MIC distributions for amphotericin B, fluconazole, itraconazole, voriconazole, flucytosine and anidulafungin and 35 uncommon pathogenic yeast species from the UK determined using the CLSI broth microdilution method

Abstract

Background

Epidemiological cut-off values and clinical interpretive breakpoints have been developed for a number of antifungal agents with the most common Candida species that account for the majority of infections due to pathogenic yeasts species. However, less-common species, for which susceptibility data are limited, are increasingly reported in high-risk patients and breakthrough infections.

Methods

The UK National Mycology Reference Laboratory performs routine antifungal susceptibility testing of clinical yeast isolates submitted from across the UK. Between 2002 and 2016, >32 000 isolates representing 94 different yeast species were referred to the laboratory. Here we present antifungal susceptibility profiles generated over this period for amphotericin B, fluconazole, voriconazole, itraconazole, anidulafungin and flucytosine against 35 species of uncommon yeast using CLSI methodologies. MIC data were interpreted against epidemiological cut-off values and clinical breakpoints developed with Candida albicans, in order to identify species with unusually skewed MIC distributions that potentially indicate resistance.

Results

Potential resistance to at least one antifungal agent (>10% of isolates with MICs greater than the epidemiological cut-off or clinical breakpoint) was evidenced for 29/35 species examined here. Four species exhibited elevated MICs with all of the triazole antifungal drugs against which they were tested, and 21 species exhibited antifungal resistance to agents from at least two different classes of antifungal agent.

Conclusions

This study highlights a number of yeast species with unusual MIC distributions and provides data to aid clinicians in deciding which antifungal regimens may be appropriate when confronted with infections with rarer yeasts.

Introduction

Early diagnosis and commencement of appropriate antifungal therapy are vital to improve outcomes in invasive Candida infections, which continue to be associated with high rates of morbidity and mortality particularly in immunocompromised hosts.^1–4 While most studies agree that Candida albicans remains the principal agent of nosocomial yeast infections, over recent years Candida glabrata, Candida parapsilosis, Candida tropicalis, Candida lusitaniae and Candida krusei have emerged as important pathogens.^1,2,4 However, due to the ever more complex surgical and medical interventions and in part due to the expanded use of antifungal agents with activity against the more common Candida species, more than 150 different yeast species distributed amongst Candida and a number of additional ascomycete and basidiomycete genera have to date been reported from mammalian infections.^5–11

Since it is well established that several of the common Candida species vary in their antifungal susceptibility profiles,^7,8,12 rapid and robust identification of the infecting species can be used to predict antifungal susceptibility in these organisms. Indeed, species-specific interpretive clinical breakpoints (CBPs) and epidemiological cut-off values (ECVs) have been proposed for a number of these common Candida species–antifungal agent combinations (reviewed in^13–17). For those species that are relatively rarely encountered in the clinical setting, the equivalent data are lacking since the establishment of wild-type distributions and ECVs requires the analysis of at least 100 independent isolates of a particular species, with MIC values obtained from several independent centres. Current accepted practice when confronted with an isolate for which ECV data and antifungal susceptibility profiles are lacking is to perform antifungal susceptibility testing of the isolate in question, and apply C. albicans breakpoints to interpret the resultant MIC results.^13–17 In such instances, where therapeutic options can be inferred at best from limited, anecdotal case reports, the analysis of MIC distributions from longitudinal studies in large reference centres can prove useful in identifying antifungal agent–organism combinations where MIC ranges are elevated compared with other species.^7,8,12

