Mycobacterium xenopi Clinical Relevance and Determinants, the Netherlands

Clinical isolation of M. xenopi represents true infection in 51% of cases; genotype is a major determinant.

M ycobacterium xenopi was fi rst described by Schwabacher in 1959; it was isolated from skin lesions in a clawed frog and named after the offi cial species designation of the frog, Xenopus laevis (1). Thereafter, these slowgrowing mycobacteria have been recovered from heated water systems in many countries and more recently from natural waters in Finland (2). Transmission to humans is believed to originate from the environment, through aerosol inhalation or ingestion. Human-to-human transmission and transmission from animal reservoirs remain controversial because these routes have not been proven by molecular typing (3,4).
Its survival in fl owing water systems and resistance to common disinfectants enables M. xenopi to contaminate laboratory samples and medical devices such as bronchoscopes, thus causing healthcare-acquired (pseudo) infections and laboratory cross-contaminations (3,6,8,9). Differentiating true infection from pseudoinfection is of paramount importance because treatment of M. xenopi infections is time-consuming and often complicated. The British Thoracic Society (BTS) trial in 2001 established that treatment for pulmonary infections should consist of a 2-year course of rifampin and ethambutol; regimens including macrolides or fl uoroquinolones are still being investigated (10). The American Thoracic Society (ATS) established general criteria for the diagnosis and treatment of nontuberculous mycobacterial, not specifi cally M. xenopi, infections. The treatment guidelines are similar to those by the BTS, although the ATS guidelines advocate macrolidecontaining regimens (5).
To assess frequency and clinical relevance of M. xenopi isolation and its determinants in the Netherlands, we performed a retrospective case study. We used the ATS diagnostic criteria available during the study period to differentiate true infection from pseudoinfection.

Methods
To determine clinical relevance, we examined medical records of all patients in the Netherlands from whom M. xenopi had been isolated from January 1999 through March 2005. The following variables were extracted from the records: sex, age, predisposing factors, symptoms, chest imaging results, treatment and outcome, and drug susceptibility and status according to the ATS diagnostic criteria (5).

Mycobacterium xenopi Clinical Relevance and Determinants, the Netherlands
Laboratory diagnosis of the isolates was made by the Dutch National Institute of Public Health and the Environment (RIVM) or by a local hospital laboratory. RIVM acts as the national reference laboratory that provides identification, drug-susceptibility testing, and genotyping of mycobacterial isolates for all hospitals and other healthcare institutions in the Netherlands. To identify a mycobacterial isolate, Hain GenoType MTBC line-blot (Hain Lifescience, Nehren, Germany) was used after PCR-based amplifi cation to determine whether an isolate was a member of the M. tuberculosis complex. If the reaction was negative, an INNO-LiPA MYCOBACTERIA (Innogenetics, Gent, Belgium) reverse hybridization multiple DNA probe assay was used to differentiate between the more common species of nontuberculous mycobacteria, including M. xenopi (11). Before 2004, 16S rRNA gene sequence analysis was performed, after ruling out membership in the M. tuberculosis complex, by using the AccuProbe MTB DNA probe kit (GenProbe, San Diego, CA, USA). The result was compared with the RIVM and BLAST (National Center for Biotechnology Information, www.ncbi.nlm.nih.gov) 16S rRNA gene sequence databases (12).
On the basis of the results at position 90 in the 151-bp hypervariable region of the 16S gene, 2 M. xenopi genotypes were discerned; a C at position 90 distinguished M. xenopi I and a T distinguished M. xenopi II. Retrospectively, the 16S rRNA genes of all M. xenopi isolates at RIVM were sequenced and assigned to their respective genotypes.
Susceptibility testing was performed by using an agar dilution method (13). Drugs in the susceptibility testing panel were isoniazid, rifampin, ethambutol, streptomycin, cycloserine, prothionamide, amikacin, ciprofl oxacin, clofazimine, clarithromycin, and rifabutin. The Pearson χ 2 test was used for statistical correlations. The local ethics committee approved the study.

