NTM patients were recruited to participate in studies of their household water systems through the auspices of Nontuberculous Mycobacteria Research and Information, Inc. Informed consent was obtained from each participating patient, and the study was reviewed by the Virginia Tech Institutional Research Board and granted exempt status. NTM isolates from the patients, if possible, were obtained through collaborating physicians and mycobacteriology laboratories. In some instances multiple patient isolates of different species were found. A questionnaire was provided to each patient to obtain information about the household plumbing.

Sterile containers and swabs were sent to each collaborating patient household. Directions for collection of hot and cold water samples (500 mL) and biofilms/sediment from water taps and showerheads by using swabs were provided. If the patient thought that infection might have occurred as a result of exposure to soil, soil samples were collected. In some cases filters (fiber, activated charcoal, and reverse osmosis) were collected. All samples were returned at ambient temperatures by express courier service to the Mycobacteriology Laboratory in the Department of Biological Sciences at Virginia Tech.

Mycobacteria in water and swab (taps and filters) samples were counted and isolated as described (27). Soil samples were processed as described (25). Most acid-fast colonies picked for identification and enumeration were small (1-mm diameter after 14 days at 37°C), unpigmented to yellow, and resembled either the transparent or opaque types previously reported (17). Acid-fast isolates were identified by nested PCR of the 16S rRNA gene (31) and PCR amplification and analysis of restriction endonuclease digestion fragments of the heat-shock protein 65 (hsp65) gene (32).

In those instances in which the Mycobacterium species from the patient and household water system isolates were the same, isolates were fingerprinted by repetitive sequence-based PCR (rep-PCR) (33). Matches were confirmed by use of GelCompar II software (Applied Maths, Inc., Austin, TX, USA).

Samples for NTM isolation were received from 31 collaborating patients throughout the United States and Canada: Arizona, California, Colorado, Connecticut, Florida, Georgia, Michigan, New Jersey, New York, Pennsylvania, Texas, Vermont, Virginia, and Wisconsin, USA; and Ontario, Canada. Six patients each had 2 residences and sent samples from each residence.

The isolates from the 31 patients with NTM infection included M. avium (9), M. intracellulare (6), MAC (11), M. abscessus (4), and M. xenopi (1). Isolates could not be obtained from 11 patients, thus preventing rep-PCR fingerprinting even in those instances where household isolates belonged to the same species. Thus, the total number of patient isolates available for fingerprinting was only 20. All putative Mycobacterium spp. isolates recovered from samples were identified, and 45% of NTM-positive households (10/22) and 1.5% of NTM-positive samples (6/394) yielded >1 NTM species (Table 1). The average number of different NTM species per household was 1.9 (range 1–5 NTM species/household). In those instances where the Mycobacterium species of the patient and their household plumbing isolates were the same (e.g., M. avium), all isolates belonging to the same species as the patient were subject to rep-PCR fingerprinting. Household isolates included M. avium (10), M. intracellulare (10), M. malmoense (5), M. szulgai (3), M. chelonae (2), M. gordonae (6), and 1 each of M. scrofulaceum, M. terrae, and M. trivale (Table 1). Samples were coded with the first 2 or 3 letters representing each patient, a letter representing sample type (W, water; Sw, swab [biofilm]; S, soil), a number for sample number from a household collection, and the final number for the isolate from the sample; thus, ML-W-6-2 is the second water isolate from the sixth sample collected from patient ML’s household.

NTM were isolated from water, biofilm, filter, or soil samples from 22 (59%) households sampled and from 109 (28%) of 394 samples. There was a positive correlation between the number of samples collected per household and the number of NTM-positive samples (r = 0.4581). In 8 households >50% of the samples yielded NTM, and in 7 households no NTM were isolated. Seventeen of the 37 household sample collections had at least 1 sample that yielded an NTM isolate that belonged to the same species as that of the patient. Among those 17 households, at least 1 NTM isolate from 7 households exhibited the same rep-PCR fingerprint as that of the patient. Specifically, the Figure illustrates matching rep-PCR band patterns of patient isolate ML-P-1 (lane 3) and shower water isolate ML-W-6-2 (lane 4) from the patient’s home and patient isolate TC-P-1 (lane 10) and tap water isolate TC-W-2-2 (lane 12) from the patient’s home. Matches were confirmed by use of GelCompar II software (Applied Maths, Inc.). Furthermore, the Figure also illustrates the relative similarity in rep-PCR band patterns of patients and their household isolates and the wide differences between isolates of different patients (compare lanes 3–4, lanes 7–8, and lanes 10–12). On the basis of diversity of band patterns and the number of bands (7–14 bands), the results confirm the discriminatory power of rep-PCR fingerprinting (32). The percentage of fingerprint matches may be an underestimate because patient isolates could not be obtained for 11/31 patients, all of whom had MAC infections.

The frequency of NTM recovery from water (47/195, 24%), biofilm (46/165, 28%), filters (4/12, 33%), and soil samples (3/17, 18%) did not differ markedly. The highest numbers of NTM, as CFUs, were recovered from biofilms (10,371 CFU/cm²), with lower numbers from filters (1,987 CFU/cm²), soils (1,500 CFU/g), and water (157 CFU/mL). Most biofilm samples were collected by swabbing either the inside of a water tap or showerhead with a sterile swab that was immediately placed in 2 mL sterile tap water. Because the samples were shipped immediately after collection, there was little opportunity for the NTM numbers to change.

Review of the responses to the NTM patient questionnaire led to identification of 2 factors that seemed to influence NTM in household samples. Households with water heater temperatures <125°C (50°C) were more likely to yield NTM (17/20, 85%) compared with households in which water temperature was >130°F (55°C) (6/15, 40%) (Table 2). That difference was significant (p = 0.0107; relative risk 2.125, by Fisher exact test). Although households with water from a public or private water system were more likely to have NTM (19/27, 70%) compared with households with water from a well (5/12, 42%) that difference was not significant (p = 0.1532; relative risk 1.689 by Fisher exact test) (Table 3).