We sought the reservoir of Fusarium species in a hospital with cases of known fusarial infections. Cultures of samples from patients and the environment were performed and evaluated for relatedness by use of molecular methods. Fusarium species was recovered from 162 (57%) of 283 water system samples. Of 92 sink drains tested, 72 (88%) yielded Fusarium solani; 12 (16%) of 71 sink faucet aerators and 2 (8%) of 26 shower heads yielded Fusarium oxysporum. Fusarium solani was isolated from the hospital water tank. Aerosolization of Fusarium species was documented after running the showers. Molecular biotyping revealed multiple distinct genotypes among the isolates from the environment and patients. Eight of 20 patients with F. solani infections had isolates with a molecular match with either an environmental isolate (n = 2) or another patient isolate (n = 6). The time interval between the 2 matched patient-environment isolates pairs was 5 and 11 months, and 2, 4, and 5.5 years for the 3 patient-patient isolate pairs. The water distribution system of a hospital was identified as a reservoir of Fusarium species.
Opportunistic fungi are an increasingly important cause of nosocomial infections; they tend to be resistant to antifungal agents and are associated with a poor outcome . The reservoirs of these fungi in the hospital environment are not well known. Identification of such reservoirs, however, would permit more effective infection-control measures. We therefore sought to identify the reservoir of Fusarium species, a representative of the opportunistic fungi, in a major oncology center with a significant incidence of patients with fusarial infections  and to assess the molecular relatedness of isolates from environmental and patient sources.
We hypothesized that the hospital water distribution system might be a reservoir for Fusarium species, on the basis of growing evidence of such contaminated water supplies worldwide , recognition that waterborne organisms can cause serious nosocomial infections, and recovery of Fusarium species from several nonmedical water systems . Herein we report the results of an investigation documenting that the hospital water system may be the source of nosocomial fusariosis.
Materials and Methods
Environmental sampling. The hospital, located in Houston, Texas, has had several patients with Fusarium infections during a period of 10 years . The hospital is served by the City of Houston municipal water supply. Municipal water enters the hospital via a cubical 75,708-liter cold-water tank (tank A), which communicates with an identical cold-water tank whose water level is passively maintained from the bottom by a pipe that originates at the bottom of the adjacent tank A. Water from the bottoms of the 2 tanks is combined and pumped through a pipe to be distributed around the hospital, including to the patient-care floors. Hot water is heated in 2 horizontal parallel tanks (5500 gallons each) that are interconnected by a pipe with a central drain valve that is located between the tanks. Heated water is then distributed throughout the hospital structures studied, including areas housing inpatients.
Because of the occurrence of infections with Fusarium species over a period of 10 years , an environmental investigation was initiated to identify a potential source in the hospital. The clinical isolates from previously infected patients had been saved as part of an ongoing evaluation of fusarial infections at that center. The study was conducted from January through June 1996. In the first part of the study, environmental sampling was conducted in areas housing high-risk patients in whom fusarial infections had been reported. Additional sampling of other areas was subsequently conducted to ascertain whether water-system contamination was localized to high-risk areas or was system wide.
Water samples were obtained from the point where the municipal water supply enters the hospital, from the cold-water and hot-water tanks, from shower and sink water in patient rooms, and from a drinking fountain. These samples were collected and analyzed according to 2 protocols, filtration and centrifugation. For filtration, 1 L of midstream water was collected from each site in sterile 1-L glass containers and passed through a 0.45-μm Millipore filter (Millipore). The filter was placed on agar medium and incubated, and the concentration of organisms was determined. For centrifugation, 25 mL of midstream water was collected from the hot-water tanks and twice centrifuged (3500 g for 15 min). The resulting pellet was resuspended in 25 mL of sterile normal saline, inoculated (0.2 mL) onto agar medium, and incubated, and the concentration of organisms was determined.
To test the water distribution system, as well as environmental surfaces and objects, 6 premoistened sterile cotton-tipped swab applicators were used to collect samples for culture. Collection of samples involved swabbing lavatory drains, shower drains, drainpipe rubber O-ring seals located under wash basins, disassembled lavatory sink faucet aerators, shower heads, toilet bowls, shower wall surfaces, water tank surfaces, and patient care—related objects. The latter included filters and vents from air-conditioning units in patient rooms, patient food, walls, false ceilings, furniture, mattresses, linens, ice machines, plants, and flowers, as well as hand soap and disinfectants and a surgical unit steam sterilizer water filter. In addition, 29 hospital employees performed direct swabbing of the bathroom sink drains in their residences located throughout Houston.
