Candida guilliermondii Fungemia in Patients with Hematologic Malignancies

ABSTRACT The microbiological, clinical, and epidemiological features of most non-Candida albicans Candida species are well known, but much less is known about species such as Candida guilliermondii, an uncommon pathogen causing a variety of deep-seated infections in immunocompromised hosts. To characterize C. guilliermondii fungemia in patients with hematological malignancies and its susceptibility to antifungal drugs, all cases of C. guilliermondii fungemia diagnosed in our department between 1983 and 2005 were retrospectively analyzed and the literature was reviewed. C. guilliermondii caused 29/243 (11.7%) candidemia episodes diagnosed during the study period. Central venous catheters were the documented sources of candidemia in 19/29 episodes (65.5%), and invasive tissue infections were documented in 2 (6.9%). In the remaining eight, the catheter was not removed and the source of the fungemia remained obscure. Seven episodes ended in death, but only one could be attributed to invasive C. guilliermondii infection. Molecular typing data reveal no evidence of common infection sources. Isolates displayed high rates of in vitro susceptibility to amphotericin B (100%), voriconazole (95%), and fluconazole (90%) and lower rates of in vitro susceptibility to flucytosine (86%), itraconazole (76%), and caspofungin (33%). Our literature review confirms that C. guilliermondii is a significantly more frequent cause of candidemia among cancer patients compared with the general hospital population. It accounted for <1% of the total number of Candida bloodstream isolates reported in the articles we reviewed, with higher rates in Europe (1.4%) and Asia (1.8%) compared with North America (0.3%).

Antifungal susceptibility tests. Twenty-one isolates of C. guilliermondii were available for in vitro antifungal susceptibility testing. Prior to testing, each isolate was passaged at least twice on Sabouraud dextrose agar.
Quality control strains C. albicans ATCC 90028 and C. parapsilosis ATCC 22019 were included in every test run.
Susceptibility to voriconazole (Pfizer, Inc., New York, NY), fluconazole (Pfizer), itraconazole (Janssen, Beerse, Belgium), amphotericin B, and flucytosine (both from Sigma, St. Louis, MO) was tested by the broth microdilution method in accordance with the M27-A2 protocol published by the Clinical and Laboratory Standards Institute (CLSI [formerly NCCLS]) (31). Results were read after 48 h of incubation at 35°C. The lowest drug concentration producing complete growth inhibition (for amphotericin B) or inhibition of 50% or more compared with control growth (for fluconazole, itraconazole, voriconazole, and flucytosine) was recorded as the MIC.
Caspofungin could not be tested by the broth microdilution method because a standard powder preparation of the drug could not be obtained from the manufacturer. Susceptibility to this drug was thus assessed by the E-test method (AB Biodisk, Solna, Sweden) and RPMI 1640-2% glucose agar (Difco Laboratories) in accordance with the manufacturer's instructions. This approach has displayed 100% concordance with the CLSI microdilution method for evaluation of caspofungin susceptibility in C. guilliermondii isolates (39).
Genotypic characterization. Genotyping was performed on 19 isolates of C. guilliermondii. Isolates from 48-to 72-h cultures were suspended in 20 ml yeast peptone glucose (1% peptone yeast extract, 2% glucose, 2% Bacto Peptone). After 24 h of incubation with agitation at 35°C, yeasts were harvested by centrifugation and genomic DNAs were extracted as described by Scherer and Stevens (58). DNA typing was performed by random amplification of polymorphic DNA (RAPD) with primers RSD12 (5ЈCCGCAGCCA3Ј) (57)  Review of the literature. We conducted a MEDLINE-based search of the English language literature published since 1966 to identify articles containing the term "candidemia," "fungemia," and/or "Candida/fungal bloodstream" in the title and abstract. All articles describing studies including at least 150 cases of candidemia were reviewed to estimate the relative weight of C. guilliermondii fungemia.

RESULTS
Prevalence of C. guilliermondii fungemia. During the 22-year period considered in this study, 247 Candida bloodstream isolates were recovered from patients treated by our staff for hematological malignancies (Fig. 1). They were responsible for a total of 243 episodes of candidemia, 29 (11.7%) of which were caused by C. guilliermondii. No case was observed in patients with nononcologic hematological diseases. Since 1988, when the first case occurred, the incidence of C. guilliermondii fungemia has been 2 per 1,000 admissions. Corresponding values for C. parapsilosis, C. albicans, and C. tropicalis fungemia are 4.5, 3.9, and 2.4 per 1,000 admissions, respectively.
