The Evolution of Drug Resistance in Mycobacterium tuberculosis: From a Mono–Rifampin-Resistant Cluster into Increasingly Multidrug-Resistant Variants in an HIV-Seropositive Population

Rifampin (RIF) and isoniazid (INH) are the 2 most effective bactericidal agents in the treatment of tuberculosis (TB) and, hence, are the cornerstone of present-day TB therapy. The loss of RIF seriously compromises the duration and cost of treatment and is associated with higher rates of mortality and treatment failure in patients undergoing short-course chemotherapy [1]. The emergence of mono–RIF resistant (RIF^R) M. tuberculosis strains is not common, given that RIF is generally administered in combination with at least INH and/or other anti-TB agents.

Two significant observations have been reported with regard to the development of mono-RIF^R. The first suggests that mono- RIF^R isolates seldom result from transmission, an observation supported by genotyping data that indicate that RIF^R isolates are often genetically distinct [2, 3]. The second indicates that the emergence of RIF monoresistance is associated with advanced levels of immunosuppression in HIV-seropositive populations [3, 4] before the administration of highly active antiretroviral therapy (HAART).

Immunocompromised hosts may allow the persistence and propagation of RIF^R M. tuberculosis strains that have mutations associated with varying levels of bacterial fitness [5]. Here, we investigate the genotypic and phenotypic characteristics of a mono-RIF^R M. tuberculosis cluster and related strains isolated from a predominantly immunocompromised cohort. Molecular markers were used to track the acquisition of drug resistance from a mono-RIF^R clone to polyresistant and various multidrugresistant (MDR; resistant to at least INH and RIF) forms. This study highlights the importance and potential consequences of mono-RIF^R M. tuberculosis in terms of transmission and accumulation of additional drug resistance, particularly within immunocompromised populations.

Methods. All MDR and RIF^R isolates from patients with TB counted by the New York City Department of Health and Mental Hygiene (NYC-DOHMH), Bureau of Tuberculosis Control, since 1 August 1993 were submitted and analyzed by IS6110- based restriction fragment-length polymorphism (RFLP) genotyping at the Public Health Research Institute Tuberculosis (PHRI TB) Center [6]. In addition, MDR and RIF^R M. tuberculosis isolates collected as part of a population-based study in the state of New Jersey between 1996 and 2001 were available for this investigation.

Patient demographic and basic clinical data were obtained for 33 patients infected with AU strains from the NYC-DOHMH, Bureau of Tuberculosis Control, and the New Jersey Department of Health and Human Services. The study protocol was approved by the Institutional Review Board of the NYCDOHMH.

Primary grouping of the M. tuberculosis isolates into the AU cluster was based on IS6110 RFLP-related patterns, using a hybridization pattern similarity of >80% (BioImage Whole Band Analyzer software, version 3.1; Genomic Solutions) [7]. Principal genetic group identification, spoligotyping, synonymous single-nucleotide polymorphism (sSNP) clustering, and mycobacterial interspersed repetitive unit analysis were used to confirm and further refine the IS6110 clustering of AU isolates and to determine their broader phylogenetic relatedness to other M. tuberculosis clusters [8].

Resistance of clinical isolates to streptomycin (STR), INH, RIF, and ethambutol (EMB) was determined using Quad Plate I and II (Bactec system; Becton Dickinson); testing of susceptibility to pyrazinamide, paraaminosalicylic acid (PAS), ofloxacin (OFL), and cycloserine (CYC), as well as MIC estimation, was performed by the diffusion assay and Etest (AB Biodisk).

For DNA sequencing of drug-target regions and putative mutator genes, comparative sequence analysis of target regions of the genes rpsL, katG, and rpoB (which are associated with drug resistance to STR, INH, and RIF, respectively)—as well as mabA-ihnA, ahpC, and the mutator genes mut2 (Rv1160), mut3 (Rv1316), mut4 (Rv3908), and ogt (Rv0413)—was performed as described elsewhere [9, 10].

Fluctuation tests in the presence of STR were done to determine the intrinsic mutation rate of isolates AU-RIF^R, AU4-RIF^S, W4, H37Rv (ATCC25618), H37Ra (ATCC35835), and CDC- 1551 as a control group of strains. The protocol for selection of spontaneous mutants was adapted from Luria and Delbruck [11] and from Morlock et al. [12]. Briefly, for each of the 6 strains, a sample was cultured under standard conditions in Sauton medium for 3 weeks, and the bacterial density was adjusted to an optical density (OD) of 1.2 at 600 nm (∼7.2 × 10⁷ cfu/mL). Then the sample was diluted to an OD₆₀₀ of 0.01, subcultured in 24 tubes (5 mL; 7H9) each for 32 days, and plated in toto on 7H11 plates containing STR (2 µg/mL). The total number of organisms plated was 1 × 10¹¹. The number of spontaneous STRR mutants was compared for all of the strains tested to determine the relative frequency at which mutants emerge for each strain.

