Mycobacterium lentiflavum in Drinking Water Supplies, Australia

Humans may acquire infection from potable water.

similarities to M. avium complex (MAC), differentiation can be diffi cult without molecular identifi cation, hence, misclassifi cation in the past is possible (2).
As with other NTM, M. lentifl avum has been isolated from soil and water samples around the world. However, links between environmental sources and human disease have not yet been demonstrated.
In Queensland, Australia (population 4.28 million), NTM disease is notifi able. A central reference laboratory performs speciation of all positive isolates. In 2008, ≈900 isolates of NTM were reported.
Strain variation within mycobacterial species is well known. Although epidemiologic studies provide useful information, molecular strain typing can be invaluable, especially if a single clone can be linked to an outbreak source. Pulsed-fi eld gel electrophoresis (PFGE) has been considered the standard for mycobacterial strain typing but is time-and labor-intensive and requires expensive dedicated equipment. Also, DNA degradation can occur during electrophoresis, generating uninterpretable banding patterns (3). Repetitive sequence-based PCR (rep-PCR) has been used to differentiate mycobacterial strains associated with disease outbreaks in mesotherapy clinics (M. abscessus and M. chelonae) (4) and in patients after surgery (M. fortuitum) (5). An automated rep-PCR system (DiversiLab; bioMérieux, Melbourne, Victoria, Australia) showed high concordance with PFGE results (6) in identifying mycobacterial strain clusters and was faster than PFGE.
We had 2 goals for this study. First, we aimed to describe the clinical signifi cance and outcomes of M. lentifl avum infection in Queensland. Second, we intended to explore the genotypic and geographic relationship between patient isolates and potable water isolates in the Brisbane area.

Methods
We reviewed the records of all patients from whom M. lentifl avum had been isolated during July 2001-November 2008. Attending physicians were contacted to establish clinical signifi cance according to American Thoracic Society (ATS)/Infectious Diseases Society of America (IDSA) criteria (2) ( Table 1). During 2007-2008, potable water was collected from 206 sites in Brisbane's drinking water system.

Laboratory Identifi cation
Human samples were digested and decontaminated by using 4% NaOH, neutralized with phosphoric acid, and centrifuged to concentrate the acid-fast bacilli (AFB). Smears were prepared from the sediment and stained by the Ziehl-Neelsen (ZN) method. We injected cells into 1 Lowenstein-Jensen slope (± pyruvate) and 7-mL mycobacterial growth indicator tube, then incubated them at 35°C until growth was detected. ZN staining of colonies confi rmed AFB. Multiplex PCR (7)

Water Sampling
Water was collected from routine sampling sites across Brisbane and processed according to described methods (8). Each 1,000-mL sample was transported at 4°C and processed within 24 hours. Half of each sample was decontaminated by using 0.005% cetylpiridinium chloride, and each 500-mL aliquot was fi ltered separately by using 45-μm cellulose nitrate fi lters (Sartorius AG, Gottingen, Germany). The fi lters were rinsed and macerated in 3 mL sterile distilled water. Aliquots (0.1 mL) were transferred in triplicate to M7H11 plates, sealed in gas-permeable plastic bags, and incubated at 32°C. Aliquots (0.5 mL) were transferred to 2 mycobacterial growth indicator tubes, 1 of which contained polymyxin, azlocillin, nalidixic acid, trimethoprim, and amphotericin B. ZN staining of colonies confi rmed AFB, and these colonies were subcultured on M7H11 plates. Multiplex PCR was performed (7) followed by 16S rRNA sequencing of mycobacterial isolates and compared by using Ribosomal Differentiation of Medical Microorganisms and GenBank databases (9,10).