Patients and predisposing factors. Tables 1 and 2 summarize the patient characteristics, clinical features, and outcomes of the 29 episodes of C. guilliermondii fungemia. As shown in Table 1, more than half the cases occurred in patients with acute nonlymphoid leukemia and almost all were diagnosed during periods of hospitalization (1 to 60 days after admission; mean, 23 days). However, four (cases 4, 8,12, and 25 in Table  2) occurred while the patient was at home (12,25,40, and 65 days after the most recent hospital discharge, respectively) and were treated on an outpatient basis. When the fungemia was diagnosed, well over half of the patients had neutropenia, which had been present for 5 to 60 days (mean, 20.8 days). Most patients were receiving antibiotic therapy, and antifungal prophylaxis was being administered in 13 cases; in 7/29 (24.1%) cases, the patient was receiving fluconazole (mean dosage, 400 mg/day) and 6 (20.7%) were taking oral nonabsorbable amphotericin B (mean dosage, 2,000 mg/day). Colonization with C. guilliermondii was documented in only one episode (3.5%) (stool culture positivity in case 2).
Clinical characteristics and outcome of C. guilliermondii fungemia. The mean duration of candidemia (from the first positive blood culture to negative blood culture or death) was 11 days (range, 1 to 82 days), and a mean of seven positive blood cultures were collected per episode (range, 1 to 25). As shown in Table 2, all 29 episodes were associated with fever. Two (6.9%) were considered secondary to invasive tissue infections, i.e., case 17, which was associated with skin lesions (culture positive for C. guilliermondii) and cellulitis at the CVC insertion site, and case 2, in which there was multiorgan failure due to disseminated C. guilliermondii candidiasis. In the other 27 (93.1%), there was no clinical or microbiological evidence of invasive C. guilliermondii tissue infection. In 19 of these episodes, the CVC was removed, and within 24 h both the fever and candidemia had cleared. Semiquantitative cultures of the catheter tips were all positive for C. guilliermondii. These 19 cases (65.5% of the total series) were considered CVC related.
In the remaining eight cases (type unknown), the CVC could not be removed and no other source of infection could be identified. In four of these, blood cultures became negative (after 1, 2, 5, and 7 days of antifungal therapy); in the remaining four, the candidemia persisted until death. Overall, fatal outcomes were recorded for 7 (24.1%) of the 29 episodes. In five, fungemia was still present at the time of death, but only one of these deaths was clearly attributed to C. guilliermondii infection (case 2) ( Table 2).
In vitro antifungal susceptibility. Table 3 summarizes the antifungal susceptibilities of the 21 C. guilliermondii isolates tested. All were fully susceptible to amphotericin B (MIC, Ͻ1 g/ml). Eighteen (86%) isolates were susceptible to flucytosine. All but two (91%) were susceptible to fluconazole. Sixteen (76%) isolates were susceptible to itraconazole. Susceptibility to voriconazole was documented in all but one strain (95%). For 14 (66%) of the isolates, the caspofungin MICs were indicative of probable resistance (Ͼ1 g/ml).
Genotypic characterization. Nineteen isolates were genotyped by RAPD with the RSD12 and OPE03 primers (Fig. 2). On the basis of the combined results obtained with the two primers, four DNA types were identified; types I and III included six strains each, four isolates were type II, and three were type IV ( Table 2).

DISCUSSION
C. guilliermondii is part of the normal flora of human skin and mucosal surfaces, but it is occasionally implicated as a cause of chronic onychomycosis, acute osteomyelitis, septic arthritis, endocarditis, fungemia, and disseminated invasive infections (56). It is one of the opportunistic fungi recovered most frequently from severely immunocompromised patients. Our literature review confirmed that C. guilliermondii is a more common cause of candidemia in cancer patients than it is in general hospital populations, but it is rarely implicated in bloodstream infections occurring in other high-risk categories, such as intensive care unit patients (62). Even among cancer patients, the actual incidence appears to be quite low. A review of 37 reports published between 1952 and 1992 revealed that C. guilliermondii was responsible for only 0.8% of all systemic Candida infections in this risk group (67). The largest reported series includes nine cases (two-thirds occurring in leukemia patients) observed over 11 years (1988 to 1998) at the M. D. Anderson Cancer Center (28).
In comparison, the rates observed in our institute appear fairly high. The first case was observed in 1988, and since then, 28 other episodes have been diagnosed. C. guilliermondii accounted for 11.7% of the Candida bloodstream isolates recovered from our patients during the 22-year study period. Its frequency was inferior only to those of C. parapsilosis, C. albicans, and C. tropicalis. It is important to note, however, that the incidence of C. guilliermondii fungemia in our institute is by no means uniform. Approximately 80% of the cases were observed between 1992 and 2001, and only one has occurred since then.
