The impact of western versus agrarian diet consumption on gut microbiome composition and immune dysfunction in people living with HIV in rural and urban Zimbabwe

Abstract

Background: People living with HIV (PLWH) suffer from chronic inflammation even with effective antiretroviral therapy (ART). A high-fat, low-fiber western-type diet has been linked with inflammation, in part through gut microbiome changes. In sub-Saharan Africa (SSA), a region with high HIV burden, urbanization has been linked with a shift from traditional agrarian towards westernized diets, and with changes in food security. To explore the relationship between diet, inflammation, and the gut microbiome in PLWH, we enrolled 1) ART Naïve PLWH who provided samples before and after 24 weeks of ART, 2) PLWH on ART at both timepoints and 3) HIV-seronegative controls. Individuals were evenly recruited from rural and urban Zimbabwe (locations were 145 kilometers/90 miles apart). Using a food frequency survey designed to measure intake of agrarian versus western-type food items in Zimbabwe, we determined how diet differs with urbanization, HIV-infection and treatment, and is related to inflammation and the gut microbiome.

Results: Individuals residing in a rural area of Zimbabwe less frequently consumed high-fat, low-fiber western type food items and had lower consumption of diverse food items overall, except for sadza- a subsistence staple-processed from home-grown grains. Consumption of a more western-type diet correlated with lower CD4+ T cell percentage in untreated and treated PLWH and with increased T cell exhaustion in PLWH on ART. PLWH on ART at time of enrollment also consumed diverse food items at a lower frequency and more often were underweight. Low food consumption correlated with muted improvements in T cell exhaustion after 24 weeks of ART. Individuals residing in the rural area had more Prevotella-rich/Bacteroides-poor microbiomes, but this did was not significantly mediated by diet. western diet consumption reduced the diversity of carbohydrate substrate degradation capabilities in the microbiome, based on predictions made using metagenomic polysaccharide utilization loci.

Conclusions: Taken together, this work supports that consumption of more high-fat/low-fiber type food items has the potential to exacerbate HIV pathogenesis in a sub-Saharan setting where HIV burden is high and reinforces the importance of nutritional support for promoting immunologic response to ART in PLWH in SSA.

Keywords: Zimbabwe; agrarian; diet; gut microbiome; immune response; rural; urban; western.

Figure 1:

Food Category Principal Component Analysis (PCA). (A) Food Consumption axis (PC1) versus Western/Agrarian diet 1 axis (PC2) and (B) Western/Agrarian diet 2 axis (PC3) versus Mixed axis (PC4). For both (A) and (B) points and ellipses are colored according to urban or rural location. There is a 95% confidence that a point from the group (rural or urban) will fall within the region of the ellipse. Both (C) and (D) represent the same PCA spaces as A and B and arrows represent Food Categories and Food Types that correlate with the PC axes. The green and brown arrows represent Food Categories and are colored by agrarian versus western assignment of the category. Grey arrows were calculated by a PCA performed on the Food Types (Table 2). Legend for numbers is shown on right. (C) Food Category and Food Type PC1 versus PC2. (D) Food Category and Food Type PC3 versus PC4. Arrow size represents food importance, determined by Euclidean distance from the origin (larger arrows indicate greater importance).

Figure 2:

Food Category PCs Food consumption and Western/Agrarian diet 1 PCs stratified by clinic location, cohort, water source and visit. (A) Food consumption stratified by clinic location. (B) Western/Agrarian diet 1 stratified by clinic location. (C) Food consumption stratified by cohort. (D) Western/Agrarian diet 1 stratified by water source. Statistical comparisons were conducted using the following mixed linear model. Food Category PC ~ Location + Cohort + Week + Water Source (1|PID). The model was tested for Food Category PC1, 2, 3, & 4. Only significant results are plotted here.

Figure 3:

Weighted UniFrac Principal Coordinates Analysis (PCoA). (A) Weighted UniFrac PCoA analysis showing PCo1 versus PCo2. Points are colored by cohort. Arrow size represents genera importance, determined by Euclidean distance from the origin (larger arrows indicate greater importance). (B) Same as A except showing PCo3 versus PCo4. (C) Same as A but points are gradient colored by the log ratio of Prevotella to Bacteroides. (D) Results of linear modeling on Microbiome PCo 1–4. PC1-PC4. Detailed plots of significant relationships in D are in Figure S4.

Figure 4:

Microbial substrate targets varied by participant confounders, (A) Bar plot of relative polysaccharide utilization loci (PULs) in baseline fecal samples determined by shotgun metagenomic sequencing colored by substrate target and split by cohort, (B) Coefficients of linear models in which columns represent outcomes and rows represent predictors, predictors that had a significant impact on the model are colored red (↑/positive) or blue (↓/negative). (C) Boxplot of number of PULs targeting Beta-Fucosides split by HIV status. (D) Scatter plot of FFQ PC3 and total substrates targeted by different PULs, points colored by cohort with trend lines generated by linear regressions and p-values determined from linear models.

Figure 5:

Detailed plots from Figure S5 models. R-squared and p-values in red are derived from models in Figure S6.

Figure 6:

Baseline Food Category PC1 (Food Consumption) by change in CD8+PD1+ (%) over time. R and p-value calculated using a Pearson correlation. Ellipses and points colored by location. There is a 95% confidence that a point from the group (rural or urban) will fall within the region of the ellipse.