Mycobacteria in Water and Loose Deposits of Drinking Water Distribution Systems in Finland

Abstract

Drinking water distribution systems were analyzed for viable counts of mycobacteria by sampling water from waterworks and in different parts of the systems. In addition, loose deposits collected during mechanical cleaning of the main pipelines were similarly analyzed. The study covered 16 systems at eight localities in Finland. In an experimental study, mycobacterial colonization of biofilms on polyvinyl chloride tubes in a system was studied. The isolation frequency of mycobacteria increased from 35% at the waterworks to 80% in the system, and the number of mycobacteria in the positive samples increased from 15 to 140 CFU/liter, respectively. Mycobacteria were isolated from all 11 deposits with an accumulation time of tens of years and from all 4 deposits which had accumulated during a 1-year follow-up time. The numbers of mycobacteria were high in both old and young deposits (medians, 1.8 × 10⁵ and 3.9 × 10⁵ CFU/g [dry weight], respectively). Both water and deposit samples yielded the highest numbers of mycobacteria in the systems using surface water and applying ozonation as an intermediate treatment or posttreatment. The number and growth of mycobacteria in system waters correlated strongly with the concentration of assimilable organic carbon in the water leaving the waterworks. The densities of mycobacteria in the developing biofilms were highest at the distal sites of the systems. Over 90% of the mycobacteria isolated from water and deposits belonged to Mycobacterium lentiflavum, M. tusciae, M. gordonae, and a previously unclassified group of mycobacteria. Our results indicate that drinking water systems may be a source for recently discovered new mycobacterial species.

Environmental mycobacteria are common heterotrophic bacteria in soils and natural waters. Some species may cause infections similar to tuberculosis in humans and animals. They pose a particular risk of severe infections in individuals with immunocompromising conditions. Compared to most other bacterial species, mycobacteria are exceptionally resistant to disinfection with chemicals such as chlorine (2, 24, 33, 46). Thus, they may survive chemical treatments at waterworks and enter the distributed water and finally tap water, where their occurrence has been known since the beginning of the 20th century (reviewed by Collins et al. [3]). It is possible that mycobacteria replicate at least in warm parts of the distribution systems, like taps and showers. Some thermotolerant species, including the important mycobacterial pathogens Mycobacterium avium complex (MAC) and M. xenopi, tolerate heating and may survive even in hot water (5, 22, 38). Due to the expanding immunocompromised population, water distribution systems have recently gained increasing attention in studies of mycobacterial reservoirs. According to the knowledge obtained, complex water systems in large buildings like hospitals are important sources of mycobacterial infections (1, 21, 53, 55), but infections have also been reported following exposure to simpler water systems, such as hot tubs (28, 56).

During the past few decades, it has been realized that biofilms formed on pipeline surfaces have a major impact on the microbial quality of system waters. Also, potentially pathogenic microbes may grow or survive in biofilms, which offer suitable nutrient conditions and protection against disinfection (34, 45). Biofilm formation by mycobacteria and the occurrence of mycobacteria in biofilms have also gained interest in recent years (13, 14, 39, 40). However, most studies have been conducted at temperatures higher than those normally prevailing in distribution system (13, 14, 39). Also, the occurrence of mycobacteria in real system deposits, a potential site of mycobacterial growth in water systems, and the factors affecting the growth of mycobacteria in water systems have received less attention (6, 40).

In the present study, we focused on drinking water systems in a boreal region where mycobacteria are common in natural waters and soils (16, 17). We studied the occurrence and species of mycobacteria in loose deposits collected from old system pipelines before and after their mechanical cleaning and examined the association between the occurrence of mycobacteria and water and deposit characteristics. We further studied the growth of mycobacteria in different parts of a distribution system using an experimental device for biofilm formation.

MATERIALS AND METHODS

Study sites.

The study sites covered 16 drinking water system areas in eight localities in Finland. Ten of the systems used surface water, four used ground water, and two mixed surface and ground water as the raw water source. All surface waterworks applied chemical coagulation and rapid sand filtration combined with disinfection with chlorine or chloramine (Table 1). Granular activated-carbon filtration and ozonation, as an intermediate oxidizing agent or posttreatment disinfectant, were added to the process in some waterworks using surface water. Two waterworks using ground water used KMnO₄ oxidation combined with flotation or rapid sand filtration. In one waterworks using ground water, rapid sand filtration was combined with pH adjustment and chlorination. One waterworks using ground water applied slow sand filtration, and one waterworks using ground water had no treatment for drinking water. Pipelines were made of polyethylene (PEH) or cast iron.

