Detection of rpoB Mutations Associated with Rifampin Resistance in Mycobacterium tuberculosis Using Denaturing Gradient Gel Electrophoresis

Abstract

Denaturing gradient gel electrophoresis (DGGE) was used to probe for mutations associated with rifampin (RIF) resistance in the rpoB gene of Mycobacterium tuberculosis. DGGE scans for mutations across large regions of DNA and is comparable to DNA sequencing in detecting DNA alterations. Specific mutations are often recognized by their characteristic denaturation pattern, which serves as a molecular fingerprint. Five DGGE primer sets that scanned for DNA alterations across 775 bp of rpoB were developed. These primer sets were used to scan rpoB for DNA alterations in 296 M. tuberculosis patient isolates from the United States-Mexico border states of Texas and Tamaulipas. The most useful primer set scanned for mutations in the rifampin resistance-determining region (RRDR) and detected mutations in 95% of the RIF-resistant isolates compared to 2% of RIF-susceptible isolates. Thirty-four different alterations were observed within the RRDR by DGGE. In addition, isolates harboring mixtures of DNA within rpoB were readily detected by DGGE. A second PCR primer set was used to detect the V146A mutation in 5 to 7% of RIF-resistant isolates. A third primer set was used to detect mutations in 3% of RIF-resistant isolates, some of which also harbored mutations in the RRDR. Only 1 of 153 RIF-resistant isolates did not have a detectable rpoB mutation as determined by DGGE and DNA sequencing. These results demonstrate the power and usefulness of DGGE in detecting mutations associated with drug resistance in M. tuberculosis.

Drug resistance is a major concern in the global tuberculosis epidemic. Multidrug resistance, defined as resistance to at least isoniazid and rifampin (RIF), is greater than 10% in many countries and appears to be more widespread than previously documented (25, 26). Resistance to RIF is almost exclusively associated with mutations in the rpoB gene that encodes the β-subunit of RNA polymerase (17, 28). Over 70 distinct rpoB mutations have been reported for RIF-resistant Mycobacterium tuberculosis isolates worldwide (7, 17, 20, 28). Approximately 95% of RIF-resistant isolates harbor mutations in the rifampin resistance-determining region (RRDR), an 81-bp region within rpoB that spans codons 507 to 533. Mutations at the serine 531, histidine 526, and aspartate 516 codons have been observed in approximately 86% of RIF-resistant isolates and therefore represent hot spots within the RRDR (17, 28). Alterations outside of the RRDR have also been reported for the N-terminal, CII, and CIII regions of rpoB (6, 28) (Fig. 1).

FIG. 1.

Regions of rpoB scanned for RIF resistance mutations. The NI, NII, RRDR, CII, and CIII fragments that were amplified to scan for mutations in rpoB are shown. Codon numbers refer to alterations that were previously reported for RIF-resistant isolates (6, 28). Primer sequences and fragment properties are described in Table 1.

Molecular techniques are receiving increased scrutiny as alternatives to traditional drug susceptibility testing for M. tuberculosis (22). Molecular techniques can detect DNA alterations in hours or days, while assaying for drug resistance by culture methods takes weeks. Although the clustering of most mutations in the RRDR simplifies molecular analysis, the heterogeneity of mutations represents a challenge in detecting DNA alterations as a predictor of RIF resistance. These genetic considerations affect both the sensitivity and the specificity of molecular assays for drug susceptibility testing. Some molecular assays detect mutations as genetic variants and often focus on individual DNA alterations that represent mutational hot spots associated with drug resistance. Alternatively, other techniques scan for DNA alterations within a wider region of a gene in which drug resistance mutations occur. One of the latter techniques is denaturing gradient gel electrophoresis (DGGE), which distinguishes mutant amplicons from their wild-type equivalents on the basis of their altered melting temperature as the DNA fragments migrate through a gradient of denaturants (23). DGGE has been reported to be even more sensitive than DNA sequencing in detecting mutations in complex DNA samples (4, 19), and it can detect point mutations, small insertions, and deletions. Different mutational variants within a DNA fragment can often be recognized based on their characteristic denaturation pattern, which serves as a molecular fingerprint. For organisms like M. tuberculosis, with only one copy of each gene, heteroduplexing of the mutant DNA to a reference DNA (usually wild-type) generates additional homoduplex and heteroduplex bands that facilitate mutation detection. DGGE has been used successfully in the detection and discovery of mutations associated with RIF resistance within the RRDR (18). In addition, DGGE has been used to detect DNA alterations in the pncA gene in pyrazinamide-resistant M. tuberculosis isolates (12).

In this report, DGGE has been used to assay for RIF resistance mutations across 775 bp of rpoB. We screened a large set of M. tuberculosis isolates to build a database of rpoB DNA alterations and to identify a suitable combination of PCR target fragments that detect RIF resistance mutations. We detected mutations in 99% of RIF-resistant isolates within two DGGE target fragments. These results suggest that DGGE is a useful technique for detecting drug resistance mutations in M. tuberculosis.

MATERIALS AND METHODS

Bacterial isolates and drug susceptibility testing.