Air samples (1 m3 each) were collected with a 2-stage Andersen bioaerosol sampler (Andersen-Grasby) as described elsewhere . Air samples were recovered from 4 patient care floors (including the 3 where highly immunocompromised patients were hospitalized), from adjacent corridors, and from nearby connecting stairwells that opened to the outdoor environment. Two modes of air testing were used. In the dry mode, patient shower stalls were tested before water usage. In the wet mode, sampling was conducted with the water running for 35 min.
Sabouraud dextrose agar (Becton-Dickinson Microbiology Systems) was used as the culture medium for all samples. Malt-extract agar (Fluke Biochemika) was used for sampling bioaerosols . Charcoal yeast extract agar (Remel) and modified Nash-Snyder medium (for selective isolation of Fusarium species, kindly provided by the Fusarium Research Center, Pennsylvania State University) were used as supplemental media for samples acquired from water storage tanks. All inoculated plates were incubated at 25°C for ≥4 days. Colony counts (expressed as cfu/mL of water) were determined by serial dilution, whereas bioaerosol samples were read as cfu/m3 of air. Fusarium species and other fungi were identified according to established methods .
Data on residual (free) chlorine levels were determined from institutional maintenance records, which indicated that the cold-water tanks maintained a mean concentration of 0.3 parts per million (ppm), and the hot-water tanks maintained a mean concentration of 0.1 ppm. Water temperatures were measured at the time of sample collection. Water from cold-water tanks had a mean temperature of 25°C, and water from hot-water tanks had a mean temperature of 48.5°C.
Organism samples. Number-coded environmental and clinical isolates of Fusarium solani and Fusaroum oxysporum were initially grown on carnation leaf agar and lyophilized. A small portion of the lyophilized sample was resuspended in normal saline plus 0.0125% Tween 20 and inoculated onto slanted text tubes containing potato dextrose agar. Conidia were harvested, filtered, and pelleted. Pellets were rinsed twice with cold PBS, resuspended, and stored at 4°C.
DNA extraction. Flasks containing 50 mL 2% Sabouraud dextrose broth were inoculated with 4.5 × 107 conidia cells/mL and incubated in a 27°C shaking water bath for 24–48 h. Cells were harvested and spheroplasts prepared as described elsewhere . DNA was extracted from spheroplasts (100–200 mg wet weight) by means of the Qiagen DNeasy Plant Mini Kit for DNA extraction (Qiagen).
Restriction fragment—length polymorphism (RFLP). For RFLP, on the basis of DNA yield (visualized on a 1% agarose gel), restriction digests were performed, according to the manufacturer's instructions, with 40 units of HaeIII restriction enzyme (Boehringer Mannheim). The samples were incubated at 37°C over night and run on a 0.8% Tris-borate/EDTA agarose gel for 20 h at 20 V.
Random amplified polymorphic DNA (RAPD). For RAPD, the oligonucleotide primer used for amplification was RP5, 5′-CGG TCA CGC T-3′ (sequence data provided by S. Weir, personal communication). Amplification reactions (50 μL of PCR reaction) were similar to those used by others [8, 9]. All oligonucleotide primers were synthesized in the National Cancer Institute laboratory (Bethesda, Maryland) with use of a DNA/RNA synthesizer (Applied Biosystems).
Interrepeat—polymerase chain reaction (IR-PCR). For IR-PCR, an oligonucleotide primer pair, 1245 (AAG TAA GTG ACT GGG GTG AGC G) and 1246 (ATG TAA GCT CCT GGG GAT TCA C), was used, and a single oligonucleotide, 1236 (TAC ATT TCG AGG ACC CCT AAG TG), was used as a primer in a separate reaction . Amplification reactions were performed in 30 μL of a solution of the following: 6.25 mM magnesium chloride, 83 mM potassium chloride, 16.7 mM ris-hydrocloride, 0.001% (w/v) gelatin, 0.33 mM of each deoxynucleotide, 1.25 units Taq DNA polymerase (Boehringer Mannheim), 2 μL of genomic DNA diluted 1 : 10 in Tris-EDTA. PCR was performed as follows: 95°C for 1 min, 25°C for 1 min, 74°C for 2 min for 35 cycles, preceded by a 5-min incubation at 95°C and followed by a 5-min extension period at 74°C, ending in a 4°C soak. The entire reaction was run on a 1.5% Tris-acetate/EDTA agarose gel.
