In Vitro Antifungal Resistance of Candida auris Isolates from Bloodstream Infections, South Africa

ABSTRACT

Candida auris is a multidrug-resistant fungal pathogen that is endemic in South African hospitals. We tested bloodstream C. auris isolates that were submitted to a reference laboratory for national laboratory-based surveillance for candidemia in 2016 and 2017. We confirmed the species identification by phenotypic/molecular methods. We tested susceptibility to amphotericin B, anidulafungin, caspofungin, micafungin, itraconazole, posaconazole, voriconazole, fluconazole, and flucytosine using broth microdilution and Etest methods. We interpreted MICs using tentative breakpoints. We sequenced the genomes of a subset of isolates and compared them to the C. auris B8441 reference strain. Of 400 C. auris isolates, 361 (90%) were resistant to at least one antifungal agent, 339 (94%) to fluconazole alone (MICs of ≥32 µg/ml), 19 (6%) to fluconazole and amphotericin B (MICs of ≥2 µg/ml), and 1 (0.3%) to amphotericin B alone. Two (0.5%) isolates from a single patient were pan-resistant (resistant to fluconazole, amphotericin B, and echinocandins). Of 92 isolates selected for whole-genome sequencing, 77 clustered in clade III, including the pan-resistant isolates, 13 in clade I, and 2 in clade IV. Eighty-four of the isolates (91%) were resistant to at least one antifungal agent; both resistant and susceptible isolates had mutations. The common substitutions identified across the different clades were VF125AL, Y132F, K177R, N335S, and E343D in ERG11; N647T in MRR1; A651P, A657V, and S195G in TAC1b; S639P in FKS1HP1; and S58T in ERG3. Most South African C. auris isolates were resistant to azoles, although resistance to polyenes and echinocandins was less common. We observed mutations in resistance genes even in phenotypically susceptible isolates.

INTRODUCTION

Candida auris is an important multidrug-resistant nosocomial pathogen in health care settings (1). Bloodstream infection is the most frequently reported form of invasive disease, with a reported crude mortality rate of 30% to 60% (2, 3). The clinical impact of antifungal resistance of C. auris is poorly defined, with no published clinical breakpoints and only tentative epidemiological cutoff values (ECVs) (4). Using breakpoints determined for closely related Candida species and based on expert opinion, the U.S. Centers for Disease Control and Prevention (CDC) published a guide to assist laboratories in the interpretation of C. auris MICs (5).

The recommended first-line antifungal agent for C. auris bloodstream infections is an echinocandin (6). However, isolates with reduced susceptibility to these agents have been reported (7–9). Acquired resistance to fluconazole and variable susceptibility to amphotericin B and other triazoles also limit treatment options for C. auris infections (1, 7, 9). Furthermore, a small minority of C. auris isolates are resistant to all classes of systemic antifungal agents in current use for candidemia (9, 10).

C. auris strains are separated into four distinct clades named for the geographic areas in which the strains were first isolated, i.e., South Asian (clade I), East Asian (clade II), African (clade III), and South American (clade IV) (11), with a potential clade V represented by a single isolate (11). Differences in genetic backgrounds, biochemical characteristics, and antifungal susceptibility patterns make each clade unique (12). Antifungal resistance seems to be clade specific; clade I, III, and IV isolates may be resistant to multiple antifungal agents, while clade II isolates are generally more susceptible (13, 14). In addition, S639F/S639P mutations in the FKS1 hot spot 1 region have been detected in echinocandin-resistant isolates (14, 15). Furthermore, distinct mutations in the ERG11 gene (F126L, Y132F, and K143F) and the TACB1 gene (A640V) have been detected in azole-resistant isolates (9, 16).

Since C. auris was first detected in South Africa in 2009, health care-associated transmission events and large outbreaks have led to this pathogen accounting for more than 1 in 10 cases of candidemia (3, 17, 18). Some isolates have been reported to be resistant to more than one antifungal agent (19). In order to determine the resistance profiles of C. auris bloodstream isolates from South Africa and thus guide empirical treatment, we performed antifungal susceptibility testing on C. auris isolates obtained from private- and public-sector hospitals in South Africa through a national laboratory surveillance program in 2016 to 2017.

RESULTS

Cases, species identification, and selection of isolates.

