INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

To date, more than 70 mycobacterial species have been identified, and occasionally, isolates with unknown characteristics, mostly from immunodeficient patients, are being described. Differentiation of pathogenic and nonpathogenic mycobacteria other than Mycobacterium tuberculosis (MOTT) from members of the M. tuberculosis complex (MTC) is needed for patient management, considering that many MOTT are resistant to the antibiotics used for treatment of tuberculosis (27).

Identification of mycobacterial isolates to the species level is performed by analysis of phenotypic and biochemical characteristics of the organisms after culture in solid media, which is a time-consuming process, or by high-pressure liquid chromatography analysis, which requires expensive equipment. Development of molecular tests has speeded up diagnosis, but most methods suffer from specific drawbacks. The AccuProbe system (Gen-Probe) differentiates only the MTC, M. avium, M. intracellulare, M. gordonae, and M. kansasii. In-house PCR-based identification systems have been developed but either identify a limited number of mycobacterial species or are difficult to use on a routine basis (4, 11). Restriction enzyme analysis of PCR products of specific genes is still widely used for identification of mycobacteria to the species level (7, 19, 23) and on clinical isolates (22, 26). Sequencing of conserved genes is sensitive and accurate but still expensive and technically demanding (9, 18, 21, 28), although commercially available sequencing systems might simplify the procedure (16). Finally, recently developed high-density DNA probe systems need specialized equipment (8).

Recently, INNO-LiPA Mycobacteria (LiPA), a commercially available assay targeting the 16S-23S rRNA spacer region, was developed for the detection of Mycobacterium spp. and identification of members of the MTC, M. kansasii, M. xenopi, M. gordonae, M. avium, M. intracellulare, M. scrofulaceum, and M. chelonae. The assay has been evaluated with BACTEC 12B bottles, and LiPA results have been compared with results obtained by using DNA probes, conventional biochemical tests, and PCR-restriction fragment length polymorphism (RFLP) analysis; the assay demonstrated clear-cut results and was rapid and easy to perform (14). All samples tested were, however, local clinical isolates, and the specificity of the test was not evaluated on a large number of different species. Furthermore, genetic variability has been observed in mycobacterial isolates with distinct geographic origins (13, 17). We therefore evaluated the assay with reference strains and clinical isolates from different regions of Brazil belonging to 42 different species.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Culture of mycobacterial strains and DNA extraction. Forty-seven mycobacterial reference strains belonging to 37 species were used in this study and are listed in Table 1. Clinical mycobacterial isolates (n = 110, see Table 2) were obtained as Lowenstein-Jensen cultures mainly from the National Reference Center Professor Hélio Fraga (Rio de Janeiro, Brazil) and from Laboratório Richet (Rio de Janeiro, Brazil). Conventional identification had been performed according to standard procedures (10), and high-quality DNAs from mycobacterial reference cultures were prepared as described earlier (20). For clinical isolates, the DNA extraction procedure was simplified by suspending a loop of mycobacterial mass taken from a culture in Lowenstein-Jensen medium in 0.5 ml of 10 mM Tris-HCl-1 mM EDTA-1% Triton X-100 and submitting it to three cycles of freezing and boiling (5 min, 70°C; 10 min, 100°C). Nucleic acids and processed cultures were stored at 20°C until further use.

PCR-RFLP analysis of hsp65. PCR-restriction enzyme analysis of a 441-bp amplified fragment of the hsp65 gene (PRA) was performed as modified by Telenti et al. (23). Briefly, 10 ng of purified mycobacterial DNA or that amount present in 2 µl of the supernatant of a frozen and boiled bacterial mass was amplified using 1.25 U of Taq polymerase by submitting the sample to 45 cycles of 1 min at 94°C, 1 min at 65°C, and 1 min at 72°C, followed by a final extension step at 72°C for 7 min. When an amplicon was present, 20 µl was digested with 10 U of either BstEII or HaeIII and the restriction pattern was analyzed by electrophoresis in a mixture of 4% agarose (Sigma) and 1% NuSieve agarose (FMC BioProducts, Rockland, Maine) and by ethidium bromide staining. Each gel had a 50- or 100-bp-ladder molecular weight marker (Gibco BRL) in at least two lanes. Patterns were compared with those published in the literature (3, 6, 22, 23) or analyzed against a pattern database constructed with GelCompar (3).

