Performance Criteria of DNA Fingerprinting Methods for Typing of Helicobacter pylori Isolates: Experimental Results and Meta-Analysis

Abstract

Typing systems are used to discriminate between isolates of Helicobacter pylori for epidemiological and clinical purposes. Discriminatory power and typeability are important performance criteria of typing systems. Discriminatory power refers to the ability to differentiate among unrelated isolates; it is quantitatively expressed by the discriminatory index (DI). Typeability refers to the ability of the method to provide an unambiguous result for each isolate analyzed; it is quantitatively expressed by the percentage of typeable isolates. We evaluated the discriminatory power and the typeability of the most currently used DNA fingerprinting methods for the typing of H. pylori isolates: ribotyping, PCR-based restriction fragment length polymorphism (PCR-RFLP) analysis, and random amplified polymorphism DNA (RAPD) analysis. Forty epidemiologically unrelated clinical isolates were selected to constitute a test population adapted to the evaluation of these performance criteria. A meta-analysis of typeability and discriminatory power was conducted retrospectively with raw data from published studies in which ribotyping, PCR-RFLP, RAPD, repetitive extragenic palindromic DNA sequence-based PCR (REP-PCR), or pulsed-field gel electrophoresis (PFGE) was used. Experimental results and the meta-analysis demonstrated the optimal typeability (100%) and the excellent discriminatory powers of PCR-based typing methods: RAPD analysis, DIs, 0.99 to 1; REP-PCR, DI, 0.99; and PCR-RFLP analysis, DIs, 0.70 to 0.97). Chromosome restriction-based typing methods (ribotyping and PFGE) are limited by a low typeability (12.5 to 75%) that strongly decreases their discriminatory powers: ribotyping, DI, 0.92; PFGE, DIs, 0.24 to 0.88. We do not recommend the use of ribotyping and PFGE for the typing of H. pylori isolates. We recommend the use of PCR-based methods.

Helicobacter pylori colonizes human gastric mucosa, causes chronic gastritis and ulcers, and is a major risk factor for the development of gastric cancers (15, 17, 26). Discrimination between closely related isolates of H. pylori is needed for epidemiologic and clinical purposes (23). Precise methods of strain characterization are necessary to monitor H. pylori infections (familial or nosocomial transmissions, treatment failures, relapses, and cocolonizations). On the basis of the high degree of genomic variability of isolates within the H. pylori species, various molecular techniques have been developed to differentiate clinical isolates (23). Some of these are based on the restriction patterns of the chromosomal DNA, such as restriction enzyme analysis (6, 8, 9, 38, 46, 67, 76), ribotyping (5, 11, 21, 24, 37, 40, 43, 44, 47–50, 52, 54, 56, 58, 68), and pulsed-field gel electrophoresis (PFGE) (27, 57, 64–66, 73). Others are based on PCR, such as sequencing of PCR products (1, 18, 33), restriction fragment length polymorphism (RFLP) analysis of PCR products (PCR-RFLP analysis) (1, 4, 9, 12, 16, 18–20, 22, 29–31, 33, 35, 36, 41, 42, 45, 52, 56, 58, 60–62, 67, 74), repetitive extragenic palindromic DNA sequence-based PCR (REP-PCR) (13, 25, 72), and random amplified polymorphism DNA (RAPD) analysis (2, 3, 7, 14, 16, 32, 33, 39, 55, 59, 67, 69–71, 73, 75, 76).

The techniques most frequently reported to be used for typing are PCR-RFLP analysis, RAPD analysis, and ribotyping. These techniques are increasingly used to type H. pylori isolates, but their performances, particularly their typeabilities and discriminatory powers, have not been completely and explicitly evaluated. Typeability refers to the ability of the method to provide an unambiguous result for each isolate analyzed (63). A good fingerprinting method requires an excellent typeability in order to be able to type all the isolates studied. Discriminatory power refers to the ability of the method to differentiate among unrelated isolates (63). To be sure that two isolates with the same fingerprint are really genetically linked, the discriminatory power must be high.

The discriminatory powers of methods used to type H. pylori have most often been assessed subjectively or by considering the number of types obtained by each method. Although this allowed some level of discrimination, the comparison of methods is difficult because several values, generally, the size and the number of groups of isolates, are compared. In 1988, Hunter and Gaston (28) used a single numerical index of discrimination, the discriminatory index (DI), based on Simpson's index of diversity, to compare the discriminatory powers of several methods used to type Candida albicans and some enteric species. The calculation is based on the probability that two unrelated isolates would be classified as the same type. Since that time, the DI has widely been used to compare the discriminatory powers of typing methods, and the determination of this index is now recommended in guidelines for the evaluation of typing methods (63). It was never used, however, to evaluate H. pylori typing methods.