Between 2016 and 2017, 6,669 cases of candidemia were reported. Of those 6,669 cases, additional species identification was performed at the National Institute for Communicable Diseases (NICD) for viable isolates from 3,020 cases and only at the diagnostic laboratories for isolates from 2,856 cases. Thus, the corresponding isolates from 5,876 cases underwent species-level identification at the NICD and/or at a diagnostic laboratory; of those, 794 cases (14%) involved a C. auris bloodstream infection. We previously described those cases in detail (3). Of the 794 cases, 450 had isolates that were identified as C. auris only at a diagnostic laboratory. Only 400 isolates from 344 cases were submitted for further testing at the NICD and confirmed to be C. auris. The clinical details of those cases are presented in Table 1. The 344 cases included 45 patients with 2 or more serial isolates. A total of 394 isolates (99%) were confirmed to be C. auris by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS); the other 6 isolates (3 with no MALDI-TOF MS identification, 2 with MALDI-TOF MS scores of <2.00, and 1 with a low-discrimination identification) were later confirmed to be C. auris by internal transcribed spacer (ITS) sequencing.

TABLE 1.

Demographic and clinical characteristics of patients with C. auris infections in 2016 and 2017 (n = 344)

Demographic or clinical feature	Value
Age (median [interquartile range]) (yr)	53 (34–64)
Sex (n = 340) (no. [%])
Male	215 (63)
Female	129 (37)
Province (n = 344) (no. [%])
Gauteng	327 (95)
Other province	17 (5)
HIV status (n = 42)
HIV seropositive (no. [%])	10 (24)
Has known CD4+ cell count (no. [%])	4 (10)
CD4+ cell count range (cells/μl)	19–86

Antifungal susceptibility distributions.

Table 2 summarizes the broth microdilution (BMD) and Etest MIC distributions and MIC₅₀ and MIC₉₀ values of nine antifungal agents for the 400 C. auris isolates. The fluconazole BMD MIC₅₀ and MIC₉₀ values for all 400 isolates were 128 μg/ml and 256 μg/ml, respectively. The amphotericin B BMD MICs ranged from 2 μg/ml to 4 μg/ml for 27% of the C. auris isolates (107/400 isolates), while 73% (293/400 isolates) had MICs that ranged from 0.25 μg/ml to 1 μg/ml. Only 6% (22/400 isolates) were confirmed to be amphotericin B resistant by Etest; the Etest MICs for those 22 isolates ranged from 2 μg/ml to 8 μg/ml. The BMD MICs for posaconazole, itraconazole, and voriconazole ranged from 0.015 μg/ml to 1 μg/ml, from 0.03 μg/ml to 2 μg/ml, and from 0.03 μg/ml to 8 μg/ml, respectively. The BMD MICs for micafungin and anidulafungin ranged from 0.015 μg/ml to 8 μg/ml and from 0.015 μg/ml to 2 μg/ml, respectively. Two isolates from a single patient had high micafungin MICs of 4 μg/ml and 8 μg/ml but low anidulafungin MICs of 1 μg/ml and 2 μg/ml, respectively. The micafungin Etest MICs for these 2 isolates were 16 μg/ml. Flucytosine MICs were relatively low (range, 0.015 μg/ml to 2 μg/ml) for all 400 C. auris isolates.

TABLE 2.

Antifungal susceptibility of 400 South African C. auris bloodstream isolates from 2016 and 2017^a

Antifungal agent	Test method	No. of isolates with MIC (μg/ml) of:																	MIC50	MIC90	MIC range	% resistant
		0.015	0.03	0.06	0.125	0.25	0.5	1	2	3	4	8	12	16	32	64	128	256
Itraconazole	BMD		3	54	204	122	12	3	2										0.12	0.25	0.03–2	0
Voriconazole	BMD		3	8	18	52	121	141	55		1	1							0.5	2	0.03–8	0
Posaconazole	BMD	26	91	156	104	19	3	1											0.06	0.12	0.015–1	0
Fluconazole	BMD										3	5		31	58	83	110	110	128	256	4–256	90
Caspofungin	BMD	5	21	198	131	26	12	3				2		2					0.06	0.25	0.015–16
Micafungin	BMD	2	12	198	159	14	8	4	1		1	1							0.06	0.12	0.015–8	0.5
Anidulafungin	BMD	2	4	95	232	54	7	5	1										0.12	0.25	0.015–2	0
Flucytosine	BMD	2		79	256	57	4	1	1										0.12	0.25	0.015–2	0
Amphotericin B	BMD					1	23	269	104			3							1	2	0.25–4	27
Micafungin	Etest													2								0.5
Amphotericin B	Etest	14	4	9	28	122	145	55	15	2	4	1							0.38	1	0.015–8	6

^aTentative breakpoints are as follows: fluconazole, ≥32 μg/ml; micafungin and anidulafungin, ≥4 μg/ml; amphotericin B, ≥2 μg/ml. Resistant isolates are highlighted in bold.