TABLE 1.

Characteristics of study sites

Study site	Raw water	Treatmenta	Disinfectionb	Piping material	No. of water samples studied	No. of deposit samples studiedc
1A	Surface water	Coagulation + RSF, SSF	Chloramination	PEH	6	1
1Bd	Surface water	Coagulation + RSF, SSF	Chloramination	PEH	4	2
2Ae	Surface water	Coagulation + RSF, GAC	Chlorination	Cast iron	5	1
	Ground water	RSF, pH adjustment	Chlorination
2Be	Surface water	Coagulation + RSF, GAC	Chlorination	Cast iron	0	2
	Ground water	RSF, pH adjustment	Chlorination
2C	Surface water	Coagulation + RSF, GAC	Chlorination	Cast iron	4	0
3A	Ground water	Liming, KMnO4 oxidation + coagulation, limestone filtration		Cast iron	3	1
3B	Ground water	KMnO4 oxidation, RSF		Cast iron	3	1
3C	Surface water	Coagulation + RSF, GAC	Chlorination	Cast iron	3	0
4A	Surface water	Coagulation + RSF, O3, GAC	UV, chlorination	PEH	3	2
4B	Surface water	Coagulation + RSF, O3, GAC	Chloramination	PEH	4	0
5A	Surface water	Coagulation + RSF	O3, chloramination	Cast iron	5	0
5Be	Surface water	Coagulation + RSF	O3, chloramination	Cast iron	7	1
	Surface water	Coagulation + RSF	O3, chloramination
6	Surface water	Coagulation + RSF	O3, chloramination	Cast iron	5	1
7	Surface water	Coagulation + RSF	O3, chloramination	Cast iron	5	0
8A	Ground water	No treatment		PEH	2	1
8B	Ground water	SSF		PEH	2	2

^aCoagulation, flotation or sedimentation; RSF, rapid sand filtration; SSF, slow sand filtration; GAC, granular activated-carbon filtration; O₃, ozonation.

^bUV, UV disinfection.

^cThe two samples from the same study site were collected in consecutive years.

^dThe biofilm formation study was carried out at this site.

^eTwo waterworks delivered water to the same distribution system.

Sampling of drinking water and soft deposits.

Water samples were collected from one or two waterworks in each locality (n = 17) and from one to five sites from the later parts of the distribution systems (n = 44). The water samples were collected in sterile 1-liter polyethylene bottles from water pumping stations, water meters, or houses close to the main line after the water was allowed to run for several minutes. The samples were transported to the laboratory in cool bags, stored at 4°C, and analyzed within 48 h of being sampled.

Within 2 weeks after these samplings, 11 distribution systems were cleaned mechanically by swabs (pipeline internal gauging), and 4 of those 11 were cleaned again 1 year after the first cleaning. The polyurethane swab was pushed into the pipeline by water pressure, and the released deposit was collected for microbiological and chemical analyses at the end of the studied line via an opened fire hydrant. The lengths of the cleaned pipeline sections varied from 300 to 4,000 m. The age of the piping was not known for every site, but it varied from at least 11 to 44 years. To evaluate the efficiency of the cleaning at the 11 sites, two sets of control water samples were collected from the waterworks and systems 2 weeks and ∼1 year after the first cleaning.

Mycobacterial colonization of surfaces.

An experimental study of the colonization of surfaces by mycobacteria was performed in one of the distribution systems (Table 1, site 1B). Colonization was followed at the waterworks and at three sites in the distribution system, where the water retention times were 10, 39, and 141 h. In the studied system, the temperature of the water leaving the waterworks was between 11 and 23°C (measured from July to September), the pH was between 7.5 and 8.2, and the concentration of assimilable organic carbon (AOC) was 23 μg/liter, the concentration of nonpurgeable organic carbon (NPOC) was 1.7 to 2.4 mg/liter, the concentration of oxygen was 5.8 to 10 mg/liter, the concentration of free chlorine was <0.03 to 0.30 mg/liter, the concentration of total chlorine was 0.21 to 0.40 mg/liter, and the concentration of heterotrophic bacteria was <10 to 55 CFU/ml.

The biofilm sampler (57) consisted of a series of polyvinyl chloride (PVC) tubes (inner diameter, 10 mm; length, 200 mm; area, 62.8 cm²) linked to each other by stainless steel loops and valves. The sampler was connected to a cold-water tap, and water was allowed to run through at a flow rate of 1 liter/min. At each sampling, starting on day 10 and ending 40 weeks after the installation, the last tube filled with drinking water was closed with valves and disconnected for analysis.

