Typing M. Tuberculosis By Cleavase Fragment Length Polymorphism

JOURNAL OF CLINICAL MICROBIOLOGY, Dec. 1996, p. 3129–3137 0095-1137/96/$04.0010 Copyright q 1996, American Society for Microbiology

Vol. 34, No. 12

Differentiation of Bacterial 16S rRNA Genes and Intergenic Regions and Mycobacterium tuberculosis katG Genes by Structure-Specific Endonuclease Cleavage MARY ANN D. BROW,1* MARY C. OLDENBURG,1 VICTOR LYAMICHEV,1 LAURA M. HEISLER,1 NATASHA LYAMICHEVA,1 JEFF G. HALL,1 NANCY J. EAGAN,1 D. MICHAEL OLIVE,1 LLOYD M. SMITH,2 LANCE FORS,1 AND JAMES E. DAHLBERG3 Third Wave Technologies, Inc.,1 and Department of Chemistry2 and Department of Biomolecular Chemistry,3 University of Wisconsin, Madison, Wisconsin Received 3 July 1996/Returned for modification 3 August 1996/Accepted 17 September 1996

related with particular differences at the nucleic acid level. When such sequence differences are confined to a few known sites, it is possible to assess a new isolate by simple methods such as allele specific oligonucleotide hybridization (6) or restriction fragment length polymorphism (RFLP) analysis. In many genes, however, polymorphisms may occur anywhere within a given region, rather than at one or a few easily examined positions. Genetic analysis of this type requires the use of methods that indicate differences between nucleic acids regardless of the identity or the position of change and without prior knowledge of the precise sequence of the molecule. While DNA sequencing (11, 18) is a reliable way of identifying such mutations, resequencing thousands of bases in search of single-base changes can be slow and prohibitively costly. The development of alternative methods to efficiently detect and characterize such changes will allow comparison studies to proceed more rapidly, as well as enable their use in the clinical diagnostic arena. Methods such as RFLP (5), single-strand conformation polymorphism (8, 13), and dideoxy fingerprinting (19) have been applied for the detection of mutations associated with drug resistance in M. tuberculosis (5, 10, 22). All of these methods are limited by either their inability to analyze long DNA fragments, their ability to examine only a few genetic loci (i.e., RFLP), or their inability to differentiate phenotypically silent mutations from mutations resulting in drug resistance. Identification and typing of bacteria have been accomplished by both phenotypic and genotypic methods. Methods in which

The characterization of microbiological pathogens on the basis of nucleic acid sequences is becoming increasingly important in clinical microbiology. In the future, detection of mutations in genes such as the Mycobacterium tuberculosis rpoB and katG genes may be used to predict resistance to rifampin and isoniazid, respectively. As the number and spectrum of nosocomial pathogens continue to expand, the ability to accurately determine the relatedness of microbial isolates has become increasingly important. Techniques for analyzing microbial relatedness have become critical (i) for identifying outbreaks of infection, (ii) for determining the mode of acquisition of a pathogen, (iii) for analyzing individual patients to determine if a series of isolates obtained over time represents relapse of infection due to a single strain or separate episodes of disease from different infections, and finally (iv) for defining effective preventive and therapeutic measures. The advent of the PCR and improvements in DNA sequencing technology have led to the rapid accumulation of primary sequence information for a number of genes, gene systems, and whole organisms. The understanding of the significance of any particular locus, however, requires the comparison of newly determined sequences to those of similar nucleic acids of known function, so that differences in phenotype may be cor* Corresponding author. Mailing address: Third Wave Technologies, Inc., 2800 S. Fish Hatchery Rd., Madison, WI 53711. Phone: (608) 273-8933. Fax: (608)-273-6989. Electronic mail address: Mary_Ann [email protected]. 3129

Downloaded from jcm.asm.org by on September 19, 2009

We describe here a new approach for analyzing nucleic acid sequences using a structure-specific endonuclease, Cleavase I. We have applied this technique to the detection and localization of mutations associated with isoniazid resistance in Mycobacterium tuberculosis and for differentiating bacterial genera, species, and strains. The technique described here is based on the observation that single strands of DNAs can assume defined conformations, which can be detected and cleaved by structure-specific endonucleases such as Cleavase I. The patterns of fragments produced are characteristic of the sequences responsible for the structures, so that each DNA has its own structural fingerprint. Amplicons containing either a single 5*-fluorescein or 5*-tetramethyl rhodamine label were generated from a 620-bp segment of the katG gene of isoniazid-resistant and -sensitive M. tuberculosis, the 5* 350 bp of the 16S rRNA genes of Escherichia coli O157:H7, Salmonella typhimurium, Salmonella enteritidis, Salmonella arizonae, Shigella sonnei, Shigella dysenteriae, Campylobacter jejuni, Staphylococcus hominis, Staphylococcus warneri, and Staphylococcus aureus and an approximately 550-bp DNA segment comprising the intergenic region between the 16S and 23S rRNA genes of Salmonella typhimurium, Salmonella enteritidis, Salmonella arizonae, Shigella sonnei, and Shigella dysenteriae serotypes 1, 2, and 8. Changes in the structural fingerprints of DNA fragments derived from the katG genes of isoniazid-resistant M. tuberculosis isolates were clearly identified and could be mapped to the site of the actual mutation relative to the labeled end. Band patterns which clearly differentiated bacteria to the level of genus and, in some cases, species were generated from the 16S genes. Cleavase I analysis of the intergenic regions of Salmonella and Shigella species differentiated genus, species, and serotypes. Structural fingerprinting by digestion with Cleavase I is a rapid, simple, and sensitive method for analyzing nucleic acid sequences and may find wide utility in microbial analysis.

3130

BROW ET AL.

