The applicability, feasibility, and validity of the WGA method to enrich microbial DNA
We first analyzed the applicability of the WGA method for characterizing oropharyngeal microbial DNA samples. We compared the DGGE profiles between WGA-amplified DNA (Fig. 1a, lanes A1, B1, C1, and D1) and the original DNA (Fig. 1a, lanes A0, B0, C0, and D0) and found that the WGA-positive samples comprised more bands than those of WGA-negative samples, suggesting that the WGA method was capable of enriching oropharyngeal microbial DNA samples for DGGE analysis.
We then determined the reproducibility of the WGA method. Two fecal microbial DNAs were diluted 1:100 (3 ng/μL) and amplified in triplicate using the WGA method. The DGGE profile analysis indicated that the WGA method amplified fecal DNA (Fig. 1b, lanes A1–3 and B1–3), the most predominant bands were similar to those of the original fecal DNA (Fig. 1b, lanes A0 and B0), and the results were reproducible. These results prove that the WGA method effectively enriched low concentrations of microbial DNA suitable for DGGE analysis.
The validity of the WGA method was further characterized using DGGE profiling and cluster analysis. Seven oropharyngeal microbial DNA samples were amplified in duplicate using WGA. Cluster analysis showed that the DGGE profiles were >90 % similar between duplicates (Fig. 1c), suggesting the validity and reproducibility of the WGA method for preparing small quantities of DNA for DGGE analysis.
Determination of the temporal stability of the oropharyngeal microbiome in a 3- week follow-up study
To investigate the associations between the oropharyngeal microbiome and disease, we first confirmed the temporal stability of oropharyngeal microbiomes of the HC and CC groups during a follow-up period. We chose randomly 32 subjects, including 10 volunteers from the HC group (three times), 10 patients from the CC group (three times, most were discharged from our hospital before the fourth appointment.), and 12 patients from the CI group (four times). The first sample from the patients in the CI group were before antibiotic treatments, the second samples were during antibiotic treatments, and the third and fourth were after antibiotic treatments.
The DGGE profiles and cluster analysis demonstrated that three follow-up samples of each individual from the HC or CC groups (n = 10) clustered together, respectively (Fig. 2a). These results suggest that the oropharyngeal mucosal microbiome of each control individual exhibited relatively stable patterns during the follow-up (Fig. 2a). In contrast, because of antibiotic treatment, four follow-up samples from patients with pneumonia from the CI group (n = 12) did not cluster together (Fig. 2b). However, the DGGE profiles of each patients with pneumonia were similar between their third and fourth visits (Fig. 2b), suggesting that the oropharyngeal microbiome was relatively stable in patients after antibiotics treatments.
Moreover, the diversity of the oropharyngeal microbiome among different groups was calculated using Past software. The predominant microbiome of patients with pneumonia from group CI had the highest diversity (CI-1), followed by control patients with liver cirrhosis (group CC) (Fig. 2c). The diversities of groups CI and CC were higher compared with that of the group HC. Notably, oropharyngeal microbial diversity was decreased in group CI-2 during antibiotic treatments versus group CI-1 before antibiotic treatments.
To investigate oropharyngeal microbial variation and identify the key bacteria associated with liver cirrhosis and pneumonia, we analyzed the DGGE profiles of the 90 subjects.
Cluster analysis of DGGE profiles
Cluster analysis of DGGE profiles indicates that almost all individuals in each group (except for three samples from the group CC) clustered together, suggesting that the microbial composition of each individual in the same group was similar to the others and that the microbial composition of patients in the group CI differed from those of both control groups. Notably, the DGGE profiles of all patients in group CI clustered together at high UPGMA coefficient values ranging from 57.7 to 94.0 % (average, 82.30 ± 9.85, Fig. 3a). These results were confirmed using MDS analysis (Fig. 3b) and PCA (Fig. 3c). For example, note the overlap of symbols representing the microbiomes of patients infected with the same pathogen in the group CI.