Current clinical practice guidelines recommend first-line treatment with an echinocandin (anidulafungin, caspofungin or micafungin) or a lipid formulation of amphotericin B, depending on the precise clinical presentation, for the management of most invasive or disseminated forms of candidiasis.³ Conversely, due to its better penetration, the triazole antifungal fluconazole, which is widely employed prophylactically in high-risk patients and neonates,^3,18 is the recommended treatment for Candida osteomyelitis and infections of the eyes and urinary tract.³ Moreover, it is proposed as an appropriate step-down therapy for candidaemia, disseminated (hepatosplenic) candidosis and CNS and intravascular infections in patients infected with isolates that are unlikely to be fluconazole resistant, or without prior fluconazole exposure.³ Additionally, for several of the most common Candida species, available data support the use of fluconazole based on the results of in vitro susceptibility testing.¹⁹ Here we have attempted to add to the limited antifungal susceptibility data that are available for less common pathogenic yeast species, and present susceptibility profiles for amphotericin B, anidulafungin, fluconazole, itraconazole, voriconazole and flucytosine with 35 such species. Profiles were generated from MIC values obtained during 15 years of in vitro susceptibility testing using the CLSI broth microdilution method²⁰ and yeast isolates from across the UK.

Materials and methods

Clinical isolates for antifungal susceptibility testing

From 2002 to 2016, more than 32 000 isolates of yeast were submitted to the UK National Mycology Reference Laboratory (MRL), Bristol, for determination of antifungal MICs. Isolates included examples of common and rarer Candida species, and a large number of other pathogenic yeasts from a wide variety of teleomorph genera. For the current study, all species less common than C. tropicalis for which more than six independent isolates had been analysed were included (Table 1). The identification methods that were employed as standard at the MRL have evolved over the study period as described below. Between 2002 and December 2007, isolates were identified using a combination of AUXACOLOR2/API 20 C AUX in conjunction with 26S rRNA gene sequencing.²¹ Between January 2008 and May 2012, isolates were identified by Pyrosequencing^© of a segment of the internal transcribed spacer region 2.⁵ Finally, from May 2012 to December 2016 all isolates were identified by MALDI-TOF MS.⁶ Numbers tested against each antifungal vary as the exact panel of antifungal agents tested against each isolate depended on the site of isolation, the clinical presentation and the antifungal agents requested by the referring laboratory. However, the majority of isolates were tested against at least one triazole antifungal (usually fluconazole ± voriconazole) and amphotericin B, with isolates from the urinary tract or CNS or those from cases of endocarditis also tested against flucytosine. Following its introduction into routine clinical usage, isolates were also frequently tested against anidulafungin, which was employed as the sentinel echinocandin for resistance detection.²² MICs were determined according to CLSI guidelines²⁰ by broth microdilution as described below.

Antifungal drugs

Antifungal drugs were obtained from their respective manufacturers as standard powders. Amphotericin B (Sigma Chemical Co., St Louis, MO, USA), anidulafungin and voriconazole (both Pfizer Central Research, Sandwich, UK) were dissolved in DMSO. Itraconazole (Janssen Research Foundation, Beerse, Belgium) was dissolved in PEG400 by heating to 70°C. Based on our laboratory experience, solubility of itraconazole is better in PEG than in DMSO as recommended by CLSI, and precipitation of the antifungal agent upon freezing is reduced. Fluconazole (Pfizer Central Research, Sandwich, UK) and flucytosine (Sigma Chemical Co.) were resuspended in sterile water. Serial 2-fold dilutions of the various drugs were prepared in RPMI 1640 medium (with l-glutamine, without bicarbonate; Sigma Chemical Co.), and buffered to pH 7.0 using a 0.165 M solution of MOPS (Sigma Chemical Co.). The antifungal drug concentration ranges tested were 0.03–16 mg/L (amphotericin B, itraconazole, voriconazole), 0.016–8 mg/L (anidulafungin) and 0.125–64 mg/L (fluconazole and flucytosine).

