During 2015–2023, the AKUH clinical microbiology laboratory identified 432 presumed Clavispora (formerly Candida) lusitaniae fungemia cases. Presumed C. lusitaniae cases detected from blood cultures increased from 10–20 cases/year during 2015–2019 to 30 cases in 2020 and 56 cases in 2021. The number of cases detected from cultures from blood specimens further increased to 108 in 2022 and to 150 in 2023 (Figure 1, panel A), representing <0.11% of the total annual blood culture specimens processed by the laboratory (Appendix Table 2). Of the 432 C. lusitaniae cases, 415 patients (96.1%) were <13 years (4,745 days) of age and 365 (84.5%) were <1 year (365 days) of age: 326 (75.4%) <3 months of age, 274 (63.4%) <1 month of age, and 68 (15.7%) <7 days of age. Reflective of the primary referral base for the AKUH laboratory, most (n = 369; 85.4%) C. lusitaniae bloodstream cases were from Karachi, and 241 (55.8%) samples were from 1 tertiary care public sector pediatric hospital.

Using the laboratory practices for isolate banking described herein, 136 presumed C. lusitaniae isolates had been banked: 21 of 31 isolates cultured in 2020, 0 of 56 isolates cultured in 2021, 29 of 108 isolates cultured in 2022, 80 of 150 isolates cultured in 2023, and 6 of 15 isolates cultured in January of 2024. To investigate those cases, we performed WGS on banked isolates, 121 of which yielded high-quality assemblies. We excluded 15 isolates because total contig size exceeded 12.3 million bases (n = 8), total contig number exceeded 800 (n = 11), genome coverage was <95% as determined by BUSCO (n = 9), or a combination of those factors. BUSCO analyses showed equivalent rates of assembly completeness, fragmentation, and missingness between the isolates sequenced at either institution. Querying the internal ITS sequences from the whole-genome assemblies against the NCBI ITS database showed that just 13 (11%) of the 121 high-quality genome sequences were true C. lusitaniae, whereas ITS sequences from the other 108 (89%) sequences had 99.3%–100% sequence identity with the ITS sequence of Scheffersomyces spartinae CBS 6059 (NCBI accession no. NR_111290.1) (Figure 1, panel B; Appendix Table 1).

The median age of S. spartinae patients was 19.5 days (range 2 days–12 years); 68.5% of cases were among infants <4 weeks of age (Table 1). For most cases, specimens were obtained in Karachi, reflective of catchment, but >8% of cases were identified in patients from other regions. Most (69.4%) cases were identified at a single hospital (hospital H) in Karachi specializing in neonatal and pediatric care. Blood cultures from that hospital constituted 1.4% of the total blood cultures processed by the AKUH clinical microbiology laboratory since 2020, and 2.9% of blood cultures from that hospital were C. lusitaniae positive by biochemical assay (Appendix Table 2).

We cultured S. spartinae from specimens received from 14 different institutions but could not determine the origin of 11% of specimens (Table 1). Of patients whose families or guardians could be contacted and consented to provide clinical information (92/108), all were hospitalized at the time of specimen collection. Treating physicians provided admitting diagnoses and brief clinical histories for 63 (58.3%) patients, 43 (68.2%) of whom were admitted for sepsis. MICs for amphotericin, azole, and echinocandin antifungal drugs were low among the 108 S. spartinae isolates tested, and MIC₉₀ values were <1 µg/mL for all agents in those classes (Table 2).

Isolate 128-CS, which we selected for long-read sequencing, came from a sample collected in Karachi in December 2022. Assembly from long and short reads of 128-CS produced 8 linear chromosomes ranging in length from 1.19 to 2.05 Mbp and 1 circular mitochondrial genome, for a total genome size of 12.27 Mbp (Appendix Table 4). Comparison of the ITS sequence of isolate 128-CS to the NCBI ITS database reconfirmed S. spartinae (Figure 2).

Phylogenetic analysis of all 108 S. spartinae sequences aligned to 128-CS revealed that most sequences belonged to a clade consisting of nearly all specimens obtained from Karachi and 2 specimens from Nawabshah (Figure 3). That large clade (clade A) contained 2 subclades, clades A-1 and A-2 (Figure 4). We also identified 2 smaller clades: clade B, consisting of 1 isolate each from Karachi and Hafizabad and all 5 isolates from Lahore, and clade C, consisting of 1 genetically distant sequence isolated from a patient in Multan (Figure 3). Although the Multan isolate differed by >213,000 SNVs from 128-CS, the ITS regions of the 2 isolates were 100% identical. All isolates in clade B were collected from infants <1 month of age; the clade C isolate was collected from a 6-month-old child (Figure 3, panel A).