Physical, chemical, and microbiological analyses.

The data on routine physical and chemical parameters of the drinking water were measured at the waterworks according to the Finnish standards (27) and are presented in Table 2. The concentration of AOC in the water was analyzed with the addition of nutrients, i.e., as AOC_potential (31, 52). The concentration of NPOC was determined with a Shimadzu (Kyoto, Japan) 5000 TOC analyzer from samples acidified with phosphorus acid and purged in a nitrogen flow for 5 min (7).

TABLE 2.

Physical, chemical, and microbiological characteristics of water and deposit samples collected from different sites of drinking water distribution systems^a

Parameter	n	Minimum	Maximum	Mean	Median
Water
Heterotrophic bacteria (R2A, 7 days) (CFU/liter)	44	<1.0 × 104	7.7 × 106	1.3 × 106	5.0 × 105
Turbidity (FTU)b	39	0.06	1.2	0.29	0.16
Water retention time (h)	38	1	140	27	18
Temp (°C)	38	5.2	19	12	12
KMnO4 no. (mg/liter)	30	2.8	15	6.8	6.4
Alkalinity (mmol/liter)	39	0.63	2.7	0.98	0.95
pH	39	7.4	8.8	8.2	8.2
NPOC (mg/liter)	44	0.5	3.8	2.3	2.4
AOC (μg/liter)	42	38	350	170	150
Iron (mg/liter)	30	<0.005	0.44	0.08	0.04
Manganese (mg/liter)	17	<0.005	0.02	0.006	0.003
Calcium (mg/liter)	25	13	35	23	22
Oxygen (mg/liter)	39	1.3	12	9.7	11
Total N (μg/liter)	19	75	880	400	470
NO3 N (μg/liter)	16	<5	2,500	370	250
Free chlorine (mg/liter)	33	<0.03	0.19	0.04	0.02
Total chlorine (mg/liter)	33	<0.03	0.6	0.18	0.12
Deposit
Heterotrophic bacteria (R2A, 7 days) (CFU/g)	11	1.8 × 105	6.6 × 109	8.3 × 108	2.2 × 108
Total carbon (%)	11	0.6	17	8.6	7.8
Total nitrogen (%)	11	0.3	1.5	0.8	0.7
Iron (%)	11	4.6	38	17	15
Manganese (%)	11	0.7	46	11	8.5
Calcium (%)	11	1.5	15	4.2	3.2
Al (%)	11	0.1	8.7	2.4	1.7
% dry matter in water sediment slurry (g/liter)	11	0.14	150	19	3.5

^aResults for the samples collected at waterworks are not included.

^bFTU, formazin turbidity units, based on nephelometric determination.

The contents of total C, total N, and metals in the 11 old deposit samples were analyzed as described in detail previously (27). The viable counts of heterotrophic bacteria in the water samples and deposits were determined by the spread plate method on R2A agar by incubation at 20°C for 7 days (36). The moisture contents of the deposit samples were determined by drying 30 ml of each sample at 65°C for 20 h. The microbial and chemical results are expressed per gram (dry weight) of deposit.

Mycobacterial analyses.

For cultivation of mycobacteria, 1 liter of sampled water was concentrated by membrane filtration and microbes were eluted from the filter pieces to the sample water as described previously (15). Fifty microliters of the concentrate was directly inoculated into two parallel slopes of egg medium supplemented with glycerol (pH 6.3; medium i) and Na-pyruvate (pH 6.3; medium ii) (19, 20). The rest of the 5-ml concentrate was decontaminated by the modified method of Schulze-Röbbecke et al. (41), in which the sample was treated for 5 min at room temperature with cetylpyridinium chloride (CPC) to reach the final concentration of 0.001% (wt/vol). After 15 min of centrifugation (8,600 × g; 4°C) (Sorvall RC-5C; Du Pont Company, Wilmington, Del.), 30 ml of sterile deionized water was mixed with the sediment, and it was centrifuged again as described previously. The sediment obtained was resuspended in 400 μl of sterile deionized water (SDW), and 50 μl was inoculated into two parallel slopes of media i and ii and into two parallel slopes of these egg media adjusted to pH 5.5 (19, 20). One set of the four parallel egg media was incubated at 30°C and another was incubated at 35°C for 3 months.

The 5- to 15-ml soft-deposit samples were decontaminated for 5 min at room temperature with CPC to make a concentration of 0.005% (wt/vol). The final sediment was resuspended in 1,400 μl of SDW, and additional dilutions of 10⁻¹ and 10⁻² were made in SDW for cultivation. Otherwise, the culture procedure was similar to that used for water samples. In the deposit samples, the number of mycobacteria was expressed per gram of dry weight.