Multidrug resistance.

Of the 400 C. auris isolates, 361 (90%) were resistant to at least one antifungal agent and, of those, 339 (94%) were resistant to at least fluconazole. Of the 339 fluconazole-resistant isolates, 19 (6%) were also resistant to amphotericin B and thus were multidrug resistant. The flucytosine MICs for those 19 isolates ranged from 0.06 μg/ml to 1 μg/ml. Two of 3 C. auris isolates from a single patient were micafungin resistant (MICs of 4 μg/ml and 8 μg/ml), fluconazole resistant (MICs of 32 μg/ml and 64 μg/ml), and amphotericin B resistant (MICs of 4 μg/ml and 2 μg/ml). This patient was a 69-year-old man who was admitted to a cardiothoracic intensive care unit at a private hospital in Pretoria, South Africa. Blood culture samples collected on 3 consecutive days in August 2016 yielded C. auris. The first bloodstream isolate was susceptible to all antifungal agents except amphotericin B (MIC of 2 μg/ml). The 2 subsequent isolates were confirmed by Etest to be resistant to micafungin (MICs of ≥16 μg/ml), anidulafungin (MICs of ≥16 μg/ml), and amphotericin B (MICs of 2 μg/ml). The flucytosine MICs for these 3 isolates remained low (0.25 μg/ml).

Clade-specific susceptibility and resistance mutations.

Of the 92 isolates with whole-genome sequencing (WGS) analysis results, 84 (91%) were resistant to one or more antifungal agents, while the remaining 8 were susceptible to all tested antifungal agents. The majority of the isolates (77 isolates) belonged to clade III, 13 belonged to clade I, and 2 belonged to clade IV (Table 3). Among the 77 clade III isolates, 70 (91%) in total had some evidence of resistance; 60/70 (86%) isolates were resistant to fluconazole alone, 7/70 (10%) were resistant to both fluconazole and amphotericin B, 2/70 (3%) were resistant to micafungin, fluconazole, and amphotericin B, and 1/70 (1%) was resistant to amphotericin B alone. Of the clade III isolates, 99% (76/77 isolates) had two substitutions (F126L and V125A; now referred as VF125AL) based on ERG11 gene mutations. These mutations were observed in 68 isolates with fluconazole MICs of ≥32 μg/ml and 8 isolates with MICs of 8 μg/ml to 16 μg/ml. These 76 clade III isolates also had N647T MRR1 substitutions. In addition, only 16 of the clade III isolates had A651P TAC1b substitutions, with a single isolate having an extra S195G TAC1b substitution. These isolates had BMD fluconazole MICs ranging from 16 μg/ml to 256 μg/ml (see Table S1 in the supplemental material). The distribution of common specific substitutions within the sequenced resistance genes in each clade is shown in Fig. 1.

TABLE 3.

Antifungal resistance and susceptibility across C. auris clades (n = 92)

C. auris isolate susceptibility profile	No. (%) in:
	Clade I	Clade III	Clade IV
Resistant (n = 84)
Fluconazole	1 (8)	60 (78)	1 (50)
Amphotericin B	0	1 (1)	0
Micafungin	0	0	0
Fluconazole and amphotericin B	12 (92)	7 (9)	0
Fluconazole, amphotericin B, and micafungin	0	2 (3)	0
Susceptible (n = 8)
Fluconazole	0	0	0
Amphotericin B	0	0	0
Micafungin	0	0	0
Fluconazole and amphotericin B	0	0	0
Fluconazole, amphotericin B, and micafungin	0	7 (9)	1 (50)
Total (n = 92)	13 (14)	77 (84)	2 (2)

FIG 1.

Distribution of known drug mutations within the ERG11 (Y132F, VF125AL, K117R, N335S, and E343D) (n = 5), MRR1 (N647T) (n = 1), TAC1b (S195G, A651P, and A657V) (n = 3), FKS1HP1 (S639P) (n = 1), and ERG3 (S58T) (n = 1) genes in 92 C. auris isolates. ERG11, MRR1, and TAC1b mutations are associated with fluconazole resistance, FKS1 mutations with echinocandin resistance, and ERG3 mutations with amphotericin B resistance.

