3.1 demographic characteristics of the participants
There were 98 participants enrolled in this study, including 33 INRs, 35 IRs and 27 HCs. The demographic characteristics of PLWH and HCs were shown in Table 1. There was no significant difference between HIV group and HCs in BMI, age, gender, and so on. Table 2 demonstrates the demographic characteristics of the INRs and IRs. There were statistically significant differences between INRs and IRs in recent CD4+ T-cell count (P < 0.001), baseline CD4+ T-cell count (P < 0.001), and nadir CD4+ T-cell counts (P < 0.001). There were no significant differences in gender, ART duration, ART regimen, and so on between the two groups(P > 0.05).
Table 1 Demographic characteristics of PLWH and HCs
|
Variables
|
HC (N=27)
|
HIV (N=68)
|
χ2/Z
|
P-value
|
Age (years)
|
29.0(25.0,54.0)
|
39.0(34.0,50.8)
|
-1.379
|
0.168a
|
BMI (kg/m2)
|
22.6(21.5,23.8)
|
22.0(20.6,25.2)
|
-0.900
|
0.368a
|
Gender
|
|
|
-
|
0.347c
|
Female
|
3(11.11)
|
3(4.41)
|
|
|
Male
|
24(88.89)
|
65(95.59)
|
|
|
Education level
|
|
|
2.441
|
0.295b
|
Junior high school and bleow
|
9(33.33)
|
13(19.11)
|
|
|
Senior high school
|
5(18.52)
|
12(17.65)
|
|
|
College or above
|
13(48.15)
|
43(63.24)
|
|
|
Marital status
|
|
|
1.685
|
0.243 b
|
Unmarried/divorced/widowed
|
14(51.85)
|
45(66.18)
|
|
|
Married/remarried
|
13(48.15)
|
23(33.82)
|
|
|
a Wilcoxon rank sum test; b chi-square test; c Fisher's exact test
Abbreviations: HCs, healthy controls
Table 2 Demographic characteristics of INR and IR
|
Variables
|
INR (N=33)
|
IR (N=35)
|
χ2/Z
|
P value
|
Age (years)
|
41.0(35.0,55.0)
|
39.0(33.0,48.0)
|
-1.173
|
0.241a
|
BMI (kg/m2)
|
22.1(19.9,24.0)
|
22.0(21.0,25.4)
|
-0.841
|
0.401 a
|
ART duration (years)
|
4.3(3.0,6.8)
|
4.7(2.7,6.8)
|
-0.202
|
0.840 a
|
Recent CD4+T-cell count (cells/μL)
|
253.0(202.0,297.0)
|
548.0(474.0,626.0)
|
-7.087
|
<0.001 a
|
Baseline CD4+T-cell count (cells/μL)
|
140.0(40.5,196.0)
|
344.5(238.5,425.8)
|
-5.502
|
<0.001 a
|
Nadir CD4+ T-cell count
|
111.0(40.5,162.0)
|
333.0(181.3,398.2)
|
-4.921
|
<0.001 a
|
ART regimen
|
|
|
-
|
0.146b
|
PIs-based
|
5(15.15)
|
12(34.29)
|
|
|
INSTIs-based
|
6(18.18)
|
3(8.57)
|
|
|
NNRTIs-based
|
22(66.67)
|
20(57.14)
|
|
|
Gender
|
|
|
-
|
1.000 c
|
Male
|
32(96.97)
|
33(94.29)
|
|
|
Female
|
1(3.03)
|
2(5.71)
|
|
|
Education levels
|
|
|
0.041
|
0.980b
|
Junior high school and below
|
6(18.18)
|
7(20.00)
|
|
|
Senior high school
|
6(18.18)
|
6(17.14)
|
|
|
College or above
|
21(63.64)
|
22(62.86)
|
|
|
Marital status
|
|
|
0.185
|
0.667b
|
Unmarried/divorced/widowed
|
21(63.63)
|
24(68.57)
|
|
|
Married/remarried
|
12(36.36)
|
11(31.43)
|
|
|
a Wilcoxon rank sum test; b chi-square test; c Fisher's exact test
Abbreviations: INRs, immunological non-responders; IRs, immunological responders; PIs, protease inhibitor; INSTIs, integrase strand transfer inhibitors; NNRTIs, non-nucleoside reverse transcriptase inhibitors.