Special attention should be paid to the selection of an appropriate set of test strains for evaluation of the typeability and discriminatory power of a typing system. The rationale for most of the reported typing studies is that different isolates of an epidemiologic cluster or successive isolates from a single patient may be clonally related, that is, are directly derived from a common parental strain. The aims of such studies are mainly to group the isolates rather than to distinguish them. Several studies have compared many epidemiologically related isolates thought to be clonal and a few (less than 10) unrelated isolates (2, 5, 6, 9, 14, 20–22, 27, 32, 44, 50, 52–57, 60, 67, 70, 71, 76). These studies have confirmed that the typing systems tested have good concordance in terms of the epidemiologic relatedness of the isolates. However, because in these studies the isolates are often epidemiologically related, such strategies are not appropriate for evaluation of the typeability and the discriminatory power of a typing method (63). A large test population of isolates correctly identified to the species level must be selected to reflect the diversity of the species as much as possible (63). The test population should include isolates that are unrelated epidemiologically on the basis of detailed clinical and epidemiological data. Thus, studies evaluating the typeabilities and discriminatory powers of typing system needs to be done with collections of isolates from unrelated patients.

Typeability and discriminatory power can be expressed quantitatively (28, 63). In reported studies in which enough unrelated isolates have been typed, some of these criteria were evaluated qualitatively but were never expressed quantitatively and thus could not be compared from one study to another. Furthermore, the performances of more than two different typing techniques were never compared by using the same H. pylori isolates.

To evaluate the discriminatory power and the typeability of ribotyping, PCR-RFLP analysis, and RAPD analysis as methods for the typing of H. pylori isolates, we studied 39 unrelated H. pylori isolates from different geographic origin and three familial isolates (from a mother and her son and daughter) considered genetically related. We also performed a meta-analysis of typeability and discriminatory power using raw data from previously reported studies of H. pylori typing systems.

MATERIALS AND METHODS

Bacterial isolates.

The 42 H. pylori isolates were isolated between 1984 and 1995 from gastric tissues collected during endoscopy of 42 dyspeptic patients of different geographic origins. These patients included 4 patients from Venezuela (kindly provided by N. Muñoz), 3 patients from the United States (kindly provided by M. J. Blaser), and 35 patients from different areas of France. Among these 42 isolates, 39 were unrelated and 3 were familial isolates isolated from the mother, the son, and the daughter of a French family.

The isolates were identified to the species level by use of standard criteria, including colony morphology, Gram staining result, and biochemical test results (positivity for urease, catalase, and oxidase). Until they were used, the isolates were frozen at −70°C in 10% glycerol.

Extraction of total DNA.

Stored isolates of H. pylori were regrown on Columbia agar with 5% sheep blood. The plates were soaked with an inoculum of a no. 5 McFarland standard and were incubated for 2 days at 37°C under microaerobic conditions. Bacterial cells were harvested from two plates and were suspended on ice in 1 ml of lysis buffer (10 mM Tris-HCl [pH 8.5], 100 mM EDTA, 1% sodium dodecyl sulfate, 0.05 mg of pronase per ml). The mixture was incubated for 60 min at 37°C. The sample was extracted with an equal volume of phenol-chloroform-isoamyl alcohol (25:24:1), and after centrifugation, the aqueous phase was collected. RNase A (50 μg/ml) was added, and the mixture was incubated for 60 min at 60°C and then ethanol precipitated. After centrifugation (13,000 × g) the pellet was resuspended in 20 μl of sterile distilled water.

The DNA concentrations and the quality of the DNA in the samples were estimated after agarose gel electrophoresis with DNA standards.

PCR-RFLP analysis.

PCR-RFLP analysis was performed as described by Foxall et al. (20) and Owen et al. (52). Four techniques were used to compare the 42 isolates: restriction of a 2.4-kb fragment containing the ureA and ureB genes by HaeIII, BamHI, or HindIII and restriction of a 1.1-kb fragment containing the ureC (now glmM) gene by Sau3A or HaeIII and HindIII.

The PCR was carried out according to the instructions supplied with the reagents (Eurogentec, Seraing, Belgium) in a Perkin-Elmer GeneAmp PCR system 2400 thermal cycler (Perkin-Elmer Cetus, Norwalk, Conn.) in 100 μl containing 1 μl of chromosomal DNA (∼20 ng), 3 mM MgCl₂, each primer at a concentration of 0.2 μM, 2.5 U of Eurotaq DNA polymerase (Eurogentec), each deoxynucleoside triphosphate (Eurogentec) at a concentration of 0.2 μM, 10 mM Tris-HCl (pH 8.3), and 50 mM KCl.

