Multidrug-Resistant Candida haemulonii and C. auris, Tel Aviv, Israel

Clinical features and experimentally deduced virulence indicate that C. auris has the greater lethal potential.

contact among patients were implicated in C. haemulonii transmission. C. auris exhibited higher thermotolerance, virulence in a mouse infection model, and ATP-dependent drug efflux activity than C. haemulonii. Comparison of ribosomal DNA sequences found that C. auris strains from Israel were phylogenetically distinct from isolates from East Asia, South Africa and Kuwait, whereas C. haemulonii strains from different countries were closely interrelated. Our findings highlight the pathogenicity of C. auris and underscore the need to limit its spread.
C andida species are leading causes of bloodstream infection (BSI) in hospitalized patients, particularly those in intensive care units who are exposed to broadspectrum antimicrobial drugs, indwelling vascular catheters, parenteral nutrition, abdominal surgery, and immunosuppressive agents (1,2). High rates of attributable death have been associated with delayed initiation of appropriate antifungal treatment (3,4). This problem is compounded by the emergence of drug-resistant Candida species, notably C. glabrata, in many hospitals (5).
C. auris is an emerging opportunistic pathogen, first reported in 2009 as an isolate from the external ear of an inpatient at a hospital in Japan (6). It has since been identified as a cause of nosocomial BSI in numerous countries in East Asia, the Middle East, Africa, and Europe (7)(8)(9)(10)(11). C. auris might be resistant to multiple classes of antifungal agents and apparently has a potential for person-to-person transmission, challenging clinicians and infection control teams (12). C. auris often is misidentified by traditional microbiological methods as C. haemulonii, a phylogenetically related drug-resistant Candida species that also is increasingly reported in healthcare facilities worldwide (13).
We report on the detection of multidrug-resistant C. auris and C. haemulonii in clinical specimens in Tel Aviv, Israel, and specifically on the emergence of C. auris as a cause of nosocomial BSI. We highlight distinct clinical and epidemiologic characteristics of these 2 species and present experimental evidence for differences in their virulence.

Materials and Methods
We undertook this study after C. auris BSI was detected in 4 patients during May-October 2014 at the Tel Aviv Sourasky Medical Center (TASMC), a tertiary-level hospital in Tel Aviv. An additional C. auris bloodstream isolate was recovered in April 2015 from a patient at the Wolfson Medical Center in Holon (southern Tel Aviv metropolitan area). No additional C. haemulonii or C. auris isolates were identified through inquiries at additional clinical microbiology laboratories in Israel.
The TASMC Institutional ethics committee approved this study. Need for informed consent was waived because of the observational and anonymous nature of the study.

Clinical Candida isolates
Candida isolates recovered from clinical specimens were identified at the TASMC Clinical Microbiology Laboratory by growth characteristics on CHROMagar Candida (CHROMagar, Paris, France) and the Vitek 2 YST ID system (bioMérieux, Marcy-l'Étoile, France). The Vitek 2 database does not include C. auris, and this species is routinely misidentified as C. haemulonii (13). We therefore reviewed all isolates identified as C. haemulonii during January 2009-August 2015. Isolates recovered during May 2014-August 2015 were stored at -20°C and subjected to further analyses. We assessed thermotolerance by plating serial dilutions of yeast culture on Sabouraud dextrose agar (SDA) plates and assessing growth after 24 h incubation at 35°C-42°C.

Phylogenetic Analyses
We aligned ITS and LSU sequences of C. haemulonii and C. auris isolates by using MUSCLE (15) and generated phylogenetic trees with the neighbor-joining method (16), using the Kimura 2-parameter method to compute evolutionary distances (17). We tested phylogeny with the bootstrap method (500 replicates) and used Schizosaccharomyces pombe strains ATCC 38366 and CBS 356 as outgroups. Evolutionary analyses were performed in MEGA7 (18).

Patient Characteristics
We retrospectively reviewed the medical records of patients from whom Vitek-identified C. haemulonii was recovered from any site and recorded patient demographics, hospital unit, co-morbidities, medications, and clinical characteristics by using a structured form. We found 40 patient-specific C. haemulonii isolates, 20 (50%) of which originated from patients receiving care at an outpatient peripheral vascular disease clinic (clinic A). We therefore conducted an investigation at clinic A, which included review of patient treatment protocols, observed patient care, and surveillance mycologic cultures from environmental surfaces, wound irrigation solutions, dressings, and the hands of medical staff. To define risk factors for C. haemulonii colonization, we conducted an unmatched case-control study using 40 noncolonized patients followed at clinic A as controls.

