Automated ribotyping as a tool for identifying of nontuberculous mycobacteria was evaluated. We created a database comprising of riboprints of 60 strains, representing 32 species of nontuberculous mycobacteria. It was shown that combined ribopatterns generated after digestion with EcoRI and PvuII were distinguishable between species of both slow-growing and rapid-growing mycobacteria. The findings were in good agreement with the 16S rRNA gene sequencing results, allowing correct identification of Mycobacterium lentiflavum isolated from clinical specimens and from biofilms growing in public water distribution system. The automated ribotyping was powerful in discriminating between M. lentiflavum and closely related species M. simiae and M. palustre. Mycobacterium lentiflavum strains from drinking water biofilms were resistant to two to four antimycobacterial drugs. The drinking water distribution system may, thus, be a source of nontuberculous mycobacteria resistant to multiple drugs.
In Finland the occurrence of nontuberculous mycobacterial (NTM) infections is close to that of the cases of tuberculosis (Anonymous, 2004). The environmental reservoirs of NTM causing infections in human are not fully understood. Water has been proposed as a source of human Mycobacterium avium infections (Primm et al., 2004). The risk of disseminated M. avium infection among AIDS patients was found to be associated with urban residence and exposure to treated water sources in Finland (Ristola et al., 1999).
Recently, mycobacteria were reported in 15 out of 16 deposits taken from drinking water distribution systems at different sites in Finland with M. lentiflavum as a dominant species (Torvinen et al., 2004). Mycobacterium lentiflavum, first described in 1996 (Springer et al., 1996), can cause a range of symptoms in children, aged persons and immunocompromised patients (Tortoli, 2003; Molteni et al., 2005). The presence of M. lentiflavum in drinking water distribution system calls for efficient methods to track their environmental reservoirs. The rapid identification of NTM may assist in recognition of their environmental reservoirs.
Automated ribotyping has been shown as being useful for the identification of various bacteria with too few easily recordable biochemical features for species identification (Busse et al., 2000; Satokari et al., 2000; Clermont et al., 2001) and successfully applied to selected fast growing NTM (Vuorio et al., 1999). In ribotyping, genomic DNA is digested with restriction enzymes and the obtained restriction fragments are hybridized with an rRNA operon targeted probe (Riley, 2004). An automated ribotyping device RiboPrinter (Qualicon, Du Pont, Wilmington, DE) generates, analyses and stores riboprint patterns of bacteria. It allows the development of custom database and comparison of any newly obtained ribopatterns with those stored in the database. Moreover, automated ribotyping was shown to be highly reproducible between different laboratories (Brisse et al., 2002).
In this study, NTM were isolated from biofilms growing in drinking water distribution system and from clinical specimens, and automated ribotyping was evaluated as a tool for identification, particularly of M. lentiflavum.
Materials and methods
Nontuberculous mycobacteria isolated from patients in Finland were identified by Mycobacterial Reference Laboratory, National Public Health Institute, Turku, Finland. The clinical strains for this study were chosen from isolates collected between 1995 and 2001 from patients living in Helsinki and Tampere areas. Strains H0394/01, H0549/99, H0376/04, H0855/04, H1171/04, H0530/04, H1025/00 originated from sputum, H0850/95 from blood, the origin of strain H0184/97 was unspecified.
Strains M2, M3, M4, M6, MH1, AGHA3, and AGHA13 were isolated from biofilms in water meters of detached and apartment houses and a children's day-care centre in the cities of Tampere and Helsinki. The meters had been in use for 4–12 years at the time of sampling.
The sampled biofilms were dispersed in 4.5 mL of Zwittergent-solution (Camper et al., 1999) or phosphate-buffered saline by shaking with glass beads for 2 h or by sonicating four times for 1 min, 0.5 mL of 50% H2SO4 was added and neutralized after 30 min with 50% NaOH. After centrifuging at 1300 g for 15 min, the pellet was washed twice with 5 mL of 0.9% NaCl, and cultivated on agar tubes Mycobacterium 1 and Mycobacterium 2 modified Löwenstein-Jensen (Orion Diagnostica, Espoo, Finland) or Middlebrook 7H10 agar medium (Difco Laboratories, Detroit, MI) at 28°C for 2–5 weeks. The colonies obtained were stained by the Ziehl-Nielsen's carbol–fuchsin method. Representatives of each acid-fast colony type were subcultured on Middlebrook 7H10 agar at 28°C and stored in glycerol at −70°C.