Analysis. Analysis of gel banding patterns was performed by visualization of images captured by the Alphalmager (Alpha Innotech). The banding patterns of gels for DNA from each clinical and environmental isolate were compared for relatedness to all other isolates. The gels were compared by investigators who were blind to the clinical or environmental identity of the isolates.
The comparative banding patterns revealed by RAPD, RFLP, and IR-PCR were assessed independently at the National Cancer Institute (Bethesda, MD). On the basis of the gel patterns, the probability of 2 isolates being related was assessed for the patterns revealed by each of the 3 methods. If the banding patterns of 2 isolates were the same as revealed by 1 method, the relationship between the 2 isolates was scored as highly probable. If the patterns were the same with the presence or absence of 1 band, the relationship was scored as probable.
To provide a stringent basis for designating the relatedness of 2 isolates, we required that evidence for relatedness was be revealed independently by all 3 methods (RAPD, RFLP, and IR-PCR). In addition, a fourth molecular biotyping procedure, RAPD with a different set of primers , P1 (GAC AGA CAG ACA GAC A) and P2 (GAG GGT GGC GGT TCT), was performed in another laboratory-based study at the University of Texas at Houston. Those results were used to further assess relatedness. After comparison of all strains by use of the 4 biotyping methods, the results were classified.
Two isolates were considered to be related if the banding patterns revealed by all 4 methods were each scored as being highly probable. The isolates were considered to be probably related if the banding patterns revealed by 2 methods were each scored as being highly probable and the other 2 as probable. The isolates were considered to be possibly related if the banding pattern for 1 method was scored as being highly probable and the banding patterns for the other 3 methods were scored as probable. Two isolates were considered to be not related if none of the above criteria was fulfilled. The combined use of all 4 molecular methods has sensitivity to distinguish ≥95% of isolates of F. solani.
Environmental cultures. Results of analysis of samples of water and samples from the water distribution system are summarized in table 1. Of 283 samples from the water distribution system that were analyzed with use of the centrifugation protocol, 162 (57%) were positive for Fusarium species, mostly Fusarium solani and Fusarium oxysporum. In addition, Aspergillus species (Aspergillus fumigatus, Aspergillus terreus), Curvularia species, and Alternaria species were recovered from the hospital hot-water storage tanks. Cultures of 2 of 4 water filters extracted from a surgical unit sterilizer yielded Fusarium sporotrichioides, as well as F. solani, F. oxysporum, Aspergillus species, and Pseudallescheria boydii.
Bioaerosol studies. Aerosolization of Fusarium species and Aspergillus species was demonstrated in the wet sampling mode; all dry mode cultures were negative. Seven (27%) of 26 samples obtained by the wet sampling mode were positive for airborne F. oxysporum (table 1).
Molecular analysis of epidemiology. Thirty-seven isolates of F. solani were tested (20 from consecutive patients for whom isolates were available and 17 from randomly selected environmental samples). When the samples were analyzed by use of RAPD alone, isolates from 4 patients had a highly probable match (i.e., identical banding patterns) with environmental isolates, and 15 had probable matches (i.e., differing by ±1 band). When samples were analyzed by use of all methods (RAPD, RFLP, and IR-PCR) at 2 laboratories, the molecular epidemiology studies still revealed that, for F. solani, 6 paired isolate combinations were probably or possibly related. According to the criterion that isolates match if results from all 4 molecular biotyping methods match, these genetically related pairs of strains of F. solani included 2 correlations for patient-environment isolate pairs and 3 correlations for pairs of patient isolates. A third patient-environment pair was related according to the results of 3 methods, but this was not classified as a match because the fourth method did not reveal similar bands. An additional pair of related environment isolates was also identified according to the high-stringency criteria.
