The total length of the de novo assembled reference contigs we showed the genome size of S. kiliense was ≈38 MB. The average sequencing depth for the reference strain was 354×, which resulted in 98% coverage across the genome. The assembly contained 1,092 contigs ranging in length from 500 bp to 1.7 million bp (average contig length 35,528 bp; N50 591,374 bp) (13).

Figure

Figure. Whole-genome single-nucleotide polymorphism (SNP) typing of Sarocladium kiliense strains, Chile and Colombia, 2013–2014. All patient (clinical) and drug (vial) isolates from these 2 countries differed by <5 SNPs, and >21,000...

Reference-based phylogenetic analysis identified ≈117,000 shared SNPs, of which ≈60% were parsimoniously informative (Figure). The cladogram showed 1 distinct clade that comprised all outbreak isolates and included isolates from the patients and the contaminated drugs. No more than 5 SNPs were detected between any patient and drug isolates from Chile and Colombia, demonstrating that these isolates had nearly indistinguishable genomes. Conversely, all control isolates from different sources clustered separately from the outbreak clade.

The genomewide SNP analysis showed greater diversity among the control isolates. Although the closest control isolate, B10764, differed from the outbreak clade by 501 SNPs, the remaining majority of the control isolates diverged from the outbreak clade by ≈21,000 SNPs. Two control isolates, B5504 and B5505, isolated from a 1996 cluster of S. kiliense infections in a US hospital, were genetically indistinguishable from each other.

S. kiliense is primarily a saprobic soil organism; it can cause opportunistic infections in humans that typically occur after inoculation of the fungus (14). Here we report the results of molecular epidemiologic investigation of an outbreak of S. kiliense BSI that affected >50 patients in 2 South American countries. The epidemiologic investigation suggested contaminated antinausea medication as a possible source of this infection. With no existing genotyping methods for this uncommon pathogen, we used WGST to understand the genetic relationships among the isolates and identify a potential source of this outbreak.

The use of WGST to investigate fungal outbreaks has become integral to epidemiologic investigations (3–5,15). During nosocomial outbreak investigations, ascertaining potential source(s) of infection based solely on the descriptive epidemiologic findings often is difficult. Molecular genotyping frequently is needed to test hypotheses generated by epidemiologic investigation: the presence of a single strain or dominant clone usually suggests a point source, whereas the presence of multiple strains is usually consistent with an environmental exposure or exposure to mixed populations from a single source. For example, in a recent outbreak of Curvularia spp. (formerly Bipolaris spp.) among cardiac surgery patients, initial epidemiologic investigation suggested a point source; however, a molecular epidemiology analysis demonstrated multiple strains, consistent with an environmental source (16,17). Similarly, in an outbreak of Fusarium spp. associated with the use of contact lenses, initial epidemiologic investigation linked the infections with a particular lens cleaning solution, suggesting a point source; however, molecular analysis indicated multiple sources (18). In our current investigation, the WGST analysis demonstrated that the patient isolates from Chile and Colombia were nearly genetically indistinguishable (<5 SNPs) from those recovered from the medication vials, indicating the likely presence of a single-source infection. Conversely, the control isolates clustered differently from the outbreak clade by ≈21,000 SNPs, and except for 2 strains from the same cluster, thousands of SNPs separated any 2 control strains (Figure).

Although BSIs with this organism are rare, case reports/clusters of S. kiliense fungal infections have been reported in the literature (19–22). In a cluster of S. kiliense infections of endophthalmitis and catheter-related BSIs, an environmental source was strongly suggested but could not be confirmed because of the lack of available typing methods (21). Isolates from this cluster of infections served as controls in our study: specifically, isolates B5504 and B5505 from 2 patients who underwent cataract extraction with intraocular lens implantation differed from each other by 1 SNP, indicating that the 2 genomes were nearly indistinguishable and suggesting a common source of infection.

In this investigation, the use of WGST identified a likely source of S. kiliense BSI in oncology patients in both countries by linking these infections to the receipt of contaminated medication. Results from this outbreak are consistent with those from other outbreaks of fungal pathogens in which WGST was used and a common source was hypothesized. For example, Litvintseva et al. investigated Exerohilium rostratum infections associated with the injection of contaminated methylprednisolone acetate by using whole-genome SNP typing; they found that all outbreak isolates were genotypically indistinguishable: no more than 2 SNPs separated the strains in the outbreak clade (3).

The genetic diversity among the S. kiliense control strains in our study was congruent to the level of diversity among control isolates found in other studies. We identified comparable levels of genetic similarities to those found among isolates of Saprochaete clavata from patients from a multicenter outbreak in France (15). Conversely, >28,000 and up to 1.2 million SNPs were observed among the control isolates of Apophysomyces trapeziformis infections associated with a tornado in Joplin, Missouri, USA (5). Although we were able to include a variety of controls from different sources, a major limitation in this study is the lack of background isolates of S. kiliense from the affected countries to assess genomic similarity among the unrelated isolates from South America.

Traditional molecular strain typing methods, such as MLST, are limited to analysis of specific genomic regions, typically protein-coding regions. However, the conservation in these regions might be insufficient to discriminate strains of certain fungi or provide the resolution needed to identify a likely source in a fungal outbreak. In the absence of established methods for fungi, the rapid development of traditional typing methods is often necessary. In this investigation, before deploying WGST, we evaluated MLST using 7 protein-coding regions to determine the relatedness among all isolates (17). Across 3,651 nt amplified, only a 1-nt polymorphism was identified in the β-tubulin gene of one of the control isolates, which did not provide sufficient resolution between outbreak and control isolates, indicating that the use of MLST was not informative in this outbreak investigation. Although conventional typing methods are rapid, the conclusions drawn from them might be confounded by either their discriminatory power or their character state conflict (i.e., homoplasy). Typically, genotyping methods with high resolution (e.g., variable-number tandem-repeat typing) have higher degrees of homoplasy, whereas methods with lower resolution (e.g., MLST) have less homoplasy; both might be at risk of misidentifying a common source. Conversely, WGST enables accurate deduction of genetic relationships among strains in fungal outbreaks for which current typing methods are ineffective or nonexistent. Furthermore, with the improvement of next-generation sequencing technologies, investigating fungal outbreaks in real time will soon be possible.

Contamination of medical products, particularly with rare fungi, poses growing concern and a public health threat, especially in vulnerable populations. Fungal clusters/outbreaks of common environmental species (rather than classical clinical pathogens) have been associated with medical products. For example, the 2002 outbreak of fungal meningitis was caused by injected steroids contaminated with Exophiala dermatitidis, and the 2012–2015 outbreak of fungal meningitis and other infections resulted from methylprednisolone contaminated by Exserohilum rostratum (3,23). Similarly, in 2009, intestinal zygomycosis resulted from ingestion of allopurinol tablets contaminated with Rhizopus microsporus (24). In 2012, an endophthalmitis outbreak was associated with use of an ophthalmologic dye contaminated with Fusarium sp. in the Fusarium incarnatum-equiseti species complex and triamcinolone contaminated with Curvularia hawaiiensis (Bipolaris hawaiiensis) (16,25,26). Increased vigilance and the use of advanced technologies are needed to rapidly identify the likely source(s) of infection to efficiently guide epidemiologic investigations and initiate appropriate control measures.

In summary, our study highlights the utility of advanced molecular methods to investigate outbreaks involving rare fungi. Next-generation sequencing and bioinformatics analyses will remain critical molecular epidemiology tools in such epidemiologic investigations.