU/ml. Liver cirrhosis was diagnosed based on histological findings or the observation of a nodular liver surface, coarse liver parenchymal texture, narrowed vessels with irregular intrahepatic distribution, and clinical and laboratory data. HCC was diagnosed on the basis of pathological findings or a typical dynamic imaging study, according to the European Association for the Study of the Liver Diseases guidelines available at the time of inclusion [24]. Baseline biochemical values, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), albumin, total bilirubin, platelet (PLT) count, and the international normalized ratio (INR), were collected at enrollment. The Child‒Pugh class and Model for End-Stage Liver Disease (MELD) scores were also recorded. Patients were excluded if they had a concurrent cause of liver disease other than HCV infection or decompensated cirrhosis; a history of antibiotics, probiotics, prebiotics, nonsteroidal anti-inflammatory drugs, or proton pump inhibitors use in the previous 60 days; concurrent infections; small bowel/colorectal diseases; a history of gastrointestinal surgery; alcohol intake ≥ 40 g/day for men and ≥ 20 g/day for women; or a body mass index (BMI) >30. Healthy individuals without a history of liver or gastrointestinal diseases, such as inflammatory bowel diseases, irritable bowel syndrome, or colitis, were enrolled among the subjects who underwent screening colonoscopy. Biopsies were collected from the ileum and sigmoid region during endoscopy after standard bowel cleansing (Selg-ESSE 4 L) and immediately frozen at −80 °C for DNA extraction and molecular analysis [25]. The study was conducted in accordance with the Declaration of Helsinki and the Declaration of Istanbul. This study was approved by the University Federico II Ethics Committee of Naples, Italy (Prot. N.220/13). Written informed consent was obtained from all participants. All methods were conducted in accordance with the relevant guidelines and regulations.

Total genomic DNA was extracted from all 84 collected biopsy tissues (42 ileum samples and 42 sigmoid colon samples) using the QIAamp DNA Mini Kit (Qiagen, Venlo, Netherlands) according to the manufacturer’s instructions.

Each tissue sample was first weighed (no more than 25 mg was used), cut into small pieces to obtain d efficient lysis reaction, and resuspended in ATL buffer (provided by the DNA purification kit). Twenty microliters of Proteinase K were added to each tube, and each sample was incubated at 56 °C until the tissue was completely lysed. After this step, 4 µl of RNase A and 200 µl of AL Buffer were added to each sample, and each tube was incubated at 70 °C for 10 min. At this point, 200 µL of ethanol was added to each tube, and three subsequent steps of centrifugation using different buffers (AW1 and AW2) were applied to spin the columns. Finally, 200 µL of distilled water was added directly to the filter to elute the DNA from each sample. Each filter was incubated at room temperature for 1 min and then centrifuged to obtain purified DNA samples. Subsequently, a quantitative assessment was performed using the NanoDrop spectrophotometer (Thermo Fisher Scientific, Massachusetts, USA), and the concentration of the samples ranged between 4 and 32 ng/µl , while the 260/280 and 260/230 values ranged between 1.5 and 2.

16S rDNA libraries were constructed via a 2-step PCR-based protocol following the Illumina 16S Metagenomic Sequencing Library Preparation guidelines (Illumina, San Diego, CA). The first PCR allowed the amplification of a custom 548 bp amplicon covering the V4-V6 hypervariable regions of the 16S rRNA gene, as previously described. The primers used in the first round of PCR also contained overhang sequences with Illumina adapters: forward primer: 5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCAGCAGCCGCGGTAAT-3’; reverse primer: 5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGGTTGCGCTCGTTGC- 3’.

In detail, the PCRs were incubated at 95 °C for 10 min, followed by 30 cycles at 95 °C for 30 s, 59 °C for 30 s, and 72 °C for 1 min. The amplicons obtained were purified via AMPure XP magnetic beads (Beckman Coulter, Milan, Italy). A second round of PCR was used to add Illumina index barcodes (unique sequences required for multiple sample pooling and for cluster generation and sequencing) to the amplicons according to the Nextera XT protocol (Illumina, San Diego, CA). The PCR conditions were as follows: 95 °C for 3 min; 8 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 30 s. A second purification step was set up, and the quality of each library profile was assessed via a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA) and quantified using Qubit 3.0 fluorometer (High Sensitivity assay). Each library was diluted to 10 nM and pooled to obtain a final pool of 10 nM. Two sequencing runs were performed via the Illumina MiSeq instrument using a paired-end (PE) layout with 600 (i.e., 2 × 300) cycles on the basis of our experience [26, 27].

Illumina adapters were removed using trim-galore! (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). Considering the overall quality of reverse reads, which was poorer than that of the forward ones [28], to obtain satisfactory overlapping between paired-end reads, reverse reads were discarded, and we focused the analysis on the forward reads, as indicated by Nardelli et al. [29].

