From January 1987 through December 1998, Penicillium marneffei infection (23 patients) or colonization (1 patient) was diagnosed in a total of 24 patients in Taiwan. Of these 24 patients, 16 (67%) had AIDS and 20 (83%) had disseminated P. marneffei infection. The majority (63%) of the infections were considered indigenous. The number of cases has increased markedly in recent years, with 17 of the 24 cases diagnosed from 1996 through 1998. Twenty preserved isolates of P. marneffei, recovered from 11 patients treated at National Taiwan University Hospital during the period of January 1996 through December 1998, were studied to determine the epidemiology of P. marneffei infections. Among the 20 isolates, a total of 8 strains (highly related isolates) were identified on the basis of tests for susceptibility to 5 antifungal agents, for chromosomal DNA restriction fragment—length polymorphism types, and for randomly amplified polymorphic DNA patterns. One of the strains (6 isolates) was isolated from 4 patients treated in 1997 and 1998. Strain spreading of P. marneffei may partially contribute to the increased number of infections caused by this organism in immunosuppressed patients in Taiwan.
Penicillium marneffei has become a well-recognized human pathogen since the first report of natural human infection in 1973 . Infection caused by this organism is frequently disseminated and is progressive in nature. The majority of reported cases involve patients with impaired cell-mediated immunity due to Hodgkin's disease, immunosuppressive therapy, or human immunodeficiency virus (HIV) infection, although infections in apparently immunocompetent hosts have been documented [2–9].
A history of recent travel in Southeast Asia, the southern part of China, or Indonesia is common among patients with P. marneffei infection [2, 4, 5, 8, 9]. The organism has been isolated from many species of bamboo rat in these countries and from soil samples from the animals' burrows; however, it remains unclear whether the bamboo rat is the only reservoir for transmission of the infection to humans in areas of endemicity [3–5, 9, 10]. Although indigenous cases of P. marneffei infection have been reported in several Southeast Asian countries other than Thailand, the environmental reservoir of the organism in these countries remains obscure [2, 5, 11].
In Taiwan, the first disseminated infection caused by P. marneffei was reported in 1994 in an ethnic Chinese man with HIV infection who had come to Taiwan from Thailand . A total of 13 cases of invasive P. marneffei infection had been reported in the literature by July 1999, and 7 of them were considered to have been acquired in Taiwan [11–16]. To understand the epidemiology of P. marneffei infection in Taiwan, we studied an additional 11 patients who were infected with or colonized by the organism, and we also studied 20 preserved P. marneffei isolates recovered from 11 of the 24 reported patients. We determined the isolates' antibiotypes, by testing their susceptibility to 5 antifungal agents. We also determined profiles of cellular fatty acid methyl esters (FAMEs), chromosomal DNA restriction fragment—length polymorphism (RFLP) types, and randomly amplified polymorphic DNA (RAPD) patterns.
Materials and Methods
Twenty preserved isolates of P. marneffei were recovered from various clinical specimens from 11 patients (patients 11, 12, 14, 16–18, and 20–24) who were treated at National Taiwan University Hospital (NTUH), a 2000-bed teaching hospital in northern Taiwan, from January 1996 through December 1998 (table 1). Clinical specimens for isolation of P. marneffei were inoculated onto Sabouraud dextrose agar (Difco, Detroit) and incubated at 25°C in ambient air. Colonies with the characteristic mycelial form and the presence of a soluble red pigment were subcultured onto brain-heart infusion agar slants (Difco) and were incubated at 37°C to yield the yeast phase. P. marneffei isolates were characterized by use of conventional methods, as described elsewhere .
Cellular fatty acid composition
For cellular fatty acid analysis, each isolate was harvested after 5 days of incubation at 37°C on Sabouraud dextrose agar plates. Lysisof cells, saponification, methylation of cellular fatty acids, and extraction of FAMEs were performed according to the manufacturers' instructions, as well as by methods described elsewhere [17–19]. FAMEs were separated by use of gas chromatography and were analyzed with the microbial identification system (Microbial ID, Newark, DE). The analysis was performed twice, with different subcultures, for each isolate.