From the PVC tubes of the biofilm sampler, the biofilms accumulated were detached by adding sterile glass beads (diameter, 2 mm) to the tubes and shaking them in a rotary mixer (Vortex Genie 2; Scientific Industries, Inc., Bohemia, N.Y.) for 20 min. The contents were poured into a sterile measuring cylinder. The PVC tube was rinsed with SDW, which was combined with the initial content, and the total volume was measured. Mycobacteria were cultured from a 5-ml portion using the 0.005% CPC described above for the soft deposits. After the final centrifugation, 300 μl of the SDW was added, and dilutions of 10⁻¹ and 10⁻² were made in SDW. Fifty microliters of the resuspended sediment and the dilutions was inoculated into one slope of each of the four egg media described above. The bacteria were grown at 30°C for 3 months.

Each colony morphology type recovered on each of the media used was examined for acid-fastness by Ziehl-Neelsen staining. The same applied to water, deposit, and biofilm samples. Acid-fast isolates recovered at 35°C from water and deposit samples were studied further in more detail. This was done by selecting representatives of every colony type on each medium and identifying them to species level by gas-liquid chromatography-mass spectrometry (Hewlett-Packard, Palo Alto, Calif.) analyses of cellular fatty acids and alcohols (48). These analyses were complemented by tests for urease, 10-day arylsulfatase, pyrazinamidase, semiquantitative catalase, nitrate reduction, and Tween 80 hydrolysis and for growth characteristics (growth at 22, 30, 35, and 42°C; growth rate; and pigment production) performed as described previously (48). If these analyses did not provide species identification, the 16S ribosomal DNA was sequenced as described previously (44).

Statistical analyses.

The highest count of viable mycobacteria recovered after decontamination with CPC and incubation at 30°C on any medium was used for statistical analyses. Due to the nonnormal distribution of the variables, the results were analyzed by nonparametric tests (SPSS for Windows version 9.0.1.) The relationships between microbial, chemical, and physical parameters in the system waters and deposits were studied by Spearman rank correlation analysis. The Mann-Whitney U test or the Kruskal-Wallis test was used to compare microbiological differences among different water types, water treatments, and piping materials. To study the association between the number of mycobacteria and the AOC content, the growth value proposed by Falkinham et al. (6) was calculated for each system by dividing the median numbers of mycobacteria at system sites by the number of mycobacteria at the waterworks.

Nucleotide sequence accession number.

The sequence data of the isolates have been submitted to the EMBL nucleotide sequence database under accession number AJ550515.

RESULTS

Mycobacteria in water samples.

In the water samples taken at the waterworks before the pipelines were cleaned, the isolation frequency of mycobacteria grown at 30°C was 35% (Fig. 1A). At the two samplings from the waterworks after the pipelines were cleaned, the isolation frequencies were 31 and 8% (mean of the three samplings, 26%). The isolation frequency increased with the distance from the waterworks; it was 80% at the most distal sites, where water arrived 4 to 141 h after leaving the waterworks (Fig. 1A). The median number of mycobacteria in the positive samples was low, but it increased by 1 order of magnitude from the waterworks (15 CFU/liter; range, 10 to 30 CFU/liter) to the distal sites (140 CFU/liter; range, 10 to 3,500 CFU/liter) (Fig. 1B). The increase in the numbers occurred in 13 of the 15 systems studied. In the remaining two systems, the mycobacterial numbers at the waterworks and in the systems were equal. In all, 14 of the 15 systems studied were positive for mycobacteria. Cultivation at 35°C resulted in lower isolation frequencies and lower colony counts than at 30°C (Fig. 1).

FIG. 1.

Occurrence (A) and numbers (B) of mycobacteria in positive samples of water and deposits.

The mycobacterial numbers were higher in the systems using surface water and applying ozonation as an intermediate treatment or posttreatment than in those distributing nonozonated surface water, ground water, or mixed water (medians, 110, 15, and 10 CFU/liter; n = 22, 13, and 9, respectively; P < 0.01). The number of mycobacteria in the system and the growth value correlated positively with the AOC concentration of the drinking water leaving the waterworks (Fig. 2 and Table 3). In addition, the number of mycobacteria recovered correlated positively with the number of other heterotrophic bacteria, the water retention time, and the Fe content and negatively with the content of free available chlorine (Table 3). The number of mycobacteria was not associated with the chlorination method used, the use of granular activated carbon filtration, or the piping material.

