Burn resuscitation strategy influences the gut microbiota-liver axis in swine

Abstract

Fluid resuscitation improves clinical outcomes of burn patients; however, its execution in resource-poor environments may have to be amended with limited-volume strategies. Liver dysfunction is common in burn patients and gut dysbiosis is an understudied aspect of burn sequelae. Here, the swine gut microbiota and liver transcripts were investigated to determine the impact of standard-of-care modified Brooke (MB), limited-volume colloid (LV-Co), and limited-volume crystalloid (LV-Cr) resuscitation on the gut microbiota, and to evaluate its' potential relationship with liver dysfunction. Independent of resuscitation strategy, bacterial diversity was reduced 24 h post-injury, and remained perturbed at 48 h. Changes in community structure were most pronounced with LV-Co, and correlated with biomarkers of hepatocellular damage. Hierarchical clustering revealed a group of samples that was suggestive of dysbiosis, and LV-Co increased the risk of association with this group. Compared with MB, LV-Co and LV-Cr significantly altered cellular stress and ATP pathways, and gene expression of these perturbed pathways was correlated with major dysbiosis-associated bacteria. Taken together, LV-Co resuscitation exacerbated the loss of bacterial diversity and increased the risk of dysbiosis. Moreover, we present evidence of a linkage between liver (dys)function and the gut microbiota in the acute setting of burn injury.

Figure 1

Limited-volume resuscitation does not exacerbate measured physiologic derangements compared with MB. (a, b) Representative CT angiogram of the (a) superior mesenteric artery (SMA) and (b) inferior mesenteric artery (IMA). Left panel: green trace indicates selected vessel with red bars indicating the location of cross-sectional measurements. Right panel: cross-section of the selected vessel with blue circle outlining the vessel lumen; red and green lines depict two measurements of vessel lumen diameter. Text indicates measurements of area (blue) and diameter (red and green). Images were acquired with the Vitrea software v6.9.87.1 (https://www.vitalimages.com/vitrea/). (c, d) Quantification of cross-sectional areas of (c) SMA and (d) IMA at baseline (BL) and 48 h post-injury. Bars depict mean and 95% confidence intervals. Images that did not show sufficient contrast could not be measured accurately and were excluded from analysis. (e) Urine output 20 h after burn injury. Experimental replicates are indicated in each respective figure. (f, g) Dimension reduction analysis of 14 clinical biomarkers visualized by PCA biplots with (f) time post-injury or (g) resuscitation strategy supplied as supplementary variables. Each small dot depicts an individual sample, larger dots indicate the centroid of the ellipse. Ellipses depict 95% confidence region. MB (n = 6 pigs), LV-Co (n = 12 pigs), and LV-Cr (n = 12 pigs) *: p < 0.05; **: p < 0.01; ***: p < 0.001, ns: p > 0.05 calculated by ANOVA with Holm family-wise error rate correction at α = 0.05.

Figure 2

Impact of resuscitation on bacteria alpha diversity. (a, b) Faith phylogenetic diversity, (c, d) Shannon diversity, (e, f) Pielou evenness, and (g, h) OTU counts. (a, c, e, g) Raw diversity values with each point depicting the mean and bars representing 95% CI. Within treatment differences between post-injury (24 h and 48 h) and respective baseline time-points were evaluated with two-sided Wilcoxon rank sum test with Holm family-wise error rate correction at α = 0.05. Statistical significance is indicated by text and symbol colors corresponding to the resuscitation strategy. *: p < 0.05; **: p < 0.01 with actual p values indicated in text. (b, d, f, h) LME model predicted values and regression line after normalization of the data. Each data point depicts the predicted value of an individual sample. Comparison of regression slopes and p value calculations were evaluated by LME models. MB (n = 6 pigs), LV-Co (n = 10 pigs), and LV-Cr (n = 11 pigs).

Figure 3

Impact of resuscitation on bacteria beta diversity. (a) Bray–Curtis and (c) generalized UniFrac distances visualized by PCoA plots. Symbol shapes depict sample time-points, whereas symbol colors represent different resuscitation strategies. Each data point represents an individual sample. The amount of variation explained by each axis are indicated in parentheses. Boxplots of (b) Bray–Curtis and (d) generalized UniFrac longitudinal first-distances. First-distances were calculated as the distance between post-injury (24 h and 48 h) time-points and its respective baseline. Each boxplot shows the median along with the first and third quartiles. Whiskers depict the standard error and data points in the boxplots represent an individual sample. *: p < 0.05; **: p < 0.01; ***: p < 0.001 calculated by LME models. MB (n = 6 pigs), LV-Co (n = 10 pigs), and LV-Cr (n = 11 pigs).

Figure 4

Altered bacterial composition after severe burn and resuscitation. (a) Stacked bar plot showing the mean relative abundance at the phylum level for LV-Co (n = 10 pigs), LV-Cr (n = 11 pigs), and MB (n = 5 pigs) resuscitation strategies. Only phyla consisting of ≥ 1% of the total bacterial community are shown; all other phyla are included in the “Other” category. (b) Differentially abundant phyla across time identified by LEfSe. Color of bars indicates enrichment of phyla at a particular time point. Only phyla with LDA score > 2 are shown. (c) Unsupervised clustering of the top 20 differentially abundant taxa scaled by relative abundance. Rows depict samples and columns indicate genus or the highest order taxon that could be classified (i.e., family or order). Letter in parentheses indicates the level of taxonomic classification: (o) order, (f) family, (g) genus. Compositional similarity of samples was assessed by hierarchical clustering. Clusters 1 and 2 are outlined by red rectangles. Each tip of the dendrogram represents an individual sample that is identified by injury time-point (orange, green, purple) and resuscitation (red, black, blue) side colors. Symbol preceding the taxa names indicate significant enrichment (Holm corrected two-sided Wilcoxon p < 0.05) in cluster 1(*) or cluster 2 (**).

Figure 5

Changes in the bacterial diversity are associated with liver (dys)function. (a–c) Correlation between generalized UniFrac and (a) AST, (b) ALT, and (c) TP. Correlation coefficients were calculated by Mantel test and indicated p values were corrected using the Holm procedure for family-wise error rate at α = 0.05. Correlations between Bray–Curtis and AST, ALT, and total protein (TP) were similar (Supplementary Fig. S6). (d, e) Five most significantly perturbed pathways for (d) LV-Co (n = 6) and (e) LV-Cr (n = 6) resuscitation, compared with MB (n = 5). (f) Hierarchical clustering of Pearson correlation coefficients between gene expression of the three most significantly perturbed pathways in common with LV-Co and LV-Cr and major differentially abundant taxa.