Impact of pathogen genetics on clinical phenotypes in a population of Talaromyces marneffei from Vietnam

Abstract

Talaromycosis, a severe and invasive fungal infection caused by Talaromyces marneffei, is difficult to treat and impacts those living in endemic regions of Southeast Asia, India, and China. While 30% of infections result in mortality, our understanding of the genetic basis of pathogenesis for this fungus is limited. To address this, we apply population genomics and genome-wide association study approaches to a cohort of 336 T. marneffei isolates collected from patients who enrolled in the Itraconazole vs Amphotericin B for Talaromycosis trial in Vietnam. We find that isolates from northern and southern Vietnam form two distinct geographical clades, with isolates from southern Vietnam associated with increased disease severity. Leveraging longitudinal isolates, we identify multiple instances of disease relapse linked to unrelated strains, highlighting the potential for multistrain infections. In more frequent cases of persistent talaromycosis caused by the same strain, we identify variants arising over the course of patient infections that impact genes predicted to function in the regulation of gene expression and secondary metabolite production. By combining genetic variant data with patient metadata for all 336 isolates, we identify pathogen variants significantly associated with multiple clinical phenotypes. In addition, we identify genes and genomic regions under selection across both clades, highlighting loci undergoing rapid evolution, potentially in response to external pressures. With this combination of approaches, we identify links between pathogen genetics and patient outcomes and identify genomic regions that are altered during T. marneffei infection, providing an initial view of how pathogen genetics affects disease outcomes.

Keywords: Talaromyces marneffei; GWAS; Vietnam; penicilliosis; population genomics; talaromycosis; whole genome sequencing.

Fig. 1.

T. marneffei centromere identification. A sliding window analysis of GC% and repetitive element content was used to predict candidate centromere locations. a) GC% for each chromosome (line plot), calculated and plotted in 500-bp windows beside each chromosome (bars). Regions of low GC (17–40%) are shown as bands on each chromosome. b) Proposed centromere regions with repetitive elements colored by color representing class, with the flanking gene IDs listed on either side (EYB25 prefix).

Fig. 2.

Phylogeny of patient isolates reveals no genetic clusters of severe outcomes. A maximum likelihood phylogeny was estimated using segregating SNP sites. Isolates separate distinctly into northern and southern clades, with both clades having 100% bootstrap support. The colored circles correspond to clade and sample type. Metadata for the clinical isolates includes initial fungal burden for a patient, indicated by the Log₁₀ CFU/ml heatmap, and patient outcomes that are indicated by colored squares in the outer perimeter of the phylogeny.

Fig. 3.

Frequencies of regional genomic duplications and deletions per chromosome. Frequency of deleted regions (left of chromosome axis) and duplicated regions (right of chromosome axis) assessed across all isolates, with frequency of deletion or duplication represented as distance from axis of the chromosomes (bars). 1000-bp windows were used to assess copy number variation. GC content across chromosomes represented by black (GC 17–40%) and red (GC 50–57%) bars plotted on the chromosome bars.

Fig. 4.

Genomic regions under selection across northern and southern clades. a) Signals of selection (CLR and TD), nucleotide diversity (Pi), and divergence (Fst and Dxy) across all chromosomes, per clade. b) Enlarged view of regions displaying elevated signals of selection, including region 1–4.