Automated Rep-PCR Strain Typing
The similarity of 16 clinical and 7 water isolates was determined by using a rep-PCR method (DiversiLab). DNA was extracted from clinical and water isolates by using the Ultraclean Microbial DNA Isolation Kit (Mobio Laboratories, Carlsbad, CA, USA). PCR mixture was prepared by using AmpliTaq polymerase and PCR buffer (Applied Biosystems, Foster City, CA, USA) and Mycobacterium DiversiLab primer mix according to the manufacturer's instructions. Rep-PCR products were separated and detected by using microfl uidic chips of the DiversiLab system. Fingerprints were analyzed with DiversiLab software version 3.4.38 by using the Pearson correlation coeffi cient and unweighted pair-group method with arithmetic means to compare isolates and determine clonal relationships.

Clinical Isolates
Forty-seven isolates of M. lentifl avum were reported from 36 patients ( Figure 1; Risk-benefit of therapy should be considered for each patient before institution of therapy Expert consultation should be obtained when NTM are recovered that are either infrequently encountered or that usually represent environmental contamination Patients suspected of having NTM lung disease but who do not meet the diagnostic criteria should be followed until the diagnosis is firmly established or excluded *Adapted from (2). NTM, nontuberculous mycobacteria. †Transbronchial or other lung biopsy.
with uncertain or nonsignifi cant disease, 26 had 1 positive specimen, 2 had 2 positive specimens from the same period, 3 had 2 positive specimens separated by 3 months, and 1 had 2 positive specimens separated by 11 months. Antimicrobial drug susceptibility tests were performed for 2 isolates (cases 1 and 2 below). Both were sensitive to clarithromycin 4.0 μg/mL and resistant to isoniazid 0.4 μg/ mL, ethambutol 5.0 μg/mL, and streptomycin 1.0 μg/mL. The case 1 isolate was sensitive to ofl oxacin 2.0 μg/mL; the case 2 isolate was resistant to rifampin 1.0 μg/mL.

Environmental Isolates
Mycobacteria were grown from 70% of water sites. The predominant isolates were M. gordonae and M. kansasii. M. lentifl avum was isolated from 13 (6.3%) sites, 2 of which were reservoirs, 1 a treatment plant, and the remainder points in the distribution system. Eleven sites shared the same groundwater source but were distributed among 10 different reservoir zones. For 12 patients living within 20 km of Brisbane central business district, the mean distance between their residential addresses and nearest positive water site was 3.49 km (range 0.9-9.8 km). The 4 persons with clinically signifi cant illness lived a mean of 2.7 km from a positive water site ( Figure 2).

Case Descriptions for Signifi cant Isolates
The 4 patients whose disease met the ATS/IDSA criteria are described below. All specimens were ZN stain negative.

Case 1: Disseminated Infection
A 43-year-old woman who smoked had a background of intravenous drug use and HIV. In 1998, granulomatous hepatomegaly developed, thought to be a reaction from injecting methadone mixed with orange juice, and resolved after she ceased this activity. A tunneled intravenous access device was placed in February 2006. In April 2007, she sought care for hepatosplenomegaly and mild pancytopenia. Liver and gastric lymph node biopsies showed granulomata. Two bone marrow biopsy samples taken 6 weeks apart showed initially scant, but then more marked, granulomata. All specimens were culture negative for AFB. A working diagnosis of sarcoidosis was made, and prednisone with highly active antiretroviral therapy (tenofovir, emtricitabine, and efavirenz) began. Azathioprine was introduced and prednisone ceased by April 2008. In June, she was admitted with massive hepatosplenomegaly, weight loss, and fever. CD4+ count was 0.14 × 10 9 /L (0.43-1.62 × 10 9 /L), and viral load was undetectable (<50 copies/mL HIV-1 RNA). Over the next month, all 4 blood cultures grew M. lentifl avum; after 15 days, mycobacteria were apparent and M. lentifl avum was confi rmed 7 days later (day 22). Bone marrow biopsy showed granulomata and grew M. lentifl avum. Urine and fecal samples were negative for any mycobacteria. She did not produce any sputum. Chest radiograph showed extensive miliary nodules, and computed tomography  (CT) showed peribronchial thickening and bronchiolitis but no lymphadenopathy. The patient was empirically given isoniazid, rifampicin, pyrazinamide, clarithromycin, and ethambutol. Oral prednisone (25 mg 1×/d) improved symptoms and liver biochemistry and decreased splenic size. She was discharged on prednisone (15 mg 1×/d), isoniazid (300 mg 1×/d), ethambutol (400 mg 2×/d), and clarithromycin (500 mg 2×/d). Her organomegaly improved over the next 6 months. The intravenous port was removed. Ten months later, she remained well and compliant with treatment.