Antifungal susceptibility testing
The MICs of amphotericin B, fluconazole, ketoconazole, itraconazole, and 5-fluorocytosine for the 20 isolates of P. marneffei isolates were determined by use of the E-test (PDM Epsilometer, AB Biodisk, Solna, Sweden). The concentrations of amphotericin B, 5-fluorocytosine, itraconazole, and ketoconazole had a range of 0.002–32 µg/mL, and the concentrations of fluconazole had a range of 0.016–256 µg/mL. For these assays, we used RPMI 1640–2% glucose agar medium supplemented with morpholinepropanesulfonic acid (MOPS; Difco). The MIC of each antifungal agent was determined according to the manufacturers' instructions for susceptibility testing of yeasts. In brief, after 5 days of growth on Sabouraud dextrose agar at 37°C in ambient air, individual colonies were suspended in normal saline. The turbidity was adjusted to match that of a 1 McFarland standard, and ∼400 µL of the suspension was directly inoculated onto a 150-mm RPMI 1640–2% glucose—MOPS agar plate. After 5 days of incubation at 37°C, the MICs were determined on the basis of the inhibition ellipse intersects of the MIC scale. Candida parapsilosis ATCC 22019 and Paecilomyces variotii ATCC 22319 were used as control strains. All MIC tests were performed in duplicate. When the MIC reading was between the traditional log2 concentrations, the result was rounded up to the next value. The antibiotypes of the isolates were considered identical if the MICs of all agents tested were identical or were within a 2-fold dilution discrepancy. Isolates having identical antibiotypes were considered as possibly related.
RAPD patterns of the 20 isolates were generated by arbitrarily primed polymerase chain reaction (APPCR). Strains were cultured on yeast extract—peptone-dextrose (Difco) agar for 5 days at 37°C. Two 10-µL loops of culture material were suspended in 300 µL of lyticase solution (Sigma, St. Louis). Chromosomal DNA of the isolates was extracted with a commercial kit (Puregene D-6000, Gentra Systems, Minneapolis, MN). Six arbitrary oligonucleotide primers selected from 3 commercial kits, A (OPA-01 to OPA-20), B (OPB-01 to OPB-20), and H (OPH-01 to OPH-20), were used: OPA-03 (5′-AGTCAGCCAC-3′), OPB-18 (5′-CCACAGACGT-3′), OPH-03 (5′-AGACGTCCAC-3′), OPH-05 (5′-AGTCGTCCCC-3′), OPH-18 (5′-GAATCGGCCA-3′), and OPH-20 (5′-GGGAGACATC-3′) (OPERON Technologies, Alameda, CA). Amplification was performed in a PTC-100 thermocycler (MJ Research, Watertown, MA) and consisted of the following steps: predenaturation at 94°C for 4 min, 40 cycles of 30 s at 94°C, 1 min at 34°C, and 2 min at 72°C, with a final extension for 5 min at 72°C. The amplification products were separated by electrophoresis in 1.5% agarose gels. The gels were photographed and interpreted by visual examination. A 1-kb ladder (Bethesda Research Laboratories, Gaithersburg, MD) was used in each gel as a DNA fragment—size marker. The entire procedure, from fungal growth to RAPD pattern recording, was repeated at least 3 times for each isolate to confirm the results. For interpreting the results of RAPD analysis, patterns having all the bands with the same mobility, with regard to intensity, were considered identical.
Macrorestriction endonuclease analysis of chromosomal DNA
The RFLP types of the 20 isolates of P. marneffei were determined by digestion of whole-cell DNA with restriction enzymes HaeIII, EcoRI, CfoI, HindIII, SmaI, and XbaI (Bethesda Research Laboratories), as described by Vanittanakom et al. .
Isolates were defined as being of the same strain (highly related isolates) if they had identical antibiotypes, RFLP types, and RAPD patterns.
Clinical characteristics of the patients
Table 1 shows clinical characteristics of the 24 patients who received a diagnosis of P. marneffei infection from January 1987 to December 1998. The annual number of cases increased markedly from 1996 through 1998 (figure 1), during which time the majority (64%) of the infections were considered to have been acquired in Taiwan. Of the 24 patients, 19 (79%) were male. Their ages ranged from 3 months to 58 years (mean, 33 years). Nine of the infections were acquired outside Taiwan (7 in Thailand and 1 each in Hong Kong and Canton), and the others were considered indigenous. All but 2 of the patients were immunocompromised: 16 (67%) had AIDS, and 3 (13%) received corticosteroids or other immunosuppressive agents. Twenty (83%) patients had disseminated infection, 3 (13%) had localized diseases (2 with cavitary pneumonia and 1 with septic arthritis and osteomyelitis), and 1 (4%) had respiratory tract colonization. Eighteen patients infected with P. marneffei received amphotericin B (1 mg/kg/day) as the initial treatment, and 2 received intravenous fluconazole therapy (400 mg/day). Six patients died of P. marneffei infection; all had disseminated disease, 3 had not received antifungal therapy, and 1 (patient 23, who received amphotericin B) had isolates with a low amphotericin B MIC (0.032 µg/mL). Of the 16 patients who survived after primary treatment with amphotericin B, 14 received maintenance therapy with oral itraconazole. No relapse of infection was found among these patients.