The resulting retained trimmed forward reads were denoised via the DADA2 procedure [30]. This strategy included chimera removal (i.e., PCR artifacts [31, 32] and PhiX removal (i.e., the PhiX phage is used during Illumina library preparation to increase nucleotide variability removal [33]. The obtained amplicon sequence variants (ASVs) were taxonomically annotated in BioMaS [34] via release 22.10 of the GreenGenes2 database [35]. In particular, the query sequences were aligned to the reference collection via Bowtie2 [36], and the resulting alignments were filtered according to query coverage (≥ 70%) and identity percentage (≥ 90%). Taxonomic classification was performed via TANGO [37]. In particular, for ASV sequences with identity percentages equal to or greater than 97%, classification at the species level was accepted [38]; otherwise, the ASVs were classified at higher taxonomic ranks. The ASV table was normalized via rarefaction [39] for diversity analysis.

The Shannon, inverse Simpson, Faith PD, and observed ASV indices (alpha diversity) were measured on rarefied ASV tables via the phyloseq R package [40], and significant differences between groups were evaluated via the Kruskal‒Wallis and Wilcoxon tests. Principal coordinate analysis (PCoA), which describes the diversity between samples (i.e., beta diversity) based on weighted and unweighted UniFrac [41] metrics, was performed via the vegan R package [42] and evaluated via PERMANOVA.

Microbiota-based metabolic pathway prediction was performed via the PICRUst2 pipeline [43].

Differential abundance (DA) analysis was performed via DESeq2, and the “poscount” option for size factor estimation was used to address the compositionality of the DNA metabarcoding data. Multiple testing corrections were performed via the Holm‒Bonferroni method to control the FDR. Statistical comparisons were performed at the phylum, class, order, family, genus, and species levels and for inferred metabolic pathways. Taxonomic and functional counts were then stratified for the ileal and sigmoid samples. Statistical comparisons were corrected for batch effects.

Among the 260 consecutive patients diagnosed with HCV, 231 were excluded according to the predefined exclusion criteria (see Figure S1).

Twenty-nine patients with HCV (6 with CHC, 9 with LC, and 14 with HCC) met the inclusion criteria. Thirteen HCs were enrolled in this study. The demographic, laboratory, and clinical findings of the study population are presented in Table 1.

There were no significant differences in terms of age, sex, or BMI between the patient and HC groups. Patients with LC and HCC had higher levels of AST (p = 0.026), ALT (p = 0.027), -GGT (p = 0.020), and bilirubin (p = 0.011) and lower platelet counts (p < 0.001) and albumin levels (p < 0.001) than did those with CHC. No significant differences in HCV genotype, viral load, ALP, INR, or LC prognostic score (i.e., Child or MELD) were observed among the groups.

In sigmoid MAMs, α-diversity did not differ significantly between CHC and HC, but it was significantly reduced in LC and HCC, irrespective of the diversity metric applied (Shannon, Simpson, or Faith’s phylogenetic diversity [FPD]; p ≤ 0.05). According to the Shannon and Simpson indices, both richness and evenness were lower in HCC compared with HC (p ≤ 0.05), while the FPD index revealed that the LC group had significantly lower diversity than the CHC group (p ≤ 0.05) (Fig. 1, upper panel).

In the ileal MAM analysis, the FPD index α diversity was significantly lower in the LC group compared with the HC group (p ≤ 0.05); according to the FPD index and observed ASVs, the microbiota diversity of the LC group was significantly lower than that of the CHC group (p ≤ 0.05) (Fig. 1, lower panel).

UniFrac distances did not highlight distinct clusters among groups (Figure S2-S5). However, according to the unweighted UniFrac metric in the sigmoid colon, we observed stratification into two main groups along the first component (approximately 18% of the explained variance). Specifically, the HC and CHC samples tended to co-cluster on the left side of the plot, whereas the LC and HCC samples were scattered on the right side. PERMANOVA based on the weighted UniFrac distance matrix revealed a significant association between the microbial community and the groups analysed (HC, CHC, LC, and HCC) in both ileal and sigmoid colon samples, accounting for 14.4% and 17.1% of the observed variance, respectively. Similarly, PERMANOVA revealed a significant association between the microbial community and disease status in both the ileum and sigmoid colon samples, accounting for 13.0% of the observed variance. Regardless of the tissue or proposed model, a significant batch effect was observed (p ≤ 0.05) (Table 2).

PERMANOVA pairwise comparisons highlighted significant differences between HC and patients with HCC, irrespective of the metric applied orthe sampling site. Furthermore, a significant difference was detected between HC and LC in sigmoid colon samples and between CHC and HCC in ileal samples based on the weighted UniFrac metric(Table 3).

In summary, α- and β-diversity results revealed that the MAM diversity was significantly decreased in the most advanced stages of liver disease compared with HC or CHC.

Phylotypes with a median relative abundance greater than 1% of total sequences were included in the analysis. The taxonomic distribution of predominant bacteria at different phylogenetic levels is shown in Fig. 2 and Table S1. Statistical taxonomic analysis was conducted at the phylum and genus levels (Table S2).