Case 2: Chronic Pulmonary Nodules and Bronchiectasis
In December 2007, an 85-year-old woman sought care for lobar pneumonia. She had never smoked and had no previous lung disease or immunosuppression. At follow-up after discharge from the hospital, she was lethargic with a persistent cough but no weight loss or fever. CT of her thorax confi rmed bilateral well-defi ned nodules up to 1 cm in diameter. Bronchoscopic washings grew mycobacteria, but the organism could not be speciated. Results of a percutaneous nodule biopsy were nondiagnostic. Surgical biopsy of the right lung found caseating granulomata, but culture was negative. At 7 months follow-up, a CT scan of her thorax showed no change in the nodules, but mild bronchiectasis had developed. Bronchoscopic lavage grew M. lentifl avum for the fi rst time. In February 2009, she began ethambutol (800 mg), rifampin (450 mg 1×/d), and clarithromycin (500 mg 2×/d). Her symptoms improved, and she completed 18 months of treatment. Bronchoscopic washings posttreatment were ZN stain and AFB culture negative.

Case 3: Bronchiectasis
A 49-year-old Taiwanese woman who had never smoked sought care in 1998 for hemoptysis. She had moved to Australia 5 years earlier. Thoracic CT showed a right middle lobe infi ltrate. Three sputum samples were culture negative for AFB. Transbronchial lung biopsy samples showed peribronchial granulomata but were culture negative. She received empirical quadruple therapy for tuberculosis. The cough continued but without hemoptysis. In 2004, a chest radiograph showed middle lobe and lingular bronchiectasis. Three sputum samples were AFB culture negative. Bronchoscopic washings were ZN negative but grew M. lentifl avum, thought to represent colonization. In 2007, an unspeciated NTM grew on 1 of 3 sputum specimens. By January 2009, the patient was well, with no exacerbations in the previous year and stable radiographic appearance.

Case 4: Cervical Lymphadenitis
A 20-month-old girl was examined for a 4-week history of bilateral cervical lymphadenopathy. The largest node (20 × 24 mm) was excised. Necrotizing granulomata were seen. M. lentifl avum was cultured. No antimycobacterial therapy was administered; she recovered fully. examination, but cytologic examination of a lymph node aspirate from the child showed lymphocytes, macrophages, neutrophils, and fragments of epithelioid histiocytes but no well-formed granulomas. From 2 other patients (35-yearold woman, chronic leg ulcer; 59-year-old woman, post thyroidectomy wound abscess), M. lentifl avum without S. aureus were cultured; the patients were treated with wound debridement and fl ucloxacillin. Biopsy samples showed no granulomata.

Nonsignifi cant Isolates
Most other isolates were cultured from respiratory samples. One isolate each was grown from ascitic fl uid and blood. Three patients with cystic fi brosis (2 with mild disease, 1 lung transplant recipient) had 1 or 2 isolates each but no evidence of disease.

