Characterization of the Catalase-Peroxidase Gene ...

196

Characterization of the Catalase-Peroxidase Gene (katG) and inhA Locus in Isoniazid-Resistant and -Susceptible Strains of Mycobacterium tuberculosis by Automated DNA Sequencing: Restricted Array of Mutations Associated with Drug Resistance James M. Musser, Vivek Kapur,* Diana L. Williams, Barry N. Kreiswirth, Dick van Soolingen, and Jan D. A. van Embden

Section of Molecular Pathobiology, Department of Pathology, Baylor College of Medicine, Houston, Texas; Laboratory Research Branch, Gillis W. Long Hansen's Disease Center, Louisiana State University, Baton Rouge, Louisiana; Tuberculosis Laboratory, Public Health Research Institute, New York, New York; Unit Molecular Microbiology, National Institute of Public Health and Environmental Protection, Biltoven, Netherlands

The resurgence of tuberculosis in the United States is well documented [1]. Commensurate with the rise in the number of disease cases has been the increased recovery of Mycobacterium tuberculosis isolates resistant to one or more of the primary antimicrobial agents (e.g., rifampin, isoniazid, streptomycin) used as therapy [2]. Increasing drug resistance, the lack of efficacious new antitubercular medications, and the need to formulate rapid strategies for susceptibility testing have provided the impetus for substantial research devoted to elucidating the molecular genetic basis of resistance [3]. Among the recent discoveries is the finding that 96% ofrifampin-resistant M tuberculosis isolates have mutations in an 8l-bp region of the gene encoding the f3 subunit of RNA polymerase (rpoB) [3]. The molecular genetic basis of resistance to isoniazid is less well understood. In 1954, Middlebrook [4] observed that highly resistant strains of M tuberculosis lacked or had greatly decreased catalase activity. This finding and biochemical evidence implicating catalase-peroxidase in the antitubercular ac-

Received 24 May 1995; revised 22 August 1995. Grant support: National Institutes of Health (AI-37004 and DA-09238 to J.M.M. and AI-35274 to D.L.W.). Reprints or correspondence: Dr. James M. Musser, Section of Molecular Pathobiology, Dept. of Pathology, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030. * Present affiliation: Department of Veterinary Pathobiology, University of Minnesota, St. Paul. The Journal ofInfectious Diseases 1996;173:196-202 © 1996 by The University of Chicago. All rights reserved. 0022-1899/96/7301-0027$01.00

tivity of isoniazid [5] led to the cloning and characterization ofthe gene (katG) encoding this enzyme [6, 7]. The M tuberculosis enzyme is an 80-kDa protein with substantial homology to hydroperoxidase I from Escherichia coli, catalase-peroxidase from Mycobacterium intracellulare [8], and other microbial catalase-peroxidases, The observation that 2 of 3 isoniazidresistant patient isolates lacked katG sequences [6] initially suggested that gene deletion would be a common resistance mechanism, but it was soon shown that most isoniazid-resistant strains had a grossly intact katG gene [9]. These observations implied that for most strains, either simple base-pair changes resulting in (for example) missense or termination mutations or small deletions were associated with resistance. The discovery that most isoniazid-resistant M. tuberculosis strains did not have gross katG deletions suggested the need to more precisely analyze the structure of katG present in susceptible and resistant organisms, and this has been accomplished by several groups [10-14]. Some investigators have reported substantial katG allelic variation in both isoniazidsusceptible and -resistant strains, and a large number of distinct missense and other mutations have been associated with resistance [11, 12]. In striking contrast, sequencing of katG from a total of? resistant and susceptible strains revealed no mutations in this gene [10]. Moreover, other investigators have reported a restricted group oflargely katG missense mutations in isoniazid-resistant M tuberculosis strains [13, 14]. Hence, two divergent views have emerged regarding the level of katG allelic variation in natural isolates of susceptible and resistant M. tuberculosis and the characteristics of katG variants associated

Downloaded from jid.oxfordjournals.org by guest on July 26, 2011

The catalase-peroxidase gene (katG) and a two-gene locus (inhA) containing mutations associated with resistance to isoniazid in Mycobacterium tuberculosis were sequenced in 34 resistant and 12 susceptible strains. Virtually all resistant organisms had amino acid changes in KatG or nucleotide substitutions upstream of inhA. A region of katG encoding two amino acids frequently altered in resistant strains (residues Ser31S and Arg463) and the inhA locus were sequenced in 10 susceptible and Sl isoniazid-resistant isolates from the Netherlands. Most (84%) of the resistant isolates had mutations in katG or the inhA locus or lacked katG. Together, ~7S% of isoniazid-resistant isolates had replacements at amino acids 31S or 463 in KatG or nucleotide substitutions upstream of inhA. All 16 strains of Mycobacterium bovis and Mycobacterium microti studied had Leu463 rather than Arg463in KatG, an observation consistent with the hypothesis that Leu463is the ancestral condition in M. tuberculosis.