Table 2 summarizes findings (according to high-stringency criteria) for the relatedness of strains of F. solani recovered from patients and the hospital environment. In total, 8 of the 20 patients with F. solani infections had isolates with a molecular match with either an environmental isolate or another patient isolate, even with use of the high-stringency criterion of matching results of multiple molecular epidemiological methods. Representative RFLP patterns for related isolates are presented in figure 1.
Patients with fusariosis. The features and outcome of all cases of invasive fusariosis in immunocompromised patients with hematological malignancy treated at the same cancer center have been reported elsewhere . The study was conducted by reviewing the medical records of patients with positive cultures for Fusarium species who received care at the cancer center from January 1986 through December 1995. Invasive organ infection was documented by both culture of samples from the involved organ and histopathological examination of the organ. Disseminated infection was defined as involvement of at least 2 noncontiguous organs by Fusarium species
A total of 38 patients were identified. The risk factors for fusariosis included relapsing underlying disease (38 patients), neutropenia (36 patients), acute myelogenous leukemia (AML; 30 patients), therapy with adrenal corticosteroids (14 patients), and bone marrow or stem cell transplantation (BMT; 12 patients; allogeneic transplant in 9). Fusarial infection developed after engraftment (median, 2 months after) in 7 of the 9 allogeneic BMT recipients, and 4 of these patients had grade ≥2 graft versus host disease (GVHD). Twenty-nine of these patients died with disseminated fusarial infection (including 12 of 14 for whom an autopsy examination was performed).
Isolates from 20 patients were available for study. Eight of these 20 patients with F. solani infections had a molecular match, either between patient and environmental isolate pairs (2 patients) or between patient isolate pairs (6 patients). The time interval between the 2 patient-environment isolate pairs was 5 and 11 months. This time interval was 2, 4, and 5.5 years for the 3 patient-patient isolate pairs.
The 2 patients had isolates that matched strains isolated from water-related environmental sources. Patient 1, a 62-year-old woman, had AML and received remission induction chemotherapy (2 cycles of fludarabine, cytarabine, and idarubicin); she had subsequent pancytopenia but no leukemia response. After the second cycle of chemotherapy, neutropenic fever, multiple nodular cutaneous lesions, and respiratory distress requiring intubation developed. A biopsy of the skin lesions revealed acute septate branching hyphae and yielded F. solani. Despite optimal supportive care, antifungal therapy, and recovery from neutropenia, patient 1 died of disseminated fusarial infection. Patient 2, a 46-year-old man, had chronic myelogenous leukemia and underwent allogeneic peripheral stem cell transplantation with successful engraftment. His hospital course was complicated by fever and multiple cutaneous lesions; 2 different specimens obtained 2 days apart yielded F. solani. Despite appropriate antifungal therapy, patient 2 died as a result of disseminated fusarial infection.
A third patient-environment pair of F. oxysporum isolates was related according to all 3 methods performed at laboratory A, but was not classified as a match because the fourth method (performed at laboratory B) did not reveal similar bands. For 1 of the 6 patients with matched patient-patient isolate pairs, F. solani was recovered from the tip of a central venous catheter; the other 5 patients had disseminated disease. Notably, fusarial infections were diagnosed several years apart in the 3 patients with matched patient-patient isolate pairs (2, 4, and 5.5 years), which suggests that the same strain of F. solani may have persisted in the hospital water system for a long period of time.
To our knowledge, this is the first study to demonstrate that (1) hospital water systems can be reservoirs for opportunistic fungi, particularly F. solani and F. oxysporum; (2) genetically diverse strains of F. solani can contaminate a water system; (3) these strains can persist in a water system for a period of years; (4) the water system can disseminate these organisms by way of aerosols from showers and sinks; and (5) the isolates of Fusarium species recovered from the water system can also cause nosocomial infections. This study also establishes that potable water can be colonized by the highly toxigenic species F. sporotrichioides.
These findings are supported by increasing reports of pathogenic fungi in nonmedical water distribution systems worldwide , as well as our more recent findings of Fusarium species and other fungi in the water distribution systems of 2 hospitals in another city . Other investigators have found pathogenic fungi in water-damaged sites within hospitals and in the water system itself [13–15]. In the latter study , filamentous fungi were recovered in 94% of all water samples of a pediatric BMT unit. Our recovery of various molds in the hospital water system in Houston, despite standard and adequate water chlorination (mean, 0.3 ppm of chlorine) is best explained by the well-known high-level resistance of various molds to the water chlorine levels recommended by the US Environmental Protection Agency .