Isolates of M. tuberculosis were cultured and DNA was extracted as previously described (16). Strain genotyping was determined by restriction fragment length polymorphism analysis of IS6110 elements and by spoligotyping as previously described (16). Two hundred ninety-six isolates from individual patients were obtained from the U.S. state of Texas and from the bordering Mexican state of Tamaulipas (16). The use of archived materials was approved by the University of Texas Health Science Center at San Antonio Institutional Review Board (protocol 011-6000-105). Drug susceptibility testing was performed by the Texas Department of State Health Services Laboratories (formerly known as the Texas Department of Health Laboratories). Drug susceptibilities were determined primarily by the BACTEC460 liquid method, although some older isolates were analyzed by the agar proportion method alone. Thirty-five isolates were retested for RIF susceptibility. Half of these isolates harbored common rpoB mutations that were either previously reported to impart resistance to RIF (7, 17, 20, 28) or synonymous RRDR polymorphisms. The remaining 18 isolates lacked identified RRDR mutations by DGGE. Retesting of isolates was performed by the Texas Center for Infectious Diseases (San Antonio, Texas) using the BACTEC460 method. These isolates were assayed with three different concentrations (0.5, 2, and 8 μg/ml) of rifampin. Resistant isolates were defined as growing in 8 μg/ml of RIF, while susceptible isolates were unable to grow in 0.5 μg/ml. Isolates were revised if susceptibility retesting was consistent with DGGE genotyping of rpoB and DNA sequencing of rpoB and if the clinical laboratory records indicated variable and inconsistent results regarding RIF and rifabutin resistance. The susceptibility revisions were done exclusively for DGGE assay development and did not affect patient treatment.

DGGE and heteroduplex analysis.

Primer sets used for the scanning of rpoB mutations by DGGE are listed in Table 1. DGGE primers were designed using the SG Primer program (21). Primers containing 18 to 20 bp of M. tuberculosis sequence were designed to amplify the region of interest. One primer in each set also contained a 40- to 50-bp GC-rich portion (GC clamp) at its 5′ end to serve as the highest melting domain (21). The GC clamps, with a melting temperature of ∼95°C, prevent the strands of a PCR fragment from separating completely. Electrophoretic mobility of the partially melted structure slows dramatically on the denaturing gel at its characteristic melting temperature, which is exquisitely sequence dependent (21, 23). A standard 50-μl PCR consisted of 20 ng of genomic DNA, 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.1 mM MgCl₂, 0.01% gelatin, 0.2 mM concentrations of each deoxynucleoside triphosphate, a 0.2 to 0.8 μM concentration of each primer, 2% dimethyl sulfoxide, and 2 U of REDTaq DNA polymerase (Sigma Chemicals, St. Louis, MO). Standard PCR amplification consisted of 1 min at 94°C, followed by eight cycles of 94°C for 1 min, 65°C for 1 min, and 72°C for 1 min, followed by 27 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 1 min, followed by 10 min at 72°C. Amplified products were analyzed on 2% agarose gels in TBE (100 mM Tris, pH 8.4, 90 mM boric acid, 1 mM EDTA).

TABLE 1.

Primers used and regions scanned for rpoB mutations

Primer and region scanned	Forward	Reverse	bpb (codons)
DGGE primera
NI	5′-cgcccgccgcgccccgcgcccggcccgccgcccccgcccgCGAGAAGGGCACGTTCATCA-3′	5′-GCGCTTGTCGACGTCAAACT-3′	465-639 (140-180)
NII	5′-cgcccgccgcgccccgcgcccggcccgccgcccccgcccgGTCGACGCTGACCGAAGAAG-3′	5′-ACCACGCCGTCGACCACCTT-3′	972-1237 (414-485)
RRDR	5′-cgcccgccgccgcccgccgcgccccgcgcccgtcccgccgcccccgcccgCCGCGATCAAGGAGTTCT-3′	5′-GCACGCTCACGTGACAGA-3′	1255-1385 (507-536)
CII	5′-TGACCCACAAGCGCCGACTG-3′	5′-cgcccgccgcgccccgcgcccggcccgccgcccccgcccgaaataataaaCGATCGGCGAATTGGCCTGT-3′	1413-1627 (561-615)
CIII	5′-GTGGCACAGGCCAATTCG-3′	5′-cgcccgccgcgccccgcgcccggcccgccgcccccgcccgGCCTGGCGCTGCATGTTT-3′	1603-1826 (623-681)
Sequencing primer
NI	5′-AGGGCACGTTCATCATCA-3′	5′-TCGTAGCGCTTCTCCTTG-3′	480-902
NII	5′-TGGACATCTACCGCAAGCTG-3′	5′-TGATCAACGTCTGCGGTGTG-3′	806-1237
NII-RRDR-CII	5′-ACCGACGACATCGACCACTT-3′	5′-GTACGGCGTTTCGATGAACC-3′	1081-1530
RRDR-CII	5′-GTCGCCGCGATCAAGGAGTT-3′	5′-GATCGGCGAATTGGCCTGTG-3′	1252-1626
CIII	5′-GGCGTGGTTAGCGACGAGAT-3′	5′-CTCCTCGATGACGCCGCTTT-3′	1546-1938

^aPrimer sequences derived from rpoB are in uppercase letters, while DGGE-specific primer regions are in lowercase letters.

^bThe 5′ nucleotides of forward and reverse primers relative to rpoB open reading frame are given. The codons scanned by each target fragment represent the range of whole codons between the 3′ ends of primers and include a gap of 2 to 5 bp.