LiPA. The same amount of the DNA sample that was submitted to PCR-RFLP was added to a 40-µl PCR mixture containing deoxynucleoside triphosphates (supplied in the kit as amplification buffer), biotinylated primers and MgCl₂ (supplied as primer solution), and 1.25 U of Taq polymerase (Amersham Pharmacia, Little Chalfont, Buckinghamshire, United Kingdom). After 1 min at 95°C, amplification was performed by submitting mixtures to 30 s at 95°C, 30 s at 62°C, and 30 s at 72°C for 40 cycles. Amplification was verified in 2% agarose and by staining with ethidium bromide, and generally, a band with a size ranging from 250 to 500 bp was obtained. A 10-µl aliquot of the amplified product was mixed with 10 µl of denaturing solution and incubated for 5 min at room temperature in a hybridization trough in a 12-trough tray (both supplied with the kit). Two milliliters of hybridization solution (at 62°C) and a membrane strip were added to each sample, and the tray was then placed in a shaking water bath (Gemini twin; Robbins Scientific, Sunnyvale, Calif.) at 62°C and incubated for 30 min. This hybridization step was followed by two 1-min incubations with 2 ml of wash solution at room temperature and one 10-min incubation at 62°C. All further incubations were performed at room temperature in an orbital shaker. The strips were rinsed with 2 ml of rinse solution, followed by incubation with alkaline phosphatase-streptavidin conjugate solution for 30 min and washing with 2 ml of rinse solution and 2 ml of substrate buffer. Finally, 2 ml of substrate solution was added and the strips were incubated for 30 min; the color reaction was stopped by washing the strips twice with 2 ml of deionized water, and identification of the sample was performed by comparing the strip against an interpretation chart supplied with the kit. In each test run, one LiPA control sample was hybridized against a strip as a control for the hybridization conditions. The LiPA test was repeated when hybridization signals suggested the presence of two mycobacterial species in a sample or when discrepant results were obtained by the different identification procedures. Also, 3 of the 12 samples that hybridized with the genus and M. avium-M. intracellulare-M. scrofulaceum probes only were resubmitted to the test and had their hybridization patterns confirmed.

Sequencing and analysis. For some isolates, the amplified 16S-23S spacer region generated as described above was purified using a QIAquick PCR purification kit (Qiagen, Chatsworth, Calif.) and sequenced using an ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit on an ABI PRISM 377 sequencer (Applied Biosystems, Foster City, Calif.). Analysis of sequences was performed by comparison against sequences in GenBank using Seqed and Fasta of the Wisconsin Package (version 9.1; Genetics Computer Group, Madison, Wis).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

No difference in signal intensity or background was observed when the LiPA assay was performed on purified DNA or on the supernatant of a heat shock-treated bacterial mass (data not shown). For all isolates analyzed by hybridization, a clear amplicon of the expected size was visible on agarose gel.

Initially, we tested 49 reference strains, including the type strains of M. ulcerans (ATCC 19423) and M. simiae (ATCC 25275), and all hybridized to the Mycobacterium genus probe (MYC). The M. ulcerans and M. simiae strains, however, gave an M. intracellulare pattern on LiPA assay, and PRA analysis demonstrated that they had been contaminated with M. intracellulare, so they were not included in the study. Twelve strains also hybridized to their corresponding complex- or species-specific probes (Table 1): reference strains of M. tuberculosis, M. bovis, and M. africanum strains hybridized correctly with the probe for the MTC, while M. gordonae, M. avium, M. intracellulare, and M. xenopi reacted with their specific probes. For M. kansasii and M. chelonae, species for which more than one probe was present on the strip, hybridization occurred with MKA-1 and MCH-1, respectively. All other reference strains, including M. fortuitum, reacted with the MYC probe only.

