Inclusion and Ethics

We anonymized all identified T. austroamericanum strains recovered from clinical samples and stored them in Cryosysteme Protect bead tubes (Dutscher, https://www.dutscher.com) at −20°C. The hospital strain bank assigned a unique number to each sample: L0221, L0385, L0399, L0419, L0445, and L0458. The hospital routinely performed antifungal susceptibility testing by using the Sensititre YeastOne microdilution method (Thermo Fisher Scientific, https://www.thermofisher.com), according to the manufacturer’s instructions.

This study was reviewed and approved by the Ethics Committee of Assistance-Publique-Hôpitaux de Marseille (approval no. CSE24-49) and the Assistance-Publique-Hôpitaux de Marseille Health Data Access Portal (approval no. PADS24-275). Patients were informed of the research and their nonopposition to the use of their data was collected. In accordance with those committees and current regulations, written informed consent was not required. All potentially identifying information was removed in compliance with International Committee of Medical Journal Editors guidelines.

Mycologic Investigation

From identified patients, we collected cryptococcal antigen in serum by using Cryptococcal Antigen Lateral Flow Assay (IMMY, https://www.immy.com) and Platelia Aspergillus Antigen (Bio-Rad Laboratories, https://www.bio-rad.com). In addition, the hospital’s infection control team initiated an environmental investigation in 2023, which they continued in 2024 until the last infected patient was identified. The investigation focused on the rooms and units where the initial kidney transplant patients had been under care from the date of admission to discharge, including radiology, the intensive care unit (ICU), the urological surgery operating theater, and the dialysis unit. The team also investigated the nephrology unit, where patients were hospitalized for a week after kidney transplantation. The team collected air samples by using AESAP1075 Sampl’air Lite microbiological air sampler (AES Laboratories, https://www.chemeurope.com), collecting 330 liters on Sabouraud dextrose agar (SDA). The team collected water samples from faucets or showers in patients’ rooms in 250-mL bottles containing sodium thiosulfate. The team also collected surface samples by using eSwab Liquid Amies Elution Swabs (Copan, https://www.copangroup.com).

Next-Generation Sequencing

We performed DNA extraction on pure subculture from SDA supplemented with gentamicin and chloramphenicol (Bio-Rad), using previously described methods (6). We sequenced genomic DNA by using a paired-end strategy. We barcoded and prepared samples by using the COVIDseq Test sample prep kit (Illumina, https://www.illumina.com) adapted for fungi, in which the tagmentation step fragmented and tagged the DNA. We used limited cycle PCR amplification (12 cycles) to complete the tag adapters and introduce dual-index barcodes. After purification on ITB beads (Illumina), we normalized libraries to the same molarity, then pooled those into a single library for sequencing on the NovaSeq 6000 (Illumina). We loaded the pooled single-strand library onto the reagent cartridge and then onto the instrument along with the flow cell. We conducted automated cluster generation and paired-end sequencing of dual index reads in a single 25-hour run of 2 × 150 bp.

We also performed nanopore sequencing on individually sequenced genomic DNA by using the PromethION 2 Solo or GridION and the LSK109 Ligation Kit on a FLO-PRO002 flow cell (all Oxford Nanopore Technologies, https://nanoporetech.com). The end-prep step fixes specific nucleotides to use for adapter ligation, after which we purified the DNA on magnetic beads (CleanNA, https://www.cleanna.com). We activated the flow cell by adding a flush buffer and a tether from Flow Cell Priming Kit EXP-FLP002 (Oxford Nanopore Technologies). We then loaded the libraries onto the flow cell for a 72-hour run.

Bioinformatics

We converted Illumina binary base call (BCL) files into fastq files by using bclconverter version 4.2.4 (Illumina). We used Trimmomatic version 0.39 (7) to trim reads to a minimum Phred quality of 33 and minimum read length of 36 bp. We trimmed nanopore raw reads by using ProwlerTrimmer (8), a Phred quality of 20–35, and minimum read length of 1,000 bp.

We obtained complete genomes from mixed de novo assembly of Illumina and nanopore reads by using Unicycler version 0.4.4 (9). We removed contigs <1,000 bp in length. We determined GC content and evaluated contamination by using ContEst16S (EZBio https://www.ezbiocloud.net). We assessed assembly quality by using BUSCO version 5.7.1 and the fungi_odb10 database (10).

We used the Burrows-Wheeler aligner (Galaxy version 2.3 plus galaxy0; https://usegalaxy.eu) (12) to map filtered reads to the reference genome CBS 17435 (11) from the CBS culture collection (https://wi.knaw.nl/fungal_table) hosted by the Westerdijk Fungal Biodiversity Institute (Utrecht, the Netherlands). We generated consensus sequences by using iVar consensus (Galaxy version 1.4.4 plus galaxy0) (13), with depth coverage of >10 reads and base quality of Q20. We aligned the resulting sequences by using MAFFT (Galaxy version 7.526 plus galaxy2) (14).