Analysis of microbial diversity
We used Past software to assess the microbial diversity of the oropharyngeal mucosa using Shannon’s diversity index, Species richness, and Shannon’s evenness index. The values of Shannon’s diversity index, Species richness, and Shannon’s evenness index were obviously higher in group CI compared with those in groups CC and HC (p < 0.01) (Fig. 3d). Further, Shannon’s diversity index and Species richness were higher in group CC compared with group HC (p < 0.01).
Phylogenetic analysis of DNAs isolated from DGGE profiles
In the 90 PCR-DGGE profiles analyzed in this study, 39 band-classes were identified (Fig. 4a). Firmicutes (20 band-classes) was the most common phylum, followed by Actinobacteria (8 band-classes), Bacteroidetes (4 band-classes), Proteobacteria (4 band-classes), and Fusobacteria (3 band-classes) (Fig. 4b). Seven band-classes were highly prevalent (median intensity in at least one of the groups was >2 %), including band-classes 30.1, 33.0, and 65.6 of Streptococcus, band-class 8.3 of Fusobacterium, band-class 59.4 of Veillonella, as well as band-classes 80.6 and 63.0 of Actinomyces, in which there was little variance among band-classes 65.6, 59.4, and 80.6 (Additional file 3: Table S3).
Identification of bacterial species that account for the variation in the oropharyngeal microbiome
To identify the key bacterial species that shifted composition, we calculated the intensity and the frequency of each band in the 90 PCR-DGGE profiles and analyzed the variation of each band-class. We found that the intensities of 19 band-classes differed significantly (Fig. 5a), and the frequencies of 14 band-classes varied (Fig. 5b) among the three groups. According to our analysis of the intensities and frequencies of the variable band-classes, six key band-classes were identified that reflected the differences between the CI and each control group. The abundances of five band-classes 4.8 (Bacteroides sp.), 36.8 (Eubacterium sp.), 43.3 (Lachnospiraceae sp.), 54.9 (Neisseria sp.), and 63.0 (Actinomyces sp.) were much higher in the group CI compared closely with those in each control group, while the abundance of band-class 30.1 (Streptococcus sp.) in group CI was significantly lower compared with each control group (P < 0.017 with modified Bonferroni correction the P –values were shown in Supplementary Table S3.)
To assess the effects of liver cirrhosis on band-class distribution, we analyzed the differences in the abundances of the band-classes between groups CC and HC and found that five band-classes, including 26.5 (Bulleidia sp.), 33.0 (Streptococcus sp.), 34.5 (Haemophilus sp.), 45.4 (Lactobacillus sp.), and 90.9 (Olsenella sp.) were higher in abundance in group CC versus group HC, while the abundances of band-classes 32.4 (Campylobacter sp.) and 63.0 (Actinomyces sp.) decreased in group CC versus group HC.
Quantitative analysis of the disease-associated differences in the original (non-WGA) DNA
To verify the key bacteria present in infected subjects, which were associated with oropharyngeal microbial variation, quantitative PCR was used to analyze the original (non-WGA) DNA samples using bacterial species-specific primers. The abundances of Bacteroides sp., Neisseria sp., and Actinomycetes sp. and those of Streptococcus sp. in the oropharyngeal mucosa of group CI were higher and lower, respectively, compared with those of groups CC and HC (3.23, IQR 2.07–4.10 versus 1.08, IQR 0–1.91 and 1.97, IQR 1.27–2.85 for Bacteroides sp.; 1.78, IQR 1.33–3.60 versus 0, IQR 0–1.10 and 0 IQR 0–0 for Neisseria sp.; 2.42, IQR 0–3.77 versus 0, IQR 0–3.02 and 0, IQR 0–3.18 for Actinomycetes sp.; 3.26, IQR 2.48–4.66 versus 4.74, IQR 4.08–6.03 and 4.03, IQR 3.01–5.44 for Streptococcus sp.) (p < 0.05) (Fig. 6). The abundances of Bacteroides sp. and Streptococcus mitis were higher in group CC versus group HC (p < 0.05).