Strain Types
DiversiLab patterns were grouped into 7 rep-PCR profi les, A-G ( Figure 3). The 8 clinical isolates of profi le A showed 97%-99% similarity. This profi le included 2 clinically signifi cant isolates (cases 1 and 3) and 6 nonsignifi cant isolates (3 respiratory samples, 2 soft tissue samples, and 1 ascites sample). Two further pulmonary isolates (profi les A1 and A2) were ≈90% similar to the profi le A isolates. The isolate from case-patient 2 was contaminated and could not be analyzed. The isolate from case-patient 4 (profi le B) had 94% similarity to a nonsignifi cant isolate from soft tissue. These 2 isolates were from patients who lived 1,800 km apart.
Profi le D comprised a pair of nonsignifi cant pulmonary isolates of 97% similarity. These isolates came from patients who lived within 80 km of each other, 450 km north of Brisbane. Profi les C and E were nonsignifi cant isolates and distinct from other rep-PCR profi les.
Five water sample isolates (profi le A3) had 97%-99% similarity and shared 90% similarity with the clinical isolates of profi les A, A1, and A2. The other 2 water isolates (profi les F and G) were distinct from all other clinical and water isolates.

Global Case Reports
In 30 cases of clinically signifi cant disease published in English (online Appendix Table, www.cdc.gov/EID/ content/17/3/395-appT.htm), disease spectrum varied from cervical lymphadenitis (8 of 9 cases in children) to acute or chronic disease usually affecting lungs and pleura (infi ltrates, cavities, nodules, effusions) but also arthritis/ discitis, bone lesions, skin ulcers, and hepatosplenomegaly. The rapid onset of cervical lymphadenitis has been noted in many reports, usually with an excellent outcome from excision alone. The mean age of adults with nonlymphadenitis disease (20 cases) was 56 years (range 23-87 years), and they were evenly split between the sexes. Eleven case-patients had associated immunocompromise. Eleven were reportedly stable or improved at follow-up, 6 died, and 3 had uncertain outcomes.

