Evaluation of Caspofungin Susceptibility Testing by the New Vitek 2 AST-YS06 Yeast Card Using a Unique Collection of FKS Wild-Type and Hot Spot Mutant Isolates, Including the Five Most Common Candida Species

Abstract

FKS mutant isolates associated with breakthrough or failure cases are emerging in clinical settings. Discrimination of these from wild-type (wt) isolates in a routine laboratory setting is complicated. We evaluated the ability of caspofungin MIC determination using the new Vitek 2 AST-Y06 yeast susceptibility card to correctly identify the fks mutants from wt isolates and compared the performance to those of the CLSI and EUCAST reference methods. A collection of 98 Candida isolates, including 31 fks hot spot mutants, were included. Performance was evaluated using the FKS genotype as the “gold standard” and compared to those of the CLSI and EUCAST methodologies. The categorical agreement for Vitek 2 was 93.9%, compared to 88.4% for the CLSI method and 98.7% for the EUCAST method. Vitek 2 misclassified 19.4% (6/31) of the fks mutant isolates as susceptible, in contrast to <4% for each of the reference methods. The overall essential agreement between the CLSI method and Vitek 2 MICs was 92.6% (88/95) but was substantially lower for fks mutant isolates (78.6% [22/28]). Correct discrimination between susceptible and intermediate Candida glabrata isolates was not possible, as the revised species-specific susceptibility breakpoint was not included in the Vitek 2 detection range (MIC of ≤0.250 to ≥4 mg/liter). In conclusion, the Vitek 2 allowed correct categorization of all wt isolates as susceptible. However, despite an acceptable categorical agreement, it failed to reliably classify isolates harboring fks hot spot mutations as intermediate or resistant, which was in part due to the fact that the detection range did not span the susceptibility breakpoint for C. glabrata.

INTRODUCTION

Three echinocandins, anidulafungin, caspofungin, and micafungin, are licensed for the treatment of invasive candidiasis and are among the preferred agents for this disease. They exert their antifungal activity through noncompetitive inhibition of β-1,3-d-glucan synthase (GS), thus disrupting the synthesis of the major and essential component of the fungal cell wall, β-1,3-glucan. GS is an enzyme complex with at least two subunits: a catalytic subunit encoded by three related genes (FKS1, FKS2, and FKS3) and a regulatory subunit, Rho1p. The inhibition of GS disrupts the structure of the growing cell wall, resulting in osmotic instability and death of susceptible cells (1).

Echinocandin resistance in species such as Candida albicans, C. tropicalis, C. glabrata, and C. krusei has been uncommon (2). However, with the increased use of echinocandins, Candida isolates with reduced susceptibility to echinocandins are increasingly encountered (3–9). In C. albicans, this is shown to be associated with a variety of single amino acid substitutions due to mutations in specific hot spot regions of the well-conserved target gene FKS1: hot spot 1 (corresponding to amino acids 641 to 649) and hot spot 2 (corresponding to amino acids 1357 to 1364) (10). Homozygous hot spot mutations confer the highest increases in MICs (8). Among isolates of C. albicans, the most significant increases in MICs have been shown to be related to amino acid changes at S645 (whereas changes at other codons confer less pronounced MIC elevations) (10). In the case of C. glabrata, a variety of amino acid changes in FKS1 and/or FKS2 are encountered, with alterations at Fks1p S629 and S663 and Fks2p F659S conferring the highest MIC elevations (11). Likewise, mutations in FKS1 for C. tropicalis and C. krusei have been linked with increases in echinocandin MICs (12, 13). C. parapsilosis and C. guilliermondii have naturally occurring “mutations” at FKS1 hot spot 1 accounting for the elevated MICs observed (14).