FIG. 1. Schematic representation of steps of CFLP pattern generation. Labeled fragments of DNA are heated to separate the complementary strands. When the samples are cooled, the single strands of DNA assume folded hairpinlike structures; subtle differences in the sequences of the fragments can cause formation of different structures. The Cleavase I enzyme cleaves at the 59 side of these structures, at the junction between duplexed and single-stranded regions. Separation and detection of the resulting fragments create signature banding patterns that can be compared to detect differences between the test molecules.

maintenance of the structure is not required after the cleavage, resolution of the products can be done quickly and reproducibly by denaturing gel electrophoresis. The data presented here show that sequence polymorphisms, whether single or multiple base, cause sufficient alteration in the folded structures of large molecules to yield polymorphisms in the structural fingerprint patterns. By analogy to RFLP and single-strand conformation polymorphism analyses, we call this method Cleavase fragment length polymorphism (CFLP) analysis. We demonstrate the utility of CFLP analysis by using it for differentiating isoniazid-sensitive and -resistant M. tuberculosis and for distinguishing bacteria at the levels of genus, species, and strain. MATERIALS AND METHODS Cleavase I unit definition. Cleavase I is an engineered enzyme consisting of the 59 nuclease domain (amino acids 1 through 281) of Thermus aquaticus DNA polymerase. Its construction and characterization will be described elsewhere. The activity of the Cleavase I enzyme was determined by digestion of an oligonucleotide substrate (59-TGGTCGCTGTCTCGCTGAAAGCGAGACAGCGT G-39) labeled at the 59 end with fluorescein, which folds to create a 12-bp stem with a 3-nucleotide (nt) loop. Reaction mixtures contained enzyme and 20 pmol of this substrate in 20 ml of 10 mM MOPS (morpholinepropanesulfonic acid [pH 7.5])–1 mM MnCl2. Reaction mixtures were incubated at 508C for 2 min. Release of a 5-nt fragment from the 59 end was monitored by electrophoresis through a 10% polyacrylamide gel (19:1 cross-link) with 7 M urea in 0.53 TBE buffer (45 mM Tris-borate [pH 8.3], 1.4 mM EDTA [17]), followed by visualization and quantitation with an Hitachi model FMBIO-100 fluorescence image analyzer fitted with a 505-nm filter. One unit of enzymatic activity is defined as the amount that digests 3.7 pmol of this substrate under conditions of vast substrate excess in 1 min. The enzyme is stored in 20 mM Tris-Cl (pH 8.0) with 50 mM KCl, 100 mg of bovine serum albumin per ml, 0.5% Tween 20, and 0.5% Nonidet P-40. Substrate DNA fragments. The DNA fragments used in this study were generated by PCR. Primers were labeled on their 59 ends with either fluorescein, tetramethyl rhodamine, or tetrachloro fluorescein or were used unlabeled, as indicated. PCR mixtures (12, 16) contained 1 to 5 ng of plasmid, 25 ng of genomic DNA or approximately 10 fmol of DNA fragment, 2.5 U of AmpliTaq DNA polymerase (Perkin Elmer, Inc., Foster City, Calif.), 25 pmol of each primer, and 50 mM of each deoxyribonucleotide (Perkin Elmer) in 100 ml of 20 mM Tris-HCl (pH 8.5)–1.5 mM MgCl2–50 mM KCl–0.5% Tween 20–0.5% Nonidet P-40. Visualization of fluorescently labeled amplification products following electrophoresis revealed a background smear of shorter products that were not visible by staining. To remove these products, the PCR mixtures were treated with 1 U of exonuclease I (Amersham International, Chicago, Ill.) for 20 min at 378C (9). The nuclease was inactivated by heating at 708C for 15 min. Reaction mixtures were brought to 2 M ammonium acetate, and the DNAs were precipitated by the addition of 1 volume of isopropanol. Amplification reactions that produced multiple double-stranded products were resolved on polyacrylamide gels (6 to 10%, with 19:1 cross-linking) gels with 7 M urea, in 0.53 TBE buffer. Gels were stained with ethidium bromide, and the bands of interest were excised and eluted by passive diffusion at 378C overnight into a solution of 0.5 M ammonium acetate, 0.1 mM EDTA, and 0.1% sodium dodecyl sulfate. Eluates were collected, and the DNA was recovered by the addition of 2.5 volumes of 100% ethanol. Precipitates were collected by microcentrifugation, washed briefly with 70% ethanol, and dried under vacuum. All DNAs were dissolved in 10 mM Tris-Cl (pH 7.5 to 8.0) with 0.1 mM EDTA. Bacterial DNA preparation. The bacteria used in this study are listed in Table 1. E. coli O157:H7 isolates were obtained from Jeff Klinger (Vysis, Inc., Naperville, Ill.). Bacteria were grown in Trypticase soy broth overnight at 378C with constant agitation, and the DNA was obtained by phenol-chloroform extraction (1:1) and ethanol precipitation. For PCR amplification of the 16S rRNA gene, a 350-bp amplicon was generated from each strain of bacteria with a set of conserved eubacterial primers designated EB-R (59-AGAGTTTGATCCTGGCTC AG-39), corresponding to nt 41 to 61, and EB-L (59-CTGCTGCCTCCCGTAG GAGT-39), corresponding to nt 388 to 408 on the E. coli consensus sequence. The intergenic regions of Salmonella and Shigella spp. were amplified with a 16S sense primer (59-CTGGGGTGAAGTCGTAACAA-39) corresponding to nt 2768 to 2788 on the E. coli 16S rRNA consensus sequence and a 23S antisense primer (59-GGGCATCCACCGTGTACGCT-39) corresponding to nt 26 to 46 on the E. coli consensus 23S rRNA sequence. DNAs from isoniazid-sensitive and -resistant strains of M. tuberculosis were the gift of Frank Cockerill (Mayo Clinic). Amplicons spanning a 620-bp region of the M. tuberculosis katG gene (codons 302 to 507) of a wild-type and three isoniazid-resistant variants (S at position 315 replaced by T [S315T], R463L, and the double mutant S315T/R463L) were generated with primers katG904:59-AG CTCGTATGGCACCGGAAC-39 and katG1523:59-TTGACCTCCCACCCGA CTTG-39 (5).