3.2 Gut Microbiome Diversity and Composition
3.2.1 α-diversity
α-diversity of the gut microbiota was indicated by the species richness indices (ACE, Chao1) and species evenness index (Shannon, Simpson). The analysis revealed that lower in species richness indices (ACE, Chao1) were observed in HIV group compared to HCs (P < 0.05), but no statistically significant differences (P > 0.05)were found in species evenness index (Shannon, Simpson) (Fig. 1). Significantly difference of gut microbial α-diversity between INRs and IRs groups was not observed (P > 0.05) (Fig. S1).
Fig 1 α-diversity of bacterial between PLWH and HCs. ***P<0.001.(A) ACE; (B) Chao 1; (C) Shannon; (D) Simpson.
3.2.2 β-diversity
The β-diversity in the HIV group and the HCs group as well as the INRs group and the IRs group were compared using PCoA of the weighted UniFrac distance. For β-diversity, the findings indicated that the composition of bacterial communities varied markedly between PLWH and HCs (R2 = 0.033, P = 0.036) (Fig. 2A), while no significant differences were observed between IRs and INRs (R2 = 0.006, P = 0.777) (Fig. 2B).
Fig 2 β-diversity of bacterial between PLWH and HCs. (A) PLWH and HCs; (B) IRs and INRs.
3.2.3 Compositional analysis of fecal microbiota
Average relative abundances data for each bacterial at phylum and genus between HIV-infected patients and HCs as well as INRs and IRs are showed respectively in Fig.3-Fig.6.
Fig. 3 Community structure of HIV-infected patients and HCs. (A) phylum level; (B) family level; (C) genus level.
At the phylum level, the top five abundant bacterial phyla in HCs were Bacteroidetes, Firmicutes, Proteobacteria, Actinobacteria, Verrucomicrobia (Fig. 3A). The top five abundant bacterial phyla in HIV-infected patients (both INRs and IRs) were Firmicutes, Bacteroidetes, Proteobacteria, Fusobacteria, Actinobacteria (Fig. 3A, Fig. 5A). The top 10 phylum of abundance were selected for comparative analysis. Fusobacteria (3.341% vs. 0.048%, P < 0.001) and Actinobacteria (1.257% vs. 1.069%, P = 0.023) resulted more abundant in PLWH than HCs, Verrucomicrobia (0.023% vs. 0.282%, P = 0.004) resulted less abundant in PLWH than HCs (Fig. 4A). At the phylum level, no statistically difference was observed between IRs and INRs (Fig. 6A).
Fig. 4 Comparison of key species differences between HCs and HIV-infected patients. (A) phylum level; (B) family level; (C) genus level. *P < 0.05, **P < 0.01, *** P <0.001.
At the family level, the top five abundant bacterial family in HCs, INRs, and IRs were Prevotellaceae, Ruminococcaceae, Bacteroidaceae, Lachnospiracea, Verrucomicrobiaceae (Fig. 3B. Fig. 5B). At the family level, Verrucomicrobiaceae (21.715% vs. 7.994%, P = 0.037), Acidaminococcaceae (4.776% vs. 2.175%, P = 0.028), Fusobacteriaceae (3.341% vs. 0.048%, P < 0.001) are more abundant in PLWH than those in HCs, while Fusobacteriaceae (3.341% vs. 0.048%, P < 0.001) are more abundant in HCs than those in HIV-infected patients (Fig. 4B). Only Sutterellaceae (0.852% vs. 0.558%, P = 0.022) were more depleted in INRs than those in IRs (Fig. 6B).
Fig. 5 Community structure of IRs and INRs. (A) phylum level; (B) family level. (C) genus level.
At the genus level, the top five abundant bacterial genera in HCs were Prevotella, Bacteroides, Faecalibacterium, Megamonas, Roseburia (Fig.3C). The five most prevalent bacterial genus in INRs were prevotella, Megamonas, Bacteroides, Megasphaera, phascolarctobacterium (Fig. 5C), as well as which in IRs were prevotella, Bacteroides, Megamonas, Fusobacterium, Faecalibacterium (Fig. 5C). At the genus level, Megamonas (15.291% vs. 4.356%, P = 0.004) and Megasphaera (3.414% vs. 0.344%, P = 0.027) are higher in PLWH than in HCs. The relative abundances of Faecalibacterium (4.129% vs. 14.478%, P < 0.001), Roseburia (2.235% vs. 3.803%, P = 0.032) and Dialister (2.198% vs. 2.601%, P = 0.002) are higher in HCs than in HIV-infected patients (Fig. 4C). At the genus level, the difference in bacterial abundance between IRs and INRs was not statistically (Fig. 6C).