Because most isolates cultured during 2020–2024 were closely related, we examined the characteristics of those 100 specimens. Within that clade, the median pairwise genetic distance across the 12.2 million base alignment was just 9 (range 0–22) SNVs. The time-scaled maximum-likelihood phylogenetic tree showed that, despite the small genetic distances between sequences, subclades A-1 and A-2 are separated with >80% bootstrap support (Figure 4). Isolates from 2020 were all in clade A-2 and clustered more closely with each other than with most later isolates; we excluded 1 isolate, 2020-498, which did not follow the fitted molecular clock. All 5 isolates from hospital B were in the same subclade; otherwise, we noted no apparent association between sampling location and either isolation date or genetic similarity.

Isolates from patients <1 month of age predominated in subclade A-2, comprising 93.6% of that clade compared with 38.5% of clade A-1 (p<0.001 by Fisher exact test^­) (Table 3). No association between patient age and genetic similarity was noted among isolates in either subclade. Only isolates from hospital H were represented in both subclades and were equally distributed (Table 3). Those results suggest that genetic distances are small among S. spartinae in that group; that isolates were diverse across time, geography, and patient ages; and that closely related but independent lineages were cocirculating within the same region and institution, potentially indicating separate sources of infection.

We describe S. spartinae as a cause of fungemia in pediatric patients in Pakistan. S. spartinae has been previously described as an environmental organism capable of surviving extreme conditions, but isolation from clinical specimens represents emergence as a human pathogen. The genus Scheffersomyces was proposed by Kurtzman and Suzuki on the basis of D1-D2 large subunit rRNA and small subunit rRNA sequencing; species assigned to the Scheffersomyces genus were originally included in the genus Pichia (39). Most Scheffersomyces species ferment xylose and are used in industrial applications (39–41). The assignment of S. spartinae to the genus was originally considered uncertain because of weak bootstrap support in the D1–D2 small subunit tree and because S. spartinae does not ferment xylose. According to our literature review, S. spartinae has not previously been reported as a cause of human infections nor cultured in healthcare facilities. We speculate S. spartinae might be emerging in humans because of introduction from environmental sources into healthcare environments. However, given the broad distribution of healthcare facilities from which we cultured S. spartinae in Pakistan, we cannot rule out nonhealthcare community sources of infection.

The reasons for the emergence of some fungi as agents of human infection have been multifactorial. Fungi often survive extreme conditions, adapt to selection pressures, and develop enhanced thermotolerance, virulence, and antifungal drug resistance. Emergence of Candida auris in human infections has been attributed to global warming (6,7); some researchers postulate that C. auris inhabits an environmental niche and has only recently become a human pathogen (4). Low virulence fungi such as Saccharomyces cerevisiae, Saprochaeta clavate, and Rhodotorula spp. can contaminate food, medical products, and healthcare environments and cause infections in susceptible hosts (10–12). Alterations in the geographic range of Cryptococcus deuterogattii from tropical regions to temperate climates have led to human infections in nonendemic areas, attributed to human activity, thermal adaptation, and climate change (5).

Another reason for increased emergence of fungal infections is environmental disruption caused by natural disasters, which can lead to a wider distribution of fungi (9). An example is increased human Coccidioides immitis infections after earthquake-related landslides and sandstorms in previously low-prevalence areas (8). Whether similar or other factors contributed to the emergence of S. spartinae as a human pathogen is unclear.

Cases of C. lusitaniae fungemia identified by the AKUH laboratory began increasing in 2020, and a further substantial rise in cases began in 2022. Most C. lusitaniae patients were children <1 year of age who were admitted to hospitals in the city of Karachi. During 2015–2019, rare Candida species were reported as causes of invasive infections in neonatal and pediatric populations in Pakistan (42). Isolation of C. lusitaniae was infrequent, however, so rising case numbers from 2022 onwards were flagged by the laboratory, which more frequently banked cultured isolates. Biochemical identification of Candida by the VITEK 2 YST ID card system inaccurately identified the organism, and most of the isolates were ultimately identified as S. spartinae by whole-genome and ITS sequencing. The high percentage of isolates subsequently identified as S. spartinae suggests that at least some of the isolates identified as C. lusitaniae from before 2020 also might have been S. spartinae, but we cannot verify that hypothesis because earlier specimens were not banked.