JID 1996;173 (January)

Genetics of Isoniazid-Resistant TB Strains

Materials and Methods Bacterial strains. The katG gene and the inhA locus were sequenced in entirety from 46 epidemiologically unrelated strains,

including 34 resistant to isoniazid. These 34 strains were recovered from patients living in diverse localities, including the United States (10 isolates), Brazil (3), Belgium (1), Rwanda (4), Yemen (4), Japan (9), and the Philippines (3). The 12 isoniazid-susceptible organisms were from Brazil (4), the Netherlands (2), and Belgium, Burundi, Hong Kong, Mexico, New York, and Texas (l each). Virtually all isolates were also resistant to rifampin. Previous characterization of a segment ofthe rpoR gene revealed that a heterogeneous array of distinct mutations was present in these organisms [17, 18]. IS6110 subtyping done on one-half of the isolates in this group revealed that each was unique. These results support the epidemiologic independence ofthe sample isolates. Resistance was determined by the 1% proportion method, usually on Middlebrook 7Hll agar; however, for some strains, susceptibility testing was done on Lowenstein-Jensen media [19]. These strains had isoniazid MICs ;;:: 1.0 j.tg/mL. A second strain sample comprised 61 isolates from patients in the Netherlands: 10 were susceptible to isoniazid, and 51 were resistant to isoniazid alone or to isoniazid plus one or more other antimicrobial agent(s). Susceptibility testing of these 61 isolates was done using the absolute concentration method with 0.2 j.tg of isoniazid/mL. For these bacteria, the entire inhA locus and katG nucleotides 900-1470 (encoding amino acids 301-490) were determined. This katG region encompasses sequences encoding amino acids 315 and 463, residues implicated in isoniazid resistance by several investigators [13, 14]. A total of 48 unique IS6110 patterns was identified in these organisms, which means that most isolates are epidemiologically unrelated. A third group of organisms studied included 45 multidrug-resistant M. tuberculosis isolates cultured from patients in New York City. These strains were resistant to isoniazid, rifampin, ethambutol, and streptomycin and have the same or closely similar IS6110 restriction fragment length polymorphism (RFLP) pattern (type "W," with 17 IS6110 copies, and "WI," with the same 17 copies, plus one additional hybridizing band) and are presumed to be epidemiologically related as a consequence of sharing a recent common ancestor. For the W and WI strains generally, MICs of isoniazid are >4 j.tg/mL, as determined by the BACTEC radiometric method [19]. Seven strains of M. bovis and 9 of Mycobacterium microti recovered from diverse sources were analyzed. The M. bovis were cultured from unassociated humans (n = 3), seals (n = 2), and deer (n = I), and the M. microti were taken from unassociated voles in England (n = 6), a rock hyrax in South Africa, and a pig in the Netherlands. One M. microti isolate was cultured from an unknown source. RFLP analysis ofthese isolates revealed a heterogeneous array of subtypes (unpublished data). Preparation of M. tuberculosis complex DNA. DNA was isolated from cultures grown in liquid or on solid media by previously described methods [20] and shipped to the laboratory at Baylor College of Medicine for analysis. Strain growth and DNA isolation were done in laboratories equipped with biosafety level 3 facilities. Automated DNA sequencing of katG and the inhA locus. The oligonucleotide primers and polymerasechain reaction (PCR) conditions used to amplify the entire katG gene and inhA locus have been described [10]. The locations of the internal primer sites used for katG sequencing are availableby request. Sequencingreactionswere done with a cycle sequencingkit (Taq DyeDeoxyterminator; Applied Biosystems, Foster City, CA), and the data were generated by an