The recovery of airborne Fusarium species in the vicinity of running water, coupled with the importance of sinopulmonary infection in fusariosis , suggest that aerosolization from water sources may be partly responsible for these infections. Nosocomial fusarial cellulitis of the lower extremities might result from patient exposure to Fusarium species during showering, through breaks in skin integrity .
The molecular relatedness among isolates of F. solani was subjected to an analysis that required agreement between the results of 3 molecular analysis methods; results from an independent laboratory that used yet another molecular technique further corroborated this finding. To our knowledge, such an approach—the use of a combination of 4 molecular systems in 2 laboratories—has not been used in molecular epidemiological analysis of nosocomial infections. Results from the 3 molecular methods used at laboratory A only also revealed a third matched patient-environment isolate pair and another matched patient-patient isolate pair.
No single molecular biotyping method can be designated as a “standard” for identification of Fusarium species. Our stringent criteria undoubtedly prevented some isolates from being designated as related. Yet those isolates that fulfilled our criteria are designated as related with more reliability than that conferred by use of 1 biotyping method alone.
However, we recognize that no single method or combination of methods is capable of distinguishing among all isolates under all circumstances. Ultimately, advanced methods and an understanding of the whole genomes of Fusarium species and other filamentous fungi will allow for more refined discrimination on the basis of individual sequenced analysis. Fully recognizing that molecular biotyping is predicated on evolving technologies, we anticipate that more refined methods will be available in the future.
Several reasons may explain why isolates from some patients were not related to environmental isolates. Isolates may have had different times of influx into the water system. Some patients may have been colonized by organisms from outside the hospital system. Indeed, the recovery of Fusarium species from drains in Houston residences suggests that community-acquired infection is theoretically possible.
Additional studies are needed to better understand the epidemiology of nosocomial fusariosis. Such studies should be conducted prospectively at several cancer centers where severely immunocompromised patients (e.g., who have undergone allogeneic BMT or who have AML) are cared for, and the studies should focus on a correlation between the presence and concentration of Fusarium species in each hospital water system and the incidence of fusariosis. Such studies should also adjust for significant variations in practices between centers that may put patients at different risk for fusariosis. These variations include differences in the selection of patients who undergo allogeneic BMT (younger, healthier, and untreated patients vs. older patients with comorbidities and extensive previous therapy), different degrees of immunosuppression (e.g., manipulated vs. unmanipulated stem cells, standard myeloablative vs. nonmyeloablative chemotherapy, and intensity of GVHD prophylaxis), and others.
The persistence of Fusarium species in the hospital water system over a long period of time is suggested by the presence of similar organisms (as determined by 3 molecular methods used) in the 3 related patient-patient isolate pairs over a period of 2, 4, and 5.5 years. This is similar to what is observed with Legionella species, another waterborne pathogen, which causes legionellosis. Indeed, Legionella species have been shown to persist in a hospital water system for as long as 10 years [17–19].
On the basis of these findings, we recommend that hospitals where cases of fusariosis occur should consider testing their water supply for the presence of opportunistic molds. If such fungi are recovered from the water system, these hospitals should then formulate policies to avoid or minimize exposure of immunosuppressed patients to tap water from any source. The most effective and least expensive approach that can be applied worldwide is the prevention of exposure of immunosuppressed patients to hospital tap water by providing sterile (boiled) water for drinking. In addition, patients should avoid showering during severe immunosuppression because of the risk of acquiring the organisms through aerosolization of contaminated hospital water. We recommend that bed baths provided with sterile disposable sponges be used instead of showering.
In conclusion, we have identified the water distribution systems of hospitals and residences as potential indoor reservoirs of Fusarium species.
We thank Edwin E. Geldreich, senior research microbiologist at United States Environmental Protection Agency, Drinking Water Research Division, and Dr. Glen Mayhall, University of Texas Medical Branch, Galveston, Texas, for their expert advice; and Jean Juba, of the Fusarium Research Center, for expert technical work in maintenance of Fusarium isolates and identification of Fusarium species.