In May 2005, a disease outbreak was investigated at a zebrafish (Danio rerio) research facility experiencing severe losses. Mycobacterium haemophilum was isolated from these fish and the disease was subsequently recreated in experimentally infected zebrafish. Fish exhibited signs characteristic of mycobacteriosis, including granuloma formation and severe, diffuse, chronic inflammation. Bacteria were observed in multiple tissues, including the central nervous system. Biofilm samples from the outbreak facility were PCR positive for M. haemophilum, suggesting biofilms might act as a reservoir for infection. Zebrafish appear to be particularly vulnerable to M. haemophilum, and measures such as quarantine and treatment of incoming water should be implemented to minimize the likelihood of introduction of this bacterium to zebrafish research facilities. Zebrafish are already a well-established laboratory animal model for genetics, toxicology and disease, their susceptibility to M. haemophilum may make them useful for the study of this bacterium in the future.
Mycobacteriosis is a common disease in zebrafish (Danio rerio) used for research (Kent et al., 2004). Mycobacterial infections in fish are often attributed to Mycobacterium marinum, Mycobacterium fortuitum, and Mycobacterium chelonae (Decostere et al., 2004), although additional species have been identified as a cause of disease in zebrafish, such as Mycobacterium abscessus, Mycobacterium peregrinum, and Mycobacterium haemophilum (Astrofsky et al., 2000; Kent et al., 2004; Seok et al., 2006; Watral & Kent, 2007). When zebrafish are experimentally infected with a moderate dose of some of these species, the severity of disease varies. For example, Watral & Kent (2007) demonstrated that M. marinum causes outright mortality in exposed fish, whereas little disease and mortality is caused by M. abscessus, M. chelonae, or M. peregrinum when injected at the same level.
In our studies of zebrafish diseases, M. haemophilum has been found to be associated with two severe and persistent disease outbreaks from facilities in Europe and the western United States (Kent et al., 2004). Better known as a human pathogen, the host range and distribution of M. haemophilum is likely underappreciated due to the specific culture conditions required for its growth (Bruijnesteijn et al., 2005). Indeed, investigation of these previous outbreaks in zebrafish relied on DNA sequence data acquired directly from infected tissues for identification (Kent et al., 2004). Thus, pathogenicity of the bacterium in experimentally infected zebrafish could not be evaluated. However, in May 2005, a disease outbreak at a zebrafish research facility provided an opportunity for isolation and characterization of M. haemophilum in zebrafish.
In May 2005, the Zebrafish International Resource Center's Pathology Service received fish for histological processing from a zebrafish research facility at the University of Georgia (UGA) experiencing significant mortality in their population (3–4 fish each day). The zebrafish system consisted of about 200 tanks on a recirculating water system. Following the identification of lesions and acid-fast staining bacteria characteristic of a mycobacterial infection, c. 30 zebrafish (Danio rerio) were sent to Oregon State University's Center for Fish Disease Research for identification and characterization of this mycobacterium. Ten fish exhibiting signs of disease (emaciation) were examined upon arrival. The remaining fish were held in isolation in a BSL-2 laboratory in static aquaria (i.e. without flowing water) in dechlorinated city water, and water temperature was maintained at 28°C. Ammonia and nitrate concentrations in the static aquaria were controlled by monitoring levels with test kits and periodic water changes. Each aquarium was equipped with a biological filter filled with porous lava rock. Fish were fed twice daily. Fish were monitored over the next 9 months and moribund or dead fish were collected and processed as follows. Moribund fish were euthanized with an overdose of tricane methanesulfonate (Argent Laboratories, Redmond, WA) and examined for signs of mycobacteriosis, presence of dermal lesions, emaciation, and raised scales. With sterile instruments for each fish, a portion of the spleen or liver was removed, which was used to make an imprint on a microscope slide and then placed in a tube for subsequent culture. An additional piece of spleen or liver was removed for molecular analysis. The remainder of the fish was placed in Dietrich's fixative for histological processing. Biofilm samples were collected from the affected facility at UGA following the outbreak event, using sterile swabs on system gutters, drains, and tank meniscuses.