with isoniazid resistance. As a consequence, there is clearly a need for additional molecular genetic data addressing the nature and extent of katG allelic variation in isoniazid-susceptible and -resistant isolates from diverse sources. A further complicating issue has arisen due to the identification of a second genetic locus conferring both isoniazid and ethionamide resistance in the laboratory when present on a multicopy plasmid [15). The inhA locus consists of two contiguous open-reading frames (ORFs), designated orfl and inhA, encoding 25.7-kDa and 28.5-kDa proteins, respectively, which may participate in fatty acid biosynthesis [16]. (As used herein, the term inhA locus denotes the two contiguous ORFs and the upstream presumed regulatory region. The preferred designation for orfJ is now mabA [mycolic acid biosynthesis].) In M tuberculosis and Mycobacterium bovis, the two genes are separated by a short (21 bp) noncoding region that lacks a readily identifiable promoter [15]. It is thought that mabA and inhA together constitute a two-gene operon that is transcribed from a promoter located upstream of mabA. The identification of an inhA gene missense mutation (Ser94- Ala94) that conferred isoniazid and ethionamide resistance in the laboratory together with the observation that most rifampin-resistant strains had missense mutations in a defined region of rpoli [3] led to the plausible hypothesis that missense mutations in the inhA gene would constitute a very abundant cause of resistance among M. tuberculosis strains recovered from patients. However, sequencing of the entire inhA gene, including the upstream presumed regulatory region, in 37 isoniazid-resistant organisms revealed that with only 1 exception, all strains had the identical wild type inhA allele, and none contained the Ser94- Ala94 substitution conferring resistance in the laboratory [10). Variation was also examined in the 744bp mabA gene in 24 resistant patient isolates, and substitutions were identified at two nucleotides flanking a presumed ribosomal binding site in 4 isolates [10). On the basis of these studies, it was suggested that the base-pair substitutions flanking the presumed mabA ribosomal binding site result in altered regulation ofMabA or InhA (or both) in some M tuberculosisresistant strains. Despite these advances, there is a need for further exploration of several questions related to isoniazid resistance in M tuberculosis. Because of the considerable variance in the reported data [10-14], more information is needed about katG allelic variation in resistant and susceptible organisms. In addition, relatively few strains have been sequenced in entirety for both katG and the inhA locus. There is also a need for characterization of katG and inhA locus variation in a more extensive geographic sampling of M tuberculosis. The goal ofthe present study was to address these issues.

197

198

Musser et al.

automated DNA sequencing instrument (model 373A; Applied Biosystems). Both strands of the target region were sequenced. The sequence data were assembled and editedelectronically with EDITSEQ, MEGALIGN, andSEQMAN computer programs (DNASTAR, Madison, WI) by a person blinded to the drug susceptibility phenotypes of the strains. All mutations not previously described by our group or otherinvestigators wereconfirmed by resequencing the gene segment containing the variant region.

Results

vious characterization of the inhA locus from 31 resistant strains revealed that 4 organisms had nucleotide substitutions in regions flanking a presumed ribosomal binding site located in the region upstream of mabA in the inhA locus [10]. Hence, sequence variation was assessed in the inhA locus in the 46 isolates for which the entire katG gene was characterized: All had the identical wild type sequence, except 1 organism that had a C-T substitution on the 5' side of the presumed ribosome binding site [10]. Variation in katG and the inhA locus in the Netherlands strain sample. To investigate the frequency of occurrence of katG and inhA locus mutations in a large sample of unassociated organisms recovered from a defined geographic area, 61 M. tuberculosis strains cultured from patients in the Netherlands were studied for nucleotide variation in a segment of katG (nucleotides 900-1470) containing codons 315 and 463 and the entire inhA locus. This area of katG was chosen for investigation because, as noted above, codons 315 and 463 were frequently polymorphic in the 34 resistant strains sequenced in entirety. Among the 10 isoniazid-susceptible organisms studied, only 1 had a sequence alteration (G-T at nucleotide 1388 in codon 463) in the katG region characterized, and none had inhA locus variation. Of 51 isoniazid-resistant M. tuberculosis strains, 3 failed to yield an amplified PCR product with multiple primer sets specific for katG, 25 had an AGC-ACC (Ser-r-Thr) change alone in codon 315, 8 had an isolated CGG-CTG (Arg-r-Leu) alteration in codon 463, 4 had both the codon 315 and 463 substitutions, and 11 had no changes in this katG region. Eleven of 51 isoniazid-resistant organisms had a C-T substitution flanking the 5' side of the presumed ribosome binding site [10]. None of the 29 organisms with a codon 315 mutation had this inhA locus alteration. Five of the strains with a codon 463 mutation had the mabA upstream substitution, and 6 strains with a wild type katG sequence in the region studied had the upstream substitution. In summary, in this sample of isoniazidresistant bacteria, 43 (84%) of 51 organisms either lacked katG or had a mutation in katG in codon 315 or 463 (or both) or a substitution flanking the presumed ribosome binding site region of the inhA locus. Variation in katG in the New York City sample: unique katG codon 315 alteration. All 45 multidrug-resistant strains from New York City patients with either the W or WI IS6l10 polymorphism pattern had an allele of katG identical to that characterized by a missense mutation at codon 315 resulting in a Ser-v'Ihr amino acid replacement. The change was an AGC-ACA dinucleotide substitution rather than the AGC - ACC (Thr) or AGC-AAC (Asn) present in all other resistant organisms in the total sample with an alteration in codon 315. The katG codon 463 polymorphism. Morris et al. [8] observed that an isolate of M. intracellulare had the sequence CTG (Leu) at codon 463, and Heym et al. [13] and Morris et


