Characterization of shifts of koala (Phascolarctos cinereus) intestinal microbial communities associated with antibiotic treatment

Abstract

Koalas (Phascolarctos cinereus) are arboreal marsupials native to Australia that eat a specialized diet of almost exclusively eucalyptus leaves. Microbes in koala intestines are known to break down otherwise toxic compounds, such as tannins, in eucalyptus leaves. Infections by Chlamydia, obligate intracellular bacterial pathogens, are highly prevalent in koala populations. If animals with Chlamydia infections are received by wildlife hospitals, a range of antibiotics can be used to treat them. However, previous studies suggested that koalas can suffer adverse side effects during antibiotic treatment. This study aimed to use 16S rRNA gene sequences derived from koala feces to characterize the intestinal microbiome of koalas throughout antibiotic treatment and identify specific taxa associated with koala health after treatment. Although differences in the alpha diversity were observed in the intestinal flora between treated and untreated koalas and between koalas treated with different antibiotics, these differences were not statistically significant. The alpha diversity of microbial communities from koalas that lived through antibiotic treatment versus those who did not was significantly greater, however. Beta diversity analysis largely confirmed the latter observation, revealing that the overall communities were different between koalas on antibiotics that died versus those that survived or never received antibiotics. Using both machine learning and OTU (operational taxonomic unit) co-occurrence network analyses, we found that OTUs that are very closely related to Lonepinella koalarum, a known tannin degrader found by culture-based methods to be present in koala intestines, was correlated with a koala's health status. This is the first study to characterize the time course of effects of antibiotics on koala intestinal microbiomes. Our results suggest it may be useful to pursue alternative treatments for Chlamydia infections without the use of antibiotics or the development of Chlamydia-specific antimicrobial compounds that do not broadly affect microbial communities.

Keywords: 16S rRNA genes; Antibiotics; Bacteria; Chlamydia; Diversity; Koala; Machine learning; Microbiome; Microbiota; Network analysis.

Figure 1. Principle Coordinate Analysis (PCoA) of microbial communities found in koala fecal samples based on analysis of rRNA gene sequences.

PCoA plots were generated using the QIIME (Quantitative Insights Into Microbial Ecology) version 1.9.1 workflow on sequencing reads following quality filtering and chimera removal. PCoA plots are shown for multiple metrics to illustrate how beta diversity is related to abundance and phylogeny: unweighted Unifrac (A), weighted Unifrac (B), and Bray-Curtis (C). Each point represents a unique sample. Coloring indicates individual koalas, shape indicates type of treatment or sample (circles—antibiotic treated, triangles—no antibiotics, squares—samples from the built environment), fill indicates whether the koala lived or died. Numbers in legend indicate the number of samples for that koala. Colors were chosen based on Martin Krzywinski’s Color Blindness Palette for improved viewing by those with color blindness.

Figure 2. Modified boxplots of alpha diversity (Shannon index) for microbial communities found in koala fecal samples based on analysis 16S rRNA gene sequences.

Data is presented for pools from different groups of koalas. The Shannon diversity index was calculated for each sample using R and averages for each group of interest were then calculated. The average alpha diversity for each sample was not significantly different between samples from koalas treated with antibiotics (t (test statistic) = 1.1413, df (degrees of freedom) = 52.692, p-value = 0.2589). The average alpha diversity for samples from koalas that lived through antibiotic treatment was found to be significantly greater than the average alpha diversity for samples from koalas that died during their admission which included antibiotic treatment (t = 2.9239, df = 38.43, p = 0.005768). Outliers (shown by open circles) represent values that are 1.5 times greater than the difference between the third and first quartiles of the data set.

Figure 3. The distribution of the top 20% of feature-importance OTUs across individual samples as determined by Random Forest Analysis.

Individual samples on the X axis are organized by the individual koalas from which they came (represented by unique colors), and divided by ‘Released’ and ‘Deceased’. The Y axis represents the taxonomic assignment (see ‘Methods’) for each of the top 20% OTUs that were the most predictive of fate according to our Random Forest Analysis. The coloring of the text for the OTU names is used to highlight specific taxa of interest flagged by network analysis (see main text). Highlighted in red are OTUs identified as Lonepinella koalarum that the network analysis identified as the most predictive OTUs of fate. Highlighted in blue are OTUs identified as ‘Cyanobacteria YS2’, which were identified as being highly predictive of fate and are of interest because they are Melainabacteria. The density of each point in the heatmap is representative of the relative abundance of each OTU for each sample.