To the Editor: Mycobacterium kyorinense is a nonpigmented, slowly growing mycobacterium that was initially isolated in 2007 from a patient with pneumonia in Japan (1,2). The sequences of the 16S rRNA, hsp65, and rpoB genes of M. kyorinense were closely related to, but different from, those of the type strains of M. celatum and M. branderi, the 2 most phylogenetically related species (1). Biochemical tests, such as those for arylsulfatase activity, tellurite reduction, and heat-stable catalase, also distinguish M. kyorinense from M. celatum and M. branderi. In our initial report, in which this species was first recognized, we described 3 strains isolated from Japanese patients (1). Recently, 1 additional case was reported in Brazil (3). Here we describe 7 newly identified patients whose infection may have been caused by M. kyorinense.

In reviewing the characteristics of these 11 patients (10 from Japan and 1 from Brazil), we found no apparent contacts among them. Nine of the 11 patients had respiratory infections, 1 had lymphadenitis, and 1 had arthritis (Table). Of these, 9 patients fulfilled the criteria for infections of clinical significance (4) and were considered to harbor infection by M. kyorinense. Of the 9 patients with respiratory infections, 4 died as a result of the infection. These data suggest that M. kyorinense belongs to a class of nontuberculous mycobacteria that are pathogenic for humans and have substantial clinical effects.

Among the 10 patients for whom precise clinical records were available, 7 patients were treated with first-line tuberculosis drugs, mainly rifampin, isoniazid, and ethambutol, but these therapies were ineffective for all patients. Six patients received a combination of antimicrobial drugs, including macrolides and fluoroquinolones, as first- or second-line chemotherapy, and infection was subdued without recurrence in 5 patients. In contrast, 4 patients with pneumonia who did not receive sufficient therapy with the latter regimen eventually died of infection (3 patients) or breast cancer (1 patient).

MICs of various antimicrobial drugs for the 9 strains of M. kyorinense were determined by the broth microdilution method as described (1). For most strains, the MICs of rifampin, ethambutol, and isoniazid were relatively high, and MICs of macrolides, aminoglycosides, and quinolones were relatively low. Notably, MICs of rifampin were remarkably high (>32 μg/mL) for all tested strains (Table).

Direct sequencing of the 16S rRNA gene, performed as previously described, revealed that 8 of the 9 available M. kyorinense isolates were identical across the entire sequenced interval (1,470 bp). The sole exception was the strain from Brazil, which showed a 4-bp substitution that the other strains did not (3). Although the other 8 strains had identical 16S rRNA sequences, all showed heterogeneity at 9 positions that had not been observed to be heterogeneous in the previous investigation (1). This observation might reflect the presence of 2 copies of the 16S rRNA gene, as has been occasionally reported for other mycobacterial species, including M. celatum (5). Direct sequencing of the entire rpoB gene demonstrated that all strains had identical sequences for this locus. The strains differed from the sequence of M. tuberculosis at 15 nt within codons 511–533. At the amino acid level, these changes were synonymous for the 2 species, with the exception of amino acid residue 531. This residue, Ser531 in the M. tuberculosis RpoB protein, was replaced by an Asp in M. kyorinense. Notably, Ser531 is the most frequent location of substitutions in rifampin-resistant strains of M. tuberculosis (6).

Why M. kyorinense has been isolated almost exclusively in Japan is not clear. This tendency may be largely caused by a reporting bias in Japan. However, M. kyorinense may have a particular geographic distribution. In this context, it is noteworthy that the sole strain from Brazil characterized in the current study differed slightly in 16S rRNA sequences from the strains isolated in Japan.

It also is notable that the M. kyorinense strains isolated so far were invariably resistant to rifampin by in vitro susceptibility testing. Rifampin appeared to have been clinically ineffective in most patients, although definite efficacy of antimicrobial drugs cannot be evaluated by this retrospective type of study. Analysis of the rpoB gene sequence of M. kyorinense revealed the replacement of aa 531 when compared to the rpoB gene sequence of the M. tuberculosis protein. This finding suggests that M. kyorinense is inherently resistant to rifampin because of the structural features of its RpoB protein. Amino acid replacement at RpoB residue 531 also has been reported in other bacterial species resistant to rifampin, such as M. celatum, Borrelia burgdorferi, and Spiroplasma citri (7–9). In any case, understanding the intrinsic resistance of M. kyorinense to rifampin is critical for appropriately treating infection by this microorganism. On the basis of the results of our study, we recommend that a combination of fluoroquinolones and macrolides and/or aminoglycosides be used for the initial treatment of infection by M. kyorinense in most patients.