Downloaded from jcm.asm.org by on September 19, 2009

phenotypic characteristics are used to assess relatedness are inherently limited. These include biotyping, antimicrobial susceptibility testing, serotyping, and multilocus enzyme electrophoresis. Characteristics which define a particular strain may be expressed differently under different environmental conditions. For example, some toxins involved in virulence due to Escherichia coli are expressed only in a host animal but are repressed on solid growth media. In other cases, point mutations which result in the abnormal expression of a gene responsible for an observed phenotype can cause related bacteria to appear different when in fact they are closely related (2). Because of the problems associated with phenotypic typing techniques, interest in the development of DNA-based or genotypic typing methods has been stimulated. Ribotyping (15, 21) has moderate discriminatory power for organisms such as E. coli and Klebsiella, Haemophilus, and Staphylococcus spp. which contain multiple ribosomal operons. Nevertheless, epidemiologically unrelated isolates frequently demonstrate the same pattern, limiting the utility of this method (4, 14, 23). For bacterial species with only a single ribosomal operon such as mycobacteria, ribotyping detects only one or two bands and has limited utility for epidemiologic purposes (3). Pulsed-field gel electrophoresis has been used extensively as a means of identifying bacterial strain differences (20). While pulsed-field gel electrophoresis has been applied to a number of bacterial species, pathogens such as E. coli or methicillinresistant Staphylococcus aureus represent a genetically restricted subset of strains within a species and consequently may have similar genotypes, which often make them indistinguishable (2). Several methods based on PCR amplification of either short extragenic repetitive sequences or random sequences have been applied to strain identification (2). However, the PCRbased methods have shown either poor reproducibility or limited discriminatory power (19, 26). We describe here a more robust and informative method of differentiating the highly individual conformations assumed by strands of nucleic acid, using enzymes that recognize and cleave these structures. The method comprises steps of (i) separation of DNA strands by heating, (ii) formation of intrastrand structures on cooling, (iii) rapid enzymatic cleavage of these structures before they are disrupted by reannealing of the complementary strands, and (iv) separation and visualization of the resulting structural fingerprint (Fig. 1). By detecting conformational changes with the use of enzymatic cleavage rather than electrophoretic mobility, sequence differences in much larger molecules are detectable. Moreover, because

J. CLIN. MICROBIOL.

VOL. 34, 1996

SEQUENCE ANALYSIS WITH A STRUCTURE-SPECIFIC ENDONUCLEASE

TABLE 1. Bacterial strains tested in this study Bacterium

ATCC no.

Escherichia coli O157:H7 Shigella sonnei Shigella dysenteriae Shigella dysenteriae Shigella dysenteriae Salmonella typhimurium Salmonella enteritidis Salmonella arizonae Yersinia enterocolitica Campylobacter jejuni Staphylococcus aureus Staphylococcus aureus Staphylococcus hominus Staphylococcus warneri

25931 29026 29027 12021 23566 13076 13314 27729 33291 13565 33591 29885 17917

Description

Clinical isolate Serotype 8 Serotype 1 Serotype 2

Enterotoxic Methicillin resistant

RESULTS Examination of bacterial 16S rRNA genes by CFLP analysis. The cleavage reactions used in CFLP analysis are partial digests producing nested sets of products from each 59-endlabeled DNA fragment. The ideal pattern, therefore, should have both a full range of band representation, from short products a few nucleotides long to residual undigested material, and a relatively uniform distribution of signal across the pattern. The DNA sequences for the amplicons derived from the 16S rRNA genes of some of the organisms examined here are shown in Fig. 2. CFLP analysis of both DNA strands of eight different isolates of E. coli O157:H7 are shown in Fig. 3 (lanes 1 through 8 of each panel). The CFLP fingerprints of these closely related strains are indistinguishable from each other. The four isolates of Shigella, which are evolutionarily close to E. coli O157:H7 (Fig. 3, lanes 9 to 12) show a close relationship

to each other and a strong familial resemblance to E. coli. Sequence differences between the 16S rRNA genes of the Shigella isolate and E. coli O157:H7 are reflected as changes in several bands (e.g., mobility shifts, appearance or disappearance of bands, or intensity changes). Salmonella species, which are evolutionarily close to both E. coli O157:H7 and Shigella spp., show a similar strong familial resemblance (Fig. 3, lanes 13 to 15) to both groups of organisms and yield related yet clearly different structural fingerprints for each species. On the antisense strand, a band of approximately 60 nt in length is unique to S. typhimurium (Fig. 3A, lane 13), while the sense strand S. arizonae pattern (Fig. 3B, lane 15) shows a characteristic band shift at approximately 180 nt. Structural fingerprints of staphylococcal species show a strong relatedness to each other yet have distinct differences from those of the gram-negative organisms. Finally, Campylobacter jejuni and Yersinia enterocolitica show patterns indicative of distinct genera, as would be expected based on the DNA sequence. Sensitivity of Cleavase I structural fingerprints to reaction conditions. To assess the reproducibility of these patterns, we examined reaction parameters that would influence the extent of cleavage (enzyme amount and reaction time) and those that would influence the formation of the structures (temperature and salt concentration). The effects of varying these reaction conditions and constituents on the development of the structural fingerprint patterns are shown in Fig. 4. Since the CFLP is a partial digest, it is desirable for the patterns to show some residual uncut material to ensure that the pattern includes the upper range of digestion products. Therefore, the DNA is considered overdigested when no uncut material remains. We examined the effects of varying the enzyme concentration and the reaction time and the degree to which these variations would cause overdigestion. All of these tests were performed with the 516-bp amplicon derived from the R422Q tyrosinase mutant spanning exons 2 through 4 (7). A standard reaction was defined as including 100 fmol of this DNA digested with 25 U of Cleavase I enzyme at 558C for 2 min without added KCl. For each variable analyzed below, the other parameters of the reaction were held to the standard. As the amount of Cleavase I enzyme was increased, more extensive digestion was observed (Fig. 4). However, optimal levels of cleavage were obtained over a broad range between 5 and 25 U of enzyme, indicating that the reactions can tolerate a fivefold variation in the amount of enzyme without significant detriment to the pattern. Variations in the time of incubation showed that the cleavage pattern development was very rapid, being nearly complete within 1 min, and that the DNAs were overdigested when the time exceeded 5 min. These data show that while care must be taken, absolute precision of time and enzyme measurement is not required to achieve optimal digestion. Longer DNAs may be more easily overdigested because they contain a greater number of potential cleavage sites. Overdigestion of such DNAs can be prevented by including 1 mM MgCl2, in addition to the MnCl2, in the reaction mixture to slow the reaction approximately fivefold (data not shown). Alternatively, the amount of enzyme may be reduced to 2 to 5 U per reaction or the cleavage time may be shortened to 0.5 to 1 min. When conditions likely to influence structure were examined, it was expected that changes in the banding patterns would be observed, but we questioned whether the changes would be abrupt and global, and therefore less reproducible, or gradual, so that patterns would be reproducible even with minor differences in the reaction conditions. To examine the effects of temperature on the cleavage pat-