To amplify the 2.4-kb fragment containing the ureA and ureB genes, we used two oligonucleotides with recognition sequences of 5′-AGGAGAATGAGATGA-3′ (base pairs 308 to 322) and 5′-ACTTTATTGGCTGGT-3′ (base pairs 2718 to 2703), respectively (34). To amplify the 1.1-kb fragment containing the ureC gene, we used two oligonucleotides with recognition sequences of 5′-TTTGGGACTGATGGCGTGAGGGGTAA-3′ (base pairs 10 to 35) and 5′-GGACATTCAAATTCACCAGGTTTTGAG-3′ (base pairs 1142 to 1116), respectively (34).

After an initial denaturation of target DNA at 95°C for 5 min, thermal cycling for each set of primers was 95°C for 1 min, 50°C for 1 min, and 72°C for 1 min for a total of 35 cycles. The final cycle included extension for 5 min at 72°C.

Ten microliters of the reaction products was analyzed by electrophoresis on a 0.8% (wt/vol) agarose gel at 110 V for 45 min.

Prior to digestion for RFLP analysis, 90 μl of the PCR product was transferred to a fresh tube, and DNA was precipitated by adding 2 volumes (approximately 180 μl) of ethanol (95%; vol/vol). After 10 min of gentle mixing, the samples were centrifuged at 14,000 × g for 15 min. The DNA pellet was washed with 70% ethanol, dried, and resuspended in 10 μl of sterile distilled water.

For unique restriction, 5 μl of the concentrated DNA was added to 11.5 μl of sterile distilled water and 15 U (1.5 μl) of enzyme (HaeIII, BamHI, or Sau3A) with 2 μl of the appropriate restriction buffer, giving a final volume of 20 μl. This mixture was incubated at 37°C for 3 h. For mixed restriction, DNA that had already been digested with the first enzyme was precipitated, washed, dried, and resuspended in 20 μl of the second buffer consisting of 16.5 μl of sterile distilled water, 15 U (1.5 μl) of the second enzyme (HindIII), and 2 μl of the appropriate restriction buffer. This mixture was incubated at 37°C for 3 h.

The digested PCR products (20 μl) were analyzed by submarine gel electrophoresis at 110 V for 45 min through a 1.8% (wt/vol) agarose gel. pBR328-BglI + pBR328-HinfI (DNA molecular weight marker VI; Boehringer Mannheim) was used as a size marker in all gels. Calculation of the sizes of the DNA bands produced by DNA fingerprinting was performed with Taxotron software (P. A. D. Grimont, Institut Pasteur, Paris, France).

Digested PCR products yielding identical numbers of bands of the same size were considered identical and were grouped in the same PCR-RFLP type designated by a numeral. Four numerals were assigned to each isolate on the basis of its four PCR-RFLP types. A similar analytical procedure was applied to the RAPD analysis profiles and to the ribotyping patterns (ribopatterns).

Ribotyping.

DNA samples (∼2 μg) were digested at 37°C for 4 h with 30 U of the HaeIII restriction enzyme (Eurogentec), which in previous studies (5, 21, 50, 53) gave the most clearly resolved patterns for analysis. DNA fragments were separated by electrophoresis through 0.8% agarose gels (length, 13 cm) at 35 V for 16 h in TBE (89 mM Tris, 89 mM boric acid, 2 mM EDTA [pH 8.3]). After electrophoresis, the gels were stained with ethidium bromide and photographed. The DNA fragments were then transferred to nylon membranes (Hybond-N; Amersham) by capillarity and hybridized with an acetylaminofluorene (AAF)-labeled rRNA (AAF-rRNA; Eurogentec) under the conditions specified by the manufacturer. A molecular weight marker (Raoul 1; Eurogentec) was loaded onto the agarose gel and was specifically hybridized with the AAF-labeled pBR322, which is visualized as AAF rRNA.

RAPD analysis.

The PCR was carried out as described above for PCR-RFLP analysis. Three arbitrary primers were used: primers 1254 (5′-CCGCAGCCAA-3′), 1247 (5′-AAGAGCCCGT-3′), and 1281 (5′-AACGCGCAAC-3′) (3). The cycling program was 1 cycle of 94°C for 2 min, 37°C for 1 min, and 72°C for 4 min and 29 cycles of 94°C for 2 min, 37°C for 3 min, and 72°C for 7 min. After PCR, 20 μl of the PCR products was electrophoresed in 2% agarose gels containing 1× Tris acetate running buffer. pBR328-BglI + pBR328-HinfI (DNA molecular weight marker VI; Boehringer Mannheim) was used as a size marker in all gels.