Antifungal susceptibility testing and determination of MICs

MICs were determined according to CLSI methodologies (CLSI M27-A4,²⁰) in round-bottomed 96-well microtitre plates with yeast suspensions prepared in saline and then diluted into RPMI 1640 and adjusted to final concentrations of 0.5–2.5 × 10³ cfu/mL. Inoculated plates were incubated for 24–48 h at 35°C. Although CLSI methodological changes over the period of the current study included a switch from 48 h reading to 24 h reading for all but the more slowly growing Candida species, our laboratory continued to read at 48 h (in addition to a 24 h read). The data presented here are from 48 h reads. This is essential for many of the rarer Candida species and non-fermentative yeasts of other genera (e.g. Pichia spp.), which are more slowly growing than C. albicans, and is the approach supported currently by CLSI for such organisms.²⁰ MICs were read as the concentration of drug that elicited 100% inhibition of growth (amphotericin B) or significant (∼50%) inhibition of growth compared with a drug-free control (anidulafungin, itraconazole, fluconazole, voriconazole and flucytosine). All assays included the control strains C. parapsilosis NCPF 8334 (ATCC 22019) and Pichia kudriavzevii (C. krusei) NCPF 3953 (ATCC 6258) (Table S1, available as Supplementary data at JAC Online). MIC ranges were determined for all Candida species that were less prevalent than C. tropicalis for which data were available for six or more independent isolates. Since the current study included isolates of C. orthopsilosis and C. metapsilosis, data for C. parapsilosis, the final member of this species complex, were also analysed for comparison, and to allow comparison with other international studies that have included this common Candida species.

Results

As we have previously described, the four most common pathogenic yeast species (C. albicans, C. glabrata, C. parapsilosis and C. tropicalis) typically account for ∼90% of all yeast isolates referred to the MRL for identification and antifungal susceptibility testing.^5,6,21 However, the remaining ∼10% of isolates encompass a diverse array of species belonging to numerous teleomorph genera. As the aim of the current study was to evaluate the antifungal susceptibility profiles of rarer yeast species, where isolate numbers from disseminated or deep infections are typically very low, we did not discriminate between organisms recovered from deep as opposed to superficial sites. However, in order to improve the reliability of the antifungal susceptibility profiles generated, and reduce the impact of organisms with outlying MICs, only species for which at least six independent isolates were available for analysis were included. Based on these criteria, 3619 of the ∼32 000 isolates received during the period 2002–16, representing 35 additional species of yeast and yeast-like fungi, were retained for analysis (Table 1).

The MIC distributions for amphotericin B, itraconazole, fluconazole, voriconazole, anidulafungin and flucytosine for these 35 species are shown in Tables 2–7, and results are summarized in Figure 1. In all tests, the MICs of the control reference strains were within the accepted limits (data not shown). Since CBPs and ECVs are not available for the majority of these less-common organism–antifungal agent combinations, we employed the C. albicans CBPs or ECVs as a means of identifying isolates/species with elevated MIC ranges. As this is a single-centre study, we also included data for >3000 C. parapsilosis isolates generated during the same period, to allow comparison with other European and US reports and with data for C. metapsilosis and C. orthopsilosis collated here (Tables 2–7). The MIC data generated here with C. parapsilosis were in excellent agreement with those previously reported from large international studies involving the USA and continental Europe,^{12,13,16,23,24} suggesting that UK isolates might be comparable to their overseas equivalents.

MIC data for amphotericin B were concordant with previously published data²⁴ that suggest resistance to this antifungal agent is uncommon in isolates of pathogenic yeast (Table 2). Species with appreciable proportions of isolates with MICs greater than the C. albicans ECV were restricted to several genera in Basidiomycota, including Rhodotorula, Naganishia (ex-Cryptococcus) and Cutaneotrichosporon, and the ascomycete genus Trichosporon (Table 2 and Figure 1). It should, however, be noted that the majority of these species were represented by relatively low isolate numbers in the current analysis. The same basidiomycete species exhibited skewed MIC distributions with anidulafungin, in keeping with the accepted lack of activity of the echinocandin antifungal agents against members of the Basidiomycota (Table 3 and Figure 1). MIC distributions for isolates of the C. parapsilosis (C. parapsilosis, C. metapsilosis, C. orthopsilosis) and Meyerozyma (Candida) guilliermondii species complexes were consistent with the higher CBPs for the echinocandin antifungals suggested with these organisms.²² The majority of other yeast species examined here exhibited MIC distributions for anidulafungin that fell below the C. albicans CBP (Table 3). However, notable exceptions included all 18 isolates of Trichosporon asahii, 17/18 isolates of Magnusiomyces capitatus and 4/6 isolates of Dipodascus geotrichum, all of which are ascomycetous yeasts.

