The fecal microbiomes of six HCV patients and eight healthy individuals (Table 1; Additional file 1: Table S1), from the city of Beni Suef in Egypt, were sequenced by high-throughput
Illumina MiSeq. High-throughput sequencing generated 651,912 reads that passed initial
quality filtering (median reads per sample = 46,219; median read length = 253 bp).
The quality of these sequence data was checked by standard quality control procedures,
(e.g., total number of input sequences; barcode not in mapping file; reads too short
after quality truncation; count of N characters exceeding limit; Illumina quality
digit; barcode errors exceeding max; median sequence length), and the filtered data
that passed quality control were used for analysis (primary taxon assignments are
provided in Additional file 2: Table S2).
Even though some samples had much fewer reads than others, rarefaction curves indicated
that the sampling depth was sufficient for data comparison, as all samples reached
an asymptote (Additional file 3: Figure S1).
Primary data validation
At first, we sought to validate the raw 16S sequence data sets. In the absence of
robust published high-throughput data specifically representing Egyptian gut microbiome
composition, we chose to compare our samples to well-documented and published gut
microbiome sequences, such as the American Gut samples (available from https://www.github.com/biocore/American-Gut), to validate the results of taxonomic assignments in our samples. Although those
samples are from individuals living in different geographical area and consuming different
type of diet, the data set is still a valid representative of human gut microbes.
The comparison of our 14 samples to the American Gut samples showed that the major
taxa usually associated with the human colon environment (e.g., phyla Firmicutes and
Bacteroidetes; genera Faecalibacterium, Ruminococcus, and Bacteroides) were shared by both data sets—yet in different proportions (Additional file 3: Figures S2, S3). For example, the overall Firmicutes-to-Bacteroides ratio in the
Egyptian samples was 1.1 (vs. 1.4 in the American Gut samples).
Alpha-diversity analysis
Using different diversity indices (e.g., Chao, Shannon, or Simpson), we found that,
regardless of the used metric, the richness and diversity of healthy individuals’
microbiota were higher than those of HCV patients (Fig. 1; Additional file 3: Figure S4). This finding agrees with several other human microbiome studies in which
chronic inflammation tends to decrease biodiversity at different microbiome sites
26]–28].
Fig. 1. Alpha diversity estimation in patients vs. control group (Chao and Shannon indexes)
Distinct core microbiomes differentiate healthy controls from HCV patients
Core bacterial taxa shared by each group (healthy controls and HCV patients) were
identified. Overall, 22 distinct OTUs were conserved among all samples, constituting
a core gut microbiome. This core set is characterized by genera Streptococcus, Ruminococcus, Clostridium, Faecalibacterium, Bacteroides, Blautia, and Lachnospira in addition to some undefined members of families Enterobacteriaceae and Clostridiaceae.
Most of these genera were differentially distributed among healthy and HCV patients
(Fig. 2a). Overall, 22 OTUs were shared by both groups, 23 distinguished healthy controls,
and 31 distinguished HCV patients (Fig. 2b).
Fig. 2. Core microbiomes of analyzed samples. a Main taxa of the core microbiome of all fecal samples and their relative distribution
in healthy controls (blue) compared to patients with HCV (orange). Left panel average relative abundance of 16S sequence reads representing core taxa (in percent).
Right panel actual values of the average proportions of 16S sequence reads representing core
taxa per group. b Venn diagram representing the core OTUs (genus level) in each analyzed group and
their intersection
Major taxonomic differences between microbiomes of healthy individuals and HCV patients
The ultimate goal of this study was identifying consistent differences in microbial
composition between the two analyzed groups (healthy controls and HCV patients).
On the phylum level, a mild but significant increase was observed in the ratios of
Bacteroidetes among HCV patients (Kruskal–Wallis p = 0.039), whereas Firmicutes were slightly more abundant in healthy controls (Fig. 3a; Additional file 3: Figure S5); yet that observed overabundance of Firmicutes is not statistically significant
(Kruskal–Wallis p = 0.301).
Fig. 3. Boxplots representing the average proportion of each 16S sequence read attributed to each
taxon between the two groups (Blue healthy control samples; Red patient samples). a On the phylum level, b on the genus level—major taxa; c on the genus level—minor taxa; d on the species level—selected taxa
Genus-level analysis, however, was more informative (Fig. 3b, c; Table 2). It revealed that genus Prevotella was clearly enriched in HCV patients (p = 0.038), possibly inflating the total Bacteroidetes abundance observed on the phylum
level. Other minor genera that were also significantly overabundant in HCV patients
are Acinetobacter, Veillonella, and Phascolarctobacterium (Table 2). In addition, Faecalibacterium was another genus with higher abundance in HCV patients than in healthy controls;
yet, Faecalibacterium abundance was less consistent among HCV patients. On the other hand, genera Ruminococcus and Clostridium were more abundant in healthy controls (Fig. 3b, c; Additional file 3: Figure S6). Interestingly, two of the healthy controls had relatively high abundance
of the probiotic genus, Bifidobacterium, which was undetected in any of the HCV patients.
Table 2. Genera that are statistically significantly different between the two groups (non-parametric
t test)
Some other sample-specific peculiarities are worth mentioning. For example, one patient
had a fair amount of phylum Fusobacteria, which has been described as a biomarker
of colon cancer 29]. Another individual exhibited an unusual overabundance of phylum Actinobacteria (Additional
file 3: Figure S5).
In sum, although only one single phylum was statistically significantly different
between patients and controls, and although only three genera were clearly differential,
the OTU differences were sufficient to separate most cases into two distinct clusters,
as revealed by principal coordinate analysis (Fig. 4). All patients, except P1, clustered together, while all healthy controls but H7
clustered together.
Fig. 4. Principal coordinate analysis representing the beta diversity estimated by the weighted
UNIFRAC method 30]. Each sphere represents one sample (Blue healthy control, H1–H8; Red patients with HCV, P1–P8). The three principal coordinates (PC1 through PC3) explain
55.68, 10.28 and 9.72 %, respectively
The clustering was mostly affected by the relative abundance of Prevotella, since patient 1 (P1) coincidentally had no detectable Prevotella OTUs while healthy control 7 (H7) had an unusually higher proportion than the rest
of the healthy control group.
However, this patient (P1)—in particular—had the highest proportion of Faecalibacterium, possibly suggesting that the combined proportion of Prevotella and Faecalibacterium may be a good biomarker/predictor of the HCV-associated microbiome.
To run a full, unbiased investigation of which OTUs can serve as biomarkers, we used
the LEfSe classification tool. This analysis was able to pick some of the minor OTUs,
which were not as obvious in the taxon chart analyses (Fig. 3; Additional file 3: Figures S5, S6), and defined a list of taxa as potential biomarkers for the healthy
vs. HCV groups. For example, the bacteroidetes can serve as biomarkers for HCV patients
on the phylum, order, and class level while a few taxa were markers of the healthy
microbiome, most prominent of which are genera Bifidobacterium, Ruminococcus and phylum Tenericutes (Fig. 5).
Fig. 5. LEfSe classification analysis