FIG. 2.

Association between number (A) and growth value (B) of mycobacteria and AOC concentration in water.

TABLE 3.

Correlation coefficients between numbers of mycobacteria^a in water samples and microbiological, physical, and chemical water parameters

Parameter	n	rb	P
Heterotrophic bacteria (R2A, 7 days)	44	0.47	<0.01
Water retention time	38	0.35	<0.05
Turbidity	39	0.01
Temp	38	−0.17
pH	39	−0.13
Alkalinity	39	−0.21
KMnO4 no.	30	−0.03
NPOC	44	0.01
AOC (at waterworks)	39	0.51	<0.01
Concn
Fe	30	0.60	<0.001
Mn	17	−0.13
Ca	25	0.26
Total N	19	−0.23
NO3 N	16	−0.09
Oxygen	39	0.25
Free chlorine	33	−0.36	<0.05
Total chlorine	33	−0.34

^aMycobacteria incubated at 30°C were used for calculations.

^bSpearman rank correlation coefficient.

The effect of mechanical cleaning on the occurrence of mycobacteria was evaluated by comparing the numbers of bacteria in the water sampled at sites downstream (i.e., after cleaning) and upstream (i.e., before cleaning) of the cleaned pipeline sections. A few weeks after the mechanical cleaning, the numbers of mycobacteria downstream of the cleaning area were higher than the numbers at the same sites before cleaning. In contrast, the numbers upstream of the cleaning area were lower than before cleaning (Fig. 3). A year after the cleaning process, the mycobacterial numbers at the downstream and upstream sites were similar (Fig. 3).

FIG. 3.

Number of mycobacteria in water samples before and after mechanical cleaning of pipelines.

Mycobacteria in deposit samples.

Mycobacteria were present in all 11 old deposits and 4 1-year-old deposits sampled from the pipelines. Their numbers (median, 1.8 × 10⁵ CFU/g; range, 4.9 × 10² to 4.2 × 10⁶ CFU/g) were several log units higher than those in the water samples (Fig. 1). The numbers of mycobacteria in the four 1-year-old deposits were only slightly lower (median, 3.9 × 10⁵ CFU/g; range, 1.0 × 10⁵ to 2.1 × 10⁶ CFU/g) than the numbers in the old deposits (median, 7.0 × 10⁵ CFU/g; range, 4.9 × 10² to 1.4 × 10⁶ CFU/g) accumulated over decades at the same four sites. Similar to the results for water, mycobacterial numbers at 35°C were lower than those at 30°C (Fig. 1). In the new deposits, mycobacteria grown at 35°C were present in two of the four samples analyzed (1.0 × 10⁵ and 4.7 × 10⁵ CFU/g).

The numbers of mycobacteria in deposits correlated positively with the numbers of other heterotrophic bacteria (R2A agar, 7 days) (r = 0.64; P < 0.05; n = 11) and negatively with Mn (r = −0.64; P < 0.05; n = 11) and Sr (r = −0.71; P < 0.05; n = 11). Similar to the results for water, the median numbers of mycobacteria in the deposits were highest in the systems using surface water and applying ozonation as an intermediate treatment or posttreatment (medians, 1.4 × 10⁶, 7.3 × 10⁵, and 5.1 × 10⁴ CFU/g in the systems using ozonated surface water, nonozonated surface water, and ground or mixed water; n = 3, 2, and 6, respectively; P < 0.05). No differences in the numbers of mycobacteria were detected between chlorination methods, use of granular activated carbon filtration, or piping materials.

Mycobacterial colonization of surfaces.

Mycobacteria emerged in the developing biofilms at the three network sites of the studied distribution system during the first 6 weeks and reached maximum counts by 10 weeks (Fig. 4A). Colonization of surfaces was most rapid and the numbers were highest at the most distal site, where water arrived 141 h (5.8 days) after leaving the waterworks. The kinetics of mycobacterial colonization of surfaces was similar to, but slower than, that of other heterotrophic bacteria, which formed biofilms at every site in 10 days (Fig. 4B).

FIG. 4.

Numbers of mycobacteria (A) and other heterotrophic bacteria (B) in biofilms developed in PVC tubes at different sites.

Mycobacterial species.