It is difficult to pinpoint specific reasons for the relatively high frequency of C. guilliermondii candidemia in our institution. Cases of C. guilliermondii fungemia were documented in patients with different hematological malignancies who underwent various chemotherapy treatments and only in a minority of cases received systemic antifungal prophylaxis. Molecular analyses of 19 isolates recovered from our patients from 1992 to 2001 do not support the possibility of a common source of infection. In fact, there were no case clusters during any of the periods considered, and even temporally related candidemia episodes were usually due to genetically different strains. Regional variations have been documented, and European rates are significantly higher than those in North America. We have no interpretation regarding this apparently nonhomogeneous geographic distribution of C. guilliermondii bloodstream infections. Clinically and epidemiologically, C. guilliermondii fungemia seems to resemble that caused by C. parapsilosis (11,66). All of our cases occurred in patients with indwelling CVCs, at least 19 of the 29 episodes (65.5%) could be classified as catheter related, and there were only two cases of deep invasive infections. In addition, both types of fungemia increased in frequency between 1991 and 1997 and decreased thereafter. Their close association with central venous access suggests that this trend might be attributed to changes in the use of CVCs in our institute. The number of CVC insertions rose progressively during the first 10 years of the study, and this increase could explain the increasing number of CVC-related candidemias. In the last 10 years, however, CVC placement rates have re-mained fairly stable and the same type of central line was used. The decreasing frequencies of C. guilliermondii and C. parapsilosis infection during this period might thus reflect improved management of these catheters. Furthermore, it should be noted that the same blood culture method in the same laboratory was used over the years of study.
Catheter removal had a major impact on the outcome of treatment. It was almost always followed by defervescence and candidemia clearance within 24 h. The only exception was a patient with disseminated infection (case 17) whose fungemia persisted for 5 days after CVC removal. It should be noted that when the CVC was removed, this patient had been receiving antifungal therapy for 2 weeks with no sign of resolution of the candidemia. Indeed, most of the other patients had been treated unsuccessfully for more than a week before the CVC was removed. In case 3, the CVC was removed after 82 days of candidemia with no evidence of secondary deep-seated foci. In four of the nine episodes in which the CVC could not be removed, the duration of candidemia was still brief, but in the other five, the fungemia persisted and was still present when death occurred. Two patients experienced a second episode of C. guilliermondii fungemia 2 months (cases 20 and 20a in Table  2) and 3 months (cases 26 and 26a) after the first. Molecular typing was done only for the two isolates recovered from the latter two cases, which were confirmed to be genetically different. Overall, fatal outcomes occurred in 7 (24%) of the 29 episodes, but only one of these deaths could be attributed to C. guilliermondii infection.
It is of interest that 14% of our patients became fungemic at home. C. guilliermondii can be part of the normal flora of the skin and mucosal surfaces, and handling of the CVC by the patient or relatives can thus be a risk for contamination.
Although there are reports of single cases of C. guilliermon- Representative DNA types (I to IV) of four C. guilliermondii isolates obtained by RAPD with primers RSD12 and OPE03. All amplifications were repeated twice, and the most intense bands in the patterns were reproducible. For comparative analyses, only high-intensity bands (lanes II and III for RSD12 and lanes I and II for OPE03) were considered. The combined results obtained with the two individual primers revealed four different DNA types (I to IV).   (8,19,49,60,63), there is no evidence of widespread polyene and azole resistance in this species. Pfaller et al. (43,45) have recently assessed the in vitro susceptibilities of rare Candida bloodstream isolates recovered in various parts of the world, including a total of 150 isolates of C. guilliermondii. The great majority (85 to 100%) were fully susceptible to amphotericin B, flucytosine, fluconazole, voriconazole, and ravuconazole, but susceptibility to itraconazole was much less common (10%). C. guilliermondii seems to be intrinsically resistant to the echinocandins. High caspofungin MICs (Ͼ1 g/ml) have been reported for more than 95% of the isolates tested (39,44), suggesting that this drug is unlikely to be effective against C. guilliermondii infections. A recent study showed that C. guilliermondii is also among the Candida species that are the least susceptible to the echinocandin anidulafungin (47). Our data confirm high rates of susceptibility to amphotericin B (100%), voriconazole (95%), fluconazole (90%), and flucytosine (86%), but our isolates displayed lower rates of resistance to itraconazole (24%) and caspofungin (66%) than those observed in previous studies.
In conclusion, C. guilliermondii is a potential cause of fungal bloodstream infections, particularly in patients with hematologic malignancies. The overall incidence of these infections seems to be very low, even in cancer patients, but their distribution is by no means homogeneous and higher frequencies may be observed in certain hematological centers. The incidence of C. guilliermondii bloodstream infections in Asia and Europe is slightly higher than it is in North America. These infections are clinically similar to those caused by C. parapsilosis, and bloodstream invasion by both fungal species is closely related to CVC placement. C. guilliermondii is intrinsically resistant to the echinocandins, but despite isolated reports to the contrary, it is also highly susceptible to amphotericin B and all of the azoles, with the probable exception of itraconazole.