Ribotyping patterns of 60 mycobacterial isolates obtained after digestion of genomic DNA with EcoRI and PvuII restriction enzymes.
Identification of isolates
The clinical isolates were subjected to partial 16S rRNA-gene sequencing as part of a routine in the Finnish Mycobacterial Reference Laboratory, using the method described by Kirschner (1993). The biofilm isolates were sequenced for the 500 bp fragment from the 5′ end using a MicroSeq 500 16S rDNA – kit (PE Applied Biosystems, Foster City, CA) and ABI PRISM 310 genetic analyser (PE Applied Biosystems) described by (Patel et al., 2000). The sequences were edited by Sequencher™ 3.0 software and compared against RIDOM – Ribosomal Differentiation of Medical Microorganisms (http://www.ridom.de), EMBL and NCBI-GenBank databases.
For the almost full-length 16S rRNA-gene sequencing, genomic DNA was amplified using primers MBUZ1 and MBUZ2 (De Baere et al., 2002). The amplification products were purified and sequenced in both directions using primers MBUZ1 and MBUZ2, and additional three or four internal primers to get three overlapping strands.
The antibiotic susceptibility testing of mycobacterial isolates was performed by the disk-elution method with the drug concentrations recommended by NCCLS (standard M24-T2) on Middlebrook 7H10 agar plates. Commercial impregnated antibiotic disks (Sensi-Disc; Becton Dickinson, Cockeyville, MD) were used with final drug concentrations of isoniazid, 0.2 mg L−1 of agar; ethambutol, 5.0 mg L−1 of agar; amikacin, 6.0 mg L−1 of agar; streptomycin, ciprofloxacin and ofloxacin, 2.0 mg L−1 of agar; rifampin and levofloxacin, 1.0 mg L−1 of agar. An isolate was considered resistant if the number of colonies growing on the plate with the drug was greater than 1% of those on the control plate. The susceptibility to pyrazinamide was tested using nicotinamide with a final concentration of 5 mg mL−1 (Brander, 1972).
Automated ribotyping and data analysis
One loop of biomass (15–20 mg) grown on solid medium was suspended in 100 μL of sterile water, heated at 80°C for 20 min. The cells were treated with solvent to remove cellular lipids as described by Vuorio (1999). The organic solvents were discarded after centrifugation and the pellet dried at 50°C for 10 min, rinsed once more with 0.5 mL of acetone, and dried at 50°C for 25 min to ensure the complete removal of acetone which might interfere with the enzymes used in ribotyping. The dry biomass was suspended in 100 μL of sample buffer (Qualicon, Wilmington, DE), lysozyme was added and incubated at 37°C for 1 h. The samples were then analysed with the robotized RiboPrinter™ (Qualicon, Du Pont, Wilmington, DE). Ribotypes were assigned by the RiboPrinter software, and further analysed as normalized files using the BioNumerics software version 4.0 (Applied Maths, Sint-Martens-Latem, Belgium). The original ribopatterns are stored in the Riboprinter software for any further comparison.
Cluster analysis was performed using the unweighted pair-arithmetic (UPGMA) clustering algorithm based on the Pearson correlation coefficients of densitometric curves with an optimisation parameter of 1.2% (identical to that used in the RiboPrinter system). Combined clustering of EcoRI and PvuII ribopatterns was performed with the same analysis settings and equal weighting.
Sizes of the bands are as given by Riboprinter software.
Mycobacteria from clinical specimens and drinking water biofilms
The average number of annual isolations of NTM from 1996 to 2003 in Finland was 351 with an annual variation of less than 30%. However, the number of Mycobacterium lentiflavum isolations from patient specimens increased through these years almost 20-fold, from 0.3% to 5.3% of laboratory notifications of NTM (Table 1).
Number of Mycobacterium lentiflavum isolates and total number of nontuberculous mycobacteria isolated from patient specimens in Finland during the years 1996–2003*
Nontuberculous mycobacterial isolations, n
M. lentiflavum, n (% of total)
* In Finnish Mycobacterial Reference Laboratory, National Public Health Institute (Turku).