Four mutator genes (mut2, mut3, mut4, and ogt) suspected to be associated with an increased hypermutable phenotype [10] were sequenced in 3 AU-mono-RIF^R, 3 AU-MDR, 2 AU-MDR hyper-INH^R, and 3 susceptible AU4 isolates.

Results. Molecular analysis groups the AU cluster into a large phylogenetic family known as the “Haarlem 1 genotype from sSNP cluster III” [8] (table 1). Combining IS6110 RFLP genotyping with drug-susceptibility data, a M. tuberculosis mono-RIF^R strain cluster and its polyresistant and MDR progenies were identified. This cluster includes isolates with similar IS6110 RFLP patterns from 29 patients, arbitrarily labeled as the “AU-RIF^R cluster.” Additionally, 61 drug-susceptible isolates shared related IS6110-RFLP profiles, collectively labeled the “AUS cluster.” These included 6 closely related (by RFLP analysis) susceptible clusters—AU4 (n = 15), AU5 (n = 10), AU6 (n = 2), AU16 (n = 4), AU17 (n = 2), and AU23 (n = 10)— and 18 isolates with unique AU-like IS6110 profiles. The AU-RIF^R and AU^S strain clusters could be distinguished from each other phenotypically by their drug-susceptibility profile and genotypically by an additional IS6110 copy in the AU4 isolates (figure 1A). Of the 29 strains from the AU-RIF^R cluster, there were 2 isolates (AU14-RIF^R and AU8-RIF^R) that acquired 1 and 2 additional IS6110 insertions, respectively; similarly, an IS6110-associated genomic rearrangement and transposition event was noted in the drug-susceptible group (figure 1A). Twenty-nine AU-RIF^R isolates were resistant to a high concentration of RIF (>256 µg/mL). Sequence analysis of the target rpoB gene associated with RIF^R revealed an unusual duplication at codon ⁵¹⁴TTC→TTCTTC (⁵¹⁴Phe). This rare mutation was found only in the 29 AU-RIF^R cluster isolates and not in either the related 61 AUS cluster variants or in other genetically unrelated RIF^R isolates from the PHRI TB Center collection that were tested.

Six of 29 AU isolates were uniquely mono-RIF^R, whereas 1 was polyresistant (STR^R/RIF^R). The remaining 22 of the 29 AU isolates were resistant at least to INH in addition to RIF (i.e., were MDR). Four of the 22 AU-MDR isolates were found to have additional resistance profiles: 1 was EMB^RR, 1 EMB^R/PAS^R, 1 CYC^R, and 1 OFL^R.

Twenty-one of the 22 AU isolates displaying a RIF^R/INH^R phenotype had a single nucleotide substitution in the katG gene (³¹⁵AGC→ACC; Ser→Thr) associated with an INH MIC of 4 µg/mL. The one remaining AU isolate with a RIF^R/INH^R phenotype had a different nucleotide substitution at the same codon (³¹⁵AGC→AAC; Ser→Arg). On the basis of these data, it is not possible to resolve whether this AU isolate (³¹⁵Ser→Arg ) arose from an additional single-nucleotide substitution in ³¹⁵katG (³¹⁵Thr→Arg) or evolved directly from a mono-RIF^R isolate (³¹⁵Ser–Arg).

Three of 4 hyper-INH^R AU-MDR isolates were resistant to >256 µg/mL INH, and 1 was resistant to 64 µg/mL (table 1 and figure 1B). To identify the possible mutations associated with an elevated level of INH^R, the mabA-inhA region (including the promoter) and the complete katG gene (including 107 bp upstream) were sequenced in all hyper-INH^R mutants as well as in 6 AU-RIF^R strains. Six isolates had an additional SNP in 359katG (³⁵⁹GCC→GTC; Ala→Val), which did not correlate with the high INH MIC given that 4 of the isolates with this SNP retained a MIC of 4 µg/mL and that the remaining 2 displayed a hyperresistant phenotype (MIC of >256 µg/mL); no additional mutations were noted in these strains (i.e., mabA-inhA and ahpC). Three of the 4 patients carrying hyper-INH^R organisms were HIV seropositive and had pulmonary TB.