PCR products from isolates were mixed with an equal volume (8 μl) of the corresponding PCR product from the reference strain, H37Rv, unless noted otherwise. After complete denaturation and renaturation, this resulted in two homoduplex and two heteroduplex products. The heteroduplex reaction consisted of 98°C for 8 min, 0.1°C decreases per min to 55°C, 30 min at 55°C, 0.1°C decreases per min to 37°C, 20 min at 37°C, and 0.1°C decreases per min to 4°C. Samples were loaded onto denaturing gradient gels, usually 40 to 90% urea-formamide-8% acrylamide (19:1) with TAE (40 mM Tris, 20 mM acetic acid, 1 mM EDTA, pH 8) or 20 to 90% urea-formamide-8% acrylamide with TAE. Samples were electrophoresed at 60°C for 16 to 18 h at 100 V in an OptiAnalyzer 8000 electrophoresis tank (CBS Scientific, Solana Beach, CA) containing TAE buffer (13). After electrophoresis, gels were stained with ethidium bromide in TAE for 15 to 20 min, and DNA was visualized on a UV transilluminator. DNA samples that did not harbor mutations in the region scanned were visualized as a single band, while DNA samples that harbored mutations in the region scanned generated up to four bands (two homoduplex and two heteroduplex fragments) that arrested in the gel at distinctive concentrations of denaturant. Since roughly equal amounts of DNA were amplified from both the isolate and the reference strain, the four bands were of equal intensities. Bands of unequal intensities, or additional bands (>4), suggested the presence of mixed DNAs, or heteroresistance, in the region scanned. The term heteroresistant applies to cultures containing a mixture of wild-type (susceptible) and mutant (resistant) cells (5), although cases of different mutant DNAs were detected in several cultures reported here. Heteroresistance was confirmed by electrophoretic separation of the PCR products on denaturing gels without the heteroduplex reaction. The presence of multiple bands was characteristic of heteroresistant DNA samples. The DNA mixtures were identified by DNA sequencing.

DNA sequencing of rpoB.

Initial sequencing of the isolates was accomplished using primers described previously (10). Subsequent sequencing was performed using the primers shown in Table 1. DNA fragments were amplified as described above, and the DNAs were purified on QiaQuick PCR columns (QIAGEN, Valencia, CA) and quantified by measuring the A₂₆₀ on a SpectraMax 190 microtiter plate reader (Molecular Devices, Sunnyvale, CA). DNA sequencing was performed by Lark Technologies (Houston, TX), a Food and Drug Administration-approved sequencing facility. Sequencing primers were the same as the amplifying primers, except that only one primer was used in the sequencing reaction. DNA changes relative to rpoB from H37Rv (Rv0667) were noted. Codon numbers are based on Escherichia coli rpoB (accession no. AAC43085) after a BLASTP alignment (1) to Rv0667. Sequence analysis was performed using the Vector NTI version 8 suite of programs (InforMax, Madison, WI).

An initial set of isolates harboring previously sequenced mutations in rpoB (2, 10-11, 14) was used to establish a collection of distinctive mutations. This reference set of isolates harbored 16 distinct RRDR mutations (including double mutations) and four distinct mutations outside the RRDR (Table 2). rpoB mutations in the Texas-Tamaulipas border isolates were predicted based on their denaturation patterns relative to the fingerprints of previously sequenced mutations run in parallel. Samples with novel denaturation patterns were sequenced to identify the mutation, and each mutation was sequenced from at least one isolate.

TABLE 2.

Mutations in M. tuberculosis rpoB gene detected by DGGE

Affected region and codon(s)a	ORF codon(s)b	Amino acid changec	Base change(s)d	No. of isolatese	Confirmed isolatesf	Fig.:laneg
NI region
146	176	Val→Phe	GTC→TTC	7	7	5A:8-9
NII region
438	351	Gly→Gly	GGT→GGG	1S	1	5C:1
482	407	Gln→Arg	CAG→CGG	Ref		5C:2
RRDR
510, 516	438, 441	Gln→His, His→Tyr	CAG→CAC, CAC→TAC	1	1	2:29
511	436	Leu→Pro	CTG→CCG	Ref		2:5
511, 516	436, 441	Leu→Arg, Asp→Tyr	CTG→CGG, GAC→TAC	Ref		2:4
511, 516	436, 441	Leu→Pro, Asp→Gly	CTG→CCG, GAC→GGC	Ref		2:33
513	438	Gln→Leu	CAA→CTA	Ref		2:25
513	438	Gln→Pro	CAA→CCA	Ref		2:26
513-517	438-442	ΔGlnPheMet∧His	ΔAA TTC ATG G	5	1	2:34
514	439	Phe→Phe	TTC→TTT	1S	1	2:17
514-516	439-441	ΔPheMetAsp	ΔTTC ATG GAC	Ref		2:1
515	440	∧Phe	∧TTC	1, Ref	1	2:3
516	441	Asp→Gly	GAC→GGC	1	1	2:35
516	441	Asp→Val	GAC→GTC	5, Ref	3	2:6
516	441	Asp→Tyr	GAC→TAC	3, Ref	2	2:7
516, 526	441, 451	Asp→Glu, His→Asn	GAC→GAG, CAC→AAC	2	2	2:30, 2:37
516, 526	441, 451	Asp→Gly, His→Asn	GAC→GGC, CAC→AAC	Ref		2:36
516-517	441-442	ΔAspGln	ΔGAC CAG	1	1	2:2
519, 529	444, 454	Asn→Lys, Arg→Gln	AAC→AAG, CGA→CAA	1	1	2:31
522	447	Ser→Leu	TCG→TTG	1	1	2:18
526 (516)	451, (441)	His→Arg, (His→asp)	CAC→CGC, (CAC→gAC)	Ref		2:9
526	451	His→Arg	CAC→CGC	4	4	2:19
526	451	His→Asn	CAC→AAC	Ref		2:10
526	451	His→Asp	CAC→GAC	10, Ref	6	2:11
526	451	His→Cys	CAC→TGC	2	1	2:12
526	451	His→Tyr	CAC→TAC	14, Ref	1	2:13
526	451	His→Ser	CAC→TCC	1	1	2:14
526	451	His→Ser	CAC→AGC	1	1	2:15
526	451	His→Leu	CAC→CTC	1	1	2:38
526, 531	451, 456	His→Ser, Ser→Ala	CAC→AGC, TCG→GCG	1	1	2:8, 16, 24, 32, 40
528	453	Arg→Arg	CGC→CGT	2S	2	2:27
528, 531	453, 455	Arg→Arg, Ser→Phe	CGC→CGT, TCG→TTC	1	1	2:28
531	456	Ser→Tyr	TCG→TAC	2	1	2:20
531	456	Ser→Leu	TCG→TTG	67, Ref	2	2:21
531	456	Ser→Trp	TCG→TGG	12, Ref	1	2:22
531	456	Ser→Phe	TCG→TTC	2	1	2:23
533	458	Leu→Pro	CTG→CCG	Ref		2:39
CII region
562	487	Glu→Leu	GAA→TTA	Ref		5C:4
564	489	Pro→Ser	CCT→TCT	Ref		5C:3
572	497	Ile→Phe	ATC→TTC	1, Ref	1	5C:5
572	497	Ile→Leu	ATC→CTC	1	1	5C:6
CIII region
622	546	Pro→Gln	CCG→CAG	1S	1	5B:3
641	567	Glu→Glu	GAG→GAA	1S	1	5B:5
642	568	Val→Ala	GTG→GCG	1	1	5B:1
646	572	Pro→Ser	CCC→TCC	1S	1	5B:2
646	572	Pro→Leu	CCC→CTC	2, 3S	5	5B:4-5, 7-9