Typeability.

Typeability refers to the ability of a method to provide an unambiguous result for each isolate analyzed. The typeability of each typing system was defined as the percentage of typeable isolates among the 40 unrelated isolates tested (39 unrelated isolates and 1 of the 3 familial isolates). A typeable isolate is an isolate for which the typing system can provide a readable result consisting of a pattern of several well-defined bands.

Discriminatory power.

The discriminatory power of each fingerprinting method was estimated both by the number of identified types among the 40 unrelated isolates and by the DI (28). The DI is the probability that two isolates randomly chosen from a population of unrelated isolates will be distinguished by that typing method. The DI was calculated by the following equation:

where N is the total number of isolates, s is the total number of types, and nj is the number of isolates belonging to the jth type. Thus, this index can be calculated from the distribution of types, i.e., the number of isolates of each type.

DI depends on the number of types and on the homogeneity of the frequency distribution of isolates into types; ideally, each isolate should have a different type (DI = 1). For the purpose of calculation, nontypeable isolates were grouped together. According to recent guidelines, a typing system should achieve a DI of >0.95 for reliable assessment of the clonal relatedness of isolates (63).

Meta-analysis.

To compare our results to published results for typeability and discriminatory power, we conducted a meta-analysis. The English-language medical literature from 1990 to 1998 was searched, using PubMed Medline, for articles about H. pylori DNA typing, H. pylori fingerprinting, RAPD analysis, PCR-RFLP analysis, PFGE, REP-PCR, and ribotyping. Related articles proposed by PubMed were also searched. Only studies that supplied primary data and that used typing methods similar enough to be statistically grouped by typing methods were included in the meta-analysis.

For each typing method typeability was calculated as the percentage of typeable isolates among all the unrelated isolates studied in the literature. Familial isolates and isolates obtained consecutively from the same patient were excluded. Reference strains tested in numerous studies were considered only one time for the meta-analysis. A pooled typeability was calculated for each typing method. The pooled typeability consists of the percentage of typeable isolates among the unrelated isolates of all the reported studies that used the same technique.

The DI was retrospectively calculated for the unrelated isolates of each study when primary data were presented and when more than 10 unrelated isolates were tested. For each typing method, a pooled DI was calculated as the mean DI for all reported studies by using the same method, but the calculation was weighted by the number of isolates tested in each study.

RESULTS

Raw data for the 42 isolates tested are presented in Table 1. The discriminatory powers and typeabilities of the eight techniques tested in this study are presented in Table 2.

TABLE 1.

DNA patterns for the 42 isolates^a

Isolate source and no.	Ribopattern with HaeIII	PCR-RFLP pattern by digestion of:				RAPD pattern
		ureAB with HaeIII	ureAB with BamHI + HindIII	ureC with HaeIII + HindIII	ureC with Sau3A	Primer 1254	Primer 1247	Primer 1281
Venezuela
V3	1	14	3	1	26			1
V5	NCb	9	4	2	18			2
V17	2	18	4	3	11			3
V25	NC	1	4	4	24			4
United States
84-182	3	1	1	6	3			5
85-456	4	6	3	5	1			6
86-338	5	4	3	6	4			7
Paris
85-B	6	9	3	7	12			8
86-G	NC	14	3	3	21	1	1	9
86-I	7	6	2	7	16			10
88-D	8	10	6	5	1			11
11-D	9	4	4	5	1			12
Family from Poitiers
93-28 (mother)	10	3	2	5	1	2	2	13
92-1 (son)	10	3	2	5	1	2	2	13
95-93 (daughter)	10	3	2	5	1	2	2	13
Unrelated isolates from Poitiers
92-3	11	14	4	4	25	1	1	14
92-4	12	15	3	5	1			15
92-7	13	8	3	2	20			16
92-8	14	3	4	3	19	3	3	17
92-9	15	19	3	6	3			18
92-10	NC	7	3	2	20			19
92-11	16	10	3	8	8			20
92-16	17	10	4	4	25	3	3	21
93-25	18	17	3	2	20			22
93-31	19	10	2	5	1			23
93-32	20	14	7	9	13			24
93-33	NC	2	3	9	14			25
93-34	21	1	3	6	15			26
93-35	22	5	4	6	15			27
93-37	NC	16	3	3	21			28
93-38	NC	11	5	7	22			29
93-40	23	4	5	4	5			30
93-41	NC	9	2	10	6			31
93-42	NC	6	3	5	2			32
94-47	24	1	3	3	23			33
94-48	25	13	6	4	7			34
94-55	26	5	3	2	17			35
94-56	27	2	2	4	27			36
94-60	28	13	4	3	23			37
94-62	NC	20	4	11	9			38
94-63	29	11	3	6	10			39
94-65	30	12	2	6	10			40