The fks hot spot mutations confer an elevated MIC compared with those obtained with wild-type (wt) strains for all three echinocandins. So far, clinical data suggesting that fks mutant isolates are as efficiently treated with standard doses of echinocandins as the wt are lacking. To the contrary, such isolates are reported in breakthrough and failure cases (3, 9), making reliable identification of fks hot spot mutants a pivotal issue. However, this has been challenging using classical susceptibility testing techniques due to overlap or poor separation of wt and fks hot spot mutant isolates (10, 11, 15, 16). Species-specific echinocandin breakpoints have recently been proposed by the EUCAST (17) (anidulafungin as a marker for the echinocandin class) and by the CLSI (18) (individual breakpoints for each of the three echinocandins) in order to better differentiate between the fks hot spot mutant and wt isolates. However, susceptibility testing using the reference methodologies is laborious and not well suited for routine clinical microbiological laboratories. To address this issue, we evaluated the performance of the automated newly introduced Vitek 2 yeast susceptibility card for caspofungin (the only echinocandin currently included in this commercial susceptibility test) with respect to detecting fks hot spot mutants among the five Candida species most commonly involved in human infections.

MATERIALS AND METHODS

Isolates.

A total of 98 isolates were used, consisting of a well-characterized set of FKS wt and hot spot mutant isolates (15) supplemented with two additional clinical C. tropicalis isolates harboring fks hot spot mutations (see below). The combined strain set included 10 FKS wt and 10 fks hot spot mutant C. albicans isolates, 1 FKS wt and 1 fks hot spot mutant C. dubliniensis isolates, 9 FKS wt and 11 fks hot spot mutant C. glabrata isolates, 13 FKS wt and 3 fks hot spot mutant C. krusei isolates, 19 FKS wt C. parapsilosis isolates, and 15 FKS wt and 6 fks hot spot mutant C. tropicalis isolates. The FKS genotype for the two additional C. tropicalis isolates was S80S/P; the genotypes of the remaining 29 fks mutant isolates have been previously reported (15). C. krusei ATCC 6258 and C. parapsilosis ATCC 22019 were included as quality control strains. All isolates were coded, and tests for susceptibility pattern were performed blinded. The isolates had been stored at −80°C before use and were cultured on CHROMagar (SSI Diagnostica, Hillerød, Denmark) plates for 18 to 24 h, checked for purity, and subcultured for another 18 to 32 h before susceptibility testing using Vitek 2 was performed.

Susceptibility testing.

Vitek 2 testing was done according to the Vitek 2 test protocol (bioMérieux, Marcy l'Etoile, France) using unbroken packages of materials and reagents. Briefly, suspensions of Candida isolates were prepared in 0.45% sterile NaCl to a density equal to 2.0 McFarland (acceptable range of 1.8 to 2.2; DensiCheck [bioMérieux]). The testing was done using Vitek 2 AST-YS06 cards (lot number 286225910), and all isolates reached a caspofungin MIC result (test range of measurement: ≤0.25 to >4 mg/liter) within 12 to 24 h; five were terminated due to insufficient growth in positive-control wells for the rest of the antifungal panel (5-fluorocytosine, fluconazole, voriconazole, and amphotericin B). The saline used for the preparations of inoculum was plated onto CHROMagar plates to control for possible contamination and showed no growth after 5 days. The Vitek 2 caspofungin MICs were within the acceptable CLSI MIC ranges (C. krusei ATCC 6258, ≤0.25 mg/liter [reference, 0.125 to 1 mg/liter], and C. parapsilosis ATCC 22019, 0.5 [0.25 to 1 mg/liter]).

Vitek 2 results were compared to EUCAST and CLSI MIC determinations previously reported (15, 19, 20), supplemented with anidulafungin MICs for the two additional C. tropicalis isolates determined by EUCAST EDef 7.2 (21). The revised CLSI breakpoints were adopted for the interpretation of caspofungin MICs: ≤0.25 mg/liter (susceptible [S]), 0.5 mg/liter (intermediate [I]), and ≥1 mg/liter (resistant [R]) for C. albicans, C. tropicalis, and C. krusei; ≤0.12 mg/liter (S), 0.25 mg/liter (I), and ≥0.5 mg/liter (R) for C. glabrata; and, finally, ≤2 mg/liter (S), 4 mg/liter (I), and ≥8 mg/liter (R) for C. parapsilosis (18).