^aCodons are given relative to E. coli rpoB (accession no. AAC43085) using a BLASTP alignment (1) against Rv0667.

^bM. tuberculosis rpoB (Rv0667) codon numbering.

^cThe first amino acid changed to the second amino acid as indicated. Δ, amino acids deleted; ^∧, amino acid inserted. Parentheses indicate minor heteroresistant codons and amino acid changes (lowercase). Only double mutations within the RRDR are noted. For other double mutations, see Table 3.

^dNucleotides of altered codons are underlined. Minor nucleotides in heteroresistant isolates are indicated in lowercase.

^eThe number of isolates was estimated by DGGE pattern. Ref, reference mutation from previously sequenced isolate (2, 10-11, 14); S, susceptible isolate(s). Other isolates are RIF resistant.

^fMutations identified by DNA sequencing of rpoB in the United States-Mexico border isolates.

^gDGGE profile is shown in indicated Figure and lane.

RESULTS

An initial PCR primer set was designed to detect mutations in the RRDR of rpoB (Fig. 1). This PCR target fragment was amplified from a reference set of M. tuberculosis isolates with previously defined mutations in the RRDR (2, 10-11, 14), and all 16 different mutations in these isolates were successfully detected. The RRDR target fragment was then amplified and tested for mutations in 296 rifampin-resistant and susceptible isolates. The cultures were isolated from the U.S. state of Texas and the bordering Mexican state of Tamaulipas (16). Twenty-four distinct DNA alterations within the RRDR were detected in the latter isolates, and a total of 34 distinct RRDR mutations were detected from both sets (Table 2). These mutations include a 3-bp insertion, small deletions (6 to 9 nucleotides), single nucleotide substitutions, alterations of two adjacent nucleotides, and double mutations in separate codons. Within the five double mutations were an additional four unique mutations. The DNA mutations in these isolates are listed together (Table 2), since the DGGE assay detects both alterations and since both changes may contribute to the drug resistance phenotype.

The DGGE denaturation profiles of the mutations detected in the two sets of isolates are shown in Fig. 2. Each mutation generated a characteristic denaturation profile that was dependent on the nucleotide alterations and the position of these alterations within the DNA fragment. A mutation could often be predicted by its characteristic denaturation profile. This empirical approach was used to estimate the frequency of each mutation (Table 2). In addition, new mutations could often be predicted by their novel denaturation profiles, and this was used to select isolates for DNA sequencing. Finally, several mutations generated denaturation patterns that were very similar to one another (Fig. 3, top panel), indicating that it was sometimes difficult to unambiguously assign mutations based on the DGGE profile alone. To address this problem, additional DGGE assays were established.

FIG. 2.

DGGE profile of rpoB RRDR mutations. Thirty-four distinct mutations in the RRDR of rpoB were analyzed on 20 to 90% urea-formamide-8% acrylamide gels. The same isolate (an H526S S531A double mutant) was run in lanes 8, 16, 24, 32, and 40 as a denaturation standard. The following lanes have the given mutation(s): lane 1, 514 to 516 deletion; lane 2, 516 to 517 deletion; lane 3, 515F insertion; lane 4, L511R; lane 5, L511P; lane 6, D516V; lane 7, D516Y; lane 9, H526R (heteroresistant; mixed with wild type); lane 10, H526N; lane 11, H526D; lane 12, H526C; lane 13, H526Y; lane 14, H526S no. 1; lane 15, H526S no. 2; lane 17, F514F; lane 18, S522L; lane 19, H526R; lane 20, S531Y; lane 21, S531L; lane 22, S531W; lane 23, S531F; lane 25, Q513L; lane 26, Q513P; lane 27, R528R; lane 28, R528R and S531F; lane 29, Q510H and H526Y; lane 30, D516E and H526N; lane 31, N519K and R529K; lane 33, L511P and D516Y; lane 34, Δ513 to 517∧H; lane 35, D516G; lane 37, D516E and H526N; lane 38, H526L; and lane 39, L533P. Numbered mutations (e.g., H526S no. 1) represent different nucleotide alterations (see Table 2 for details).

FIG. 3.

Mutation determination using different template DNAs for the heteroduplexing reaction. The rpoB RRDR was amplified from nine different isolates. The DNAs were heteroduplexed to either the wild type (top panel), an S531L mutant (middle panel), or an F514F mutant (bottom panel). Identity between the template and the isolate is indicated by a single band on the denaturing gel. The wild type was identified in lane 9 of the top panel, S531L mutations were identified in lanes 6 and 8 of the middle panel, and the F514F synonymous polymorphism was identified in lane 1 of the bottom panel.