Results
We found 49 patients with nontuberculous mycobacterial infection (Table 1) Clinical signs and symptoms varied widely and were not associated with fulfi llment of the ATS diagnostic criteria (Table 1). Chest radiographs were taken for all patients, except for 2 who had spinal infection (Table 1). Cavitation was the only radiographic fi nding signifi cantly associated with fulfi llment of the ATS diagnostic criteria (OR 14.3, 95% CI 2.7-75.6, p = 0.001). Results of additional computed tomography scanning, performed for 27 patients, were not associated with fulfi llment of the ATS diagnostic criteria (data not shown).
We found 4 cases of extrapulmonary disease, 2 cases of pleural M. xenopi infection, and 2 cases of spondylodiscitis (in HIV-co-infected patients). The pleural infections were diagnosed by biopsy of pleural tissue for 1 patient and repeated culture of pleural fl uid for the other, after chest radiograph demonstrated pleural thickening and fl uid collection. The spinal infections were diagnosed by bone biopsy. In the pleural and bone biopsy specimens, granulomatous lesions with central necrosis were observed.
Treatment was started for 25 of 49 patients, of whom 19 met the ATS diagnostic criteria. Therapy consisted of medication for 21 patients, surgery for 2, or both for 2. Surgery consisted of lobectomy, pulmonary wedge resection, Clagett pleurostomy, and vertebral surgery with psoas muscle abscess drainage. Medication regimens varied widely but generally included rifampin, isoniazid, ethambutol, clarithromycin, ciprofl oxacin, and pyrazinamide in various 3-to 4-drug combinations. Duration of therapy varied between 5 days and 2.5 years, with a mean duration of 9 months. Macrolides were included in regimens for 58% and quinolones for 37% of the patients who met the ATS diagnostic criteria and received drug treatment.
Antimycobacterial treatment cured 11 (58%) patients who met the ATS diagnostic criteria: 7 with M. xenopi II, 2 with M. xenopi I, and 2 with M. xenopi I and II. We defi ned cure as resolution of symptoms and negative cultures after fi nishing treatment, until the end of our study period (range 0-60 months, median 25 months). Treatment failure, defi ned as protracted culture positivity for M. xenopi during and after adequate treatment, was noted for 4 (21%). Four other patients died. Treatment failure or death was not associated with genotype, susceptibility pattern, predisposing conditions, or radiographic imaging results.
Although they fulfi lled the ATS diagnostic criteria, 4 patients did not receive treatment. Of these, 1 recovered spontaneously; 2 remained positive for acid-fast bacilli, culture, or both throughout the study period; and 1 died. These outcomes were not associated with specifi c genotypes or patient factors.
Susceptibility testing was performed for 47 isolates from 42 patients. For 5 patients, cultures failed to grow for testing; for 2 others, cultures were not available. Results for isoniazid, rifampin, and ethambutol are shown in Table  2. Isolates were susceptible to all other compounds tested. Susceptibility testing of follow-up cultures was performed for 5 patients. M. xenopi bacteria in 2 patients treated with rifampin became resistant to rifampin and to ethambutol in 1 patient. For 9 patients, susceptibility testing results infl uenced the treatment regimens, mostly by inclusion or exclusion of rifampin and ethambutol or by adding a quinolone or macrolide agent.
For 6 patients with cavitation visible on chest radiograph, of whom 4 met the ATS diagnostic criteria, fungi were cultured simultaneously (Aspergillus fumigatus from 4, A. fl avus from 1, and Scedosporium apiospermum from 1). Antifungal treatment was initiated for 4 patients, which meant true nontuberculous mycobacteria infection was left untreated for 2.
Four patients who had received antimycobacterial treatment for M. xenopi infection before (mean duration 8 months) had relapses; the mean interval between discontinuation of drug treatment and relapse was 28 months (range 12-39). We found no evidence of geographic clustering, which suggests nosocomial transmission or a pseudo-outbreak. The number of new isolates per year remained steady at ≈8 per year. At least 5 patients were treated in preventive isolation for 2-15 days until M. tuberculosis complex infection was excluded by PCR.