Performance calculations were initially done using the FKS genotype as the “gold standard,” i.e., the capability to identify fks mutant isolates as I or R and wt isolates as S (defined as the categorical agreement [CA]). Very major errors (VMEs) were registered when fks mutant isolates were classified as S, major errors (MEs) when wt isolates were classified as R, and minor errors (mEs) when wt isolates were classified as I. Due to the truncated detection interval of Vitek 2 MICs (the lowest limit of detection being ≤0.25 mg/liter and thus not spanning the CLSI S breakpoint for C. glabrata [≤0.125 mg/liter]), isolates with a Vitek 2 MIC of ≤0.25 mg/liter were pragmatically evaluated as being S. The performance of Vitek 2 was compared to that of each of the two reference methods, CLSI broth microdilution using caspofungin and EUCAST using anidulafungin as a marker for the class evaluated, with the FKS genotype as the gold standard as described above and using the revised species-specific breakpoints (17, 18) for the MICs previously reported for these isolates (15, 19, 20).

Vitek 2 evaluation was also performed using the CLSI method as a gold standard. The essential agreement (EA) was defined as instances with a maximum of ±2 dilutions between Vitek 2 endpoints and caspofungin MICs obtained by the CLSI method. Vitek 2 results outside the detection range (≤0.25 mg/liter and ≥4 mg/liter) were regarded to be in essential agreement with the CLSI caspofungin MIC when the latter were within the ranges of ≤0.008 to 0.25 mg/liter and 4 to >8 mg/liter, respectively. The CA was defined as instances with concordance between susceptibility classification by Vitek 2 and the reference method. VME and ME were calculated as described above with the CLSI values as the gold standard. Minor errors were defined as instances when one of the methods found the isolate intermediately susceptible and the other found it resistant or susceptible.

FKS gene sequence analysis for the original strain collection was described previously (15). For the analysis of the additional C. tropicalis isolates, genomic DNA was extracted as described previously (22), and the following primers were applied as PCR amplification and sequencing primers targeting two hot spot regions in the FSK1 target gene encoding 1,3-β-d-glucan synthase subunit 1 of C. tropicalis: GSC_1F (TCATTGCTGTGGCCACTTTAG) and GSC_1R (TAGAATGAACGACCAATGGAGA) and GSC_2F (ATTGCTCCTGCCGTTGATTG) and GSC_2R (GGTCAAATCAGTGAAACCG). Sequencing was performed at Macrogen (Amsterdam, The Netherlands), and the obtained DNA sequences were aligned with the FKS1 reference sequence of C. tropicalis (GenBank accession no. EU676168), using the bioinformatics software CLC DNA Workbench (CLC, Aarhus, Denmark) (22).

RESULTS

General observations.

Vitek 2 provided an MIC in all 98 cases, including for one C. glabrata isolate that repeatedly failed to grow in the EUCAST medium. The caspofungin MICs for wt and fks mutant isolates using Vitek 2 and CLSI M27-A3 are shown in Fig. 1. The two C. dubliniensis isolates (1 wt and 1 fks mutant) were both correctly classified by all three methods (Fig. 1).

Fig 1.

Caspofungin MICs determined by Vitek 2 (above the x axis) and the CLSI method (below the x axis) for a collection of wild-type (WT) and FKS mutant isolates of Candida. Gray areas border the truncated MIC interval for Vitek 2. The dashed vertical lines indicate the interpretative breakpoints (left dotted line separating susceptible and intermediate categories and right line separating intermediate and resistant categories).

With the FKS genotype as the gold standard, the overall categorical agreement for Vitek 2 was 93.9% (when considering C. glabrata isolates with an MIC of ≤0.25 mg/liter as susceptible) (Table 1). Species-specific differences were noted with the highest CA for Vitek 2 classifications of C. krusei and C. parapsilosis (100%) and the lowest for C. tropicalis (85.7%). Six errors were observed for Vitek 2, which were all VMEs leading to 19.4% of the isolates harboring fks mutations being misclassified as susceptible (Table 1). This was the case in 50% (3/6) of C. tropicalis, 20% (2/10) of C. albicans, and 9.1% (1/11) of C. glabrata fks hot spot mutant isolates. The genotypes of the affected isolates are shown in Table 2. These isolates were all correctly classified as R using EUCAST anidulafungin testing, whereas one isolate was misclassified as S by CLSI caspofungin testing (C. tropicalis F76S).

Table 1.