To augment mutation assignment, a second DGGE assay was developed using mutant DNA as the heteroduplex template. Eight isolates with similar denaturation profiles were selected when wild-type (H37Rv) DNA was used as the heteroduplex template (Fig. 3, top panel). As expected, H37Rv heteroduplexed to itself produced only one band on the denaturing gradient gel (Fig. 3, top panel, lane 9). For the bottom two panels of Fig. 3, two different mutations were used as the heteroduplex template. Using an S531L mutation as template (the most common rpoB mutation), the isolates from lanes 6 and 8 generated a denaturation profile consisting of a single band (Fig. 3, middle panel). This indicates sequence identity (i.e., the absence of DNA mismatches) between these isolates and the template DNA. Hence, the isolates from lanes 6 and 8 harbor the S531L mutation. Similarly, a F514F synonymous polymorphism was used as the template in Fig. 3, bottom panel, and the isolate from lane 1 was identified with the F514F RIF-susceptible polymorphism. Since there are more DNA mismatches when different mutant DNAs are heteroduplexed with one another, there is a wider spread of the bands in these samples, and the denaturation profiles appear to be more distinctive. This represents another level of mutation typing that is possible with DGGE.

The RRDR and CII region from 59 isolates were sequenced in order to identify mutations (Table 2). Isolates were chosen primarily by unique denaturation patterns, and this strategy was largely successful at identifying novel mutations. Other mutations were sequenced to confirm the DNA alteration. In addition, this region was sequenced in RIF-resistant isolates without DGGE-detectable mutations. DNA alterations were observed in only 3 of 143 RIF-susceptible isolates. One isolate harbored the synonymous F514F polymorphism, while two isolates harbored a synonymous R528R change. Another isolate harbored the R528R polymorphism in conjunction with an S531F alteration; however, this isolate was resistant to RIF. Eight RIF-resistant isolates lacked RRDR or CII mutations.

Detection of heteroresistance or mixed cultures by DGGE.

DGGE detected five cultures that contained mixtures of DNAs within the RRDR (Fig. 4). While a homogeneous mutant DNA generates four bands on denaturing gradient gels after it has been heteroduplexed to a wild-type template (Fig. 4A, lane 6), the mixed samples (Fig. 4A, lanes 2 to 5 and 7) generated more than four bands, and the band intensities were variable from one to another. When DNA from a heteroresistant culture was run on a denaturing gel without the heteroduplexing reaction (Fig. 4B), the DNA still generated multiple bands. In contrast, wild-type and homogeneous mutant DNAs generated only a single band (Fig. 4B, lanes 1 and 6, respectively). This indicates that the heteroresistant cultures contain a mixture of DNAs. The mixed nature of these samples was confirmed. First, genotyping data revealed minor IS6110 bands in two cultures of isolates collected from human immunodeficiency virus-infected individuals, suggesting the presence of more than one strain. Second, DNA sequencing of the five samples revealed the presence of minor peaks at specific nucleotides. Interestingly, while two samples contained a mixture of wild-type and mutant DNAs, three samples contained a mixture of two different mutant DNAs. In some cases, the mutations were more consistently detected by DGGE than by DNA sequencing. Published reports indicate that DGGE is between 4- and 17-fold more sensitive than DNA sequencing in detecting mutations in mixed DNA samples (4, 19).

FIG. 4.

Heteroresistant isolates. (A) Heteroresistant samples display more than four bands of variable intensities when heteroduplexed to H37Rv. Heteroduplex of H37Rv to itself produces only one band (lane 1). In lane 6, an S531L mutation produces a characteristic four-band denaturation pattern. (B) PCR products of rpoB RRDR by DGGE without heteroduplexing reaction. Homogeneous DNAs in lane 1 (H37Rv) and lane 6 (S531L) produce a single denaturation band. Heteroresistant isolates in lanes 2 to 5 and 7 produce multiple bands, indicating that the PCR amplified DNA with a mixture of sequences. DNA mixtures were confirmed by DNA sequencing.

DGGE scanning for rpoB polymorphisms outside of the RRDR.

Four additional DGGE primer sets were developed to scan for DNA sequence alterations outside of the RRDR (Fig. 1). The NI primer set was designed to detect alterations between codons 140 and 180 (Table 1). The V146F mutation, found in this region, was reported in up to 1% of German RIF-resistant isolates (6). The NII primer set was designed to probe for DNA changes between codons 414 and 485 in front of the RRDR. The CII and CIII primer sets were designed to scan for DNA polymorphisms between codons 561 and 615 and codons 623 and 681, respectively. DNA alterations within these target regions were detected within the appropriate reference isolates (Fig. 5 and Tables 2 and 3). These primer sets were then used to scan for DNA alterations in rpoB of 210 M. tuberculosis isolates (105 RIF-resistant and 105 RIF-susceptible isolates) from the United States-Mexico border region. As DNA alterations in these regions were detected, they were identified by DNA sequencing. All of these isolates had previously been tested for mutations in the RRDR (Table 2). Importantly, these isolates included all eight RIF-resistant isolates in which RRDR mutations were not observed.

FIG. 5.