Antifungal Susceptibility Testing
We determined MICs of fluconazole, itraconazole, voriconazole, posaconazole, amphotericin B, anidulafungin, micafungin, caspofungin, and flucytosine by broth microdilution using Clinical and Laboratory Standards Institute methods (19). Results were read after 48 h for azoles, amphotericin B, and flucytosine and after 24 h for echinocandins.

Rhodamine 6G Efflux
To assess ABC-type drug transporter activity, we determined glucose-induced efflux of rhodamine 6G, as described previously (20,21). We grew Candida isolates to log-phase in liquid yeast extract glucose at 35°C. We then collected yeast cells by centrifugation, transferred 10 9 cells to 20 mL fresh yeast extract glucose, and incubated them at 27°C for an additional 2 h. Next, we collected yeast cells by centrifugation, washed them twice in phosphate-buffered saline (PBS), and added 10 mL PBS containing 15 µM rhodamine 6G without glucose to the pellets. Suspensions were vortexed and incubated at 27°C for 90 min to enable rhodamine 6G uptake under carbon source-depleted conditions. We then collected cells by centrifugation, washed them twice in PBS, and suspended them in 750 µL PBS in microfuge tubes. To start rhodamine 6G efflux, we added 250 µL PBS with 8 mmol/L glucose. We prepared control tubes with glucose-free PBS, removed them after 5, 15, and 25 min of incubation at 35°C, and measured fluorescence in 200-μL aliquots of supernatant by using a spectrophotometer at excitation 527 nm and emission 555 nm (Synergy HT, BioTek, Winooski, VT, USA).

Mouse Model of Disseminated Candidiasis
To assess the relative virulence of C. haemulonii and C. auris, we determined the lethality and tissue fungal loads of representative strains in a mouse model of hematogenous disseminated candidiasis. Experiments were approved by the TASMC Institutional Animal Care and Use Ethics Committee. We used cyclophosphamide (150 mg/kg intraperitoneally) to immunosuppress 6-week-old female BALB/c mice weighing 16-20 g (Harlan, Rehovot, Israel) 3 days before and on the day of infection. Candida cells were collected from log-phase culture on the day of infection and washed twice in sterile PBS. Mice were infected in groups of 10 with C. haemulonii strain TA001-14, C. auris strain TA005-14, and C. albicans strain CBS 8837. We injected 100 µL of PBS containing 7 × 10 7 yeast cells intravenously into the lateral tail vein of each animal. A control group received intravenous injection of cell-free PBS. Death was assessed over 30 days. Kidney tissue fungal loads were determined in separate experiments where mice were similarly immunosuppressed and infected intravenously with 4 × 10 7 yeast cells/100 μL PBS. Seven days after infection, mice were killed by CO 2 inhalation, and kidneys were excised aseptically, weighed, and homogenized in a TissueLyser (QIAGEN). Homogenates were serially diluted 10-to 1,000-fold in sterile saline and plated on SDA. We calculated fungal loads (CFU per gram of tissue) from colony counts after 48 h incubation at 35°C.

Statistical Analyses
We compared continuous variables between case and control patients using the Student t test for normally distributed variables and the Wilcoxon rank-sum test for non-normally distributed variables. We compared dichotomous variables using Fisher's exact test. Rhodamine 6G efflux, expressed as relative fluorescence units, was computed for each Candida strain and compared by using 1-way analysis of variance. We used Dunnett's multiple comparisons test to compare relative fluorescence unit values of specific C. haemulonii and C. auris strains with averaged control values of 5 C. glabrata strains. Survival curves of mice infected with different Candida strains were plotted by using the Kaplan-Meier method and compared with the log-rank test. We considered 2-tailed p values <0.05 statistically significant.

Sequence-Based Identification
We identified 40 patient-specific Candida strains as C. haemulonii by the Vitek-2 YST ID system during January 2009-July 2015. Isolates were recovered from wounds (n = 24), urine (n = 9), blood (n = 5), and central venous catheter tips (n = 2). Of these, 9 isolates recovered during May 2014-May 2015 were available for analysis; we identified 6 (including the 5 blood isolates) as C. auris and 3 as C. haemulonii by ITS and LSU sequencing (Table 1). Sequences were 100% identical among strains of each species. C. auris strains from Israel shared 98.6% and 98.3% similarity of ITS and LSU sequences, respectively, with the C. auris type strain CBS10913 T . C. haemulonii strains were 100% identical to C. haemulonii CBS5149 T on the basis of ITS and LSU sequences.
Phylogenetic trees based on ITS and LSU sequences showed that the C. auris isolates from Tel Aviv are distinct from other isolates from East Asia, Africa, and the Middle East. Specifically, isolates from Israel showed 98.6% similarity of ITS and LSU sequences with the India clone, represented by CBS12768, 96.2% similarity with the South Korea clone, and 96.7% similarity with strain CH1 from Kuwait. In contrast, ITS and LSU sequences from Israel C. haemulonii strains were 100% homologous with C. haemulonii from South Korea, Brazil, and Kuwait, suggesting worldwide predominance of a single C. haemulonii clone (Figure 1).