Comparison of the performance of Vitek 2 (caspofungin), CLSI (caspofungin), and EUCAST (anidulafungin) susceptibility testing using the presence (I or R) or absence (S) of FKS hot spot mutations as a gold standard for susceptibility classification^a

Method	Species	No. of isolates, total (no. fks mutant/no. wt)	No. of misclassified isolates/total (%)
			CA	VME	ME	mE
Vitek 2	C. albicans	20 (10/10)	18/20 (90.0)	2/10 (20)	0/10	0/20
	C. glabratab	20 (11/9)	19/20 (95)	1/11 (9.1)b	0/9	NA
	C. krusei	16 (3/13)	16/16 (100)	0/3	0/13	0/16
	C. tropicalis	21 (6/15)	18/21 (85.7)	3/6 (50)	0/15	0/21
	C. parapsilosis	19 (wt)	19/19 (100)	NA	0/19	0/19
	All (including C. dubliniensis [1/1])	98 (31/67)	92/98 (93.9)	6/31 (19.4)	0/67	0/78
CLSI	C. albicans	20 (10/10)	20/20 (100)	0/10	0/10	0/20
	C. glabrata	19 (10/9)	16/19 (84.2)	0/10	1/9 (11.1)	2/19 (10.5)
	C. krusei	16 (3/13)	12/16 (75.0)	0/3	1/13 (7.7)	3/16 (18.8)
	C. tropicalis	19 (4/15)	17/19 (89.5)	1/4 (25)	0/15	1/19 (5.3)
	C. parapsilosis	19 (wt)	17/19 (89.5)	NA	1/19 (5.3)	1/19 (5.3)
	All (including C. dubliniensis [1/1])	95 (28/67)	84/95 (88.4)	1/28 (3.5)	3/67 (4.5)	7/95 (7.4)
EUCAST	C. albicans	20 (10/10)	20/20 (100)	0/10	0/10	0/20
	C. glabrata	19 (10/9)	19/19 (100)	0/10	0/9	0/19
	C. krusei	16 (3/13)	15/16 (93.8)	1/3 (33.3)	0/13	0/16
	C. tropicalis	21 (6/15)	21/21 (100)	0/6	0/15	0/21
	C. parapsilosisc	19 (wt)	NA	NA	NA	NA
	All (including C. dubliniensis [1/1])	97 (30/67)	77/78 (98.7)	1/30 (3.3)	0/48	0/78

^aCA, categorical agreement; VME, very major error; ME, major error; mE, minor error; NA, not applicable.

^bFor the CA, VME, and ME calculations for Vitek 2, C. glabrata isolates with an MIC of ≤0.25 mg/liter were categorized as S, as this is the lowest MIC which can be reported for this system (implying that there is no I category for C. glabrata).

^cSusceptibility interpretation is not recommended for C. parapsilosis according to the EUCAST reference method E.DEF7.2. C. parapsilosis is therefore excluded from the calculations of misclassifications for the EUCAST method.

Table 2.

FKS genotype, EUCAST anidulafungin MIC, and CLSI caspofungin MICs for the six isolates harboring fks hot spot mutations which were misclassified as susceptible using Vitek 2

Species	FKS genotype (amino acid alteration)a	EUCAST Anidulafungin MIC (mg/liter)a	CLSI caspofungin MIC (mg/liter)a	Vitek 2 caspofungin MIC (mg/liter)
C. albicans	D648Y	0.063	0.5	≤0.25
C. albicans	P649H	0.063	1	≤0.25
C. glabrata	Fks2p S663F	0.25	4	≤0.25
C. tropicalis	F76S	0.125	0.25	≤0.25
C. tropicalis	S80S/P	0.25	ND	≤0.25
C. tropicalis	S80S/P	0.25	ND	≤0.25

^aData compiled from references 15 and 22. ND, not determined.