Fig. 6 Comparison of key species differences between IRs group and INRs group. (A) phylum level; (B) family level; (C) genus level. *P < 0.05.
In order to identify specific microbiota taxa between PLWH and INRs, the bacterial composition between PLWH and HCs as well as INRs and IRs were compared using the LDA effect size (LEfSe) algorithm. The threshold of two on effect size was used. Twenty-five microbiota taxa were identified to distinguish HIV patients and thirty-one were identified to distinguish HCs. Figures 7A, 7B show taxonomic cladograms representing the microbiota structure and predominant bacteria in PLWH and HCs. The predominant bacteria of PLWH and HCs in phylum, class, order, family, and genus are presented in Table S1. Six bacterial taxa were identified to distinguish INRs and four were identified to distinguish IRs. Figures 7C, 7D show taxonomic cladograms representing the microbiota structure and predominant bacteria in INRs and IRs. The predominant bacteria of INRs and IRs in phylum, class, order, family, and genus are presented in Table S2.
Fig. 7 Taxonomic differences between gut microbiota of PLWH and HCs as well as IRs and INRs. Differentially abundant bacterial taxa quantified as LEfSe and only taxa with an LDA score >2.0 are shown. (A) Taxonomic differences between PLWH and HCs; (B) Taxonomic cladogram of the data shown in panel A; (C) Taxonomic differences between IRs and INRs; (D) Taxonomic cladogram of the data shown in panel C.
3.4 Exploring the potential of the gut microbiota as a diagnostic biomarker for INRs
Screening of the intestinal flora as a diagnostic marker of INRs by means of Random Forest (RF) model. In the study constructed ROC based on the intestinal flora signature that screened five OTUs of Clostridium_XlVa, Streptococcus, Roseburia, parabacteroides for inclusion in the model as optimal diagnostic biomarkers of INRs. The area under the ROC curve was 69.78% (95% CI: 54.23%-86.33%) (Fig. 8A). Diagnostic efficacy of microbial markers for INRs were validated using this data set, and the area under the ROC curve as 68.52% (95% CI: 44.87%-92.17%) (Fig. 8B).
Fig. 8 ROC curves for the OTU-based diagnostic biomarker of INRs. Diagonal lines represent random classification (AUC=0.5). AUC, area under the curve. (A) training set RF model. (B) test set RF model.
3.5 Correlation between gut microbiota and translocation biomarkers, inflammation cytokines, clinical index
To investigate the correlation between differential gut microbiota and translocation biomarkers, inflammation cytokines, clinical index by spearman correlations. The overall correlation between cytokine and taxa enriched or reduced in PLWH is shown in (Fig. 9). TNF-ɑ is negatively correlated with the abundances of Dialister (rs = -0.278, P = 0.022). IL-6 is negatively correlated with the abundances of Dialister (rs = -0.457, P < 0.001), Parabacteroides (rs = -0.262, P = 0.031) and Actinomyces (rs = -0.251, P = 0.039). CD54 is negatively correlated with the abundances of Dialister (rs = -0.306, P = 0.011) and Subdoligranulum (rs = -0.265, P = 0.029). LBP is negatively correlated with the abundances of Dialister (rs = -0.290, P = 0.016). CRP is negatively correlated with the abundances of Dialister (rs = -0.274, P = 0.024), Sporobacter (rs = -0.264, P = 0.030) and Veillonella (rs = -0.246, P = 0.043). A positively correlated between the abundance of Butyricimonas and Parabacteroides with recent CD4+ T-cell count and baseline CD4+ T-count. A negatively correlated between the abundances of Veillonella (rs = -0.247, P = 0.043) and Rothia (rs = -0.309, P = 0.010) with recent CD4+ T-cell count. Baseline CD4+ T-cell count is positively correlated with the abundances of Odoribacter (rs = 0.250, P = 0.039) (Fig. 9).