Tissues and biofilm swabs were decontaminated overnight in 1% cetyl pyridinium chloride (Kent & Kubica, 1985). Initially, cultures were grown on Middlebrook (MB) 7H10 and Lowenstein–Jensen (LJ); however, no growth was observed on these media. Following preliminary genetic identification as M. haemophilum, cultures were grown on MB 7H10 agar supplemented with OADC and 60 µM hemin, incubated at 29°C.
Tissue imprints were stained using the Kinyoun's acid fast stain (Kent & Kubica, 1985) and examined for the presence of acid fast bacilli. Histological specimens were prepared using standard methods and stained with hematoxylin and eosin, Ziehl-Neelsen's acid fast, or Kinyoun's acid fast.
Genomic DNA from cultures, biofilms, and spleen or liver tissue was extracted using the DNeasy Tissue Kit (QIAGEN, Valencia CA). PCR was used to amplify various regions of the small subunit (SSU) and internal transcribed sequence (ITS) rRNA gene, and heat shock protein 65 (hsp65) genes from the DNA extracted from infected tissues. PCR primers and conditions were as described in our previous work (Kent et al., 2004; Poort et al., 2006). Additional hsp65 primers of Selvaraju (2005) and for RNA polymerase B (rpoB) as described by Adékambi (2006) were also used. PCR products were gel-extracted using the QIAquick Gel Extraction Kit (Qiagen Inc., Valencia CA). Direct sequencing was carried out with the ABI BigDye Terminator Cycle Sequencing Ready Reaction Kit v3.1, using the ABI PRISM® 3730 DNA Analyzer (Applied Biosystems, Foster City CA). For direct PCR testing of tissues and biofilms, a M. haemophilum ITS-specific PCR was conducted using standard conditions and primers MycoITS1F of Whipps (2003) and reverse primer MhITS1R (5′-TGAACACGCCACCATTAC), modified from Bruijnesteijn (2005).
All fish used in this study were obtained from an ongoing project that was approved by the Oregon State University Institutional Animal Care and Use Committee. Zebrafish (AB strain, c. 8 months) were used. Fish were divided into two groups of 23 and placed in static aquaria and maintained as described above. The inoculum of M. haemophilum‘SD18’ was prepared by suspending fresh cultures in PBS+0.05% Tween 80. The fish were anesthetized with tricaine methane sulfonate (Argent Laboratories, Redmond, WA) at 130 µg mL−1 and injected intraperitoneally with a 26 gauge needle attached to a Eppendorf Combitip on a repeat pipettor. The concentration of the inoculum was estimated using McFarland's turbimetric standards, and diluted in PBS to a predicted value of 4 × 106 CFU per milliliter. Each fish was injected with 25 µL of the inoculum, with a target of 1 × 105 CFU fish−1. Verification of concentrations was determined by colony counts on MB 7H10 supplemented with 60 uM hemin. Moribund or dead fish were collected for PCR, culture, and histological evaluation (preserved in Dietrich's solution), as described above. Survival in exposed and control groups was analyzed by Kaplan–Meier survival analysis and compared for differences by log-ranked, Wilcoxon, and Tarone–Ware tests.
During the outbreak at the zebrafish research facility at UGA, c. 3–4 dead fish were collected daily (postoutbreak background mortality was c. 1–2 fish monthly). Upon receipt of 30 animals at the Center for Fish Disease Research, 10 exhibited overt signs of disease such as emaciation and lethargy, and were examined immediately. Nine of 10 were shown to be infected with M. haemophilum based on both the appearance of M. haemophilum-like AFB in tissue smears and by DNA sequencing of the mycobacterial hsp65 gene directly from infected tissues. Over the following 9 months, additional fish were collected as moribund or upon death, nine of which were examined. All nine were infected with M. haemophilum as identified by DNA sequencing directly from infected tissues.
Routine culture at 29°C was unsuccessful on both MB 7H10 and LJ media. Following the identification of M. haemophilum by DNA sequencing, MB 7H10+hemin medium was used and growth was seen after 4–6 weeks. Numerous small colonies were observable with the aid of a dissecting microscope after 4 weeks and visible to the naked eye after c. 5 weeks. DNA sequencing directly from these cultures verified the identity of the organism as M. haemophilum. Sequences of SSU and ITS (GenBank accession number: DQ851570), rpoB (DQ851569), and hsp65 (DQ851571) were identical to one another and to those obtained from previous outbreaks (Kent et al., 2004).