It is widely accepted that isolates of P. kudriavzevii (C. krusei) are innately resistant to fluconazole. However, using C. albicans interpretive breakpoints, a further 23 species in the current study had skewed fluconazole MIC distributions (>10% of isolates with fluconazole MICs of ≥8 mg/L; Table 4), with >50% of isolates for 17 of those species exhibiting MICs suggestive of resistance to this antifungal agent (Figure 1). This was the case for the additional five species of Pichia examined here (P. cactophila, P. fermentans, P. kluyveri, P. mandshurica and P. norvegensis), all three species of Rhodotorula (R. glutinis, R. dairenensis and R. mucilaginosa), Naganishia albida, Naganishia diffluens and Cutaneotrichosporon curvatum (all previously classified in Cryptococcus), several species that were classified or are currently retained in Candida (C. blankii, C. zeylanoides, Diutina catenulata, Wickerhamiella pararugosa) and Magnusiomyces capitatus. With itraconazole, most isolates of many of the species analysed here had MICs below the ECVs proposed for common Candida species²⁴ (Table 5). Organisms with heavily skewed MIC distributions included Rhodotorula spp., Saccharomyces cerevisiae, C. blankii, C. zeylanoides and Yarrowia (Candida) lipolytica. However, since itraconazole is principally employed for the treatment of superficial, mucosal infections, collection of robust MIC data for many of the organisms with this antifungal agent is hindered by low isolate numbers submitted to the MRL for testing. Moreover, since such mucosal infections can often be recurrent or recalcitrant, it is also possible that the data presented here for itraconazole are biased towards resistance due to the likelihood that many of the patients will have had prior or repeated exposure to this agent. For voriconazole, skewed MIC distributions suggestive of frank resistance were only seen with Rhodotorula spp. and N. albida, although 10%–37.5% of isolates of C. blankii, C. zeylanoides, C. curvatum, D. catenulata, M. capitatus, N. diffluens and Pichia cactophila had MICs above the 0.5 mg/L CBP proposed for C. albicans (Table 6). Thus, when analysing the data for the triazole antifungal class as a whole, C. blankii, C. zeylanoides, C. curvatum, D. catenulata, M. capitatus, Naganishia spp., P. cactophila, Rhodotorula spp. and S. cerevisiae exhibited significant potential resistance to at least two of the three antifungal agents tested, and Rhodotorula spp. had uniformly elevated MICs of fluconazole, itraconazole and voriconazole (Figure 1). Finally, MIC distributions observed with flucytosine are shown in Table 7. Organisms that had systematically elevated MICs when compared with the 48 h C. albicans ECV of 1 mg/L included Cutaneotrichosporon mucoides, D. geotrichum, Naganishia spp., Pichia spp., T. asahii and Y. lipolytica.

Discussion

Here we have presented antifungal susceptibility data generated at the MRL over a 15 year period for 35 species of less-common yeast and yeast-like fungi isolated from clinical samples in the UK. For such organisms, the development of CBPs and ECVs is hindered by insufficient isolate numbers for robust MIC distribution evaluation, and limited clinical trial data. The present study, whilst not geared towards developing ECVs, is intended to contribute to existing literature and potentially aid their future development. A limitation of the current study is that isolate numbers with certain antifungal combinations for a proportion of the yeast species considered here were too low (<10 isolates per species) to reliably predict the likely susceptibility profile of the species as a whole. However, in the absence of clinical data suggesting therapeutic benefits, or susceptibility testing of the individual isolate in question, these observations may guide clinicians confronted by extremely rare yeast species in deciding whether an antifungal agent may be appropriate.

