The 2 M. tuberculosis strains were isolated from a 68-year-old man from Portugal. He did not have HIV and was treated as a confined inpatient, limiting the possibility that this was not a true in vivo mixed infection. Using a novel approach, we correlated in vitro growth and treatment characteristics for the patient strains with the in vivo strain predominance and persistence of a less-fit, drug-resistant strain. All samples were obtained with approval from St. Mary Hospital’s (London, UK) Research Ethics Center (no. 07/H0712/85) and with the patient’s written informed consent.

Details of the patient samples are in the Table. The initial bronchoalveolar lavage smear sample was positive for acid-fast bacilli (AFB); culture results were positive for fully sensitive M. tuberculosis. Treatment with isoniazid, rifampin, ethambutol, and pyrazinamide was begun. Because of the patient’s alcohol use, his treatment was managed on an inpatient basis in a single-patient, negative-pressure room. Two months later, repeat sputum smears were positive for AFB, and culture results were positive for fully sensitive M. tuberculosis. After 4 months of treatment, the patient’s clinical signs had not improved, and his sputum smear was still positive for AFB. Culture results for the sputum sample were positive for M. tuberculosis resistant to isoniazid and ethambutol; a modified treatment regime resolved the infection, and the patient was released the following month, by which time his smear and culture results were negative.

Figure 1

Figure 1. . Mycobacterium tuberculosis co-culture competition experiment in a study of the amplification of multidrug resistance induced by first-line treatment of a mixed M. tuberculosis infection. The results suggest competitive advantages in...

We molecularly characterized the strains by using mycobacterial interspersed repetitive unit–variable number tandem repeat (MIRU-VNTR) typing (12); results showed that the drug-sensitive and drug-resistant M. tuberculosis strains were 2 distinct strains (Table) rather than 1 sensitive strain that had become resistant through mutagenesis. Because the patient was isolated while an inpatient, exogenous reinfection with a primary drug-resistant strain was ruled out. In addition, treatment compliance was directly observed, so in vivo development of drug resistance caused by poor compliance was also ruled out. Thus, it is highly likely that the patient was harboring a mixed infection of drug-sensitive and drug-resistant strains when he initially sought care at the clinic. Such a co-infection would not have been detected because single-colony or broth cultures are commonly used for strain typing, and these techniques would give the fastest growing strain a competitive advantage. Thus, we devised a competitive growth assay to determine if the patient had a mixed infection and to provide correlating in vitro and in vivo evidence of mixed infection (Figure 1).

Figure 2

Figure 2. . Analyses of the amplification of multidrug-resistant Mycobacterium tuberculosis during treatment of a drug-sensitive (S) strain in a mixed infection (i.e., infection with drug-resistant [R] and S strains). In the presence...

For the in vitro growth analyses of the 2 strains, we inoculated broth cultures and measured growth at an optical density of 600 nm, characterized the dominant strain by using MIRU-VNTR, and quantified colony-forming units on agar plates in the presence and absence of isoniazid. At several points during logarithmic growth, the drug-sensitive strain grew substantially faster than the resistant strain (Figure 2, panel A), suggesting that without the selective pressure of isoniazid, the sensitive strain would be most prevalent in a mixed infection.

For the in vitro competition assays, the strains were mixed (1:1), and isoniazid (0.2 μg/mL) was or was not added before measurement of growth and determination of the dominant strain. In the presence of isoniazid, the growth rate was lower, suggesting that the drug-resistant strain outcompeted the drug-sensitive strain to become the dominant strain (Figure 2, panel B). This was confirmed by MIRU-VNTR typing and growth analyses (Figure 2, panel C). These results indicate that 1) the drug-sensitive strain had a competitive growth advantage, causing this strain type to be identified as the sole infecting strain, and that 2) the drug-resistant strain gained the competitive advantage when isoniazid was added and became the predominant strain after treatment. These findings correlate precisely with the patient data (Table) and, we believe, is representative of the in vivo host infection.

We show the selection and subsequent clinically relevant emergence of a drug-resistant M. tuberculosis strain after treatment of a drug-sensitive strain in a patient with an initial mixed infection. This case illustrates the prospect of treatment failure for TB caused by mixed infection with strains with different drug susceptibility and growth rates. The proportion of cases of secondary multidrug resistance caused by such initial mixed infections is not known; however, the ability of the resistant strain to outcompete the sensitive strain under treatment and then to potentially transmit further may have substantial implications for the control and prevention of multidrug resistance.

The case also highlights the urgent need for improved diagnostic techniques that can routinely identify mixed populations of tubercle bacilli. Given the difficulty of detecting TB co-infections by using routine diagnostic microbiology techniques, co-infection is likely underrecognized. Co-infection can currently be ruled out only by using specialized techniques, such as molecular analysis of original sample (pre-culture); analysis of multiple colonies; or the GeneXpert assay (Cepheid, Sunnyvale, CA, USA) (13). Rapid detection of mixed infections with distinct drug susceptibility profiles would enable suitably tailored drug regimens from the start of treatment, which could prevent treatment failure and emergence and transmission of drug-resistant strains.