Downloaded from jcm.asm.org by on September 19, 2009

Tyrosinase gene. Genomic DNAs and cDNA clones of a wild-type and a mutant human tyrosinase gene were used as controls for the CFLP reaction and were a gift from Richard Spritz (University of Wisconsin, Madison). The tyrosinase mutant R422Q has a G-to-A substitution in codon 422 (7). A second mutant version of the tyrosinase gene, R278X, was created from the wild-type cDNA by a single base substitution 525 nt from the 59 end of the sense strand, by recombinant PCR methods (9). R278X has a C-to-T substitution that introduces a stop codon at position 278 (24). A 516-bp fragment spanning exons 2 through 4 of the wild-type gene was made through the use of primers Tyr1 (59-CTGAATCTTG TAGATAGCTA-39) and Tyr3 (59-GCCATCAGTCTTTATGCAAT-39); 1,059bp fragments spanning exons 1 to 4 were created with primers Tyr1 and Tyr4 (59-GCAAGTTTGGCTTTTGGGGA-39). Hepatitis C virus 5* noncoding region (NCR). A cDNA of the 59 noncoding region (NCR) of the hepatitis C viral (HCV) genome was a gift from Manual Altamirano (University of British Columbia, Vancouver) (1). The cDNA was cloned into a TA vector (In Vitrogen, St. Louis, Mo.). A nested set of fragments were amplified with the primers described below, from 10 ng of template fragment. All fragments were labeled on their 59 ends with tetrachloro fluoresceinlabeled HCV-1 primer (59-CTGTCTTCACGCAGAAAGCG-39). Primers HCV-2 (59-CACGGTCTACGAGACCAC-39), HCV-3 (59-GCGGGGGCACGCCCA AAT-39), HCV-4 (59-TCCAAGAAAGGACCCGGTC-39), and HCV-5 (59-AT TCCGGTGTACTCACCGG-39), all unlabeled, were used with the HCV-1 primer to make products of 281, 186, 145, and 117 bp, respectively. Cleavage reactions. Unless otherwise noted, each cleavage reaction mixture was assembled as follows. Approximately 100 fmol of the DNA substrate in 15 ml of water was heated to 958C for 5 to 10 s, cooled to the reaction temperature, and mixed with 5 ml of a prewarmed solution containing 12.5 to 25 U of Cleavase I, 2 ml of 103 CFLP buffer (100 mM MOPS [pH 7.5], 0.5% Tween 20, 0.5% Nonidet P-40), and 2 ml of 2 mM MnCl2. Reactions were stopped after 1 to 5 min by the addition of 16 ml of 95% formamide with 10 mM EDTA (pH 8.0) and 0.02% methyl violet. The cleavage fragments were resolved by electrophoresis in 6 or 10% polyacrylamide gels (19:1 cross-link) containing 7 M urea in 0.53 TBE. Detection of structural fingerprint patterns. Polyacrylamide gel cassettes were transferred to an Hitachi model FMBIO-100 fluorescence analyzer and were scanned. A 585-nm filter was used for scans of tetramethyl rhodamine-labeled DNA; 505 nm was used for fluorescein-labeled DNA.

3131

3132

BROW ET AL.

J. CLIN. MICROBIOL.

Downloaded from jcm.asm.org by on September 19, 2009

FIG. 2. DNA sequences of amplicons derived from amplification of 16S rRNA genes of representative bacteria used in this study. The sequence of the E. coli O157:H7 amplicons is given as a reference. Deleted bases are indicated by asterisks.

tern, the reactions were performed at temperatures ranging from 35 to 758C (Fig. 4). These results show that the patterns change gradually with temperature variation, with particular bands emerging and receding in accordance with the different

stabilities of the causative structures. As would be expected, bands were progressively less abundant at higher temperatures, as less stable structures melted out. While these results do demonstrate that DNAs to be compared should be digested

VOL. 34, 1996

SEQUENCE ANALYSIS WITH A STRUCTURE-SPECIFIC ENDONUCLEASE

3133

Downloaded from jcm.asm.org by on September 19, 2009

FIG. 3. CFLP structural fingerprinting of 16S rRNA genes of bacteria. A 350-bp amplicon was generated by PCR amplification of the 16S rRNA gene of each strain as described in Materials and Methods. The amplicons were subjected to CFLP analysis. (A) Antisense strand (primer EB-L labeled). Lanes 1 to 8, E. coli O157:H7 strains; 9, Shigella sonnei; 10, Shigella dysenteriae type 2; 11, Shigella dysenteriae type 8; 12, Shigella dysenteriae type 1; 13, Salmonella typhimurium; 14, Salmonella enteritidis; 15, Salmonella arizonae; 16, Staphylococcus aureus, (ATCC 13565); 17, Staphylococcus aureus, (ATCC 33591); 18, Staphylococcus hominus; 19, Staphylococcus warneri; 20, C. jejuni; 21, Y. enterocolitica. (B) Sense strand (primer EB-R labeled). Lanes: 1 to 15, same as those in panel A; 16, Y. enterocolitica; 17, Staphylococcus aureus (ATCC 13565); 18, Staphylococcus aureus, (ATCC 33591); 19, Staphylococcus hominus; 20, Staphylococcus warneri; 21, C. jejuni.