For comparison, the previously published CLSI and EUCAST data set for this strain collection were evaluated using the revised CLSI breakpoints for caspofungin (as recommended by the CLSI) and the anidulafungin EUCAST MIC results and associated breakpoints as a marker for the echinocandin class (as recommended by the EUCAST). Using the FKS genotype as the gold standard, the CAs were 88.4% for the CLSI reference method and 98.7% for the EUCAST methodology (Table 1). Again, species-specific differences were observed. Thus, for CLSI the CA for C. krusei (75%) was especially low, whereas it was 90 to 100% for C. albicans, C. tropicalis, and C. parapsilosis (Table 1 and Fig. 1). For the EUCAST method, the categorical agreement was 100% for all species except for C. krusei (93.8%). Errors were observed for a total of 11 isolates using the CLSI method and for 1 isolate using the EUCAST method. One was a VME for both methods—CLSI, 1/4 C. tropicalis isolates (F76S), and EUCAST, 1/3 C. krusei isolates (F655F/C)—leading to VME rates of 3.5% and 3.3%, respectively (Table 1). For the CLSI, additionally, three MEs (4.5%) and seven mEs (7.4%; all species represented except C. albicans) were found.

If caspofungin CLSI broth microdilution was selected as the gold standard, an EA of 92.6% and a CA of 96.5% would have been obtained for Vitek 2 (Table 3). However, it is noteworthy that the MICs found by Vitek 2 testing in general were lower than the CLSI MICs (Fig. 1). This was the case for 87.9% (29/33, 1 to 4 2-fold dilutions) of the isolates for which an exact MIC was provided by Vitek 2, whereas the MIC was identical for three isolates and the CLSI MIC was lower than the Vitek 2 MIC for one isolate only. Consequently, eight isolates (all 5 major species represented) were classified as I by the CLSI method but S by Vitek 2, seven of which were wt. Additionally, five isolates were classified as R by the CLSI method but S by Vitek 2 (VME), three of which were wt isolates (one C. glabrata, one C. krusei, and one C. parapsilosis isolate).

Table 3.

Performance of Vitek 2 with CLSI microdilution results as the gold standard^a

Species	No. of isolates, total (no. fks/no. wt)	EA			CA (%)	VME rate (%)	ME rate (%)
		fks (%)	wt (%)	Total (%)
C. albicans	20 (10/10)	10/10 (100)	10/10 (100)	20/20 (100)	18/20 (90)	1/9 (11.1)	0/10 (0)
C. glabrataa	19 (10/9)	5/10 (50)	9/9 (100)	14/19 (73.7)	15/19 (78.9)	2/11 (18.2)	0/6 (0)
C. krusei	16 (3/13)	3/3 (100)	13/13 (100)	16/16 (100)	12/16 (75.0)	1/4 (25.0)	0/9 (0)
C. tropicalis	19 (4/15)	4/4 (100)	15/15 (100)	19/19 (100)	18/19 (94.7)	0/3 (0)	0/15 (0)
C. parapsilosis	19 (wt)	NAc	18/19 (94,7)	18/19 (94.7)	17/19 (89.5)	1/1 (100)	0/17 (0)
All isolatesb	95 (28/67)	22/28 (78.6)	66/67 (98.5)	88/95 (92.6)	82/95 (96.5)	5/29 (17.2)	0/58 (0)

^aWhen calculating the CA, VME, and ME of C. glabrata Vitek 2 results, an MIC of ≤0.25 mg/liter was classified as S (as this was the lowest result reported; hence, differentiation between S and I is impossible). Consequently, the mE does only reflect two isolates classified as I by the CLSI method.

^bIncluding the 2 C. dubliniensis isolates (1 fks mutant and 1 wt).

^cNA, not applicable.

DISCUSSION

Vitek 2 operates with MIC determination for one echinocandin only (caspofungin). As no interpretative breakpoints for caspofungin have been put forward by the EUCAST due to batch differences noted (23), the only internationally approved breakpoints available for interpretation are those provided by CLSI. Adopting these breakpoints for Vitek 2, MIC result classification into S, I, and R is possible for all Candida species except C. glabrata, as the lowest limit of the truncated MIC interval is ≤0.25 mg/liter and thus above the selected CLSI breakpoint for this species. Hence, it is not possible to discern between S and I isolates without using an alternative susceptibility test.