Nine randomly chosen patient strains of M. lentiflavum from the cities of Tampere and Helsinki were subjected to sequencing of 1400 bp fragment of the 16S rRNA gene. The identification based on partial sequences of seven out of nine isolates was confirmed. Two strains originally identified as M. lentiflavum (H0184/97 and H0850/95) were found to be identical to M. palustreT and M. simiaeT, respectively (Table 2).
↵† Identified as M. lentiflavum based on 16S rRNA gene hypervariable region B sequencing.
Water meters of residential buildings and a children's day care centre from the same cities were analysed for viable mycobacteria. The obtained isolates were identified by 16S rRNA-gene sequencing as M. lentiflavum, M. gordonae and related to M. montefiorense (Table 2).
All M. lentiflavum isolates were checked for susceptibility to antimycobacterial drugs. It was found that all clinical and environmental isolates were resistant to two to six drugs in vitro (Table 3). All strains identified as M. lentiflavum by full-length 16S rRNA-gene sequences, were resistant to isoniazid and ethambutol, and sensitive to pyrazinamid and ciprofloxacin, whereas resistance to the other four drugs varied from strain to strain.
The clinical and water biofilm isolates were ribotyped using automated riboprinter instrument. To see what was the discrimination power of this method for NTM, reference strains of several species of NTM were included (Fig. 1). Sixty strains, representing 32 species of NTM were ribotyped using two restriction enzymes, EcoRI and PvuII. Reproducible clear patterns with fragments ranging between 1.1 and 25 kb for all strains were obtained. In independently repeated runs the bands over 25 kb in size were poorly reproducible, possibly because of partially digested DNA. Therefore the bands >25 kb were excluded from clustering analysis.
The commercial ribotyping system has so far no database available for mycobacteria. The present database (DUP 2004) named M. avium ss. avium strains after digestion with EcoRI and PvuII as Comamonas testosterone and Micrococcus lylae, respectively. M. celatum and M. terrae were also incorrectly named as Micrococcus luteus.
Within the eight species of NTM that were represented by two or more isolates the ribopatterns differed only slightly not impairing the identification.
Digestion of M. lentiflavum with EcoRI resulted in three ribopatterns. One group comprised all clinical and one water biofilm (AGHA13) isolate belonging to sequevar 1 (identical to the type strain). Four water biofilm isolates M2, M3, M4, and M6, that belong to sequevar 2 (Table 2), had an additional faint 5 kb fragment. However, beside of 100 % similarity in 16S rRNA gene, the strain M 6 lacked the 13 kb band and had a more intense 2.0 kb fragment. Digestion with PvuII generated the identical patterns for the 13 M. lentiflavum strains, excepting the strain CCUG 42423 that had more intense 2.4 kb band.
The five M. avium ss. avium strains had identical, or nearly identical (DSM 44156T), EcoRI and PvuII ribopatterns.
Four of the M. murale strains were identical when digested with EcoRI or with PvuII (Fig. 1). The fifth strain MA-168-96 differed by one band in EcoRI pattern only.
The three M. chlorophenolicum strains differed from each other both in EcoRI and in PvuII ribopatterns. 16S rRNA-gene sequence analysis revealed that the strains were only distantly related (Table 2). The strains CG-1 and CP-2 were 98.7 % similar in 16S rRNA genes, suggesting that these strains may represent a new species within fast growing mycobacteria.
Our data illustrate that neither EcoRI nor PvuII alone generated ribopatterns that distinguished between different species. But when two enzymes were used, the combined ribopatterns were unique for all the 32 species displayed in Fig. 1.
Clustering analysis based on combined EcoRI and PvuII ribopatterns of 30 mycobacterial strains is shown in Fig. 2. It is based on the species that were represented by at least two isolates with confirmed identification by 16S rRNA-gene sequencing. The dendrogram reveals that the clustering is in good agreement with the identification based on the 16S rRNA-gene sequencing.