Species misidentification is concerning and implies that phenotypic identification systems cannot reliably identify emerging pathogens. Misidentification underscores the need for using molecular approaches, such as ITS sequencing, for species identification. Decreasing costs and increasing availability of WGS promise improved species identification as well as strain typing and genomic epidemiology, but the technology remains difficult to access in many low-resource areas.

Although evidence implies that cases of S. spartinae infection represent the emergence of a new pathogen, the possibility of a pseudo-outbreak (i.e., clustering of positive cultures because of contamination of clinical specimens incorrectly attributed to infections) must also be considered. Compared with bacterial pseudo-outbreaks, fungal pseudo-outbreaks have rarely been reported. Documented pseudo-outbreaks involving Candida guilliermondii were related to inadequate sterile technique during blood culture collection (43), contamination of heparin solution used to flush needles before blood draws (44), and an anaerobic holding chamber in the clinical microbiology laboratory (45). Pseudo-outbreaks reported with other Candida species include C. versatilis contamination in olive oil used for culture media supplementation (46) and C. parapsilosis contamination of a salt solution used for grinding tissues (47).

Unlike prior pseudo-outbreak reports that involved temporally clustered samples from single institutions, we report isolates collected over the course of 4 years from >14 healthcare institutions across Pakistan. Furthermore, we noted considerable demographic similarity between patients from which the positive specimens were collected: >68% were <1 month of age, and >80% were <1 year of age (Table 1). Although all the isolates were cultured and identified by a single clinical microbiology laboratory at AKUH, all processes and equipment used for processing clinical specimens in the laboratory were independent of the age of the patient from which the specimen was collected. If contamination in the laboratory explained most or all observed cases, we would expect detection in cultures from patients across all age ranges and a larger number of specimens. Even when C. lusitaniae cases increased during 2022 and 2023, those cases constituted only 0.57% of 16,505 blood cultures processed from children <1 year of age in 2022 and only 0.60% of 18,051 cultures from children in that age group in 2023 (Appendix Table 2).

The genetic diversity of specimens across and within geographic locations (Figures 3, 4) reduces the likelihood of a single infection or contamination source. Although genetic diversity between isolates within a geographic area of sampling was low, that diversity was not inconsistent with the degree of genetic variation among geographically clustered infections caused by other yeast species, such as C. auris (48,49). Until we learn more about the genomic variability and evolution of S. spartinae, the degree of expected genetic difference among epidemiologically linked or unlinked specimens will remain unclear; however, cocirculation of at least 2 lineages in Karachi over the course of >4 years suggests multiple introductions into the region.

One limitation of this study is that, before 2022, AKUH prioritized its finite laboratory resources for investigating and banking invasive antimicrobial-resistant isolates and isolates from known pathogens of epidemiologic concern. As a result, isolates identified as C. lusitaniae were not consistently banked, and few isolates from before 2022 were available for species verification and genomic analysis. Another limitation is that detailed patient clinical and demographic information was either unavailable or difficult to obtain because of variability in record-keeping and data availability across the many institutions that provided specimens to AKUH. Many treating physicians could not be contacted because in Pakistan laboratory tests are paid out of pocket by the patients or their guardians, who often provide their own contact details rather than their physicians’ contact information at sample collection. That practice greatly hinders epidemiologic investigations and transmission reconstructions to determine common exposures or shared risk factors among affected persons, to establish patients’ clinical manifestations, and to assess responses to treatments or outcomes. Efforts are ongoing to establish cooperative agreements and study protocols for the most affected institutions in Pakistan for standardized data collection and sharing for S. spartinae fungemia cases or other unusual yeasts.

In summary, our findings underscore the value of epidemiologic monitoring for identifying infection clusters and of genomic and molecular surveillance for identifying rare and emerging pathogens. Future studies will be directed at characterizing mediators of pathogenicity and virulence factors that could contribute to S. spartinae emergence as a human pathogen, as well as exploring potential environmental reservoirs or other sources of infection. Nonetheless, this study demonstrates that building capacity for specimen identification and banking, WGS, and bioinformatic analysis in low- and middle-income countries like Pakistan is imperative for early detection and study of emerging infectious disease threats.