The phyla Bacteroidetes, Firmicutes_A, and Proteobacteria dominated all study groups, together accounting for approximately 90% of sequences. The relative abundance of Bacteroidetes transiently increased in CHC but declined in more advanced stages (LC and HCC). Conversely, Firmicutes_D (class Bacilli) and Proteobacteria increased in LC and HCC compared with both HC and CHC. Fusobacteria, nearly absent in HC and CHC, reached median relative abundances of 2.5% in LC and 3.3% in HCC.

Differential abundance analysis using DESeq2 (Fig. 3) revealed that Firmicutes_A and Firmicutes_D were significantly enriched in LC and HCC, while Desulfobacterota_I and Verrucomicrobiota were reduced. Compared with CHC, HCC was enriched in Firmicutes_D, Actinobacteria, and Proteobacteria, whereas Verrucomicrobiota was depleted. Between LC and HCC, Desulfobacterota_I, Actinobacteria, and Firmicutes_D were higher in HCC, while Firmicutes_A and Verrucomicrobiota were lower.

At the genus level, Bulleidia, Pantoea, Clostridium_Q, Rothia, and Streptococcus were consistently enriched in HCC, whereas Akkermansia, Butyricimonas, and Victivallis were more abundant in HC (Fig. 3A–F). Collectively, these data suggest that microbial communities in advanced liver disease are characterised by loss of beneficial taxa and enrichment of potentially proinflammatory genera.

The log₂-fold-change values relative to pairwise comparative analysis performed at the genus level between the groups are reported in Table S2.

The metagenomic functional content, which was determined based on the microbial community profiles from the 16 S rRNA gene sequences, was obtained via PICRUSt2. The results are presented in Fig. 4 and Table S3.

Overall, significant differences were observed in 103 pathways, with the main differences observed between HCC patients and LC patients (n = 48), followed by HCC patients and HC (n = 25).

The proportion of genes involved in degradative pathways (toluene, L-arabinose, taurine, and chitin derivatives) was significantly higher in HCC than in earlier stages of HCV-related liver disease. Compared with CHC and LC, HCC samples also showed greater representation of nutrient-processing pathways, including the ethylmalonyl-CoA pathway, taurine degradation, and glycine betaine degradation.

The taxonomic distributions of the predominant bacteria among the study groups at the different phylogenetic levels are shown in Fig. 5 and Table S4. Taxonomic analysis was conducted at the phylum and genus levels to determine the composition of the prevalent microbiota. (Table S5)

The taxonomic composition of sigmoid MAM mirrored that of the ileum, with Bacteroidetes, Firmicutes_A, and Proteobacteria representing the predominant phyla (Fig. 5, Table S4). Bacteroidetes transiently increased in CHC but declined in LC and HCC. In contrast, Proteobacteria, Firmicutes_D, and Fusobacteriota increased progressively from HC to HCC.

Overall, differential abundance analysis highlighted significant differences among groups in four phyla, three classes, eight orders, eight families, 42 genera, and 50 species (Fig. 6 and Table S5), with the main differences observed in the comparisons between HCs and LCs (n = 75 taxa), followed by HCCs (n = 62 taxa).

At the genus level, Rothia, Peptostreptococcus, and JAAFIP01 were enriched in CHC, whereas Clostridium_Q_135853, Clostridioides_A, Pantoea_A_680069, and Rothia were enriched in LC. HCC samples displayed marked enrichment of JAAFIP01, Merdimonas, Rothia, Anaerococcus, Blautia_A_141780, Clostridium_Q_135853, Clostridioides_A, Catenibacterium, and Streptococcus. Conversely, Akkermansia, Eubacterium_F, and Victivallis were more abundant in HC and CHC.

These findings indicate a progressive restructuring of MAM with disease severity, characterised by loss of commensal taxa and enrichment of genera involved in inflammation and mucosal disruption.

The log₂-fold-change values relative to pairwise comparative analysis performed at the genus level between the groups are reported in Table S5.

Overall, 398 of the 403 predicted metabolic pathways were observed in the sigmoid samples (Fig. 7 and Table S6), with the main differences observed in the comparison between CHC and HCC (n = 74). Eight out of the 13 pathways were degradative pathways, and a common trend was not appreciable, with the only exceptions being the S-methyl-5-thio-alpha, D-ribose 1-phosphate degradation and L-methionine salvage cycle III pathways, which were less abundant in HC than in the other three stages. Enrichment of pathways such as L-arabinose degradation IV, L-rhamnose degradation II, carbohydrate metabolism, S-methyl-5-thio-alpha, D-ribose 1-phosphate degradation, and L-methionine salvage cycle III was observed in HCC when compared to LC.

When we compared the ileal and sigmoid colon MAM compositions, we did not observe significant differences at higher taxonomic levels (i.e., phylum and class). Micrococcaceae and its genus Rothia were more abundant in the ileum than in the sigmoid colon (Table S7).

Functional pathway comparison revealed only one significant difference: the starch degradation III pathway was more highly expressed in the sigmoid colon (log₂ fold change = 20.9, adjusted p = 0.0001) (Table S8).