Identification of the isolates
After 3 days of incubation on Sabouraud dextrose agar at 25°C, the P. marneffei isolates grew as mold, with a soluble red pigment diffusing into the agar. The reverse side was red or pink. When seen through a microscope, the isolates had typical penicillial features. When incubated on Sabouraud dextrose agar and brain-heart infusion agar at 37°C for 3 days, the colonies were convoluted and smooth, without red pigment. P. marneffei yeast cells appeared tubular or oblong, and some had a cross-septum.
Cellular fatty acid profiles
All the 20 isolates studied had similar FAME profiles (figure 2), and all had the 5 major FAMEs (>3% of total FAMEs): 16:0 (hexadecanoic acid, 19%–30%); 18 : 2 (cis-9, 12, octadecanoic acid, 17%–45%); 18 : 1 ω9c (cis-9-octadecanoic acid, 23%–36%); 18 : 1 ω9t (cis-11-octadecenoic acid, 5%–16%); and 18:0 (octadecanoic acid, 4%–15%).
The MICs of the 2 control strains in response to the 5 antifungal agents fell within the ranges provided by the manufacturer. Ketoconazole and itraconazole were the most highly active of the azole antifungal agents tested against the 20 P. marneffei isolates, with very narrow MIC ranges (table 2). Fluconazole was less active, with a MIC at which 90% of the isolates were inhibited (MIC90) of 4 µg/mL. Ketoconazole (MIC90, 0.032 µg/mL) and itraconazole (MIC90, 0.012 µg/mL) were much less active. The MIC90 of amphotericin B was 0.38 µg/mL (range, 0.023–1.5 µg/mL). All isolates were highly susceptible to 5-fluorocytosine, with a range of 0.064–0.38 µg/mL. We identified 7 antibiotypes (antibiotypes I–VII; table 2). All isolates recovered from each patient had an identical antibiotype. Isolates from 5 patients (patients 16–18, 22, and 24) all belonged to antibiotype IV.
Of the 6 enzymes, only HaeIII and CfoI could digest the DNA of the 20 isolates of P. marneffei. Digestion with HaeIII identified only 2 RFLP types (types I and II; table 3); type I isolates had 6 bands, and type II isolates had 7 bands. All 20 isolates of P. marneffei had the same RFLP type with the enzyme CfoI. The estimated sizes of the bands in these 3 RFLP types were identical to those reported from Thailand . The majority (75%) of the isolates belonged to RFLP type I when digestion with HaeIII was done.
The RAPD pattern of each isolate was qualitatively identical in each of the triplicate assays, but the band intensity varied. Eight RAPD patterns were identified by use of the 6 primers (table 3 and figure 3). The RAPD profiles generated by OPB-18 and OPH-18 were similar to those generated by OPH-03 and OPH-05. Isolates from patients 16 (isolates D1 and D2), 17 (isolate E), 18 (isolate F), and 22 (isolates I and J) had identical RAPD patterns. RAPD patterns of multiple isolates from the same patients were identical. RFLP type I isolates had 6 distinct RAPD patterns, whereas RFLP type II isolates had 2 RAPD patterns. RAPD patterns detected among the 18 isolates by use of the 4 primers (OPA-03, OPH-03, OPH-05, and OPH-20) are shown in figure 3.
Identification of strains
A total of 8 strains (strains 1–8) were identified (table 3). A major strain (strain 4), consisting of 6 isolates, was recovered from 4 patients (patients 16–18 and 22). Of these 4 patients, 2 had AIDS, 1 had colonization by P. marneffei in the respiratory tract, and all were considered to have acquired the organisms in Taiwan. Isolates from patients 12, 14, and 21, who had traveled to endemic areas outside Taiwan, belonged to different strains (strains 2, 3, and 6, respectively).
This study demonstrated a remarkable increase in the number of P. marneffei infections in Taiwan since 1994, particularly among AIDS patients. Although several strains were found in the 20 preserved isolates recovered from 11 NTUH patients, a unique strain seemed to have spread, in 1997 and 1998, to 4 patients who had not traveled to endemic areas. For strain differentiation, RAPD analysis with APPCR seems to be superior to macrorestriction endonuclease analysis of chromosomal DNA. In addition, the 2 RFLP types found in this study were identical to those found in isolates disseminated in Thailand; however, it remains unclear whether this finding indicates that the Thai strain has spread to Taiwan.
