Population structure in a fungal human pathogen is potentially linked to pathogenicity

Abstract

Aspergillus flavus is a clinically and agriculturally important saprotrophic fungus responsible for severe human infections and extensive crop losses. Here, we analyze genomic data from 300 (117 clinical and 183 environmental) A. flavus isolates from 13 countries, including 82 clinical isolates sequenced in this study, to examine population and pan-genome structure and their relationship to pathogenicity. We use single nucleotide polymorphisms to build a phylogeny, analyze admixture, and perform discriminant analysis of principal components. We identify five A. flavus populations, including a new population, D, corresponding to distinct clades in the genome-wide phylogeny. Strikingly, > 75% of clinical isolates were in population D and <5% in population B. We also use orthogroup clustering to identify core and accessory genes within the pan-genome. Accessory genes, including genes within biosynthetic gene clusters, were significantly more common in some populations but rare in others. Our functional annotations show that population D is enriched for genes associated with carbohydrate metabolism, lipid metabolism and certain types of hydrolase activity, whereas a non-clinical population is depleted in genes related to zinc ion binding. In contrast to previous results from the major human pathogen Aspergillus fumigatus, isolation of A. flavus from human specimens is associated with population structure, providing a promising system for future investigations into the contributions of population-specific genetic differences to human infection.

Fig. 1. Population structure of Aspergillus flavus reveals genetic isolation reflecting five populations, including a new population, D.

Analyses are based on 1,941,481 biallelic single nucleotide variants. A Estimates of individual ancestry, with K = 5, conducted using the software package LEA, which estimates individual admixture coefficients from a genotypic matrix. We estimated admixture for K = 2 through 10, with K = 5 providing the best fit for our data according to the cross-entropy criterion^,. B Discriminant analysis of principal components shows admixture among populations A, C, and D as well as clear separation of populations B and S-type. Dots represent individuals and ellipses indicate group clustering of individuals. Populations are color coded as indicated in the top right. The discriminant analysis eigenvalues are shown on the bottom left, with the darker bars showing eigenvalues retained. C A principal coordinates analysis displaying relative genetic distances of individual isolates, here represented by dots, using Nei’s genetic distance matrix. Axes indicate the two principal coordinates retained and the percentage of variance explained by each coordinate. Populations A, C, and D varied primarily along PC2 rather than PC1; population B showed genetic differentiation from other populations and varied primarily along PC1. Single nucleotide polymorphisms can be inferred from the alignment provided as our Source Data file.

Fig. 2. Maximum likelihood phylogeny supports the existence of five populations and non-random distribution of clinical isolates across populations.

Filled in circles along the outer track indicate clinical isolates; empty circles indicate environmental isolates. Branch colors correspond with population assignment based on the discriminant analysis of principal components (DAPC; Fig. 1): the S-type population is indicated in orange; population A in yellow; population B in blue, population C in pink, and population D in purple. The outgroup, Aspergillus minisclerotigenes, is represented in black. Apart from the outgroup, each tip represents an A. flavus isolate and branch lengths denote sequence divergence. Ultrafast bootstrap values are indicated below nodes. The phylogeny was constructed using 2,018,259 SNPs from a dataset of 300 A. flavus isolates and the outgroup A. minisclerotigenes. Source data are provided as a Source Data file.

Fig. 3. The pan-genome of Aspergillus flavus is closed and contains 17,676 orthogroups.

A Rarefaction curve of number of orthogroups added with each additional genome, excluding singletons. Data are presented as mean of 100 permutations with random ordering of genomes added with standard deviation indicated. B Histogram of orthogroup frequency determined by number of genomes in which each orthogroups is present. The core genome contains 10,161 orthogroups. The accessory genome of A. flavus contains 7515 orthogroups. C Heatmap of normalized abundance of gene ontology (GO) annotations with significant differences in abundance among populations. Significance determined by one-way ANOVA. Bonferroni-corrected, p < 0.05. The bar chart shows the mean number of genes containing each GO term across all genomes. Source data in the form of predicted proteomes are provided in the associated Figshare repository and Supplemental Data 6.

Fig. 4. Gene ontology (GO) terms more prevalent in population D than other populations include zinc ion binding and hydrolase activity, among others.

Boxplots indicate the number of genes annotated with each GO term per isolate, by population assignment. The Y axis scale is adjusted for each GO term to better show differences among populations. In the box-and-whisker plots, the horizontal line in the box indicates the 50th percentile and the box extends from the 25th to the 75th percentile. The whiskers encompass the lowest and highest values within 1.5x the interquartile range. Individual dots represent statistical outliers. Statistical significance was determined using a one-way analysis of variance followed by post hoc testing using Tukey’s honest significance test (one-sided). The letters denote significances as a compact letter display where groups that are not significantly different from each other are indicated with the same alphabet letter; P < 0.05. After removing clones, low quality assemblies, and the S-type isolates, statistics were calculated for 232 isolates: population A (n = 81), population B (n = 69), population C (n = 29), population D (n = 53). Source data are provided in Supplemental Data 6.