Twelve (92%) of 13 clade I isolates were resistant to both fluconazole and amphotericin B, and 1 isolate was resistant to fluconazole alone. One of the 2 clade IV isolates was resistant to fluconazole but not to any other agent (Table 3). The 13 clade I isolates all had Y132F ERG11 substitutions, while 8 clade I isolates had A657V TAC1b substitutions and only 2 isolates had A651P TAC1b substitutions. The 2 clade IV isolates with fluconazole MICs of 16 μg/ml and 64 μg/ml had M351V and A27T ERG9 substitutions and K177R, N335S, and E343D ERG11 substitutions. The fully susceptible clade IV isolate also had 12 uncommon substitutions within the MRR1 gene (S30T, N70S, E76_P77delnsDS, D80E, N133S, K138E, K167N, L211V, R249K, R280G, R413K, and K534N), while the resistant clade IV isolate had a A651P TAC1b substitution.

Three clade III isolates from a single patient, 2 with Etest micafungin MICs of ≥16 μg/ml and 1 with a BMD micafungin MIC of ≥2 μg/ml, had an S639P substitution due to a mutation within the FKS1 hot spot 1 region. Another 3 clade I echinocandin-susceptible isolates (anidulafungin BMD MICs of ≤1 μg/ml and micafungin BMD MICs of 0.5 μg/ml) from different patients had an uncommon D642Y substitution due to a FKS1 hot spot 1 region mutation, while a single clade III echinocandin-susceptible isolate (anidulafungin and micafungin BMD MICs of 0.06 μg/ml) also had three different uncommon substitutions (T125I, C1253fs [fs = frameshift], and G1250S) due to mutations within the FKS1 hot spot 1 region (see Table S2).

Both clade IV isolates (amphotericin B BMD MICs of ≤2 μg/ml and Etest MICs of ≤0.75 μg/ml) had uncommon S58T substitutions due to mutations within the ERG3 gene. No mutations were observed within the ERG6 and ERG10 genes for any of the 92 isolates. Of the 8 fully susceptible isolates, only 1 clade III isolate with a micafungin BMD MIC of 0.5 μg/ml, an anidulafungin BMD MIC of 0.12 μg/ml, a fluconazole BMD MIC of 4 μg/ml, and an amphotericin B Etest MIC of 0.38 μg/ml did not have any mutation within any genes and was considered a wild-type strain.

Clinical treatment and outcomes.

Clinical data on antifungal treatment and in-hospital outcomes were available for 25% of the patients (87/344 patients) with C. auris candidemia. Table S3 shows data for the 75 patients with documented receipt of systemic antifungal therapy, including MIC data and in-hospital outcomes. The overall in-hospital case fatality rate was 47% (35/75 patients).

Of the 35 patients who died, 11 were treated with amphotericin B alone and 15 were treated with amphotericin B plus either an echinocandin (n = 11) or azole (n = 4). The remaining 9 patients were treated with either an echinocandin (n = 7), fluconazole (n = 1), or an unspecified agent (n = 1). Thirty-two of the 35 isolates from patients who died were resistant to fluconazole alone (MICs of ≥32 μg/ml), while 1 isolate was resistant to both fluconazole and amphotericin B (MIC of 3 μg/ml). The remaining 2 isolates were susceptible to all tested antifungal agents. Sixteen of the patients who died had been previously treated with fluconazole (n = 5), micafungin (n = 3), anidulafungin (n = 2), amphotericin B and fluconazole (n = 2), posaconazole (n = 1), amphotericin B, caspofungin, and micafungin (n = 1), caspofungin and anidulafungin (n = 1), or posaconazole, amphotericin B, fluconazole, itraconazole, voriconazole, micafungin, and anidulafungin (n = 1).