P. marneffei infection causes morbidity and mortality among immunocompromised hosts, particularly among HIV-infected patients living or traveling in Southeast Asia or the southern part of China [2–8]. In the last 2 decades the incidence of penicilliosis marneffei has risen markedly in Thailand, where this infection is now the third most frequent AIDS-defining infection [5, 8].
The environmental niche and geographic distribution of P. marneffei are not well known [5, 8]. Experience with the disease in Thailand suggests that humans may acquire P. marneffei conidia via inhalation from a contaminated reservoir in the environment (e.g., soil) or from contact with animals (e.g., bamboo rats) infected with the fungus [3, 5, 8, 9].
In Taiwan, the increased number of P. marneffei infections since 1994 is partly due to the increase in the number of AIDS cases during this time. Most cases of penicilliosis marneffei in Taiwan are diagnosed at NTUH, because nearly two-thirds of HIV-infected patients are treated at this hospital. Of the 8 patients with HIV infection, 1 (patient 16) had no obvious underlying impaired immunity. Previous reports clearly demonstrated that P. marneffei may be considered a primary pathogen in persons with normal immunity [4, 5]. Interestingly, and contrary to previous findings [2–8], 1 of our AIDS patients had asymptomatic and transient colonization of the respiratory tract by P. marneffei and did not develop disease, even after 2 years of observation.
Although RAPD analysis has various limitations, it has been used for typing several fungal isolates [18, 20]. The microbial identification system, which is based on cellular FAME analysis by use of gas chromatography, is established for identification of bacterial and fungal species and has been used successfully in the identification of clustering among bacterial and fungal strains [17, 18]. In the present study, 8 distinct RAPD patterns were elucidated by >6 primers (initially screened with 60 primers: OPA-01 to OPA-20, OPB-1 to OPB-20, and OPH-1 to OPH-20). Although the E-test has not yet been approved in the United States for clinical use in the identification of molds (but is restricted to use in research laboratories), we were able to identify 8 antibiotypes by using this method. The presence of identical antibiotypes among unrelated isolates (strains 4 and 8) indicates that antibiotyping alone is not an ideal method for typing P. marneffei. Coupling the 2 molecular typing methods with E-test antibiotyping enabled us to clearly identify 8 strains among the 20 P. marneffei isolates in this study. In accordance with our previous, unpublished data—but contrary to the findings by Rath et al.  on cellular fatty acid composition of Exophiala isolates—cluster analysis of FAMEs failed to provide good discriminatory power for typing P. marneffei isolates (data not shown).
The in vitro susceptibilities of the 20 P. marneffei isolates from our patients were similar to those reported elsewhere [21, 22], in that both azole agents and 5-fluorocytosine were more active than amphotericin B. Previous studies indicated that patients treated with itraconazole or high doses of amphotericin B respond better than those treated with fluconazole , and the risk of relapses in AIDS patients can be significantly reduced by lifelong administration of itraconazole . However, there are some discrepancies between the in vitro activity and the in vivo efficacy of amphotericin B and fluconazole . The clinical outcomes of our patients supported the findings reported elsewhere [22, 23], although our patient number was limited. Voriconazole has been demonstrated to be highly active against P. marneffei isolates in vitro; however, the efficacy of this agent in treating disseminated infection needs further investigation .
P. marneffei is thought to be endemic throughout Southeast Asia , and the possibility of widespread strain dissemination deserves attention. Interestingly, the predominant RFLP types (types I and II) among the endemic Thai strains of P. marneffei  were the only types found among our isolates. However, we could clearly differentiate 8 strains (6 RFLP type I strains and 2 RFLP type II strains), which argues against the possibility of strain dissemination from Thailand. Moreover, the isolates from our 3 patients who had traveled to endemic areas were distinct from one another, as well as from the isolates from the 8 patients without a history of travel. However, isolates from 4 of our patients with indigenously acquired infection appeared to be of the same strain, providing evidence of strain dissemination within Taiwan. More-extensive molecular identification and typing studies should be performed to examine the spread of P. marneffei within, as well as among, various regions of Southeast Asia [19, 24]. The natural reservoirs of this organism in Taiwan should be sought more actively.
The impressive increase in the number of P. marneffei infections and the diversity of strains among immunosuppressed patients in Taiwan suggest that investigation of the ecological niche of the organism is necessary for better understanding of the transmission and control of infection. Strain spreading of P. marneffei may partly contribute to the increasing number of patients with infections caused by this organism in Taiwan.