Mycobacterium haemophilum was detected in biofilms collected from the UGA facility following the outbreak. Tank meniscuses (4/4) and a tank drain were PCR positive for M. haemophilum. Zebrafish food (2/2) was negative for the bacterium as were recently replaced system gutters (3/3).
Estimated by McFarland's standards and verified via plate count, zebrafish in the exposed group were injected with 1.2 × 105× CFU fish−1. Little mortality was observed for several weeks, but at c. 8 weeks, the number of moribund fish increased (Fig. 1). At 10 weeks postexposure, mortality had reached 34.8% (8/23) compared with 13.6% (3/22) in the controls, and the experiment was terminated. Kaplan–Meier survival analysis did not reveal a significant difference in survival between groups (P=0.10>α=0.05). However, there were significant differences when infection and disease were the end points. Acid-fast bacteria were observed in tissue imprints from all 23 zebrafish from the exposed population and all were also PCR positive for M. haemophilum. All PBS-injected fish, including mortalities, were negative for mycobacterial infection based on tissue imprint, histology, culture, and PCR. An infection with the microsporidian Pseudoloma neurophilia was seen in 72.7% exposed fish and 77.3% control fish.
Twenty-two of 23 M. haemophilum-exposed fish were examined by histology. All of the fish exhibited severe mycobacteriosis (Fig. 2), characterized by multiple granulomas (Fig. 2a) and severe, diffuse, chronic inflammation in the visceral organs (Fig 2b). Acid-fast bacteria were readily detected in granulomas (Fig 2c), and massive numbers of bacteria were found throughout the regions of diffuse inflammation (Fig. 2d and f–h). Essentially all organs were involved, including the kidney, spleen, liver, pancreas, heart and somatic muscle. The intestine exhibited particularly heavy infections in most fish (Fig. 2h), with massive numbers of phagocytes replete with bacteria occurring through all cell layers of the intestine. Acid-fast bacteria were also observed in the lumen of the intestinal tract (Fig. 2g). Two of the moribund/dead fish exhibited infections in the central nervous system, including severe involvement of the meninges and neural tissue of the brain and spinal cord. Multiple colonies of acid-fast bacteria, with little inflammatory reaction, were observed throughout the spinal cord (Fig. 2e).
A total of 11 of the 15 surviving fish exhibited the severe, diffuse inflammation type of infection, and three of these also had multiple granulomas in the visceral organs. The other four fish had infections where bacteria were only detected inside the granulomas.
Mycobacterium haemophilum was found to be the cause of significant mortality in a UGA zebrafish research facility experiencing an ongoing disease outbreak. Of the fish sent to and examined at the Center for Fish Disease Research, M. haemophilum was identified by DNA sequencing in almost every dead fish or those showing clinical signs of disease. We were able to recreate the disease by injecting the M. haemophilum into zebrafish and then reisolate the organism. Infections were severe, involving multiple tissues throughout the fish, including the central nervous system. When mortalities alone were compared between groups (Fig. 1), no statistical difference was found between exposed and control fish. However, because severe disease was observed in surviving fish, more mortalities would have been expected had the experiment run longer. Indeed, it took 9 months for all 30 of the zebrafish from the UGA outbreak to eventually perish, demonstrating the chronic nature of this disease. Comparison of mortality alone was likely confounded by an underlying infection with P. neurophilia, another common pathogen of laboratory zebrafish (Matthews et al., 2001). The role P. neurophilia infections may have played in fish exposed to M. haemophilum is unclear, i.e. a coinfection may have sped the time to death. Mortalities that occurred in PBS-injected controls may be attributed to infections of the microsporidian parasite, as PCR, whole-fish histology, and culture were negative for mycobacteria in all controls. Ultimately, many more mortalities were recorded for exposed fish than control fish, and histological evaluation shows massive mycobacterial disease in the mortalities exposed to M. haemophilum. In addition, the clinical and histological presentation of disease was the same in outbreak fish and experimentally exposed fish, demonstrating that M. haemophilum was responsible for the disease outbreak at the UGA zebrafish facility.