Based on the polyphasic taxonomic identification methodology applied, 225 of the 351 isolates (64%) recovered from the water and deposit samples belonged to three established species, M. lentiflavum, M. tusciae, and M. gordonae (Table 4). Of the remaining isolates, 25 (7.1%) were unidentifiable isolates representing both rapidly and slowly growing mycobacteria and 101 (29%) formed a homogeneous group of slowly growing mycobacteria (AJ550515) which could not be identified with any known species. The colonies of this new group were smooth, transparent, and pale yellow. They grew as scotochromogenic colonies at 22 and 30°C after 4 weeks of incubation but did not grow at 35°C after primary isolation. Phenotypically, they resembled M. lentiflavum. The tests for urease, 10-day arylsulfatase, pyrazinamidase, nitrate reduction, and Tween 80 hydrolysis were all negative. Gas-liquid chromatography fatty acid analysis of the isolates revealed a unique profile. The isolates had tetradecanoic (14:0), hexadecenoic (16:1), hexadecanoic (16:0), octadecenoic (18:1), octadecanoic (18:0), tetracosanoic (24:0), and haxacosanoic (26:0) acids but lacked 2-methyloctadecanoic (tuberculostearic) acid. The percentage of 24:0 was higher that that of 26:0. 16S ribosomal DNA sequencing of the isolates showed the same unique sequence. It was most closely related to M. triplex and M. lentiflavum but differed from both of them by 14 to 17 nucleotides in the whole 16S gene (9, 43).

TABLE 4.

Mycobacterial species found in drinking water and deposit samples

Species	No. of isolates (%)	Study site	Sample typea	Raw water
M. lentiflavum	192 (55)	1A	W, D	Surface
		1B	W, D	Surface
		2A	W, D	Mixed
		5Ab	W	Surface
		5B	W, D	Surface
		6	W, D	Surface
M. tusciae	22 (6.3)	4A	W, D	Surface
		6	W	Surface
		3A	D	Ground
M. gordonae	11 (3.1)	1A	W	Surface
		8Ac	W	Ground
		8B	W, D	Ground
Homogeneous group of previously unknown slowly growing mycobacteria	101 (29)	4A	W	Surface
		4Bb	W	Surface
		2A	W	Mixed
		5B	W, D	Surface
		6	W, D	Surface
		5Ab	W	Surface
		7b	W	Surface
		1B	D	Surface
Total	326 (93)d     of total 351

^aW, water; D, deposit.

^bNo deposit was collected.

^cDeposit yielded no growth at 35°C.

^dIn addition, 25 heterogeneous strains (7.1%) of unknown rapidly and slowly growing mycobacteria were isolated.

The same species were usually isolated from the deposit and water of a system (Table 4). Geographically, the most widely distributed species was the previously unclassified group of slowly growing mycobacteria (AJ550515), which was recovered from half of the systems studied. There were two 1-year-old deposits that yielded growth at 35°C. In both, the species isolated were the same as in the old deposits of the same systems, i.e., M. lentiflavum at site 1B and M. gordonae at site 8B. In addition, mycobacteria of the previously unclassified group of slowly growing mycobacteria (AJ550515) were recovered from the 1-year-old deposit of site 1B.

DISCUSSION

Our data showed that the occurrence of mycobacteria in drinking water increased from the waterworks toward the distal portion of the system. There was a >2-fold increase, to 80%, in the isolation frequency and a 10-fold increase, to 140 CFU/liter, in the numbers compared to the levels at the waterworks. The isolation frequency and the number obtained from incubations at 35°C were 38% and 10 CFU/liter, respectively. These figures cannot be compared directly to those obtained in other studies because isolation techniques differed between studies and might influence the results. Recent large studies, each covering several systems in different municipalities, have published isolation frequencies of from 15 to 39% of samples for mycobacteria recoverable at 35 to 37°C (4, 6, 23, 42). In contrast, in certain cities isolation frequencies of as high as 92 to 100% of the samples have been reported (10). The mycobacterial regrowth in distribution systems in buildings has been known and regarded as accounting for nosocomial mycobacterial infections (1, 21, 53, 55). Several studies (6, 11; R. Schulze-Robbecke, K. Behringer, W. Hadnagy, C. Hagenau, B. Ilg, R. Stiller-Winkler, and P. Harteman, Abstr. 1st World Water Congr. IWA, abstr. HRM-4, 2000), including the present study, indicate that mycobacteria also proliferate in the unheated main lines of drinking water distribution systems. In our material, the water temperatures remained below 19°C, which appeared to be high enough for the growth of mycobacterial species adapted to a cool boreal environment.