katG variation among 46 strains sequenced in entirety: basepair substitutions. A total of 22 allelic variants was identified in the 46 strains characterized for the entire katG gene (figure 1, table 1). Compared with a published katG sequence [7] deposited in GenBank under accession number X68081, 22 individual nucleotide sites were polymorphic due to simple base-pair substitutions (figure 1). The sequence of three of the sites (nucleotides 701, 2152, and 2181) differed from the deposited sequence in all strains examined in our sample. These variant nucleotides will not be considered further in this paper. Of the remaining 19 nucleotide sites that we consider to be polymorphic, 17 are characterized by missense substitutions (i.e., base-pair changes that result in amino acid variation). One polymorphic site (nucleotide position 1431) had a G- A substitution in 1 strain, which results in formation of a TGA stop codon from the wild type TGG (Leu) codon, and one site (bp 1653) had a GCC-GCT alteration that does not affect the primary amino acid sequence. Only four of the polymorphic nucleotide sites (bp 281, codon 94; 944, codon 315; 1388, codon 463; and 2129, codon 710) were variable in > 1 isolate. Twenty strains had either an AGC-ACC (Ser-r-'Ihr; n = 17), AGC-AAC (Ser-r Asn; n = 2), or AGC-ACA (Ser-r-Thr; n = 1) alteration in the codon for amino acid 315. In addition, 13 strains had a CGG-CTG (Arg-r- Leu) alteration at nucleotide position 1388 in codon 463, 2 had a GTC-GCC (Val-Ala) mutation at nucleotide 2129 in codon 710, and 2 had a GAC-GCC (Asp-Ala) mutation at nucleotide 281 in codon 94 (figure 1). katG variation among 46 strains sequenced in entirety: one deletion and one insertion identified. In addition to allelic variation due to simple nucleotide substitutions, the analysis identified one episode each of a deletion and insertion of short base-pair stretches (figure 1). Strain RRl, recovered from a patient in Japan, had a 12-bp deletion (GCCGGGGGCGGC) encoding amino acid residues 122-125. Strain YE7 from Yemen had a 42-bp insertion located between nucleotides 1533 and 1534, which is an exact copy of nucleotides 1534-1575 in the GenBank katG sequence (X68081). inhA locus variation among 46 strains sequenced in entirety. Although substitution of amino acid 94 (Ser-r- Leu) in the InhA protein confers resistance to isoniazid and ethionamide in the laboratory [15], only 1 patient isolate has been described that has an inhA locus coding region alteration [10]. However, pre-

JID 1996;173 (January)

BP Pos

281

296

codon

94

99

pub bp

strain 026 RR3 RR6 RR7 RRI0 029 565 2965 3046 3194 3662 8651 8873 P.03 P.52 3237 5095 RR4 YE67 YF:1 YE68 15015 P.33 4183 4325 4967 8893 9004 021 RRI RR8 RR9 RRll YEI6 15273 15402 3976 9052 T443 M4330 HK540 brl66 brl82 br265 br483 9072 codon change AA change

A

G

364375 122125 12bp

412

413

419

425

479

514

539

944

945

982

1388

1431

1533

1543

1653

1700

1778

2129

138

138

140

142

160

172

180

315

315

328

463

477

511

515

551

567

593

710

A

A

G

A

C

G

C

G

C

T

G

C

G

C

T

G

R

R R R R R R R R R R R R R

R R R R R R R R R R R R R R R R R R R S S S S S S S S S S S S

~

-

'C)

T

..0\

- .l v.>

INH R

§

'