Only 3 of the 35 patients had their isolates sequenced. These 3 isolates had fluconazole MICs of ≥32 mg/ml and VF125AL mutations within the ERG11 gene. In addition to fluconazole resistance, 1 isolate had an amphotericin B MIC of >3 μg/ml; however, no mutations were detected within the ERG3, ERG6, and ERG10 genes. For the 2 fluconazole-resistant isolates, 1 patient had been previously treated with a combination of antifungal agents, including posaconazole, amphotericin B, fluconazole, voriconazole, micafungin, and anidulafungin, before being treated with amphotericin B for 18 days during the episode of candidemia. The other patient had been previously treated with fluconazole and then was initially treated for the episode of candidemia with amphotericin B for 1 day, followed by fluconazole for 24 days. The third patient with a fluconazole- and amphotericin B-resistant isolate had no prior antifungal treatment and was treated for the episode of candidemia with amphotericin B for 10 days.

Of the 40 patients who recovered, 9 were treated with amphotericin B alone, while the others were treated with more than one antifungal agent (see Table S3). Of the 40 C. auris isolates from the patients who recovered, 38 (95%) were resistant only to fluconazole (MICs of ≥32 μg/ml). Nine of the patients had been previously treated with micafungin (n = 4), fluconazole (n = 3), voriconazole (n = 1), and amphotericin B (n = 1). One of the 9 patients had the isolate sequenced. That isolate had a fluconazole MIC of ≥32 μg/ml and had VF125AL mutations within the ERG11 gene. The patient had received prior fluconazole treatment before being treated for the episode of candidemia with anidulafungin for 11 days.

DISCUSSION

We performed antifungal susceptibility testing on 400 South African C. auris bloodstream isolates from national surveillance. Ninety percent of the 400 isolates were resistant to fluconazole, but only 5% were amphotericin B resistant and fluconazole resistant. Two isolates from a single patient were resistant to three antifungal classes. Of 92 isolates that were sequenced, 84% belonged to clade III, 14% belonged to clade I, and 2% belonged to clade IV. A much larger proportion of clade I isolates were multidrug resistant (>90% resistant to fluconazole and amphotericin B), compared with clade III isolates. Mutations were observed within resistance genes in both susceptible and resistant clade III and clade IV isolates.

C. auris was the third most common Candida species isolated from South African patients with bloodstream infections (3). The increase in the number of C. auris infections reported worldwide is of major concern because this fungus is difficult to identify using standard identification methods, is often multidrug resistant, and can cause large outbreaks in acute hospital and long-term health care settings (9, 20). Recently, Magobo et al. reported the emergence of multidrug-resistant isolates (8%) among 85 tested South African C. auris isolates (19). However, our study sample was derived from 2-year national surveillance of candidemia and provides a much more representative picture of the antifungal susceptibility profile (19). The vast majority of our study isolates were fluconazole resistant. Fluconazole must be avoided as first-line empirical treatment for candidemia in hospitals and units in which C. auris is endemic. We also observed resistance to amphotericin B and echinocandins in our setting, albeit at a much lower relative frequency. We found that resistance to azoles and amphotericin B was the most common resistance combination (5%), while 2 isolates (0.5%) were resistant to azoles, amphotericin B, and echinocandins. A multicenter study of 350 Indian C. auris isolates collected between 2009 and 2017 reported that 14% were resistant to both azoles and flucytosine, 7% to both azoles and amphotericin B, and 2% to azoles and echinocandins (2). In contrast, all of our isolates had low flucytosine MICs. A combination of amphotericin B and flucytosine is a potent regimen and is known to be efficacious for other serious fungal infections, such as cryptococcal meningitis (21). This combination may potentially be useful in resource-limited settings for treatment of invasive C. auris infections, and this should be explored in prospective studies.