Clinical Features
Eight of 9 patients with sequence-validated isolates were hospitalized at TASMC (Table 1). An additional patient with C. auris infection was hospitalized at the Wolfson  Medical Center, but was receiving regular care for HIV infection at TASMC. All 3 C. haemulonii isolates were recovered from chronic leg ulcers of patients with peripheral vascular disease, 2 of whom were treated at vascular outpatient clinic A. Five of 6 C. auris isolates represented BSI: 3 patients had vascular catheter-related candidemia, and 2 had primary nosocomial candidemia of unclear origin. Two of 5 patients with C. auris BSI died during hospitalization. We reviewed the medical records of 40 patients with Vitek-identified C. haemulonii cultures. Thirty-three (83%) were male. Median age was 74 years (range 37-91 years). Nineteen (48%) had peripheral vascular disease, 20 (50%) had diabetes mellitus, 22 (55%) had ischemic heart disease, and 11 (28%) had end-stage renal disease. Twenty patients (50%) were receiving regular care at clinic A, representing 8% (20/261) of all clinic patients. In all 20 patients, C. haemulonii had been recovered from chronic leg ulcers, and none had documented wound infection at the time of culture. Cultures of environmental surfaces, medical devices, dressings, irrigation solutions, and hands of medical staff were negative for yeast. Compared with 40 control patients who were not carriers of C. haemulonii, carriers were older, had a lower glomerular filtration rate, and were more likely to be male and to have ischemic heart disease ( Table 2). Observations revealed a practice among medical staff of routinely applying topical miconazole cream to chronic ulcers without evidence of infection. Periodic wound cultures were obtained regularly, irrespective of signs of ulcer inflammation or purulence. Multiple social interactions were noted among patients in a single room where wound care was performed.

Thermotolerance
Survival and growth at physiologic temperature are prerequisites for microbial invasion and pathogenicity. C. haemulonii isolates grew well at 35°C, but growth at 37°C was poor or absent, and no growth occurred at 40°C and 42°C. In contrast, growth of C. auris isolates at 37°C and 40°C was similar to that of C. albicans, and 4 of 6 isolates grew at 42°C (Figure 3).

Virulence in a Mouse Model of Disseminated Candidiasis
We compared the virulence of C. auris and C. haemulonii isolates in a mouse model of hematogenous disseminated candidiasis. C. haemulonii was completely nonvirulent in this model; 100% of mice survived 12 days after inoculation with no visible signs of illness. In contrast, inoculation with C. auris resulted in rapid death and only 20% survival 5 days after infection (p = 0.0002, log-rank test). Death of mice infected with C. auris was significantly less rapid than that of mice infected with C. albicans (median survival 4 d  (22); NA, odds ratio is not applicable for continuous variables. and 1 d, respectively; p = 0.01; Figure 4, panel A). Kidney tissue fungal load correlated with survival rates. Specifically, we recovered no viable yeast cells from kidneys of mice inoculated with C. haemulonii, whereas infection with C. auris and C. albicans yielded median tissue loads of 5.9 × 10 4 CFU/g and 7.1 × 10 5 CFU/g, respectively (p<0.0001; Figure 4, panel B). Histopathologic analysis showed yeast cell aggregates in kidneys of C. auris-inoculated mice, distinct from tissue invasive hyphae observed in C. albicansinfected kidneys (Figure 4, panel C).