Discussion
M. lentifl avum disease can be diffi cult to diagnose, as the cases in this report exemplify. The clinical information we gathered was largely retrospective, which poses certain limitations; however, the case-patients 1 and 2 were current patients undergoing active treatment at the time of writing (June 2009). M. lentifl avum was isolated occasionally from patients colonized or undergoing treatment for MAC and in patients with S. aureus soft tissue infections. Certainly in some patients multiple NTM can grow at the same or different times, and M. lentifl avum may be no different in this respect. M. lentifl avum has been cultured from sputum containing MAC and from sputum containing M. tuberculosis, but these cases may represent colonization/contamination rather than infection (23,28). Concurrent isolation of M. lentifl avum and S. aureus, which probably represents contamination or colonization, has not been reported as far as we are aware. Co-infection of S. aureus and M. tuberculosis has been reported, possibly as superimposed staphylococcal infection in tuberculous tissue (29,30). Although no samples were taken for histology, cytologic examination of lymph node aspirate from the 2-yearold child with lymphadenitis is intriguing because the infl ammatory cells were predominantly lymphocytes/ macrophages with epithelioid histiocytes. Treatment using fl ucloxacillin with or without drainage affected a complete cure in all cases. M. lentifl avum is a rare isolate and an unusual cause of disease in humans. As with other NTM, it can be isolated from contaminated samples: clinical signifi cance should be assessed before any treatment is considered (2). In 2005, of 488 patients with pulmonary NTM isolates in Queensland, only 26.6% were considered to have clinically signifi cant disease (31). The proportion was higher for M. intracellulare (39.4%), M. avium (33.3%), and M. kansasii (52.6%) and much lower for species traditionally thought to be more likely contaminants, e.g., M. gordonae (11.1%). In our study, isolates for 4 (11%) of 36 were clinically signifi cant, similar to published estimates of 10%-21% (16,22). This proportion may be an underestimate given that we could not determine clinical signifi cance in 4 patients. Worldwide, M. lentifl avum may be underreported and incorrectly identifi ed as other, more familiar species, especially if access to molecular identifi cation is limited.
M. lentifl avum has been isolated from water distribution samples. Torvinen et al. isolated NTM from up to 80% of sites across Finland (32); M. lentifl avum was the second most common species (38% of sites). Laboratory isolation of M. lentifl avum from clinical specimens in Finland has increased independently of speciation methods, but details of patients with disease are lacking (33). In South Korea, Lee et al. found mycobacteria in 26% of 84 drinking water sites. Sixty-fi ve percent of isolates were M. lentifl avum (34). In our study, mycobacteria were isolated from 70% of sites, but M. lentifl avum from only 6.3%. The diffi culties in isolating mycobacteria from potable water are well recognized and relate to mycobacterial growth characteristics and the need for specimen decontamination to reduce bacterial and fungal overgrowth. Decontamination reduces mycobacterial yields; hence, the prevalence of mycobacteria in potable water samples is believed to substantially underestimate the true fi gure (8). Culture-based techniques may be less sensitive than direct PCR. However, detecting mycobacterial DNA does not necessarily prove the presence of viable organisms that are able to cause infection; detection of M. lentifl avum by culture-based methods is noteworthy with respect to human health. Case-patient 1 had long-term intravenous access, which may have allowed direct exposure to contaminated water through illicit drug administration. In this report, we have geographically associated culture-positive water samples and clinical disease.
DiversiLab strain typing showed that profi les A and A3 were most prevalent among clinical and water isolates and shared ≈90% similarity. The criteria for interpreting rep-PCR typing results have been established for some mycobacterial species. For example, Cangelosi et al. found high concordance between restriction fragment-length polymorphism and rep-PCR, reporting 93% similarity as the cutoff value for clustered M. tuberculosis isolates and 92% for M. avium (6). The analysis of M. abscessus by Zelazny et al., the largest study of rep-PCR in NTM, used rep-PCR to successfully cluster M. abscessus strains that were clonally related by PFGE analysis (35). Four of the water samples constituting profi le A3 and 1 unrelated strain (profi le G) came from sites that shared a groundwater source. These fi ndings suggest a dominant environmental strain closely related (90%), but not identical, to strains found in human specimens and as a cause of human disease. The theory of dominant local environmental strains is supported by the fi nding of a different strain type from 2 patients living near each other but 450 km from Brisbane (profi le D).
Profi le A contained clinically signifi cant and nonsignifi cant isolates. Profi le B also contained a pair of highly similar isolates (94%) of which 1 was clinically signifi cant. Although the residential addresses of these patients were 1,800 km apart, nothing is known about the duration of residence or travel or work habits of these case-patients. Thus, the infection may not have been acquired locally. Conclusions cannot be drawn about the pathogenicity of different strains; a larger study is required to address this question.
The fi nding of different, less common strain types (profi les E, F, G) confi rms the validity of using automated rep-PCR (DiversiLab) as a tool for strain typing this species. Variation in M. lentifl avum strain type has been demonstrated. Buijtels et al. (20) reported 55 M. lentifl avum isolates from 149 specimens obtained from 38 patients at 1 hospital in Zambia. Illness of 2 patients defi nitely fulfi lled ATS/IDSA criteria for signifi cant disease. Because this species is a rare cause of disease, the authors performed molecular identifi cation on a subset of 12 isolates to investigate the possibility of laboratory contamination. Six strain types were identifi ed; the Zambian strains clearly differed from comparator Dutch strains. The fi nding of a dominant strain probably represented the local endemic strain, but laboratory or point-of-collection contamination could not be entirely excluded. Because isolates in our study came from multiple laboratories statewide at different times, contamination is unlikely to explain their presence in multiple clinical specimens.
The optimal treatment for M. lentifl avum disease is not established; a wide variety of regimens has been used in previous case series. Although evidence does not support the use of any specifi c regimen, we achieved symptomatic and radiologic improvement in case-patients 1 and 2 with rifampicin/ethambutol/clarithromycin at 12 months and isoniazid/ethambutol/clarithromycin at 18 months, respectively. More detailed reporting of treatment regimes and outcomes will help establish optimum therapy.