A majority of our isolates belonged to clade III, while smaller proportions belonged to clade I (14%) and clade IV (2%). We found that a much larger proportion (92%) of clade I isolates were multidrug resistant (resistant to both fluconazole and amphotericin B), compared to 13% of clade III isolates. Chow et al. also reported a larger proportion of clade I isolates (45%) as multidrug resistant, compared with clade III (8%) and clade IV (10%) (14). Furthermore, 2 of our clade III isolates were pan-resistant. To date, only 3% of clade I isolates have been reported as pan-resistant isolates (14). Fluconazole resistance has been reported for clade I isolates (in India and Pakistan), clade III isolates (in South Africa), and clade IV isolates (in Venezuela) (9). In contrast, a very low prevalence of fluconazole resistance has been reported for clade IV Colombian isolates and clade II Japanese isolates (2, 22, 23). Fluconazole resistance is associated with clade-specific mutations within the ERG11, ERG9, MRR1, and TAC1b genes. Some of our clade I, clade III, and clade IV isolates had mutations similar to those reported previously, while other mutations were uncommon (9, 14, 16). Uncommon mutations within the susceptible C. auris isolates may be related to natural evolutionary divergence, rather than a mechanism of resistance. Ninety-nine of the clade III isolates had ERG11, MRR1, and TAC1b mutations. In Candida albicans, MRR1 and TAC1b are zinc-cluster transcription factors reported to play a role in the regulation of the expression of the multidrug resistance-related genes MDR1 and CDR1, respectively, while ERG11 encodes a microsomal and membrane-bound protein that functions as a lanosterol 14,α-demethylase of the cytochrome P450 family (24). All clade I isolates had ERG11 mutations but no MRR1 mutations. MRR1 mutations have been noted only in clade III isolates (16, 25). However, 1 of the clade IV isolates had 12 MRR1 mutations and a fluconazole MIC of 16 μg/ml. It is difficult to establish whether these mutations contributed to the elevated MIC, since the other clade IV isolate did not have these mutations but had a MIC of 64 μg/ml. Both clade IV isolates had three mutations within the ERG11 gene, which are commonly reported among Colombian isolates but are not associated with fluconazole resistance (22). All clade I isolates had fluconazole MICs of ≥128 μg/ml and had ERG11 mutations. Of the clade III isolates with ERG11 and MRR1 mutations, 7 were considered susceptible, with fluconazole MICs ranging from 8 μg/ml to 16 μg/ml. A larger number of clade III genomes with MICs of ≤4 μg/ml should be sequenced to determine whether these isolates have mutations. A proportion of clade I, III, and IV isolates had TAC1b mutations. Mutations in TAC1b are associated with fluconazole resistance in C. auris isolates (16). The A657V TAC1b substitution was reported for 15 clade I isolates with the Y132F ERG11 substitution and elevated fluconazole MICs, which is similar to our findings (16). However, 2 of our clade I isolates with MICs of 256 μg/ml and a single clade IV isolate with a MIC of 64 μg/ml had a A651P TAC1b substitution. This A651P TAC1b substitution was most common among our clade III isolates with MICs of 64 μg/ml to 256 μg/ml. Rybak et al. reported 16 clade IV isolates harboring A651T TAC1b substitutions (16). None of our isolates harbored the A640V TAC1b substitution, which is most common in clade I isolates with fluconazole MICs of >64 μg/ml and K143R ERG11 substitutions. Our isolates also lacked the K247E, M653V, A15T, and P595L/H TAC1b substitutions, which have been found to occur naturally in C. auris (16).

Amphotericin B resistance is not commonly described among Candida species (11, 26). However, Chow et al. reported that 47% of clade I isolates and 11% of clade IV isolates were resistant to amphotericin B, while all clade III and clade II isolates were susceptible by the Etest method (14). Based on the Etest method, we confirmed resistance in 6% of South African isolates (22/400 isolates). The Etest method yields a much wider MIC range, compared to BMD, partly because the strip includes lower antifungal concentrations (27). Twelve clade I and 10 clade III isolates did not have mutations within the ERG3, ERG6, and ERG10 genes despite having high amphotericin B Etest MICs. In a large clade I C. auris outbreak involving 72 patients in the United Kingdom, Rhodes et al. reported 5 amphotericin B-resistant isolates with MICs of 2 mg/ml, none of which had any ERG3, ERG5, or ERG6 mutations (28). Of the 22 amphotericin B-resistant isolates in our study, none had any mutations. Only the 2 clade IV isolates had a single mutation within the ERG3 gene; these 2 isolates had amphotericin B Etest MICs of 0.75 μg/ml and 0.5 μg/ml and would be considered susceptible. We do not know the relevance of these mutations. In fungi, ERG3 encodes a C-5 sterol desaturase that is involved in one of the final reactions in the ergosterol biosynthesis pathway (24). A missense mutation in the ERG3 gene results in azole resistance in some clinical isolates of Candida albicans and Candida parapsilosis (29, 30). In this study, both the fluconazole-resistant and fluconazole-susceptible clade IV isolates had a mutation within the ERG3 gene.