3134

BROW ET AL.

J. CLIN. MICROBIOL.

at the same reaction temperatures, they also suggest that minor changes in temperature, such as might be caused by instrument variation, would have little impact on the reproducibility of the pattern. The temperature also had an influence on the extent of digestion. As the reaction temperatures were increased from 35 to 508C, the patterns showed progressively more extensive digestion. This is consistent with our observation that the activity of this enzyme increases as the temperature is raised. Overdigestion at higher temperatures is prevented by the gradual melting out of the cleavage structures, leading to a wellbalanced pattern. The effect of salt concentration on the cleavage patterns was also examined (Fig. 4). An inhibitory effect is reflected in the reduced cleavage seen with increasing KCl concentrations (Fig. 4). The rate of cleavage by Cleavase I decreases with increasing concentrations of KCl. The cleavage activity is reduced by approximately 65% at 50 mM KCl, when assayed with an oligonucleotide whose structure should not be affected by salt (see activity assay in Materials and Methods) (unpublished data). Reannealing of the strands may also contribute to the suppression of cleavage, as suggested by the appearance of increasing amounts of reassociated material, seen as a band migrating above the single-stranded uncut material at the higher salt concentrations. Strikingly, the addition of salt did not cause perceptible reorganization of the structures recognized by the enzyme, indicating that minor differences in the salt content of these reaction mixtures will not adversely affect

reproducibility. Taken together, the results shown in Fig. 4 demonstrate that the structural fingerprint patterns generated by the Cleavase I enzyme are quite resistant to minor variations in the reaction conditions. Identification and positioning of mutations associated with isoniazid resistance in M. tuberculosis. CFLP patterns of three variants of the M. tuberculosis katG gene were compared with that of the wild type gene. Analysis of the antisense strand revealed the appearance of two prominent bands (A and B) due to cleavage approximately 127 and 130 nt from the 59 end which are representative of the R463L mutation in both the R463L and S315T/R463L mutants (Fig. 5a). The actual nucleotide change is located 135 nt from the 59 end. In addition, the S315T mutation is visible as a faster migrating band due to cleavage approximately 580 nt from the 59 end of the S315T and S315T/R463L mutants, again corresponding closely with the actual position of the mutation located at 579 nt (Fig. 5b). Similar data on the detection and localization of the mutations were obtained for the sense strand (data not shown). When determined alongside the corresponding sequencing ladder, the pattern variations observed in the CFLP analysis have been within about 50 nt of the actual sites of mutation (sequence data not shown). This localization effect suggests that most or all of the structures that are cleaved are formed by local interactions, rather than by interactions between distant regions of the nucleic acid strand. Disruptions in local structures result in the observed differences in CFLP structural fingerprints. If the structures re-

Downloaded from jcm.asm.org by on September 19, 2009

FIG. 4. Effect of reaction condition variation on cleavage pattern development. The integrity of the structural fingerprint patterns was tested under a variety of solution conditions. The DNA template was the 516-bp tyrosinase gene fragment spanning exons 2 through 4 of the R422Q mutant human tyrosinase gene, labeled on the 59 end of the antisense strand with tetrachlorofluorescein. In each panel, the lane labeled S indicates that the standard reaction was reproduced precisely. (A) Reactions were performed with 1, 5, 10, 25, 50, 100, or 250 U of Cleavase I. (B) In separate reactions, the incubations were stopped after 0.1, 1, 2, 5, 10, 15, 20, 30, 60, or 120 min. In the lane labeled 0.1 min, the stop solution was added immediately after the enzyme addition. (C) Following denaturation at 958C, the DNAs were cooled and the cleavage was performed at 35 to 758C in 58C increments. (D) The reaction mixtures were supplemented with 0, 1, 5, 10, 15, 20, 25, 30, or 50 mM KCl (in addition to the 2.5 mM derived from the enzyme storage buffer). The positions of the uncut DNA and the 72-nt band are indicated.

VOL. 34, 1996

SEQUENCE ANALYSIS WITH A STRUCTURE-SPECIFIC ENDONUCLEASE

cognized by the enzyme are in fact local, their formation should be relatively insensitive to the presence of remote flanking sequences. To examine this point, we compared the CFLP patterns of the 516- and 1,059-bp tyrosinase gene fragments which start from the same labeled 59 end but which extended either 516 or 1,059 nt to their 39 ends (Fig. 6). Comparison of the banding patterns of these samples showed that they were identical up to the point at which the shorter pattern ends. This result supports the model that the bands generated in this reaction are primarily due to close interactions along the DNA strand and that the extra length on the 39 end of the 1,059-mer has little influence on these 59-proximal structures.

FIG. 6. Nested fragments with a common 59 end give nested cleavage patterns. Fragments with different sizes were amplified from a human tyrosinase cDNA clone. A 516-bp fragment, spanning exons 2 to 4, and an overlapping 1,059-bp fragment, covering exons 1 to 4, were amplified from cDNA clones of the wild-type human tyrosinase gene. The antisense strand of each fragment was labeled on the 59 end with TET. Lane M, size markers, with fragment sizes (in nucleotides) indicated on the left.