To determine whether the AU strains are predisposed to higher rates of mutations, we estimated the frequency of spontaneous mutations, as determined by the fluctuation test [11]. No significant differences in the number of emerging STRR mutants (7.× 10-7 for STR^R) between the 6 test strains were observed. These results suggest that the AU strains are not hypermutable and, hence, that the sequential acquisition of mutations associated with drug resistance is not strain dependent but rather represents the outcome of other environmental factors (e.g., the host-immune response or drug-drug interaction).

The isolates were collected from patients with TB at 18 institutions located in 5 New York City boroughs. Epidemiological data were available for 33 patients, including the 29 infected with AU-RIF^R organisms. Although epidemiological links could not be thoroughly assessed in this study, most of the patients shared common risk factors, including HIV seropositivity (79%), drug abuse (55%), and alcoholism (34%). All but one of the AU-RIF^R strains was isolated between 1993 and 1996, before the introduction of HAART [13]. The last isolate was identified in 1998.

Discussion. There are limited reports on mono-RIF^R TB cases in the literature. The reasons for the paucity of patients harboring or transmitting mono-RIF^R M. tuberculosis is not well understood. It is possible that mono-RIF^R organisms readily acquire additional resistance to become polyresistant orMDRand, therefore, are rarely identified. Here, the identification of a RIF^R M. tuberculosis cluster with a rare mutation in rpoB among a primarily immunocompromised population enabled us to study the clonal expansion of a mono-RIF^R strain into polyresistant, MDR, and INH-hyperresistant forms.

Sequencing of 29 AU-RIF^R isolates revealed a trinucleotide duplication (⁵¹⁴TTC→TTCTTC; ⁵¹⁴Phe) in rpoB. To our knowledge, this mutation has been reported in only 3 RIF^R strains, 2 originating from this same collection [9] and the other a laboratory mutant [12]. The combination of closely related IS6110 RFLP profiles and other genotyping markers with a unique rpoB mutation found in all 29 AU-RIFR isolates suggests that the acquisition of RIF^R took place once, before the dissemination and development of additional drug resistance, including INH-hyperresistant phenotypes. Other AU-related isolates (n = 61) were pansusceptible (AU^S) and had a wild-type rpoB gene sequence.

Population-based analysis of all MDR or AU-RIF^R M. tuberculosis isolates in New York City indicated that the frequency of AU-MDR or AU-RIF^R strains has since dropped (NYCDOHMH and PHRI TB Center, unpublished data). In contrast, their drug-susceptible counterparts (AU^S) have continued to disseminate and evolve (on the basis of IS6110 genotyping) while remaining drug susceptible. In contrast, mono–AU-RIF^R strains not only evolved into polyresistant and MDR forms but also developed hyperresistance to INH. However, the INH hyperresistance could not be explained by polymorphisms in katG or mabA-inhA or by in vitro mutational frequency.

The empirical association between HIV seropositivity and MDR-TB has led some to suggest that the immunocompromised state facilitates the emergence of drug resistance. Although host factors (e.g., malabsorption and disseminated disease) are often cited as reasons for this association, it is possible that, in immunocompromised patients with TB, bacterial mutations conferring drug resistance that typically result in the loss of relative microbial fitness are allowed to flourish [5, 14]. That is, drug-resistance-conferring mutations that have an accompanying physiological cost typically rendering the organism more susceptible to host killing may occur considerably less efficiently in an immunocompromised host. As a consequence, low-fitness bacilli as well as fit bacilli may be allowed to proliferate, and the former possibly develop compensatory changes that restore fitness [15]. Although the relative fitness cost of the rpoB ⁵¹⁴TTC duplication was not determined, its rarity in clinical isolates, coupled with a predominantly immunocompromised-patient profile, raises the possibility that host (i.e., suboptimal immune response) and bacterial (i.e., relative fitness) factors are involved, not those that are intrinsically mutable.

In conclusion, most patients in this retrospective study harboring an AU-RIF^R and/or AU-MDR strain were coinfected with HIV, suggesting that additional drug resistance may readily develop in HIV-seropositive individuals harboring RIF^R M. tuberculosis bacteria, thus contributing to the global rise of MDR-TB.

• Received October 8, 2007.
• Accepted January 16, 2008.

• © 2008 by the Infectious Diseases Society of America.