We calculated pairwise single-nucleotide polymorphism (SNP) distances by using SNP distance matrix (Galaxy version 7.526 plus galaxy2) and identified SNPs by using the Find SNP Sites command (Galaxy version 2.5.1 plus galaxy0) (15). We aligned concatenated SNP lists and analyzed with IQ-TREE version 2.4.0 (http://www.iqtree.org) plus galaxy1 to generate a maximum-likelihood phylogenetic tree, which we visualized on iTOL (https://itol.embl.de). We calculated genome coverage with a custom Python script (Python Software Foundation, https://www.python.org).

We isolated mitochondrial genomes into a single circular contig ≈40-kb in size and aligned contigs by using mauve version 2.4.0 (16). We oriented contigs from the cytochrome c oxidase 1 start codon. When necessary, we generated reverse complements using an online tool (https://reverse-complement.com/terms.html), then annotated genomes by using GeSeq (https://chlorobox.mpimp-golm.mpg.de/geseq.html).

We mapped trimmed Illumina reads against the reference IGS1 sequence (CBS 17435) using Burrows-Wheeler aligner (12) and obtained consensus sequences by using sam2consensus (https://github.com/edgardomortiz/sam2consensus). We used MAFFT (https://mafft.cbrc.jp/alignment/server/index.html) (13) to align sequences and performed subsequent phylogenetic analysis by using IQ-TREE with 1,000 bootstraps (17).

Maximum-Likelihood Core-Genome Phylogenetic Analysis

We constructed a core-genome tree by using the Galaxy platform. We masked repeats with RepeatMasker (Galaxy version 4.1.5 plus galaxy0) and annotated the assemblies by using the Maker annotation pipeline (18) with the Augustus predefined prediction model Cryptococcus neoformans gattii (Galaxy version 2.31.11 plus galaxy2). We derived the aligned core-genome from the resulting annotated general feature format files by using Roary in Galaxy version 3.13.0 plus galaxy3, then collected SNPs by using the Find SNP Sites tool (Galaxy Version 2.5.1 plus galaxy0). We used those SNPs to build the maximum-likelihood tree on IQ-TREE version 2.3.6 (built August 4, 2024) (17), under ModelFinder (19). We visualized trees by using the iTOL platform.

Patient Data

By June 2024, a total of 6 patients with T. austroamericanum infection had been identified (Table 1). The patients’ median age was 65.5 (range 55–83) years; 3 (50%) were male and 3 (50%) were female. Four patients were kidney transplant recipients who had clinical signs of an infection 2–4 months after transplantation; all 4 had scar dehiscence and subcutaneous abscesses, suggesting that the scar was the portal of entry.

The 2 other patients were immunocompetent. One (sample no. L0419) had been hospitalized in the cardiology department and was referred for a nephrology consultation during the time the outbreak was occurring. That patient had not undergone recent surgery. The other patient (sample no. L0445) experienced shock and acute respiratory distress syndrome after gastrointestinal surgery.

Of the 25 T. austroamericanum–positive samples, 16 (64%) were blood cultures from 2 patients, 3 were scar swab samples from 1 patient, 3 were abscess aspiration samples from 3 patients, and 3 were respiratory samples (2 bronchial aspirate and 1 bronchoalveolar lavage fluid) from 1 patient. All T. austroamericanum isolates exhibited similar antifungal susceptibility profiles (Table 2), and we noted no significant difference related to date of isolate collection.

Mycological Investigation

The infection control team conducted an environmental survey in hospital A during April 20, 2023–July 19, 2024. In total, they collected 145 surface samples, 33 air samples, and 17 water samples in the urological surgery operating theater and the following patient units: the nephrology ICU, the kidney transplantation unit, the nephrology daycare, the radiology unit, and the dialysis unit. Samples were mainly collected from patient rooms, medical device storage area, and the decontamination room. All 195 samples were negative for T. austroamericanum. However, a T. austroamericanum strain was isolated from an air sample taken as part of a routine surveillance in a pediatric ICU of hospital B.

Next-Generation Sequencing

Multilocus Sequence Typing

Figure 1

Figure 1. Maximum-likelihood phylogenetic tree of IGS1 sequences from study of Trichosporon austroamericanuminfections among hospitalized patients, France, 2022–2024. The tree includes strains isolated from 6 patients (red font) and 1...

We initially performed WGS on the first 3 reported cases, which MALDI-TOF mass spectrometry had identified as T. inkin. To investigate genetic variations, we applied a multilocus sequence typing approach. We mapped sequences of cytb, rpb1/2, and tef1 genes and the SSU and D1/D2 regions against the reference genome of T. inkin (GenBank accession no. MT801082). Although the sequences were identical among the isolates, notable differences from T. inkin emerged among isolates from those 3 cases. In addition, we observed high divergence between the strains from our investigation and the T. inkin reference mitochondrial genome and IGS1 marker, characterized by numerous point mutations, deletions, and insertions (Appendix Figure 1). The consistent mutation patterns across the strains, along with their divergence from T. inkin, strongly supported the classification of the isolates as a distinct taxonomic entity. That preliminary analysis served as the starting point for further investigation into the IGS1 region of the 6 patient-derived isolates. Phylogenetic analysis, which included 3 reference sequences provided by the National Reference Center for Invasive Mycoses and Antifungals (CNRMA) at Institut Pasteur (https://www.pasteur.fr), confirmed that the 7 strains (6 patient and 1 environmental) belonged to T. austroamericanum species (Figure 1).