The full antifungal susceptibility profiles for 33 species are summarized in Figure 1. The two less common members of the C. parapsilosis species complex (C. orthopsilosis and C. metapsilosis) were omitted since the susceptibility profile for C. parapsilosis is well documented, and the other two members of the complex were equally susceptible to all antifungal agents examined (Tables 2–7). Figure 1 (centre left panel) summarizes the susceptibility profiles of the various species based on the proportions of isolates for which MICs fall above the breakpoint for resistance. A number of the species summarized in Figure 1 show potentially alarming susceptibility profiles with MICs suggestive of resistance to several different agents, and even agents from several antifungal classes. Elevated MICs of all three triazole agents were observed for C. blankii, C. zeylanoides, N. albida and Rhodotorula spp. With Rhodotorula spp. in vitro resistance extended to include anidulafungin, with a proportion of isolates of R. glutinis also with elevated MICs of amphotericin B and flucytosine. In a similar manner, the two Naganishia species included in the current study demonstrated in vitro resistance to at least two of the triazole antifungals, plus flucytosine and anidulafungin. More than 20% of N. diffluens isolates also exhibited elevated MICs of amphotericin B. Combined fluconazole/anidulafungin resistance was seen with isolates of M. capitatus, D. geotrichum and to a lesser extent Y. lipolytica. For D. geotrichum and Y. lipolytica, resistance also extended to include flucytosine and anidulafungin in a proportion of isolates. Finally, all of the Pichia species examined demonstrated in vitro resistance to fluconazole and flucytosine, and T. asahii appeared resistant to flucytosine in addition to anidulafungin. Since breakpoints for fluconazole, itraconazole and voriconazole include a ‘susceptible-dose dependent’ category, we also analysed the MIC data in Tables 4–6 on the basis of the proportions of isolates with MICs that fall above the cut-off for susceptibility with C. albicans. The results of this more cautious approach, which might better capture species unlikely to respond to conventional antifungal drug dosing regimens, are summarized in the right-hand panel of Figure 1. Using these interpretive criteria, the vast majority of species exhibit skewed MIC distributions of both fluconazole (28/33 species) and itraconazole (29/33 species), whereas 10/33 species had MIC distributions of voriconazole that are still suggestive of susceptibility to this antifungal agent (Figure 1).

The identification of antifungal agent–organism combinations where MIC ranges are always elevated compared with other species may aid in eliminating particular therapeutic approaches. Although the current study employed organisms isolated exclusively in the UK, a number of different international reports have alluded to similar MIC distributions in non-UK isolates,^7,8 and such susceptibility profiles are reflected in recent published guidelines for the diagnosis and management of rare invasive yeast infections (²⁵ and references therein). Support for the elevated fluconazole MICs described for many of the organisms analysed here and their potential impact on therapeutic success have been discussed previously.²⁶ For flucytosine, several previous reports concur with our data indicating skewed MICs with P. kudriavzevii (C. krusei),^25,27,28,M. guilliermondii,^28,T. asahii²⁹ and Naganishia spp. (^25,30,31 and references therein). Similarly, in vitro cross-resistance of Rhodotorula spp. to the triazole antifungals,^32,33 echinocandin resistance in Trichosporon,^34,Magnusiomyces³⁵ and Dipodascus spp.,³⁶ and elevated amphotericin B MICs with Trichosporon spp.^37,38 have been reported anecdotally elsewhere, or in large-scale studies.³³ For the antifungal agents included in the current study, skewed MIC distributions for a particular organism do not necessarily indicate resistance per se, and may instead be reflective of the need for higher species-specific CBPs for that organism–antifungal agent combination. However, limited clinical data suggest that several of the organism–antifungal agent combinations highlighted as unusual in Figure 1 may indicate genuine resistance rather than the need for alternative CBPs. For example, breakthrough infections due to T. asahii,^38–40,M. capitatus^35,41,42 and D. geotrichum⁴² have all been reported in patients receiving echinocandin antifungal therapy. Similarly, infections with T. asahii that proved refractory to amphotericin B have been recognized.⁴³ More clinical data will be required to establish whether the other unusual in vitro antifungal susceptibility profiles described above correlate with therapeutic failure in vivo.