Mycobacterium haemophilum is better known as an infrequently encountered human pathogen, associated with cutaneous or subcutaneous skin lesions in immunocompromised individuals and lymphadenitis in otherwise healthy children (Straus et al., 1994; Saubolle et al., 1996; Dobos et al., 1999; Elsayed & Read, 2006). Detection of the bacterium in other animals is uncommon, reported from a python with a lung infection (Hernandez-Divers & Shearer, 2002), a hospital dwelling cockroach (Pai et al., 2003), bison (Jacob et al., 2006), and our prior studies with zebrafish (Kent et al., 2004). As specific culture conditions are required for detection of M. haemophilum, it is likely overlooked by routine screening methods and its geographic distribution and host range may be underappreciated (Bruijnesteijn et al., 2005). Genetic fingerprinting studies of human cases show some geographical clustering, suggesting a common environmental source of infection (Kikuchi et al., 1994; Yakrus & Straus, 1994). Although there are no reports of M. haemophilum isolation directly from water (Saubolle et al., 1996; Smith et al., 2003), isolation from biofilms growing on pipes in water distribution systems has been successful (Falkinham et al., 2001). We also detected the bacterium in biofilms from the zebrafish facility at UGA and suspect that biofilms might act as a reservoir for mycobacteria and in a recirculating zebrafish water system, providing a continuous source of infection to which fish are exposed. Indeed, the number of mortalities at the affected facility decreased sharply following the adjustment of the UV sterilization system, replacement of algae coated tanks, and thorough cleaning of the drain gutters.
Previous outbreaks attributed to M. haemophilum at zebrafish facilities have been described using DNA sequencing but the organism had not been cultured (Kent et al., 2004). Here, we were able to culture the bacterium and reproduce the disease observed in the original outbreak, including mortality. Of those evaluated by Watral & Kent (2007), few mycobacteria of zebrafish caused this same kind of severe disease and mortality when injected at moderate levels into healthy fish. Those that do cause mortality, such as isolates of M. marinum (Kent et al., 2006) and M. haemophilum (described here), are of serious concern to zebrafish researchers. We recognize that injection of zebrafish is not a natural route of exposure. However, such a model has been used successfully to evaluate the relative pathogenicity of other mycobacteria in fishes (Prouty et al., 2003; Broussard & Ennis, 2007; Watral & Kent, 2007). Until other methods of exposure are proven to be effective and reproducible, exposure by injection will likely continue to be employed. Even though the route of exposure and dose are not identical to a natural situation, our goal was not to evaluate routes of exposure, but to assess whether the same isolate of M. haemophilum from a zebrafish disease outbreak could cause the same disease in susceptible animals. Indeed, histological evaluations clearly show the same severe disease in exposed zebrafish, including the neurological involvement, which were seen in the other outbreaks.
The description of M. haemophilum from three apparently unrelated zebrafish facilities (western USA, eastern USA, and England) suggests that either these infections were acquired independently in each location, or a single population of fish acquired the infection and the disease was spread via transplantation. The sharing of zebrafish between facilities is a common, but poorly documented, practice. Here, the UGA facility investigated had no history of sharing zebrafish between labs. Facilities are usually supplied with dechlorinated city water, which is a potential source. Nevertheless, contaminated objects, human or insect vectors cannot be ruled out. We are unable to decipher the relatedness of the M. haemophilum strains by way of culture-based strain typing techniques, as the organism was not cultured in the first two outbreaks. However, the DNA sequences examined (SSU, ITS, rpoB, and hsp65) from all three outbreaks were identical. We are confident that our observation of M. haemophilum is not due to contamination in our laboratory as histological findings were consistent in all three cases, independent of PCR or culture results, and specific to M. haemophilum infection as described by Kent (2004).
As a final note, zebrafish appear to be very susceptible to infections by M. haemophilum and have a body temperature permissive for its growth. Zebrafish–mycobacteria model system has been proposed for M. marinum (Prouty et al., 2003; Pozos & Ramakrishnan, 2004) and we are encouraged that with further development, zebrafish may be useful for the study of M. haemophilum infection and disease. In particular, our observation of infections in the CNS of artificially exposed animals and those from outbreaks warrants further study on the mechanism by which M. haemophilum is able to cross the blood/brain barrier, a characteristic we have not observed in other mycobacteria of zebrafish.
V.G. Watral is gratefully acknowledged for assistance with injections. Thanks also to J.L. Matthews at the University of Oregon for assistance in the initial case submitted to the pathology service. Financial support was provided by the National Institutes of Health grants 5R24RR017386-02 and P40 RR12546.