Detection of DNA alterations outside the RRDR. (A) Detection of DNA alterations in the NI and CIII regions of M. tuberculosis isolates. The NI and CIII target fragments were coamplified from isolates, and the heteroduplex products were resolved on the same denaturing gel. The following lanes show the given mutation(s): lanes 1 to 2 and 4 to 7, no mutation; lane 3, P646L in CIII; lane 8, V146F in NI; and lane 9, V146F in NI and V642A in CIII. (B) Typing CIII DNA alterations by heteroduplex analysis. CIII target fragments were amplified from nine isolates. After heteroduplexing to H37Rv (top panel), most of the isolates display a similar denaturation profile. However, after heteroduplexing to isolate 2 (bottom panel), five denaturation patterns can be distinguished. See Table 3 for complete genotypes and RIF phenotypes of these isolates. The following isolates have the given CIII alteration: isolate 1, V642A; isolate 2, P646S; isolate 3, P622Q; isolates 4 to 5 and 7 to 9, P646L; and isolate 6, E641E. (C) DNA alteration in NII and CII regions of rpoB by DGGE. The NII and CII fragments were amplified from isolates, and the heteroduplex products were resolved on denaturing gels. See Table 3 for complete genotypes and RIF phenotypes of these isolates. The following lanes show the given mutation: lane 1, G438G in NII; lane 2, Q482R in NII; lane 3, P564S in CII; lane 4, E562L in CII; lane 5, I572F in CII; and lane 6, I572L in CII.

TABLE 3.

Genotypes at rpoB of isolates with DNA alterations outside of the RRDR^a

NI-NII Region		RRDR		CII-CIII region		No. of isolatesb	Resistancec
Amino acid change	Codon change	Amino acid change	Codon change	Amino acid change	Codon change
V146F	GTC→TTC					5	R
V146F	GTC→TTC	H526S	CAC→AGC			1	R
V146F	GTC→TTC			V642A	GTG→GCG	1	R
G438G	GGT→GGG					1	S
Q482R	CAG→CGG	S531L	TCG→TTG			Ref	R
		L511P	CTG→CCG	E562L	GAA→TTA	Ref	R
		S531L	TCG→TTG	P564S	CCT→TCT	Ref	R
				I572F	ATC→TTC	1	R
		D516G	GAC→GGC	I572L	ATC→CTC	1	R
		D516Y	GAC→TAC	I572L	ATC→CTC	1	R
				P622Q	CCG→CAG	1	S
		H526D	CAC→GAC	E641E	GAG→GAA	1	R
		D516V	GAC→GTC	P646L	CCC→CTC	1	R
		H526Y	CAC→TAC	P646L	CCC→CTC	1	R
				P646L	CCC→CTC	3	S
				P646S	CCC→TCC	1	S

^aMutations are described in more detail in Table 2. Codon changes are summarized using amino acid abbreviations, with the wild-type amino acid preceding the codon number and followed by the altered amino acid. Altered nucleotides are underlined.

^bThe number of isolates with a specific rpoB genotype was determined by DGGE and DNA sequencing. Ref, genotype of reference isolate (2, 10-11, 14).

^cR, RIF resistant; S, RIF susceptible.

The NI and CIII fragments were coamplified from each isolate and screened together (Fig. 5A). The V146F mutation was detected within the NI fragment in 7 of 105 RIF-resistant isolates. Two of the isolates harboring this alteration have the same IS6110 fingerprint and spoligotype, suggesting that they represent a genotypic cluster. Another isolate also contained a V642A polymorphism (Fig. 5A, isolate 9). Isolates harboring the V146F mutation with and without the V642A alteration were resistant to RIF in both initial and repeat testing. Since the V642A alteration has not been reported previously, it is not clear whether it contributes to drug resistance. Another isolate harboring the V146F mutation also contained a mutation in the RRDR, H526S (CAC→AGC) (Table 3). To our knowledge, this is the first report of V146F and RRDR double mutations. The V146F mutation was not observed in any of the other RIF-resistant isolates with previously identified RRDR mutations. Finally, no DNA alterations in the NI region were detected in any of the 105 RIF-susceptible isolates.

DNA alterations were detected within the CIII target fragment in 9 of the 210 isolates (Tables 2 and 3). A V642A mutation was detected in one RIF-resistant isolate that also harbored a V146F mutation, as described above. A P622Q alteration was observed in one RIF-susceptible isolate. The other seven CIII mutations produced similar denaturation patterns when heteroduplexed to H37Rv (Fig. 5B, top panel). However, further analysis revealed that these seven isolates harbored three different mutations. When the CIII fragment from one of the seven isolates was used as the heteroduplex template, three distinct patterns were observed (Fig. 5B, bottom panel). DNA sequence analysis confirmed that these three DGGE patterns corresponded to three distinct DNA alterations in the CIII region. One RIF-resistant isolate harbored an E641E synonymous polymorphism; this isolate also harbored an H526D mutation in the RRDR. A P646S alteration was observed in a RIF-susceptible isolate. Finally five isolates harbored a P646L alteration. Two were RIF resistant and also harbored RRDR changes (D516V and H526Y). The remaining three isolates were RIF susceptible. These CIII DNA alterations do not appear to have been reported previously (7, 20, 28).

Three RIF-resistant isolates harbored alterations in the CII region (Fig. 5C, lanes 3 to 6). Two isolates harbored I572L mutations. These isolates also contained RRDR mutations (D516G and D516Y). Another isolate harbored an I572F alteration. Interestingly, this isolate did not harbor a defect in the RRDR. Similarly, a reference isolate harboring the I572F defect lacked RRDR alterations. Isolates harboring this mutation were previously reported by three other groups (3, 6, 27); none reported an additional RRDR mutation. Reference isolates harbored mutations in both the RRDR and the CII region. One isolate harbored an L511P and E562L combination, while the other contained S531L and P564S alterations. Finally, CII DNA alterations were not detected from any of the RIF-susceptible isolates.

A single alteration was detected in the NII region in the 210 isolates scanned. A G438G synonymous polymorphism at codon 438 was detected in one RIF-susceptible isolate (Fig. 5C). In addition, a Q482R mutation was detected in a reference sample that also harbored an S531L defect.

Revised RIF susceptibility based on DGGE assays.