^aFor each technique a particular pattern is designated by a numeral. Only patterns containing several isolates are designated by a numeral; patterns containing a unique isolate are not mentioned and the cell is left blank. 

^bNC, not cleavable. 

TABLE 2.

Performances of the three typing methods tested with 40 unrelated isolates

Typing method	Typeability (%)	Discriminatory power
		No. of types	DIa
Ribotyping with HaeIII	75	30	0.94
PCR-RFLP analysis by digestion of:
ureAB with HaeIII	100	20	0.96
ureAB with BamHI + HindIII	100	7	0.71
ureC with HaeIII + HindIII	100	11	0.89
ureC with Sau3A	100	27	0.97
RAPD analysis with:
Primer 1254	100	38	0.99
Primer 1247	100	38	0.99
Primer 1281	100	40	1

^aDI according to Hunter and Gaston (28). 

Typeability.

Among the 40 unrelated isolates, 10 were not typeable by ribotyping because DNA samples were not cleaved by HaeIII (typeability = 30 of 40 = 75%). The PCR-RFLP analysis protocols successfully amplified 2.4- and 1.1-kbp fragments of the ureAB genes and the ureC gene, respectively, from all the 40 unrelated H. pylori isolates examined. Each of these two PCR products were submitted to two restriction protocols, generating four PCR-RFLP patterns for each of the 40 isolates. All 40 unrelated isolates could be typed by these four methods (typeability of PCR-RFLP = 100%). The three primers used for RAPD analysis generated DNA fingerprints for all 40 unrelated H. pylori isolates examined (typeability of RAPD analysis = 100%).

Discriminatory power.

The eight patterns obtained for each isolate by the eight tested techniques are given in Table 1 (for each technique, a particular pattern is designated by a numeral).

For ribotyping, all 30 typeable isolates were easily distinguished within 30 ribopatterns. By considering the 40 unrelated isolates tested and gathering the 10 noncleavable isolates into one particular group, the DI was 0.94. The ribopatterns each had 5 to 10 bands. The majority of the isolates exhibited ribopatterns with two common bands, at approximately 800 and 1,000 bp (Fig. 1).

FIG. 1.

Ribopatterns (HaeIII digests) of DNAs from 12 H. pylori isolates (isolates 11-D, 94-60, 94-48, 94-55, 94-47, 93-37, 93-38, 92-4, 86-6, 85-B, V17, V3). Different problems with typeability are shown: DNA was not cleaved by HaeIII (lanes 6, 7, and 9), insufficient intensities of DNA bands (lanes 3 and 11), degradation of chromosomal DNA (lane 10), and incomplete digestion of chromosomal DNA (lane 12). Lanes M, molecular size marker Raoul 1 (bands of 48.5, 18.5, 14.9, 10.6, 9.0, 7.4, 5.6, 4.4, 4.0, 3.6, 2.9, 2.3, 1.8, 1.4, 1.2, 1, 0.9, 0.7, and 0.6 kbp, from top to bottom, respectively).

For PCR-RFLP analysis, four techniques were tested. PCR with the two primers specific for ureAB amplified a 2.4-kbp fragment from within the urease A and B genes of the 40 unrelated isolates. Restriction enzyme analysis of these PCR products with HaeIII yielded two to three bands, depending on the isolate, with sizes of between 250 bp and 2 kbp. A 650-bp fragment was most commonly found. Twenty different profiles were identified among the 40 unrelated isolates, with a DI of 0.96. Mixed restriction analysis with BamHI and HindIII of the 2.4-kpb fragment amplified from within the ureAB genes yielded two to four bands ranging from 200 to 2,000 bp. A 500-bp fragment was found to be present in the majority of the patterns. This mixed restriction allowed the 40 unrelated isolates to be clustered into seven groups on the basis of the presence of one, two, or three sites and their relative positions. The DI was 0.71. PCR with the two ureC-specific primers amplified a 1.1-kbp fragment from within the ureC genes of the 40 unrelated isolates. Mixed restriction analysis with HaeIII and HindIII of the 1.1-kpb fragment yielded two to eight bands ranging from 8 to 1,000 bp. This mixed restriction allowed the 40 unrelated isolates to be clustered into 11 groups on the basis of the presence of one to seven sites and their relative positions. The DI was 0.89. Restriction enzyme analysis of these PCR products with Sau3A yielded three to six bands, with sizes of between 8 and 600 bp (Fig. 2), depending on the isolate. This RFLP analysis with Sau3A yielded 27 fragment profiles among the 40 unrelated isolates, giving a DI of 0.97. The DIs of all PCR-RFLP analyses combined are given in Table 3. The highest DI (0.99) was obtained by combining the results of two single restriction techniques: digestion of ureC with Sau3A and digestion of ureAB with HaeIII.