FIG. 7. Individual regions of DNA can be involved in multiple cleavage structures. (A) Differential responses of proximal bands to deletion of flanking sequences. A nested set of fragments were amplified from the 59 NCR of the HCV genome (1). All fragments were labeled on their 59 ends with a TETlabeled HCV-1 primer. The sizes of the uncut fragments (281, 186, 145, and 117 nt) and the predominant cleavage products (95, 71, and 69 nt) are indicated on the left. (B) Multiple band changes caused by a single-base sequence change. The cleavage pattern from the 1,059-nt segment of the wild-type (wt) human tyrosinase gene is compared with the pattern from the R278X mutant, which is different at a single site. Sense and antisense strand patterns were compared. The distance of the mutation from the 59 end of each labeled strand is indicated below each mutant-containing lane. Lane M contains markers, with the uncut DNA and marker fragment sizes (in nucleotides) indicated on the left.

Several aspects of the data presented here suggest that the pool of molecules in the cleavage reaction assume a multiplicity of competitive structures and that the cleavage pattern is a composite picture of individual cleavage events. This is suggested by the proximity of the bands in some patterns, which are closer than the cleavage of colinear structures would allow. The existence of multiple, competing structures is more strongly indicated by the observation that single-base changes can cause multiple band changes within the structural fingerprint patterns. To test if close bands in a fingerprint pattern arise from multiple structures, a nested set of DNA fragments sharing a common labeled 59 end were amplified from the 59 NCR of HCV. Cleavage in this region produces strong bands only 2 nt apart at 69 and 71 nt (Fig. 7A). We questioned whether these products arose from two points of cleavage on the same structure or from single cuts at the ends of two competing structures. The banding pattern of the 186-nt fragment is indistinguishable from the full-length pattern over the region shared by the two fragments. As the fragment is truncated further to 145 nt, two new bands occur, presumably because of the deletion of previously preferred secondary structural sequences in the 186- and 281-nt fragments and subsequent creation of alternative secondary structures in the 145-nt fragment. When the 69- and 71-nt bands are compared in the fingerprint patterns of the successively shorter DNAs, it is clear that deletion of the sequences beyond 145 nt causes a substantial reduction in the band at 71 nt and an enhancement of the band at 69 nt. This differential response to the deletion of flanking sequences shows that multiple coexisting structures are being cleaved. The point is further illustrated by examination of Fig. 7B, which shows the cleavage patterns for the sense and antisense strands of two 1,059-nt fragments of the tyrosinase DNA that differ at a single base near the middle of the fragment. The

Downloaded from jcm.asm.org by on September 19, 2009

FIG. 5. Identification and positioning of mutations associated with isoniazid resistance in M. tuberculosis. (a) Amplicons, generated with a 59 labeled primer and an unlabeled primer, were heat denatured for 15 s, rapidly cooled to 608C, and digested with 25 U of Cleavase I for 2 min. The samples are shown, with M indicating DNA size markers (the sizes are given on the left). Variant bands distinguishing mutant from wild-type (WT) DNA are marked A and B. (b) A second gel electrophoresis was performed to expand the 500- to 600-nt region. Bands C in the wild type and R463L mutant are similar while the same band migrates faster in variants containing the S315T mutation.

3135

3136

BROW ET AL.

fingerprints of the sense and antisense strands of the R278X fragment have several differences in the 500-nt region, compared with the wild-type patterns. Thus multiple changes in the CFLP pattern can result from a single point mutation. Intergenic region. We examined the potential of the CFLP method for differentiating bacteria. In order to provide the greatest utility, a bacterium-typing method should be able to differentiate both species and strains. The genotype of the 16S rRNA genes, while useful for general classification, would not be expected to necessarily allow identification to the species and strain levels. For these purposes, we examined a 550-bp amplicon derived from the intergenic region of several isolates of Shigella and Salmonella spp. (Fig. 8). The structural fingerprints generated by Cleavase I treatment clearly differentiated all species of Salmonella and Shigella (Fig. 8). Again, a familial resemblance is clear in all of these evolutionarily related bacteria, yet differences in the structural fingerprints of each genus and species are clearly visible. The power of CFLP analysis is even more striking when the analysis of Shigella spp. is examined (Fig. 8, lanes 1 to 4 and 8 to 11). Shigella sonnei was clearly differentiated from the three Shigella dysenteriae isolates examined. Furthermore, although serotypes 1, 2, and 8 of Shigella dysenteriae yielded related structural fingerprints, the antisense patterns for the individual serotypes could, however, be distinguished. Therefore, CFLP analysis can potentially differentiate microorganisms to the genus, species, and strain levels. DISCUSSION We have shown here that structure-specific endonucleases can be used as sensitive tools for the analysis of the highly