Thirty-five isolates were retested for RIF susceptibility. Half of these isolates harbored common rpoB mutations that were either previously reported to impart resistance to RIF (7, 17, 20, 28) or synonymous RRDR polymorphisms. The remaining 18 isolates lacked identified RRDR mutations by DGGE. The isolates were tested by the BACTEC460 method at three different RIF concentrations: 0.5, 2, and 8 μg/ml. Isolates with either V146F or I572F mutations grew in 8 μg/ml, confirming that these alterations contribute to the drug resistance phenotype. However, the susceptibility results of 10 isolates were revised upon retesting. One isolate that contained a mixture of DNA in the rpoB RRDR grew in 2.0 μg/ml RIF, one isolate with a mixture of DNA grew in 8 μg/ml RIF, while a third isolate with mixed DNA was unable to grow in 0.5 μg/ml RIF and was reclassified as susceptible. Finally, seven of eight isolates without detectable mutations in the amino-terminal region, RRDR, CII region, or CIII region were reclassified as RIF susceptible. An examination of the clinical laboratory records indicated variable and inconsistent results regarding RIF and rifabutin resistance for most of these isolates. The revised susceptibility was justified, since it was consistent with the absence of DNA alterations in rpoB, which was predicted by DGGE and confirmed by DNA sequencing. Once the mutation and susceptibility data were reconciled, only 1 of 153 RIF-resistant isolates lacked an identified rpoB DNA alteration.

DISCUSSION

In the present study, we used DGGE to identify mutations associated with resistance to rifampin in the rpoB gene of M. tuberculosis. Five DGGE primer sets were used to scan 775 bp of rpoB for DNA alterations. Two groups of isolates were evaluated. First, isolates harboring previously sequenced rpoB mutations were used to test target fragments and to establish denaturation patterns for commonly reported defects. These isolates were collected from around the globe and have been reported previously (2, 10-11, 14). Second, isolates from the United States-Mexico border states of Texas and Tamaulipas (16) were scanned to expand the database of observed mutations and to identify the best combination of PCR target fragments to accurately detect RIF resistance mutations. Forty-seven different rpoB DNA alterations were detected in 16 reference isolates and 296 isolates collected from the United States-Mexico border region. However, seven of these were synonymous DNA changes, appearing in either RIF-susceptible isolates or both resistant and susceptible isolates. This finding suggests that these alterations are not associated with drug resistance. Forty different mutations were observed only in RIF-resistant isolates. Five of these mutations were observed in conjunction with previously defined RIF resistance mutations, and it is unclear whether these new mutations contribute to drug resistance. As a group, these rpoB alterations underscore the complexities of RIF resistance in M. tuberculosis.

The most useful DGGE primer set scanned for mutations in the RRDR of rpoB, and 94.8% of the RIF-resistant isolates contained mutations in this fragment (Table 4). This was similar to the accumulated reported global summation derived primarily from DNA sequencing (7, 17, 20, 28). Every RIF-resistant RRDR mutation was detected from the two sets of isolates used in this report. This can be stated with confidence, since this region has been sequenced from every RIF-resistant isolate lacking a DGGE-detectable RRDR mutation, and no DNA alterations were missed. This confirms the sensitivity of the DGGE assay in detecting mutations (4, 19). Using DGGE in combination with DNA sequencing, we have identified 34 different DNA alterations within the RRDR. These mutations are representative of the RRDR defects reported globally (7, 17, 20, 28), although several double mutations appear to be novel.

TABLE 4.

DGGE detection of rpoB DNA alterations in M. tuberculosis isolates

DGGE region	No. of RIF-resistant isolates with DNA alteration			No. of RIF-susceptible isolates with DNA alteration
	Presenta	Absent	% Present	Present	Absent	% Present
NI	7b	98	6.7	0	105	0
NII	0	105	0	1	104	1.0
RRDR	145c	8	94.8	3	140	2.1
CII	3d	102	2.9	0	105	0
CIII	4e	101	3.8	4	101	3.8
RRDR + NIf	151	2	98.7	3	140	2.1
RRDR + CIIf	146	7	95.4	3	140	2.1
RRDR + NI + CIIf	152	1	99.3	3	140	2.1

^aIsolates were scanned for DNA alterations in the rpoB regions indicated. DNA alterations were either detected (Present) or not detected (Absent) in the isolates. For details of specific DNA alterations detected, see Tables 2 and 3.

^bOne isolate also harbors an RRDR mutation and another isolate also harbors a CIII alteration; a lower estimate of mutation frequency (4.6%) is calculated relative to 153 total RIF-resistant isolates tested.

^cOne isolate also harbors an NI alteration, two isolates also harbor CII alterations, and three isolates contain CIII alterations.

^dTwo isolates also harbor RRDR mutations.

^eAll CIII DNA alterations in RIF-resistant isolates also contain RRDR or NI mutations.

^fCombinations estimate detection, since not all isolates were tested with all DGGE primer sets.

A second DGGE primer set was designed to detect the V146F mutation. This mutation was not observed in 105 RIF-susceptible isolates. Of a total of 153 RIF-resistant isolates, 8 isolates lacked mutations in the RRDR, and the V146F mutation was observed in 6 (75%) of these isolates. Another isolate harbored the V146F mutation as well as an H526S mutation. However, the V146F mutation was not observed in the other 96 RIF-resistant isolates that also harbored RRDR mutations. Heep et al. (6) first proposed that the V146F mutation imparts resistance to RIF in M. tuberculosis, and in this report we have confirmed that isolates harboring this defect are RIF resistant. This mutation was originally reported in about 1% of RIF-resistant isolates from Germany (6). We observed the V146F mutation in 5 to 7% of the RIF-resistant isolates from the United States-Mexico border region. The small range reflects variation in the sample sets in which this mutation was scanned as well as the genotypic clustering of two isolates with this alteration.