All human clinical isolates reacted with the genus-specific Mycobacterium probe (MYC) (Table 2). For species or species groups that have specific probes in the LiPA assay, concordant results were found: isolates of MTC (n = 1), M. kansasii (n = 15), M. gordonae (n = 16), M. scrofulaceum (n = 2), M. abscessus (n = 11), and the M. avium complex (MAC) (n = 37) all hybridized with the appropriate probes on LiPA assay. The M. kansasii isolates were positive with the MKA-1 probe, while the M. abscessus isolates reacted with the MCH-1 and MCH-2 probes. For one M. kansasii isolate, a weak cross-reaction was visible with the MAV probe in the absence of the M. avium-M. intracellulare-M. scrofulaceum signal. The LiPA test could not be repeated on this isolate, and we consider this result a hybridization artifact that was probably due to a small variation in hybridization temperature, which is critical during the assay, or due to the fact that the strip was not completely submerged into the assay buffers. Clinical isolates of M. fortuitum (n = 10), M. flavescens (n = 2), M. terrae (n = 1), M. lentiflavum (n = 2), M. duvali (n = 1), M. szulgai (n = 5), M. triviale (n = 1), M. simiae (n = 4), and one unidentified Mycobacterium species were identified to the genus level only. One isolate identified as M. gordonae by conventional methods gave different results during different test runs: in two tests, only the MYC probe was positive, and in one other run, an M. avium isolate was found on LiPA assay.

All isolates were also characterized by PRA, as summarized in Table 2. All clinical M. kansasii isolates had the M. kansasii I-type PRA pattern, except for one isolate that had the M. kansasii III pattern. Of the 16 M. gordonae isolates, 3 had the M. gordonae I PRA pattern, 1 had the M. gordonae II pattern, 6 had the M. gordonae III pattern, 1 had the M. gordonae VII pattern, and 5 were characterized by three new PRA patterns for this species (data not shown). At the time of the study, no conventional identification procedures were performed to differentiate between isolates of M. chelonae and M. abscessus, but among the strains identified as the M. chelonae-M. abscessus complex, three had the PRA pattern of M. abscessus I and eight had the pattern of M. abscessus II. The two strains that reacted with the M. scrofulaceum probe on LiPA assay had a so-far-undescribed PRA pattern (data not shown), and their identities were confirmed twice by conventional testing. Among the isolates of the M. fortuitum-M. peregrinum complex, two had the M. fortuitum I PRA pattern, two had the M. fortuitum III pattern, three had the M. peregrinum I pattern, one had the M. peregrinum II pattern, and two had unpublished PRA patterns (data not shown). One strain had the M. duvali pattern, one had the M. terrae pattern, one had the M. flavescens pattern, and two M. szulgai strains had a so-far-undescribed PRA pattern (data not shown).

Thirty-seven clinical isolates that had been identified phenotypically as MAC were identified by the LiPA assay as M. avium (n = 18), M. intracellulare (n = 7), or M. avium-M. intracellulare-M. serofulaceum (probes MYC and MAIS positive; n = 12). From the 18 isolates hybridizing with the specific M. avium probe on LiPA assay, 14 were M. avium I, 3 were M. avium II, and 1 was M. avium III on PRA. Among the seven isolates reacting with the specific M. intracellulare probe on LiPA assay, five were identified as M. intracellulare I on PRA and two had this pattern but without the HaeIII 60-bp band (data not shown). From the 12 strains that reacted with the MAIS probe only, 6 were M. avium I, 4 were M. avium III, and 2 were M. intracellulare I.