The molecular mechanism of echinocandin resistance is highly specific and is not affected by multidrug transporters (31). In C. auris, echinocandin resistance is associated with mutations in the hot spot regions of the FKS genes (FKS1, FKS2, and FKS3), which encode β-1,3-d-glucan synthase (32). In this study, only 3 clade I isolates and 4 clade III isolates had mutations within the FKS1 hot spot 1 region. Three of the clade III isolates were from a single patient, but only 2 were resistant to micafungin. However, all 3 isolates had S639P substitutions caused by FKS1 hot spot 1 mutations. The S639P substitution has been reported for echinocandin-resistant isolates from clades I and IV but has not been reported previously for clade III isolates (14). Clade I and III echinocandin-resistant isolates have been reported to have either the S639F or S639Y substitution (7, 14). Three of the clade I isolates had a D642Y substitution due to a FKS1 hot spot 1 mutation, although their anidulafungin and micafungin MICs were ≤1 and 0.5 mg/ml, respectively. The relevance of this D642Y substitution in C. auris needs further investigation. Furthermore, another clade III isolate had three mutations in the FKS1 hot spot 1 region; however, that isolate had anidulafungin and micafungin MICs of 0.06 μg/ml. It is possible that these mutations may be related not to resistance but to other phenotypic characteristics of the organism.

The clonal expansion or transmission of pan-drug-resistant or multidrug-resistant isolates may severely compromise the treatment of C. auris infections. We found that 33% of patients had been exposed to antifungal agents before they were treated for C. auris infection. This might have resulted in the development of resistance and poorer clinical outcomes (33). However, we did not perform a multivariable analysis to look at the association between antifungal resistance and in-hospital outcomes in our study. The Federation of Infectious Diseases Societies of Southern Africa (FIDSSA) guidelines recommend echinocandins as a first-line treatment option for patients with invasive C. auris infection and amphotericin B as an alternative agent in clinical settings where echinocandins are unavailable (6). These agents are still good empirical treatment options in South African hospitals, although robust infection control and antifungal stewardship programs are essential to limit further emergence of resistance. This large study provides a representative national antifungal profile of C. auris strains in South Africa and a baseline for monitoring emerging resistance to the approved antifungal agents. We used CDC tentative breakpoints to interpret MICs, which allows comparison of our results with other published studies. We performed WGS for only a subset of resistant isolates; therefore, the clade distribution may not be completely representative.

Conclusions.

C. auris isolates from national surveillance were almost all resistant to fluconazole, with a smaller proportion resistant to amphotericin B or echinocandins. We observed mutations within resistance genes even in susceptible C. auris isolates, and further studies are required to understand the mechanism of resistance and the relevance of mutations within genes among South African isolates using a larger WGS data set.

MATERIALS AND METHODS

Isolate information and case definition.

We conducted national laboratory-based surveillance for candidemia from 1 January 2016 through 31 December 2017. Clinical microbiology laboratories affiliated with the National Health Laboratory Service (NHLS) or a private-sector pathology practice were requested to send Candida species isolated from blood culture specimens to the Mycology Reference Laboratory at the NICD. Isolates were accompanied by a laboratory report that included species identification and patient demographic details. Individual patients with >1 serial isolate were also included. At sentinel surveillance sites, we collected additional clinical information by chart review and/or interview (3).

Identification of C. auris.

Candida isolates were submitted to the NICD on Dorset transport medium (Diagnostic Media Products, NHLS, Sandringham, South Africa) and, to obtain a presumptive species identification, were inoculated onto chromogenic agar (MAST ID CHROMagar Candida; Mast Diagnostics, Amiens, France) upon receipt. We used MALDI-TOF MS (Bruker Corp., Billerica, MA, USA) to confirm species identifications. We extracted DNA from the isolates using a fungal/bacterial miniprep kit (Zymo Research, Inqaba Biotec, South Africa) if repeated analysis on the MALDI-TOF MS instrument resulted in no peaks, yielded a score of <2.00, or produced no clear identification. DNA amplification and sequencing of the ITS region of the ribosomal gene were then performed using the ITS1 and ITS4 primers (34). We used the National Center for Biotechnology Information (NCBI) Basic Local Alignment Search Tool (BLAST) database to identify the species based on pairwise sequence alignment (http://blast.ncbi.nlm.nih.gov/Blast.cgi). We included only confirmed C. auris isolates.

Antifungal susceptibility testing.