According to our results, the numbers of mycobacteria in water were higher in the systems using ozone than in those not applying this agent. Ozone is known to degrade organic matter, increasing the fractions that can be assimilated by microbes (30, 51). Also in our study, the highest AOC concentration was found at the waterworks using ozone (data not shown). We found a strong positive correlation between the number of mycobacteria and their growth value and the AOC concentration of the treated drinking water leaving the waterworks (Fig. 2). Thus, our results support the recent result reported by Falkinham and coworkers (6) concerning the association between mycobacterial growth and AOC in the system. In contrast to some previous results, there was also a positive correlation between the numbers of mycobacteria and those of other heterotrophic bacteria (4, 10). Our results suggest that mycobacterial growth in the system is also associated with substrate availability, similar to other heterotrophic bacteria. Coliform bacteria were not detected in the water samples in the present study (58), supporting previous results showing that fecal or total coliforms do not indicate the presence of mycobacteria (6, 10, 50).

Ozone can favor the growth of mycobacteria and other heterotrophic bacteria by increasing the AOC content and also by increasing the amount of microbially available phosphorus (26). In some drinking waters, including the Finnish waters, with high organic-matter contents, phosphorus has been shown to be the nutrient limiting microbial growth (29, 37). A test to analyze microbially available phosphorus was developed (25) after the present experimental work, so it was not possible to examine whether there is an association between microbially available phosphorus and the number of mycobacteria. Further studies are definitely needed to clarify the association among mycobacteria, AOC, and microbially available phosphorus in water distribution systems.

In this study, the number of mycobacteria in water also correlated with the iron content, a typical element accumulating in the Finnish systems. The iron content increased with water retention time, probably as a result of the corrosion of iron pipes, as did the mycobacterial number. It is therefore difficult to conclude that there is a causal relationship between iron content and numbers of mycobacteria. In addition, the concentrations of mycobacteria at the system sites correlated positively with the water retention time and negatively with the free chlorine content. Mycobacteria are more tolerant of chlorine than many other microbes (2, 24, 33, 46), and they have been isolated from drinking water with free chlorine contents of up to 2.5 mg/liter (4). It has also been reported that terminal chlorination at waterworks had no effect on the number of mycobacteria (11), and in previous drinking water studies, no association was found between the numbers of mycobacteria and chlorine contents (4, 6, 10). Our results indicate that mycobacteria may multiply in the distal portions of the system, where chlorine levels are minimal (maximum, 0.19 mg of free chlorine/liter in our material). It is, however, unlikely that the low chlorine content is the main cause for the elevated concentrations of mycobacteria.

Our results showed that the key site of mycobacterial occurrence in the systems lies in the deposits. Mycobacteria were found in all deposit samples analyzed, and their numbers there, if expressed per liter, exceeded those in the water samples by a factor of 2.5 × 10⁴. Surprisingly, the deposits collected a year after the cleaning of the pipelines had numbers of mycobacteria similar to those recovered from the old deposits. Also, the species diversities in the young and old deposits and the water samples were rather similar. An explanation of these results may be that the old deposits also have a surface layer that maintains active microbial growth. The biofilm experiment carried out in one system showed that mycobacterial colonization of surfaces of boreal drinking water systems is rapid. Viable mycobacteria reached their maximum density in biofilms as quickly as other heterotrophic bacteria, i.e., during the first 10 weeks. The other heterotrophic bacteria colonized biofilms more rapidly than mycobacteria, but otherwise the kinetics of mycobacterial colonization of biofilms was similar to that of other heterotrophic bacteria. There are only a few previous studies showing the high occurrence of mycobacteria in drinking water pipeline biofilms (6, 40), and the reported numbers are very similar to those of the present study (10² to 10³ CFU/cm²). The rapid colonization of biofilms by mycobacteria has been explained by the hydrophobicity of the mycobacterial cell. It has been proposed that mycobacteria could be the first colonizers on the pipelines, since they can form biofilms without the presence of other microbes and also under low-nutrient conditions (14, 40). We did not identify the mycobacterial strains isolated from the biofilms to species level, but their growth characteristics were very similar to those of the isolates recovered and identified from water specimens, indicating slowly growing species. This probably also explains why mycobacteria needed more time than other heterotrophic bacteria to appear in biofilms.

Similar to the water samples, the highest content of mycobacteria in the deposits were detected in the systems using ozonated surface water. This further supported the hypothesis that ozone promotes mycobacterial growth. In general, the pipe material may affect microbial attachment and the formation of biofilms (34). Previous results indicate that mycobacteria more easily colonize organic surfaces, such as plastics and rubber, than inorganic surfaces, such as copper and glass, perhaps because they can use the components present in the organic materials (40). We did not find a difference between the mycobacterial counts in PEH and cast iron pipes, nor did Falkinham and coworkers (6) find a difference between brass or bronze and galvanized or plastic surfaces. In our study, the deposits covered the pipes as a thick layer consisting of inorganic and organic substances accumulated over decades. Obviously, this may have diminished the significance of the original materials, as suggested by Zacheus et al. (58).