Initially, for seven isolates from patients, discrepant results obtained by different identification procedures were caused by mixed cultures: in two cultures, the presence of at least two different species was detected by genetic analysis, while in five others, the presence of different species in a single culture was suggested upon repetition of conventional identification. Among the cultures that had been characterized as a single species by conventional procedures but had mixed signals by the LiPA assay, one isolate was an M. gordonae strain that hybridized two times with the MGO, MAIS, and MAV probes on LiPA assay and had a combination of M. gordonae III and M. avium III patterns on PRA. Another isolate, phenotypically identified as M. fortuitum, was initially positive on two LiPA assays for the MYC, MAIS, MAV, and MIN probes but had a PRA pattern suggestive of M. fortuitum I with background; upon subcloning, the isolate was reidentified as M. fortuitum, had a clear M. fortuitum I PRA pattern, and was MYC probe positive only on LiPA assay. For five clinical isolates, discrepant results were obtained when conventional identification results were compared to the results obtained by PRA and the LiPA assay. One isolate that had been identified as M. scrofulaceum by a conventional method was identified as M. gordonae III and was positive for the MGO probe by the LiPA assay, an isolate identified as M. gordonae was identified as M. kansasii I and was positive for the MKA-1 probe by the LiPA assay, and three isolates conventionally identified as MAC gave genetic patterns for another species. One of the isolates of the last group had patterns for MTC by both genetic assays, one was M. peregrinum I and Mycobacterium species by PRA and the LiPA assay, respectively, and the other was M. szulgai and Mycobacterium species by PRA and the LiPA assay, respectively. For these isolates, the biochemical identification was repeated on subculture, confirming the results obtained by PRA and the LiPA assay.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we evaluated the LiPA line probe assay on both clinical isolates and reference strains. Recently, Miller et al. (14) examined the performance of the LiPA test, demonstrating correct identifications for 59 of 60 Mycobacterium isolates, but that study concentrated on cultures received in a single lab in the United States, half of which were M. tuberculosis. Genetic variability of mycobacteria has been associated with differences in geographic origin (5, 13), and although MOTT in Brazil are mostly MAC, M. kansasii, and M. fortuitum (1), all assays required for conventional identification to the species level are not routinely performed, so the number of infective or colonizing Mycobacterium species in the country is underestimated (2, 3). Also, an earlier study using PRA on a relatively small number of clinical isolates suggests the existence of particular Brazilian allelic types of mycobacteria (2). The LiPA assay in this study was performed on clinical isolates obtained from a laboratory that receives clinical specimens from different regions of the country.

All mycobacterial samples reacted positively with the Mycobacterium genus probe on LiPA assay (100% sensitivity), so, in contrast to what occurs in acid-fast staining, a single technology determines both the genus and clinically important species. All isolates for which species- or complex-specific probes are present on the strip were correctly identified by a positive reaction of the specific probes, except for one M. gordonae isolate. The LiPA assay for this isolate was performed three times, once on a DNA sample that yielded an unpublished PRA pattern that was observed in two other isolates that were positive for the MGO probe on LiPA assay. We have no simple explanation for this result, but PCR contamination or DNA degradation could be involved. The total accuracy for identification is therefore 99.4% (156 of 157 samples correctly identified).

The LiPA assay was able to further identify to the species level 25 out of 37 strains that had been conventionally characterized as MAC; these isolates were identified by the LiPA assay as M. avium and M. intracellulare and had PRA patterns corresponding to these species (Table 2). Twelve strains, besides being positive for the MYC probe, were positive for the MAIS probe only, and although four of these had the M. avium III PRA pattern that has been detected so far only in strains from Brazil (12), six had M. avium I and two had M. intracellulare I PRA patterns. Three out of the 53 isolates that could not be correctly identified by Miller et al. (14) were M. intracellulare strains (from a total of 15 MAC strains) that were identified by the LiPA assay as M. avium-M. intracellulare-M. scrofulaceum. The observation that one out of three MAC strains in our study could be identified to the species level by PRA but to the M. avium-M. intracellulare-M. scrofulaceum complex level only by the LiPA assay could be explained by the different genetic targets of both assays. Some strains may be particular to Brazil since certain PRA genotypes of MAC isolates have so far been reported only in this country (12, 15). We suspect these strains to be M. avium-M. intracellulare intermediates, based on the large number of MAC strains that have been sequenced for development of the species-specific probes (data not shown). Nonetheless, the fact that these variants have been positive for the MAIS probe demonstrates that the assay is sufficiently sensitive to pick up genetic variants within MAC.

