The incidence of individual infections and outbreaks associated with NTM has risen dramatically over the past decade establishing these organisms as significant human pathogens. Traditionally, the identification of mycobacteria to the species level has relied upon biochemical tests, which require three to six weeks to complete. Biochemical identification, even when performed by skilled microbiologists, may yield uncertain or even misleading results because (a) the tests used are inherently poorly reproducible; (b) the expected phenotypes are not an absolute property of the species, but may exhibit substantial variability; and (c) the database of phenotypic characteristics is limited to common species .
In recent years, DNA-based techniques have greatly facilitated identifying the species of NTM isolates and enabled a number of new species to be documented as infecting agents [35–39]. These approaches can be applied to a single isolated colony and a definitive result can typically be obtained within a day. PRA-hsp65, first described by Telenti et al., is based on detection of restriction fragment polymorphisms in the hsp65 gene and thereby resolving the species of a mycobacterial isolate .
In the present study, 434 NTM isolates from clinical specimens were analyzed by conventional phenotypic methods and by PRA-hsp65; further, those isolates for which the results from the two methods were discordant were analyzed using nucleotide sequencing of the hsp65 gene. For 63 (14.5%) isolates phenotypic methods could not provide a species identification and for almost a third of these isolates even the apparent Runyon classification proved inconsistent with conventional expectations. For an additional 33 (7.6%) isolates the phenotypic identification proved incorrect. Phenotypic variability among fresh clinical isolates has been observed in other studies [10, 40, 41].
In contrast, PRA-hsp65 correctly identified over 90% of evaluable isolates using currently available databases of restriction digest patterns. For most of the remaining isolates, the PRA-hsp65 pattern observed was not previously reported. There were only 4 (1.2%) clinically significant isolates for which the current PRA algorithm indicated an incorrect species.
PRA-hsp65 has proven similarly effective in other studies. Hafner et al. used 16S rDNA sequencing to analyze 126 isolates selected at random from a larger collection . The hsp65 method correctly identified 120 (95.2%) of these isolates. They also sequenced 10 additional isolates from the larger collection that gave PRA-hsp65 patterns not previously reported. All these isolates represented environmental species rarely associated with clinically significant disease.
Among our 434 isolates, 30 (6.9%) provided 13 PRA-hsp65 profiles not previously reported. Our series represents isolates cultured from varied clinical specimens collected in the metropolitan and surrounding areas of the city of Sao Paulo, Brazil. Most of the isolates with new PRA-hsp65 patterns were cultured from sputum. Many represented species typically considered non-pathogens; clinical correlation was not available and these isolates may reflect colonization by environmental organisms. Previous studies have similarly documented considerable species diversity as well as the genotypic diversity among mycobacteria isolates in Brazil [42, 43]. Sequence analysis confirmed that the new profiles were allelic variations within the species, consistent with previous studies [13, 17, 20]. Of interest, four profiles were represented by more than one isolate, suggesting that they are potentially prevalent lineages rather than singular mutation events.
The most commonly identified new profile (designated NP1) was observed in 11 isolates, representing 20% of all M. gordonae in this collection. Comparison to the prototype M. gordonae sequence indicated two point mutations that resulted in the loss of two HaeIII sites and the addition of 95-bp fragment to the profile [GenBank:EF601222]. A similar profile was assigned to M. gordonae by da Silva Rocha et al. , although sequence confirmation was not reported. Hafner et al. also noted that M. gordonae is a particularly polymorphic species .
The NP17 profile, demonstrated for five isolates, was identified by sequencing as M. arupense, a recently described species related to the M. terrae complex . The NP14 profile, observed for three Mycobacterium simiae isolates, was similar profile to the M. simiae 3 pattern reported by Legrand et al.  as well as to the prototype M. simiae 1 pattern . Sequencing confirmed that the nucleotide sequence is intermediate between those two strains. The sequence also matches that recently reported by Selvarangan et al., who proposed that their isolates represented a new species (M. sherrisii sp. nov) based on a distinct pattern of cellular fatty acids and a unique 16S rRNA gene . The NP11 profile, represented by two isolates of M. terrae, was similar to a PRA-hsp65 pattern described by McNabb et al.  with the addition of a unique HaeIII restriction site [GenBank:EF601223].
We would concur with Hafner et al. that additional work is required to define and standardize the most effective electrophoresis conditions for resolving hsp65 digests of mycobacteria . In a recent multicenter study evaluating PRA-hsp65, variations related to gel preparation, running conditions and documentation tools all complicated the interpretation of digestion patterns .
The ever-increasing amount of data available and the identification of new profiles make the analysis more complex. We present an updated PRA-hsp65 algorithm, which includes 174 patterns among 120 species and sub-species and have the basic cultural characteristics (rate of growth and pigment production). These core phenotypic traits can be readily determined and, as emphasized in a recent statement by the American Thoracic Society , can assist in confirming the molecular identification, detecting mixed cultures, and classifying species with indistinguishable PRA-hsp65 patterns.
Despite the complexities noted above, PRA-hsp65 analysis proved both more rapid and more reliable than phenotypic methods; it was particularly effective at resolving the most common pathogenic species. Commercial DNA probes are available only for a very few species and their expense may be prohibitive in some settings. DNA sequencing is more definitive, but sequencing capability is not yet widely available in clinical laboratories.