individual conformations assumed by single strands of DNA. Cleavage of the folded structure produces a fixed and reproducible record of the conformation of local regions of DNA. Because such structures are responsive to the sequence of the DNA, the pattern of the cleavage products serves as a structural fingerprint of the molecule. Comparison of the structural fingerprints of related DNAs is an extremely sensitive and rapid means of determining if they are identical. We have been able to distinguish a number of bacterial genera and species on the basis of CFLP analysis of appropriate markers. The method is suitable for analysis of 16S rRNA genes, intergenic regions, or virulence factors such as toxin, pilin, or antibiotic resistance genes. CFLP analysis of the 16S rRNA genes clearly distinguished E. coli and Salmonella, and Shigella spp. in contrast to a recent report by Widjojoatmoto et al. (25) in which analysis of this same amplicon by single-strand conformation polymorphism could not distinguish E. coli and Shigella spp. Because the structural fingerprint relies on the cleavage of intrastrand structures, we tested the influence of changes in reaction conditions that are likely to influence these structures. The patterns were remarkably robust, changing relatively little in response to two- to threefold changes in such variables as the time of incubation, enzyme concentration, and salt concentration. As expected, alterations in temperature did change the pattern of fragments, but the change was gradual, requiring several degrees shift to alter any particular cleavage structure. The fact that the influence of these different conditions is not abrupt makes the patterns reproducible, even when carried out on different days or in different laboratories. The use of a thermostable endonuclease, Cleavase I, allows the cleavage reactions to be performed at elevated temperature, which is fundamental to realizing the full benefit of this assay. For example, if a particularly stable secondary structure is assumed by the DNA, a single nucleotide change is unlikely to significantly alter that structure or the cleavage pattern it produces. Elevated temperatures can be used to bring structures to the brink of instability, so that the effects of small changes in sequence are maximized and revealed as alterations in the cleavage pattern. Furthermore, the use of high temperature reduces long-range interactions along the DNA strands, thereby increasing the likelihood that observed cleavage differences will reflect the locations of the base changes. The potential for multiple localized pattern changes in response to a single sequence change, as shown in Fig. 7B, presents an additional advantage of this method of mutation detection over other methods in use. Direct sequencing of this region would provide a single variant peak on which to base a conclusion. In contrast, the multiplicity of effects seen here provides redundant confirmation of the base change, allowing multiple checkpoints within the pattern to be compared. This is especially useful in the cases of heterozygosity, or the presence of drug-sensitive and drug-resistant bacteria in a single sample. While the analyses shown above have been done with nonisotopic labels, the method is fully compatible with radiolabeled DNAs and standard autoradiography. When staining or uniform labeling are used, the patterns produced are more complex, as all fragments are visible, and the localization feature is lost because bands cannot be measured from a discrete labeled end, yet variations in the cleavage patterns are readily detected. We have also found that CFLP can be performed on differently labeled wild-type and mutant alleles in the same reaction, with no discernible effect of the patterns of each. When fragment size analyses are performed on a fluorescencebased DNA sequencer, the wild-type pattern can serve as an

Downloaded from jcm.asm.org by on September 19, 2009

FIG. 8. Examination of 16S-23S intergenic region of Salmonella and Shigella spp. by CFLP analysis. The intergenic regions were amplified with a 16S sense primer and a 23S antisense primer to generate an amplicon of approximately 550 bp. The CFLP reactions were performed under identical conditions for each DNA strand. CFLP reactions were performed at 558C for 2 min. Lanes: M, size marker DNA (with sizes (in nucleotides) indicated on the left); 1 through 7 (sense strand analysis) S. dysenteriae type 8, S. sonnei, S. dysenteriae type 1, S. dysenteriae type 2, S. typhimurium, S. arizonae, and S. enteritidis, respectively; 8 through 14 (antisense strand), S. dysenteriae type 8, S. sonnei, S. dysenteriae type 1, S. dysenteriae type 2, S. typhimurium, S. arizonae, and S. enteritidis, respectively.

J. CLIN. MICROBIOL.

VOL. 34, 1996

SEQUENCE ANALYSIS WITH A STRUCTURE-SPECIFIC ENDONUCLEASE

ACKNOWLEDGMENTS We thank Frank Cockerill, Jeff Klinger, Richard Spritz, and Manual Altamirano for DNA samples and bacterial cultures. Cleavase I reagents may be obtained from either Boehringer Mannheim Biochemicals (Indianapolis, Ind.) for CFLPScan and CFLPScan-XL kits or Life Technologies, Inc. (Gaithersburg, Md.) for PowerScan kits. This work was supported by grants from the National Institutes of Health (1R43 GM51704-01) and the National Institutes of Standards and Technology Advanced Technology Program (94-05-0012). REFERENCES 1. Altamirano, M., A. Delaney, A. Wong, J. Marostenmaki, and D. Pi. 1995. Mutations in the catalase-peroxidase gene from isoniazid-resistant Mycobacterium tuberculosis isolates. J. Infect. Dis. 171:1034–1038. 2. Arbeit, R. 1995. Laboratory procedures for the epidemiologic analysis of microorganisms, p. 190–208. In P. R. Murry, E. J. Baron, M. A. Pfaller, F. C. Tenover, and R. H. Yolken. Manual of Clinical Microbiology. American Society for Microbiology, Washington, D.C. 3. Arbeit, R. D., A. Slutsky, T. W. Barber, J. N. Maslow, S. Niemczyk, J. O. Falkinham III, G. T. O’Conner, and C. F. von Reyn. 1993. Genetic diversity among strains of Mycobacterium avium causing monoclonal and polyclonal bacteremia in patients with AIDS. J. Infect. Dis. 167:1384–1390. 4. Arthur, M., R. D. Arbeit, C. Kim, P. Beltran, H. Crowe, S. Steinbach, C. Campenelli, R. A. Wilson, R. K. Selander, and R. Goldstein. 1990. Restriction fragment length polymorphisms among uropathogenic Escherichia coli pap-related sequences compared with rrn operons. Infect. Immun. 58:471– 479. 5. Cockerill, F. R. I., J. R. Uhl, Z. Temesgen, Y. Zhang, L. Stockman, G. D. Roberts, D. L. Williams, and B. C. Kline. 1995. Rapid identification of a point mutation of the Mycobacterium tuberculosis catalase-peroxidase (katG) gene associated with isoniazid resistance. J. Infect. Dis. 171:240–245. 6. Connor, B. J., A. A. Reyes, C. Morin, K. Itakura, R. L. Teplitz, and R. B. Wallace. 1983. Detection of sickle cell beta S-globin allele by hybridization