FIG. 2.

PCR-RFLP patterns of the amplified ureC gene products from 12 H. pylori isolates (isolates 93-32, 93-33, 93-34, 93-35, 93-37, 93-38, 93-40, 93-41, 93-42, 94-47, 94-48, and 94-55) whose amplified DNAs were digested with Sau3A. Two unrelated isolates (isolates 93-34 and 93-35 [lanes 3 and 4, respectively]) yielded identical pattern. Lanes M, molecular size marker pBR328-BglI + pBR328-HinfI (bands of 2,176, 1,766, 1,230, 1,033, 653, 517, 453, 394, 298, 234, 220, and 154 bp, from top to bottom, respectively).

TABLE 3.

DIs of combinations of PCR-RFLP typing methods

Typing method	DI
	ureAB digested with HaeIII	ureAB digested with BamHI + HindIII	ureC digested with HaeIII + HindIII	ureC digested with Sau3A
ureAB digested with HaeIII	0.96	0.99	0.99	0.99
ureAB digested with BamHI + HindIII		0.71	0.97	0.99
ureC digested with HaeIII + HindIII			0.89	0.97
ureC digested with Sau3A				0.97

For RAPD analysis, the profiles generated with the three primers contained several bands (three to nine bands for primer 1254 and two to five bands for primers 1247 and 1281) of various intensities. The bands ranged from 400 to 2,200 bp in size (Fig. 3). We did not observe any bands in lanes in which amplified blanks were run. With primer 1254 or primer 1247, the 40 unrelated isolates were gathered in 38 banding patterns (Table 1). Primer 1281 yielded 40 distinct RAPD patterns among the 40 unrelated isolates examined. The DI was 0.99 for primers 1254 and 1247, and the DI was 1 for primer 1281.

FIG. 3.

RAPD patterns of the amplified DNAs of 12 H. pylori isolates (isolates 85-B, 86-G, 92-3, 92-4, 92-7, 92-8, 92-9, 92-10, 92-11, 92-16, 93-25, and 93-31) obtained by RAPD analysis with primer 1254. Two pairs of unrelated isolates (lanes 2 and 3 and lanes 6 to 10, respectively) yielded identical patterns. Lanes M, molecular size marker pBR328-BglI + pBR328-HinfI (bands of 2,176, 1,766, 1230, 1033, 653, 517, 453, 394, 298, 234, 220, and 154 bp, from top to bottom, respectively).

Meta-analysis.

The typeabilities and DIs of the different techniques presented in the literature and determined in our study are presented in Tables 4 to 6.

TABLE 4.

Typeabilities and DIs of chromosomal restriction-based techniques (ribotyping and PFGE) determined with data from selected studies

Technique	Authors (reference no.)	No. of unrelated patients	Typeabilitya	No. of typesb	DI
HaeIII ribotyping	Owen et al. (47)	28	25/28	23	0.99
	Owen et al. (49)	93	77/93	78	0.98
	Owen et al. (51)	13	8/13	9	0.87
	Owen et al. (50)	17	13/17	14	0.95
	Prewett (53)	15	11/15	12	0.94
	Georgopoulos et al. (24)	18	13/18	14	0.93
	Salaün et al. (58)	39	28/39	29	0.92
	Burucoa et al. (this study)	40	30/40	31	0.94
	Pooled results		205/263 (78)		0.94
HindIII ribotyping	Owen et al. (48)	46	46/46	18	0.90
	Tee et al. (68)	107	107/107	77	0.99
	Marshall et al. (40)	29	29/29	15	0.94
	Rautelin et al. (54)	11	11/11	8	0.94
	Salaün et al. (58)	39	39/39	25	0.94
	Pooled results		232/232 (100)		0.92
NotI PFGE	Taylor et al. (66)	30	25/30	25	0.98
	Takami et al. (64)	24	12/24	12	0.76
	Takami et al. (65)	24	15/24	15	0.87
	Wada et al. (73)	16	9/16	9	0.82
	Hirschl et al. (27)	18	12/18	12	0.90
	Salama et al. (57)	20	15/20	15	0.95
	Pooled results		88/132 (66)		0.88
NruI PFGE	Taylor et al. (66)	30	19/30	19	0.87
	Hirschl et al. (27)	18	3/18	3	0.31
	Salama et al. (57)	20	13/20	13	0.89
	Pooled results		35/68 (51)		0.73
ApaI PFGE	Wada et al. (73)	16	9/16 (56)	9	0.82
KpnI PFGE	Wada et al. (73)	16	2/16 (12.5)	2	0.24