Discussion
Concern about the international emergence and spread of C. auris as a cause of invasive infection in hospitals stems from 3 characteristics of this opportunistic pathogen (12,13): 1) resistance to multiple antifungal drugs and possibly to all major classes of systemic antifungal drugs; 2) horizontal transmission among hospitalized patients, leading to nosocomial outbreaks (8,10,11,13); and 3) high associated death rates (7,8,10). C. auris and C. haemulonii are phylogenetically related species in the Metschnikowia clade that share a propensity for multidrug resistance. We identified C. auris and C. haemulonii in 2 hospitals in Israel and highlighted clinical and experimental evidence for differences in the drug-susceptibility patterns, drug efflux activity, pathogenicity, and global phylogenetics of these 2 species.
In our study, C. auris and C. haemulonii had high MICs of azoles and amphotericin B. Echinocandin MICs were within the susceptible range. An amphotericin B epidemiologic cutoff value of 2 mg/L previously was established (23), but clinical correlation between amphotericin B MIC and treatment outcomes is lacking (24). Compared with C. auris, C. haemulonii isolates had higher amphotericin B MICs. The relevance of these resistance patterns to treatment strategies remains to be determined.
ABC-type efflux activity, as evidenced by Rhodamine 6G transport, was significantly greater among C. auris than C. glabrata isolates. This observation provides a mechanistic basis for the intrinsic resistance of C. auris to azoles and is consistent with the identification of multiple putative transporter-encoding genes belonging to the ABC and major facilitator gene families in the C. auris genome (25).  Of Vitek-identified C. haemulonii isolates at TASMC, 50% were wound cultures from patients cared for at clinic A. That most of these isolates were not available for sequencing is a limitation of our study. However, we identified the 2 Candida isolates from clinic A patients that were available by ITS and LSU sequencing as C. haemulonii, and all 3 sequence-identified C. haemulonii isolates were recovered from leg ulcers of patients with peripheral vascular disease. Colonization of patients treated in close proximity in 1 room strongly suggests person-to-person transmission and supports interim guidelines for contact isolation (26). However, we were unable to identify an environmental reservoir of C. haemulonii. We suggest that topical application of miconazole to wounds most likely caused selective pressure and facilitated the overgrowth of C. haemulonii. After this investigation and termination of routine topical azole use, no additional cases of C. haemulonii were detected in clinic A during April 2015-July 2016.
We recovered 5 of 6 sequence-identified C. auris isolates from patients with nosocomial BSI. In contrast, all C. haemulonii isolates were cultured from superficial wounds. This observation reflects the global epidemiology of these species. C. haemulonii has been isolated from chronic leg ulcers of patients in India and Brazil (27,28). C. auris has caused outbreaks of BSI in the United Kingdom (13), India (8), Kenya (11), South Africa (9), and South Korea (29), whereas reports of C. haemulonii as an agent of BSI have been infrequent (27,(29)(30)(31)(32). Moreover, C. auris fungemia is associated with high death rates (8,10), contrasting with reports of patients surviving prolonged C. haemulonii fungemia (31). Fatal C. haemulonii fungemia, although rare, has been reported in neonates and in patients with cancer and neutropenia (27,32).
In our study, C. auris, but not C. haemulonii, grew at 37°C-42°C and exhibited lethality and tissue invasion in a mouse model of invasive candidiasis only slightly less than those of C. albicans, the prototypical pathogenic Candida species. Both C. auris and C. haemulonii are unable to form hyphae, which contribute to virulence in C. albicans. Formation of large aggregates resulting from failure of budding yeast to separate has been noted in some C. auris isolates (33). We observed distinct yeast cell aggregates in the kidneys of mice with lethal C. auris infection, which suggests that aggregation might be a mode of immune evasion and persistence in tissue. The C. auris genome contains C. albicans gene orthologs, such as secreted proteinases and mannosyl transferases, which might have roles in pathogenesis (25). However, C. auris has a genome that is highly divergent from those of other Candida species, and most of its genes have not yet been characterized (25).
Ribosomal DNA sequences were identical among C. haemulonii strains from Israel, Kuwait, East Asia, South America, and the United States. In contrast, the global phylogenetics of C. auris demonstrate distinct clones for each country, indicating greater genomic diversity for this species. Further study is needed to establish whether the divergence of C. auris clones translates into country-specific patterns of invasiveness, virulence, and drug resistance. Our findings affirm the need for intensified vigilance and mobilization of infection control measures to limit the spread of C. auris.

J.B. was supported by the European Research Council Advanced
Award 340087 (RAPLODAPT).
Dr. Ben-Ami is head of the Infectious Diseases Unit and the National Center for Clinical Mycology at the Tel Aviv Medical Center, Israel. His research focuses on the epidemiology, drug resistance traits, and virulence determinants of medically important fungi. showing significantly shorter survival of mice infected with C. albicans than C. auris and no death among mice infected with C. haemulonii. B) Kidney fungal load (CFU per gram of tissue) shown to be significantly higher in mice infected with C. albicans than in those infected with C. auris, whereas no viable yeast was cultured from kidneys of mice infected with C. haemulonii. C) In mouse kidneys, C. auris cells formed aggregates and no hyphae (top) whereas C. albicans formed extensive tissue-invasive hyphae (bottom); C. haemulonii was not detected in tissue sections (middle). Grocott methenamine silver staining, original magnification ×100 for panels, ×400 for insets.