Finally, here we have chosen to recognize the polyphyletic nature of many yeast genera by adopting teleomorph names, where known, for yeasts with Candida anamorphs, and also by applying the most up-to-date taxonomic revisions and nomenclatural updates for other genera.⁴⁴ Whilst this approach may initially appear confusing, we believe it is clinically justified in part based on some of the data presented here. Effectively, the six Pichia species included here [P. cactophila (C. inconspicua), P. fermentans (C. lambica), P. kluyveri, P. kudriavzevii (C. krusei), P. mandschurica and P. norvegensis (C. norvegensis)] share similar antifungal susceptibility profiles typified by uniformly elevated MICs of both fluconazole and flucytosine, but little in vitro resistance to any other antifungal agent (Figure 1) - profiles that are unique amongst the 35 species examined in the current study. Similarly, the flucytosine MICs observed with isolates of N. albida and N. diffluens (both ex-Cryptococcus) appear more elevated than those seen with those Cryptococcus species that have been either retained in the genus (C. neoformans) or subsequently accommodated in Cutaneotrichosporon (C. curvatum) (Table 7 and Figure 1). Thus, it appears possible that a revised taxonomy that reflects phylogenetic approaches may correlate better with the unusual antifungal susceptibility profiles observed with many of the less common pathogenic yeast species. Further studies will be required to fully investigate this possibility.

In summary, the current study has underscored the epidemiological complexity of yeasts with pathogenic potential, and the large number of uncommon species from diverse taxonomic groups that can be associated with clinical samples. Our MIC data, which reveal many apparently species-specific and genus-specific trends in antifungal susceptibility, also highlight the paramount importance of accurate identification of yeasts in the clinical setting.

Limitations of the current study

All isolates tested here were from human hosts. However, we have not attempted to differentiate isolates based on site of isolation (sterile versus superficial site), primarily because isolation site was often not available, and also because isolate numbers from disseminated or deep infections are typically very low for the rarer yeast species. However, using this combined approach, total isolate numbers more often exceeded the 15 recommended by EUCAST as being the minimum number required per centre for defining wild-type distributions. Isolates from superficial sites are more likely to have had prior antifungal exposure. In addition, the paucity of information regarding prior exposure to antifungal therapy that might have resulted in acquired antifungal resistance for isolates from systemic, usually sterile sites is likely to skew MIC distributions towards non-susceptibility. Moreover, given the number of isolates included in the current study and the study duration, prior antifungal history and clinical outcome data were not available for the vast majority of patients from whom isolates were recovered. Methods for identification of yeast isolates have improved considerably over the time period of the study, resulting in the description of new genera and species and identification of several species that were difficult to identify by means of previous methodologies. Here, we have tried to limit the potential impact of erroneous identifications on MIC distributions by excluding data for difficult species when the original isolates are no longer available for confirmation of identification by current technologies. All the data presented here were obtained using CLSI broth microdilution methodologies. However, previous studies have demonstrated that MICs obtained by the CLSI and EUCAST methods often show a reasonable correlation for yeast isolates and many of the antifungal agents tested in this study. Even though absolute MIC values generated using EUCAST methodologies tend to be slightly lower, it is likely that both methods would similarly identify species with non-wild-type distributions. Thus, we believe that the trends reported here should be broadly applicable in laboratories that employ EUCAST methodologies for antifungal susceptibility testing. Finally, in the current study we have chosen to employ C. albicans-specific ECVs/CBPs to identify those rare yeast species with MIC distributions that are potentially skewed towards resistance. Although such cut-offs are species specific, and cannot be arbitrarily applied to other species, we believe that this approach is reasonable here. Since wild-type populations of C. albicans are universally susceptible to the antifungal agents tested in the current work according to both CLSI and EUCAST criteria, assessing unusual MIC distributions based on C. albicans ECVs/CBPs is the most cautious/stringent approach available.

Acknowledgements

We thank the other members of the UK National Mycology Reference Laboratory past and present for their assistance in all aspects of processing of pathogenic yeast isolates, and in particular Martin Gough, who tirelessly prepared the drug trays for antifungal susceptibility testing, and Phillipa Brown. We are also grateful to the numerous microbiologists and clinicians across the UK who refer their unusual isolates to us for identification and susceptibility testing.

Funding

This study was carried out as part of our routine work.

Transparency declarations

None to declare.

References