^aData indicate number of typeable isolates/total number of isolates tested (percent). 

^bNontypeable isolates were grouped together in one type. 

TABLE 6.

Typeabilities and DIs of RAPD analysis with primer 1254 or primer 1281 and REP-PCR with data from selected studies

Technique	Authors (reference no.)	No. of unrelated patients	Typeabilitya	No. of typesb	DI
RAPD analysis with primer 1254	Akopyanz et al. (3)	64	64/64	64	1
	Dzierzanowska et al. (16)	40	40/40	40	1
	Schütze et al. (59)	18	18/18	18	1
	Weel et al. (75)	20	20/20	20	1
	Burucoa et al. (this study)	40	40/40	38	0.99
	Pooled results		182/182 (100)		1
RAPD analysis with primer 1281	Akopyanz et al. (3)	64	64/64	64	1
	Dzierzanowska et al. (16)	40	40/40	40	1
	Jorgensen et al. (32)	17	17/17	17	1
	Wada et al. (73)	16	16/16	16	1
	Weel et al. (75)	20	20/20	20	1
	Burucoa (this study)	40	40/40	40	1
	Pooled results		197/197 (100)		1
REP-PCR	Go et al. (25)	70	70/70	68	0.99
	Dore et al. (13)	12	12/12	11	0.98
	Van Doorn et al. (72)	32	32/32	31	0.99
	Pooled results		114/114 (100)		0.99

^aSee footnote a of Table 4. 

^bSee footnote b of Table 4. 

Table 4 shows the low typeabilities of chromosomal restriction-based techniques: HaeIII ribotyping (78%), NotI PFGE (66%), NruI PFGE (51%), ApaI PFGE (56%), and KpnI PFGE (12.5%). HindII ribotyping has good typeability (100%). The DIs of all ribotyping and PFGE techniques are less than the recommended value of 0.95, indicating a low discriminatory power.

Table 5 shows the excellent typeabilities of PCR-RFLP methods (100%). The higher DIs were obtained by amlplification of ureC (1.1 kb) digested with Sau3A (DI = 0.97) and by amplificaiton of ureAB digested with HaeIII (DI = 0.96).

TABLE 5.

Typeabilities and DIs of PCR-RFLP analysis of ureAB (2.4 kb) digested with HaeIII, ureC (820 or 1,100 bp) digested with Sau3A, or ureC (820 bp or 1.1 kbp) digested with HhaI determined with data from selected studies

Gene and enzyme	Authors (reference no.)	No. of unrelated patients	Typeabilitya	No. of typesb	DI
ureAB (2.4 kb) digested with HaeIII	Akopyanz et al. (4)	60	60/60	27	UCc
	Foxall et al. (20)	21	21/21	10	0.87
	Hurtado and Owen (29)	44	44/44	30	0.98
	Romero Lopez et al. (56)	13	13/13	12	0.99
	Owen et al. (52)	12	12/12	10	0.97
	Salaün et al. (58)	39	39/39	26	0.98
	Wang et al. (74)	13	13/13	10	0.97
	Burucoa et al. (this study)	40	40/40	20	0.96
	Pooled results		242/242 (100)		0.96
ureC (820 bp) digested with Sau3A	Stone et al. (61)	72	72/72	42	UC
		22d		10	0.86
	Shortridge et al. (60)	81	81/81	44	UC
		18d		8	0.89
	Stone et al. (62)	83	83/83	12	0.86
	Pooled results		236/236 (100)		0.86
ureC (1.1 kb) digested with Sau3A	Burucoa (this study)	40	40/40	27	0.97
ureC (820 bp) digested with HhaI	Fujimoto et al. (22)	25	25/25	10	0.86
	Stone et al. (61)	72	72/72	42	UC
		22d		8	0.70
	Shortridge et al. (60)	81	81/81	44	UC
		18		5	0.53
	Stone et al. (62)	83	83/83	12	0.71
	Pooled results		261/261 (100)		0.71
ureC (1.1 kb) digested with HhaI	Li et al. (36)	19	19/19	11	0.94

^aSee footnote a of Table 4. 