The unweighted pair-arithmetic clustering dendrogram based on combined EcoRI and PvuII ribopatterns of a subset of 30 mycobacterial isolates representing seven species. The identification of the isolates used in clustering analysis was confirmed by nearly full-length 16S rRNA sequencing. Clustering was done using Pearson correlation coefficient. The scale bar indicates the percentage of similarity.
The amount of M. lentiflavum isolated from patients increased in Finland many folds since 1996. The data collected in Mycobacterial Reference Laboratory represent the nationwide situation as the laboratory collects all mycobacterial isolates from patients for identification. The contribution of M. lentiflavum has been growing in proportion to the total amount of NTM. The observed increase of M. lentiflavum is not caused by the change in identification methodology, because identification of the isolates in Finnish Mycobacterial Reference Laboratory was always done by partial sequencing of the 16S rRNA-gene according to (Kirschner et al., 1993). Thus, the increase is true, but unexplained.
We analysed the biofilms growing in public water distribution system on presence of mycobacteria and found M. lentiflavum, M. gordonae, and an isolate related to M. montefiorense.
The applicability of automated ribotyping for the identification of NTM was evaluated. We constructed a database containing riboprints of 60 NTM isolates, including 28 type strains, generated after digestion with two restriction enzymes, EcoRI, PvuII. To our knowledge, this is the first reported database of NTM riboprints.
16S rRNA-gene sequence analysis is a standard in bacterial classification (Stackebrandt et al., 2002). However, within the genus Mycobacterium the 16S rRNA genes are exceptionally conserved, some species differ by only a few nucleotides or and some with none (Tortoli, 2003; Devulder et al., 2005). At the same time, the areas within intergenic spacers and outside of the ribosomal operon may be highly diverse. The probe used in the robotized ribotyping is rrnB rRNA operon of Escherichia coli. The fragments generated in automated ribotyping include the entire ribosomal operon and a flanking area up to 50 kb. As a result, the ribopatterns give more possibilities than 16S rRNA-gene analysis alone for detecting differences between closely related groups of mycobacteria.
Mycobacteria have one or two rRNA (rrn) operons per genome (Bercovier et al., 1986; Parish & Stoker, 1998; Reischl et al., 1998; Menendez et al., 2002), explaining the low number of bands obtained in ribopatterns if only one restriction enzyme is used. Therefore, two restriction enzymes were needed to discriminate between species. We showed that two enzyme ribotyping was effective for the identification of both slow-growing and rapid-growing mycobacteria. Specifically, the ribotyping clearly distinguished between M. lentiflavum and those belonging to closely related species M. simiae and M. palustre. Clustering analysis grouped the M. lentiflavum isolates into two subgroups conform to the sequevar grouping based on the 16S rRNA-gene sequences. Different ribopatterns obtained for the two strains of M. celatum may also reflect the differences between sequevars.
We showed here that M. lentiflavum isolates from drinking water system were multiple resistant to antimycobacterial drugs. The majority of strains resistant to rifampin or isoniazid were shown to posses only single-locus mutation in genes for rpoB or catalase-peroxidase (katG), respectively, in M. tuberculosis (Zhang et al., 1992; Telenti et al., 1993; Ramaswamy et al., 2003). Chlorination of drinking water results in many compounds mutagenic for bacteria (DeMarini et al., 1995). It may be argued that exposure of a growing mycobacterial biofilm to a mutagenic environment may be one reason leading to a higher frequency in drug resistant mutants. Thus, the drinking water distribution system needs attention as a source of bacteria of clinical importance.
In conclusion, we showed that automated ribotyping with two restriction enzymes was useful for identifying of several species of NTM, in particular M. lentiflavum, which is emerging pathogen from water environment.
TEKES (National Agency for Technology Advancement), Academy of Finland (grant 52798) and EnSTe Graduate School are acknowledged for financial support. The authors thank J. Puhakka, P. Vuoriranta, and Finnish Water Industries for collaboration; Pirjo Torkko for providing strain Mycobacterium botniense E347T. We also thank Viikki Science Library and Faculty Instrument Centre for expert support for this research.
(1994) Transfer of polychlorophenol-degrading Rhodococcus chlorophenolicus (Apajalahti et al 1986) to the genus Mycobacterium as Mycobacterium chlorophenolicum comb. nov. Int J Syst Bacteriol 44: 485–493.