with synthetic oligonucleotides. Proc. Natl. Acad. Sci. USA 80:278–282. 7. Geibel, L. B., R. K. Tripathi, R. A. King, and R. A. Spritz. 1991. Organization and nucleotide sequences of the human tyrosinase gene and a truncated tyrosinase-related segment. J. Clin. Invest. 87:1119–1122. 8. Hayashi, K. 1994. Manipulation of DNA by PCR, p. 3–13. In K. B. Mullis, F. Ferre, and R. A. Gibbs (ed.), The polymerase chain reaction. Birkhauser Press, Boston. 9. Higuchi, R. 1989. Using PCR to engineer DNA, p. 61–70. In H. A. Ehrlich (ed.), PCR technology: principles and applications for DNA amplification. Stockton Press, New York. 10. Kapur, V., L.-L. Li, S. Iordanescu, M. Hamrick, A. Wanger, B. N. Kreiswirth, and J. M. Musser. 1994. Characterization by automated DNA sequencing of mutations in the gene (rpoB) encoding the RNA polymerase b subunit in rifampin-resistant Mycobacterium tuberculosis strains from New York City and Texas. J. Clin. Microbiol. 32:1095–1098. 11. Maxam, A. M., and W. Gilbert. 1980. Sequencing end-labeled DNA with base-specific chemical cleavages. Methods Enzymol. 57:499–561. 12. Mullis, K. B., and F. A. Faloona. 1987. Specific synthesis of DNA in vitro via a polymerase-catalyzed chain reaction. Methods Enzymol. 155:335–350. 13. Orita, M., Y. Suzuki, T. Sekiya, and K. Hayashi. 1989. A rapid and sensitive detection of point mutations and genetic polymorphisms using polymerase chain reaction. Genomics 5:874–879. 14. Poh, C. L., C. C. Yeo, and L. Tay. 1992. Genome fingerprinting by pulsed field gel electrophoresis and ribotyping to differentiate Pseudomonas aeruginosa serotype 011 strains. Eur. J. Clin. Microbiol. Infect. Dis. 11:817–822. 15. Rabkin, C., W. Jarvis, R. Anderson, J. Govan, J. Klinger, J. LiPuma, W. Martone, H. Monteil, C. Richard, S. Shigeta, A. Sossa, T. Stull, J. Swenson, and D. Woods. 1989. Pseudomonas cepacia typing systems: collaborative study to assess their potential in epidemiologic investigations. Rev. Infect. Dis. 11:600–607. 16. Saiki, R. K., C.-A. Chang, C. H. Levinson, T. C. Warren, C. D. Boehm, H. H. Kazazian, and H. A. Erlich. 1988. Diagnosis of sickle cell anemia and b-thalassemia with enzymatically amplified DNA and non-radioactive allelespecific oligonucleotide probes. N. Engl. J. Med. 319:537–541. 17. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed., p. 6.36–6.46. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. 18. Sanger, F., S. Nicklen, and A. R. Coulson. 1977. DNA sequencing with chain-terminating inhibitors. Proc. Natl. Acad. Sci. USA 74:5463–5467. 19. Sarkar, G., H. Yoon, and S. S. Sommer. 1992. Dideoxyfingerprinting (ddF): a rapid and efficient screen for the presence of mutations. Genomics 13:441– 43. 20. Saulnier, P., C. Bourneix, G. Pre´vost, and A. Andremont. 1993. Random amplified polymorphic DNA assay is less discriminant than pulsed-field gel electrophoresis for typing strains of methicillin-resistant Staphylococcus aureus. J. Clin. Microbiol. 31:982–985. 21. Schwartz, D. C., and C. R. Cantor. 1984. Separation of yeast chromosomesized DNAs by pulsed field gradient gel electrophoresis. Cell 37:67–75. 22. Telenti, A., P. Imboden, F. Marchesi, D. Lowrie, S. Cole, M. J. Colston, L. Matter, K. Schopfer, and T. Bodmer. 1993. Detection of rifampin-resistant mutations in Mycobacterium tuberculosis. Lancet 341:647–650. 23. Tenover, F. C., R. D. Arbeit, G. Archer, J. Biddle, S. Byrne, R. Goering, G. Hancock, G. A. Hebert, B. Hill, R. Hollis, W. R. Jarvis, B. Kreiswirth, W. Eisener, J. Maslow, L. K. McDougal, J. M. Miller, M. Mulligan, and M. A. Pfaller. 1994. Comparison of traditional and molecular methods of typing isolates of Staphylococcus aureus. J. Clin. Microbiol. 32:407–415. 24. Tripathi, R. K., S. Bundey, M. A. Musarella, S. Droetto, K. M. Strunk, S. Holmes, and R. A. Spritz. 1993. Mutations in the tyrosinase gene in IndoPakistani patients with type 1 (tyrosinase-deficient) oculocutar (OCA). Am. J. Hum. Genet. 53:1173–1179. 25. Widjojoatmoto, W., A. C. Fluit, and J. Verhoef. 1995. Molecular identification of bacteria by fluorescence-based PCR-single-strand conformation polymorphism analysis of the 16S RNA gene. J. Clin. Microbiol. 33:2601–2606. 26. Woods, C. R., Jr., J. Versalovic, T. Koeuth, and J. P. Lupski. 1992. Analysis of relationships among isolates of Citrobacter diversus by using DNA fingerprints generated by repetitive sequence-based primers in the polymerase chain reaction. J. Clin. Microbiol. 30:2921–2929.

Downloaded from jcm.asm.org by on September 19, 2009

internal control for enzyme digestion and as an internal reference that allows accurate discrimination of pattern differences in the mutant samples. When calibrated against a sequence ladder or other known markers, the wild-type pattern also allows precise sizing of the mutant bands. With the CFLP patterns normalized in this way, experimental samples can be compared day-to-day and lab-to-lab, with a high degree of confidence. Lastly, the ability to rapidly perform an independent analysis of both DNA strands with the use of differently labeled PCR primers facilitates complementary confirmatory information analogous to sequencing both DNA strands. Analysis of DNAs through the probing of their structures with the Cleavase I enzyme promises to be a powerful tool in the comparison of sequences. As demonstrated here, the method is able to detect single base changes in DNA molecules over 1 kb long. In addition to the work discussed here, we have been able to apply CFLP analysis to detection of mutations responsible for rifampin resistance in M. tuberculosis and for determining the genotype of HCVs and have successfully analyzed DNAs as long as 2.7 kb. The low cost, rapidity, and reliability of the analysis make this method very suitable for the rapid screening of DNAs, not only for the examination of disease-associated mutations but also in applications such as tissue typing, genetic identity, bacterial and viral typing, and mutant screening in genetic crosses.

3137