A third primer set scanned for mutations in the CII region of rpoB, and mutations were detected in three RIF-resistant isolates. Two harbored an I572L alteration in conjunction with different alterations at codon 516 in the RRDR, D516G or D516Y. The third isolate harbored an I572F substitution but did not harbor a mutation in the RRDR. Identification of alterations at codon 572 contributed in a small way to the successful detection of rpoB mutations associated with RIF resistance. While the I572F mutation alone may be sufficient for drug resistance, the I572L alteration has been detected only in the presence of RRDR mutations. Further testing will be required to determine if the latter mutation directly contributes to RIF resistance.

Another DGGE target fragment was used to identify five distinct DNA alterations within the CIII region between codons 622 and 646. Alterations in this region were previously reported for RIF-resistant isolates (6, 28). However, only one of the CIII polymorphisms identified in our study, V642A, was uniquely associated with resistance to RIF. However, it is unclear whether the V642A mutation contributes to drug resistance, since it was observed in an isolate that also contained a V146F mutation. The other CIII polymorphisms were either synonymous changes or were not observed exclusively in resistant isolates. Similarly, a single synonymous alteration at codon 438 in a RIF-susceptible isolate was detected in the NII region. Since the DNA alterations detected within these two regions do not appear to be uniquely associated with drug resistance, these rpoB alterations may have limited usefulness in predicting drug resistance. However, the NII and CIII target fragments may be useful in detecting rpoB DNA alterations in other populations of M. tuberculosis strains with a different distribution of mutations.

Since each mutation generated a distinct and reproducible denaturation profile, we tested whether this property could be used to predict mutations. The successful application of this approach was dependent on three main factors: the DGGE target fragment, the number of different mutations detected within the target fragment, and the heteroduplex template. For the CII region, only four different mutations were observed, and these could be readily distinguished by denaturation profile (Fig. 5C). On the other hand, three of the five CIII DNA alterations generated similar denaturation profiles that could not readily be resolved with a single heteroduplex template (Fig. 5B). Prediction was even more complicated when comparing RRDR mutant profiles from 34 different alleles (Fig. 2 and 3). For both the CIII and RRDR target fragments, mutation prediction was greatly facilitated by using mutant DNA as a heteroduplex template. In both Fig. 3 and Fig. 5B, mutations that displayed similar denaturation profiles with H37Rv as a heteroduplex template could be distinguished when mutant heteroduplex templates were used. Since there is a greater degree of mismatches between two mutant DNAs, they can be expected to have a larger difference in melting temperature than mutant-wild-type heteroduplex templates. It was also possible to define mutations that were identical to the mutant heteroduplex template, since the homologous mutations generated a single band in the gels. Further resolution can be achieved with additional mutant templates (Fig. 3). It is expected that this type of analysis could be utilized to assign mutations using a limited number (e.g., two to four) of heteroduplex templates. Identifying the DNA alteration within the RRDR by denaturation profile is of interest, since it would help identify polymorphisms that are not associated with drug resistance, such as the F514F and R528R polymorphisms. This could be used to increase the specificity of a DGGE clinical assay to predict RIF susceptibility.

Using DGGE, we have detected five separate cultures that contain mixtures of DNAs. These heteroresistant cultures may have arisen in several ways (5). First, they may represent the emergence of drug-resistant cells within a strain (9, 15). Second, they may represent multiple infections of an individual by different M. tuberculosis strains (24). Third, they may represent laboratory contamination of one strain by another (8). Fourth, they may represent a heterogeneous population of bacilli with different resistance mutations (15). Regardless of their origin, it is difficult to detect these types of cultures. In this respect, DGGE proved especially useful and, as we have demonstrated, is capable of providing additional information. We observed instances of wild-type and resistant rpoB as well as mixtures of mutations within the RRDR, although the exact causes of these specific mixed cultures have not been determined.

Only 1 RIF-resistant isolate out of 153 scanned did not have a detectable rpoB DNA alteration as analyzed by DGGE and confirmed by DNA sequencing. By scanning in the RRDR and NI regions, mutations in 98.7% of RIF-resistant isolates could be detected (Table 4), and by adding the CII fragment, 99.3% of the resistant isolates had identifiable mutations. Specificity is also high, since only 2.1% of RIF-susceptible isolates harbored DNA sequence variants that might be interpreted as resistant, producing a “false positive” interpretation. All of the DNA alterations in RIF-susceptible isolates were detected in the RRDR. By learning how to recognize sequence variants in the RRDR, it should be possible to improve the specificity of the DGGE assay. Equally impressive is the ability of DGGE to detect inconsistencies in drug susceptibility testing. Seven isolates without rpoB mutations were revised to RIF susceptible upon reanalysis. This resulted in a 5 to 6% improvement in the sensitivity of DGGE assay.

We have reported similar results using DGGE to probe for pyrazinamide resistance mutations in the pncA gene of M. tuberculosis isolates (12). This suggests that DGGE can be used to detect mutations that impart resistance to other antituberculosis drugs. Scarpellini et al. (18) also analyzed RIF-resistant isolates using DGGE, and they were able to detect 14 different mutations in 81 RIF-resistant M. tuberculosis isolates. They were able to detect RRDR mutations in clinical samples, such as sputum, suggesting that the direct diagnosis of drug susceptibility may be possible using a DGGE assay. The RRDR and NI fragments can be coamplified and can be resolved on the same denaturing gel. Many other steps in the DGGE assay lend themselves to high-throughput screening. It will be interesting to test whether the RRDR and NI target fragments can be used to accurately predict RIF susceptibility in blinded studies, where the sensitivity and specificity of the DGGE assay can be compared directly to more traditional culture assays.