Essentially as reported by Miller et al. (14) but in contrast to a recent study performed in Italy (24), all isolates that hybridized with the MCH-1 and MCH-2 probes were M. abscessus and all M. kansasii strains (except one) that were positive for the MKA-1 probe on LiPA assay were determined to be M. kansasii I on PRA. We tested only a single clinical isolate of M. tuberculosis, but all 26 strains tested by Miller et al. (14) hybridized with the correct probes.

The fact that M. fortuitum cannot be differentiated from other species for which no specific probes are present on the strip may be a limitation of the assay because specific therapeutic schemes for this organism are available (27). In Brazil, 11% of MOTT of pulmonary origin isolated are M. fortuitum strains (1), and about 50% of the clinical isolates that reacted only with the species-specific probe in this study were M. fortuitum strains. This species, however, is genetically heterogeneous (W. Mijs, L. Rigouts, F. Portaels, and R. Rossau, Abstr. 21st Annu. Congr. Eur. Soc. Mycobacteriol., 2000), so inclusion of species-specific probes in the LiPA assay may be technically complicated.

Although 18 out of the 110 clinical isolates had so-far-unpublished PRA patterns, we feel confident about the identities of 109 samples because conventional identification procedures were performed in a reference center whose staff has considerable experience in the matter. Our results also demonstrate that discordance between the results of conventional and genetic identification procedures was mostly due to the presence of two different mycobacterial species in a culture. In the present study, no single colonies were picked for resubmitting to the different identification procedures but colonies with different morphologies were sometimes observed during conventional identification (data not shown). The LiPA assay was able to detect mixtures when mycobacterial species for which specific probes are present on the strip were involved, as demonstrated by the mixed signals described in Results; the assay did not, however, allow simultaneous identification of M. fortuitum and another species because no species-specific probe was present for the former species. We imagine that the result obtained by different identification procedures on mixed cultures will be influenced by the relative quantity of each species present in the clinical sample, by the growth rate of the organisms, and by the fraction of the culture that is being submitted to subculturing or DNA extraction. At least for species that have a specific probe, the LiPA assay seems to be more sensitive than PRA for detecting mixed cultures. This conclusion is in agreement with the reported higher sensitivity of the line probe assay than those of sequencing and phenotypic assays for detecting mixed populations of human immunodeficiency virus (25). In any case, we suggest that preferentially clonal organisms should be used for comparison of identification procedures.

When this paper was in preparation, Tortoli et al. (24) reported an assessment of the performance of the LiPA line probe assay on 238 strains isolated in seven Italian laboratories. The kit correctly identified 99.6% of the strains tested, and the discrepant result was for an unresolved MAC strain identified by the LiPA assay as a MAC intermediate.

Conclusion. Evaluation of the LiPA assay was performed on a large number of clinical isolates and on reference isolates from different Mycobacterium species, which were characterized by phenotypic methods and PRA. Discordance between the results of different identification procedures was due to the presence of two different organisms in a culture. The LiPA assay, which targets the 16S-23S rRNA spacer region, was able to correctly identify 156 out of 157 isolates or strains, resulting in an accuracy of 99.4%. One M. gordonae isolate showed ambiguous results. A correlation was found between identifications obtained by the LiPA assay, which targets the 16S-23S rRNA spacer region, and PRA, which targets the hsp65 gene, although allelic differences were observed. In general, the LiPA assay is a reliable genetic test for the identification and differentiation of Mycobacterium species.