In this study, we have pragmatically chosen to accept an MIC of ≤0.25 mg/liter as S for C. glabrata, as otherwise the entire wt population would be misclassified as I, leading to a much higher mE rate. By this approach, the CAs appear acceptable (93.9%) and no wt isolates were misclassified as I or R, which would potentially lead to echinocandin treatment being abandoned in cases where such would have been appropriate. However, when evaluating the ability to correctly identify the fks mutant isolates as nonsusceptible, Vitek 2 testing showed cause for concern. Indeed, almost a fifth of the fks mutant isolates of the more common species C. albicans, C. glabrata, and C. tropicalis were misclassified as susceptible. Reliable separation between wt and fks mutant isolates is challenging using biological susceptibility testing (15, 23, 24). Importantly, not all fks mutants are alike. Among the codons associated with the highest level of echinocandin resistance are those for C. albicans S645 or F641, C. glabrata Fksp1 F629 and Fksp2 F659 and D666, and C. tropicalis S80P, and none of the mutants with alterations at these codons were misclassified (10, 11, 15, 25). It thus appears that Vitek 2 successfully identified the most resistant genotypes as R, giving more ambivalent answers regarding mutant isolates with lower levels of echinocandin resistance. In this context, it is interesting that recent studies have shown that such isolates also in animal models are associated with a less pronounced loss of in vivo susceptibility (26, 27). However, we still feel that the main reason for performing susceptibility testing in the clinical context is to identify isolates with elevated MICs for which the clinical outcome may differ unless alternative treatment is chosen (higher doses or an alternative drug class).

To our knowledge, only one previous study has reported an evaluation of Vitek 2 with regard to echinocandin testing (28). In this study, using 867 Candida isolates and CLSI broth microdilution as a gold standard, an excellent EA of 99.5%, a CA of 99.8%, and a low VME rate of 0.2% (caspofungin) were reported using the original CLSI breakpoints of ≤2 mg/liter defining susceptibility. However, if interpreted by the revised CLSI breakpoints for Candida spp. (excluding C. glabrata, as no revised breakpoint was available at that time), much lower species-specific CAs of 72.9% for C. albicans, 84.8% for C. tropicalis and 30% for C. krusei were noted, primarily attributable to the CLSI reporting intermediate susceptibility for many isolates categorized as S by Vitek 2, as was the case in our study. In the same study, the VME rate was considerably lower than in our study. However, there are several plausible reasons for this. First, the genotypes of the isolates were not determined. Second, the collection appeared to include significantly fewer truly resistant (fks-containing) isolates, and hence, a lower VME rate is inevitably determined and should not be cautiously interpreted regarding if the method is suitable for the identification of such isolates. Finally, we had a comparable EA of 98.5% when evaluating only our wt isolates but a much lower EA of 78.6% when evaluating only our fks isolates (and even lower 50% among our C. glabrata fks mutants separately) (Table 3). It is thus important to include resistant as well as susceptible isolates, if the true performance of a susceptibility test is sought.

As no biological susceptibility test is perfect, it is relevant to evaluate the performance of the Vitek 2 system in comparison with that of both reference methodologies to obtain correct categorization of wt and mutant isolates. We therefore reevaluated our reference MICs for the EUCAST and CLSI methods by applying the breakpoints that have been established since our data set was first published (15, 17, 18). Clearly, EUCAST testing with anidulafungin as a marker for echinocandin resistance was superior to that of CLSI. MIC variations between different lot numbers of caspofungin have previously been described and could be a factor compromising CLSI performance (23). Notably, the CLSI MICs in this study were generated using a caspofungin lot which has since been associated with higher MICs (23). In this context, it was expected that the majority of misclassifications obtained with the CLSI method and caspofungin involved wt isolates misclassified as I or R, suggesting that the inferior performance was a caspofungin pure substance potency issue rather than a methodological issue.

In conclusion, Vitek 2 correctly identified wt isolates as S (ME rate, 0%) and, compared to CLSI method results, no wt isolates were misclassified as I. However, a high number of fks mutants were misclassified as S. The CLSI and EUCAST methods excellently identified fks mutants as nonsusceptible. For the CLSI method, a proportion of wt isolates were also categorized as nonsusceptible. For the wt populations of all species, any minor MIC change could not be detected using Vitek 2, as the lowest MIC reported is that of the susceptibility breakpoint for C. albicans, C. krusei, and C. tropicalis and differentiating between C. glabrata isolates as being S or I according to the new CLSI breakpoints is not possible. As data from a recent study have shown emerging echinocandin resistance of 11.1% in C. glabrata isolates with azole resistance (29), this deficiency is a valid concern.