Discussion
Clinical relevance of M. xenopi isolation, defi ned by fulfi llment of the ATS diagnostic criteria, was likely in 51% of patients; mycobacterial genotype II was a major determinant. To our knowledge, this phenomenon and its causal mechanisms have not been described. If further evidence emerges, 16S rRNA gene sequencing may become a relevant addition to the diagnostic algorithm of M. xenopi infection.
The ATS diagnostic criteria are designed for M. avium, M. kansasii, and M. abscessus infections, although the authors state "there is no reason to believe these criteria would not be applicable to other species" (5). Because the BTS statement focuses on management rather than specifi c diagnostic criteria (14), the ATS diagnostic criteria are recommended for the clinical setting. Of the main ATS diagnostic components, 2 were each signifi cantly associated with true infection in our study (cavitary lesions on chest radiograph and acid-fast bacilli on primary samples), thereby supporting the ATS criteria.
The ATS diagnostic criteria, however, have 1 limitation. Patients with pre-existing cavitary lesions are likely to have respiratory symptoms; they meet the radiologic criteria and are more likely to harbor mycobacteria in the cavity, which are not necessarily responsible for their symptoms and cavity formation. Cavity characteristics cannot reliably predict the cavity's origin or pathogenesis (15). The uncertainty is compounded when fungi are cultured simultaneously, which suggests matching requirements for in vivo success of these microorganisms or selective impaired immunity. Determining which organism causes disease in the patient is diffi cult.
The undertreatment and overtreatment that we noted indicates a relative lack of knowledge in physicians, mainly those specializing in pulmonary conditions, concerning nontuberculous mycobacteria infections. Unnecessary drug treatment could harm the patient in terms of adverse effects and costs (16), and undertreatment of patients who fulfi ll the diagnostic criteria is potentially harmful to the patients' health.
The baseline characteristics of our study group are similar to those in studies of nontuberculous mycobacteria patients in societies with low HIV prevalence (5,10). The North American series included more HIV-infected patients, which lowered the mean age of patients in these studies (17,18). Pre-existing pulmonary diseases are major predisposing factors and may be causally associated with isolation of M. xenopi. However, M. xenopi may have been isolated more often because physicians were more focused on mycobacterial cultures for this category of patients.
Minor predisposing factors were HIV infection or other causes of impaired immunity, which, in addition to the causal relationship, probably refl ect the low prevalence of HIV infection compared with chronic pulmonary disease in the Netherlands. Also, because most HIV-infected patients receive highly active antiretroviral therapy, fewer cases of severe HIV immunosuppression and its co-infections are seen. HIV infection predisposes patients to extrapulmonary M. xenopi infection, especially spinal infection (6,19). Those with HIV-associated spinal M. xenopi infection had rising CD4 counts after starting or changing highly active antiretroviral therapy regimens (Table 1). Possibly, their M. xenopi infection was an expression of immune reconstitution infl ammatory syndrome (20). We found no previous reports of pleural infections.
Although the treatment regimens recorded in our study were not in accordance with the current standards of the BTS (14) and ATS (5), cure rates were high. Although partly the result of the restricted defi nition of "cure" resulting from our research methods, this fi nding does bring the validity of the ATS diagnostic criteria into question. Despite meeting these criteria, some patients might have cleared M. xenopi infection without treatment. This possibility is endorsed by the spontaneous recovery of a minority of patients who met the ATS criteria but were not treated. Alternatively, the regimen of 24 months of rifampin and ethambutol advised by the BTS may not be better than similar regimens of shorter duration. The addition of macrolides and quinolones to therapy regimens might also account for the high cure rates after relatively short treatment durations.
Susceptibility testing results were similar to those published previously (21), but their value in clinical practice is uncertain. Interpretation of the laboratory results is diffi cult because of discrepancies between in vitro susceptibility and in vivo response to treatment (5,21). In our study, results rarely infl uenced treatment regimens; when they did, choice of regimen was controversial and not in accordance with ATS and BTS guidelines (5,14). Increasing use of PCR to rule out M. tuberculosis infection can be valuable for preventing or shortening patient isolation.
In conclusion, clinical isolation of M. xenopi was relevant for 51% of the patients; mycobacterial genotype was a major determinant. Currently, the ATS diagnostic criteria are the best tool for determining clinical relevance. We strongly recommend increased awareness of these diagnostic criteria and management guidelines by ATS and BTS.