Tsintzou et al. (50) recognized that the occurrence of mycobacteria in drinking water decreased after the replacement of an old distribution system. In our study, mechanical cleaning of the pipelines did not have a clear effect on the occurrence of mycobacteria in water. This indicates that cleaning of one pipeline fraction at the distal site of a system is not enough to diminish the occurrence of mycobacteria which have colonized the system for years.

Most mycobacterial isolates recovered belonged to the recently discovered species M. lentiflavum and M. tusciae and to a previously unclassified group of mycobacteria. Only one widely distributed species in drinking water and biofilms, M. gordonae (4, 6, 8, 23, 35, 40, 50), was isolated among them. Surprisingly, the most widely distributed species geographically was a previously unclassified group of mycobacteria that was found in 8 of the 16 systems analyzed (50%). It also accounted for one-third of the isolates identified. Based on the polyphasic taxonomy we used, the new group resembled M. triplex but was different from it in both biochemical characteristics and 16S rRNA gene sequence. It represents a possible new species and needs further study in order to be positioned taxonomically. The second most common species was M. lentiflavum, which was recovered from 6 of the 16 systems studied (38%), and it alone accounted for over half of the isolates recovered. The species was originally found in clinical specimens and as a contaminant of a bronchoscope (43). So far, the only report of the occurrence of M. lentiflavum in drinking water is that by Schulze-Röbbecke and coworkers (Schulze-Robbecke et al., Abstr. 1st World Water Congr. IWA) who found it in 8 out of the 79 samples studied (10%) in Germany. The species was found only after water treatment, not in raw water or during water treatment, which indicates that it tolerates the treatment processes well, can grow in the system, and/or benefits from the shifts of the mycobacterial flora detected during water treatment. This may also explain why so far we have not recovered this species from Finnish natural waters (18; P. Torkko, personal communication). There is increasing evidence that M. lentiflavum is a potentially pathogenic species (12, 32, 43; our unpublished results). M. tusciae was described as a new species by Tortoli et al. in 1999 (49). The authors found it in a lymph node of an immunocompromised child and also in tap water. The present knowledge of both M. lentiflavum and M. tusciae thus suggests that water supply systems are their one likely reservoir. MAC, the most important pathogen among environmental mycobacteria in Finland, was not found in the drinking water samples analyzed. There are some previous observations of its high rate of occurrence in drinking water systems in Finland (54). We collected our samples from very near the main pipelines, not inside large buildings, such as hospitals, as did von Reyn and coworkers (54). This may be a reason for the lower recovery of MAC in our study, since the greatest occurrence of mycobacteria has been observed in large buildings, such as hospitals, with likely dead ends in their systems, and the lowest occurrence has been found in private residences (4). Though we did not find MAC in this study, it is possible that it occurs in drinking water occasionally or in very low numbers. It may thus contaminate the warm-water systems, including taps, where its presence after multiplication may pose a greater risk of infection. The species variations observed in the present study were very similar in the deposits and water samples, and usually the species found in the deposit could also be found in the water from the same system. The comparison between the species from the deposits before and after cleaning of the pipelines indicates that within a year the same species recolonized the systems. This rapid recolonization is understandable, because only a fraction of the whole pipeline system was cleaned in our study. The peculiar mycobacterial flora observed in Finnish systems could reflect the boreal environment, known to be a rich source for mycobacteria (18, 47). However, the species found in the drinking water have not been isolated from our natural waters so far. Schulze-Röbbecke and coworkers have proposed that there is a shift in mycobacterial flora during treatments at waterworks (Schulze-Robbecke et al., Abstr. 1st World Water Congr. IWA). The species found in purified drinking water may occur in raw water occasionally, or in concentrations below the detection limit, but increase afterward within the system.

In conclusion, mycobacteria are common in boreal drinking water systems, especially in the deposits. Their high isolation rate indicates that mycobacteria belong to the natural microbial flora of the systems. Their increased occurrence in the distal portion of the systems suggests that they also grow in the systems. Our results demonstrate how the use of ozone at waterworks may enhance mycobacterial growth in the systems, e.g., by increasing the AOC content. The mycobacterial flora recovered indicates that drinking water may be a reservoir for infections caused by recently described mycobacterial species.