The MICs of nine antifungal agents (amphotericin B, fluconazole, voriconazole, itraconazole, posaconazole, caspofungin, anidulafungin, micafungin, and flucytosine) were determined using dried BMD panels containing alamarBlue (Thermo Fisher Scientific, Cleveland, OH, USA) and following Clinical and Laboratory Standards Institute recommendations (35). All plates were incubated at 35°C, and wells were visually evaluated for growth following 24 h of incubation. The MICs for echinocandins and azoles were defined as the lowest antifungal concentration that caused 50% growth inhibition, compared to the positive control, while the MIC for amphotericin B was defined as the lowest concentration at which there was 100% inhibition of growth. We used CDC tentative breakpoints, which were developed using C. auris MIC distribution data, known molecular mechanisms of resistance, and pharmacokinetic/pharmacokinetic data from a neutropenic mouse model of infection, to interpret MICs. Isolates with an amphotericin B MIC of ≥2 mg/ml, with a fluconazole MIC of ≥32 mg/ml, or with an anidulafungin/micafungin MIC of ≥4 mg/ml were considered resistant to that agent. Micafungin and/or anidulafungin resistance was considered a surrogate marker of resistance to the entire echinocandin class. Caspofungin MICs were not categorized due to the previously reported interlaboratory MIC variability noted for Candida species as a result of batch-to-batch variations in the powder’s potency (36, 37). Multidrug resistance was defined as resistance to more than one antifungal class. There are no breakpoints to interpret itraconazole, posaconazole, voriconazole, and flucytosine MICs. C. parapsilosis ATCC 22019 and Candida krusei ATCC 6258 were run on all days of testing, and MICs were found to be within the required quality control range. We also determined amphotericin B MICs by Etest (bioMérieux, Marcy l'Etoile, France) on RPMI 1640 plates containing 2% glucose (Diagnostic Media Products), according to the manufacturer’s instructions. We used the Etest method since it generates a much wider range of amphotericin B MIC values than those yielded by BMD testing; this may facilitate distinguishing resistant and susceptible isolates (38). Isolates with resistance to echinocandins by BMD testing were retested by Etest for confirmation only. We calculated the range, MIC₅₀, and MIC₉₀ for each distribution.

WGS of C. auris isolates.

We selected all echinocandin- and amphotericin B-resistant isolates and a random sample of fluconazole-resistant isolates. In total, 92 C. auris isolates were selected for WGS; 62 were resistant to fluconazole, 19 were resistant to both amphotericin B and fluconazole, 2 were pan-resistant (resistant to amphotericin B, fluconazole, and echinocandins), 1 was resistant to amphotericin B alone, and 8 were fully susceptible. DNA extraction from these yeast isolates was performed as described above. Paired-end libraries were prepared using the Nextera DNA Flex library preparation kit, followed by 2 × 300-bp sequencing on a MiSeq instrument (Illumina, San Diego, CA, USA). The sequenced paired-end reads were quality controlled and filtered (Q scores of >20 and lengths of >50 bp) using fastqc v0.11.8 and Trim Galore v0.6.4_dev (https://github.com/FelixKrueger/TrimGalore), respectively. The clean reads for each C. auris isolate were analyzed for the detection of mutations using a custom targeted gene approach workflow on the CLC Genomics Server v20 (Qiagen, The Netherlands). We determined the presence of mutations within the ERG11 gene, the ERG9 gene, and transcriptional regulators of efflux pumps (MRR1 and TAC1B genes) responsible for azole resistance, the FKS1 gene hot spot 1 region associated with echinocandin resistance, and ERG3, ERG6, and ERG10 genes associated with amphotericin B resistance in C. auris. Reference sequences and annotations for these genes were obtained from the susceptible clade I reference strain C. auris B8441 (GenBank accession number PEKT00000000) and the clade III C. auris B11221 isolate (GenBank accession number PGLS00000000) (1, 9). Results tables for synonymous or nonsynonymous mutations were populated to MS Excel Workbooks using a custom R script and compared. We compared our findings with those reported previously by Lockhart et al., Chow et al., and Rybak et al. (9, 14, 16).

Ethics.

NICD obtained annual approval for GERMS-SA laboratory-based surveillance from the research ethics committees of several South African universities (University of Witwatersrand, University of KwaZulu Natal, University of the Free State, Stellenbosch University, University of Cape Town, University of Pretoria, Sefako Makgatho University). Patients (from whom surveillance data were collected prospectively through interviews) provided written informed consent.

Data availability.

Genomic data for the 92 C. auris isolates can be found under BioProject accession number PRJNA737309.