WGS Typing

We conducted WGS typing to assess the clonal nature of the T. austroamericanum outbreak. Of the 7 genomes, 6 demonstrated excellent assembly quality and had contigs ranging from 12 to 31 (Table 3). Those assemblies showed consistent genome lengths of 20.8 to 21 Mb, and GC content was close to 61.31%. The BUSCO scores were all >96% (range 96.5%–97.4%), indicating highly complete and accurate assemblies. However, the assembly for strain L0445 stood out because it had a much higher number of contigs (1,051), a larger genome size of 26 Mb, and a higher GC content of 62.59%. That assembly also showed contamination with bacterial Achromobacter xylosoxidans 16S sequence, so we excluded it from further analyses.

Figure 2

Figure 2. Core-genome phylogenetic relationships and SNP distance matrix from study of Trichosporon austroamericanuminfections among hospitalized patients, France, 2022–2024. A) Phylogenetic tree based on the core genome of the analyzed...

The core genome consisting of 11,957 genes enabled construction of a phylogenetic tree from the extracted SNPs, which formed sequences with 542 nt sites and 196 parsimony-informative positions (Figure 2, panel A). Of the 542 sites, we observed 129 distinct site patterns, indicating a high level of sequence variation across the strains. Strains L0453 and L0221, separated by only 85 SNPs (Figure 2, panel B), were the closest in the tree. L0453, from the ICU air duct, also appeared near L0458, with a difference of 97 SNPs. In addition, we identified 7,668 shell genes, representing genes shared by a subset of strains, indicating genetic diversity beyond the core genome.

Whole-Genome SNP Analysis

We performed whole-genome SNP calling by mapping the Illumina reads of our strains against T. austroamericanum reference genome CBS 17435 (11). That method enabled us to include strain L0445, for which high-quality mapping data was available. Read mapping demonstrated high genome coverage quality across all 8 chromosomes and strict coverage exceeding 97% for all strains (minimum coverage 97.30% +SD 0.00368%) and reaching up to 99.98% (+SD 0.00043%).

Figure 3

Figure 3. Whole-genome SNP-based phylogeny and distance matrix from study of Trichosporon austroamericanuminfections among hospitalized patients, France, 2022–2024. A) Whole-genome phylogenetic tree of the analyzed strains. Blue font indicates reference...

The resulting whole-genome SNP distance matrix revealed patterns consistent with the core genome analysis (Figure 3, panel A). Distances between the CBS 17435 reference and the clinical strains ranged from 447 to 1,085 SNPs. The smallest pairwise distance was 105 SNPs between L0385 and L0399, which were 2 isolates collected within 3 months of each other from the same hospital, supporting their close genetic relationship. The second closest group consisted of L0221, L0453, and L0458, which clustered together despite being collected in different years (Table 1). Most other pairwise distances ranged from ≈400 to >1,100 SNPs, including between L0445 and the rest of the isolates (≈1,070–1,186 SNPs), suggesting that L0445 is genetically more distant, consistent with its location at a hospital site geographically distant from the others.

Mitochondrial Genome

As a complementary approach, we analyzed the mitochondrial genome to assess its potential for discriminating between strains. We used mapping to obtain mitochondrial genomes of ≈44 kb from 3 T. austroamericanum reference strains from the CNRMA and the L0445 strain, then generated genomes of other 6 genomes through de novo assembly (Appendix Figure 2). Of note, synteny was identical between the de novo assembled mitogenomes.

Figure 4

Figure 4. SNP map and phylogenetic analysis of mitogenomes from investigation of Trichosporon austroamericanum infections among hospitalized patients, France, 2022–2024. T. austroamericanumisolates. A) SNP map of mitochondrial genomes...

The alignment of those mitochondrial genomes against the de novo assembled L0221 sequence revealed 4 distinct mutational profiles, including 3 among Marseille strains (Figure 4). Those profiles included 3 key mutated positions in the mitochondrial genome of the cox3 (at position 22240), apocytochrome b (at position 31837), and trnL (tRNA-Leu) (at position 39897) genes. The mutations provide clear genetic distinctions among the strains. For instance, strain CNRMA15 exhibited a unique mutation, C at position 22240, whereas other strains, such as L0221, have a G at that position (Figure 4, panel A). That analysis underscores the value of mitochondrial genome sequencing for identifying and distinguishing between T. austroamericanum strains involved in this outbreak and future outbreaks.

The phylogenetic placement of mitochondrial genomes showed 4 distinct clades. L0221 and CNRMA15 formed a separate branch because of their unique SNP profiles. L0458 and L0453 were closely related, as seen in the core-genome analysis. Finally, L0385, L0419, L0399, L0445, CNRMA22, and CNRMA16 cluster together, and CNRMA 15 branches off the others (Figure 4, panel B).