^bSee footnote b to Table 4. 

^cUC, uncalculable. 

^dFor these data from two studies (60, 61), limited raw data allowed us to calculate the typeability with data for all isolates and the DI with data for a limited number of isolates. 

Table 6 shows the excellent typeabilities and DIs obtained by RAPD analysis (typeability = 100%; DI = 1) and REP-PCR (typeability = 100%; DI = 0.99).

DISCUSSION

Typeability.

The excellent typeabilities of both PCR-RFLP and RAPD analyses as assessed from the results of our study are confirmed by the meta-analysis in which the pooled typeability is 100% for the PCR-based techniques (Tables 5 and 6). The low typeability of ribotyping with HaeIII observed in our study (75%) is confirmed by the meta-analysis (78%). Despite this low typeability, HaeIII is the most commonly used enzyme since it gave the best-resolved patterns for analysis and the best discriminatory power (21). In the literature, the typeability of ribotyping is variable according to the restriction enzyme used (5, 47, 48, 50, 51, 53, 58). The meta-analysis of published studies in which ribotyping was used yielded pooled typeabilities of 78% when HaeIII was used and 100% when HindIII was used (Table 4). In studies in which two enzymes were tested, some isolates, from which DNA was not digested with HaeIII could be well digested with HindIII (5, 48, 58). A good typeability could then be obtained, but two or three consecutive assays were required. Moreover, as shown in Fig. 1, several other problems can occur with this technique: insufficient DNA yield (lanes 3 and 11), DNA degradation (lane 10), and noncomplete digestion of chromosomal DNA (lane 12). Because of these problems several assays were required to achieve a readable result. PFGE analysis of DNA from H. pylori isolates is based on the same principle as ribotyping: the generation of restriction fragments from the chromosomal DNA (23). The meta-analysis of published studies in which PFGE was used yielded low typeabilities, ranging from 12.5 to 66% (Table 4). The low typeabilities for both these two restriction techniques (ribotyping and PFGE) could be due to protection from restriction endonuclease digestion by the production of endogenous methylases (66).

Discriminatory power.

The highest DIs were obtained for RAPD analysis: a DI of 1 with primer 1281, which yielded 40 distinguishable patterns among the 40 unrelated isolates, and DIs of 0.99 with primers 1254 and 1247, which yielded 38 distinguishable patterns. The meta-analysis of published studies in which RAPD analysis with primers 1254 and 1281 was used confirmed our results, yielding pooled DIs of 1 for four and five studies, respectively (Table 6). Interestingly, two primers (primers 1247 and 1254) yielded different but dependent patterns in our study. Each primer gave 38 patterns for the 40 unrelated isolates. The banding pattern obtained with one of the primers was clearly different from the one obtained with the other primer, although the two pairs of indistinguishable isolates were the same with these two primers (Table 1, RAPD patterns 1 and 3). These two pairs of isolates (isolates 86-G and 92-3 and isolates 92-8 and 92-16) were easily distinguishable by the other typing systems (ribotyping, PCR-RFLP analysis, and RAPD analysis with primer 1281). This observation highlights the possibility of achieving a false identity by RAPD analysis, even if two primers are used.

With another primer, Cho et al. (7) distinguished only 11 RAPD types among 108 unrelated isolates. The calculated DI was low: 0.83. This discrepancy in the discriminatory powers of RAPD typing systems highlights the difficulty in appreciating the level of discrimination when this technique is applied for the first time in a laboratory. Since the interlaboratory reproducibility of RAPD analysis is very low (10, 41, 69), it is impossible to predict the discriminatory power of this typing method even if the technical conditions tend to reproduce the published conditions. This problem of reproducibility is of great importance since a single variation in one of the multiple technical conditions of RAPD analysis could change the discriminatory power of this typing method.

In our study the DIs of PCR-RFLP methods ranged from 0.71 to 0.97 (Table 2). The higher DIs were obtained with the Sau3A restriction digest of the PCR-amplified ureC genes (DI = 0.97) and the HaeIII restriction digest of the PCR-amplified ureAB genes (DI = 0.96), and the DIs were greater than the recommended value of 0.95 (28, 63).

The meta-analysis allowed us to compare the DIs of three PCR-RFLP techniques. For PCR-RFLP analysis of ureAB (2.4 kb) digested with HaeIII, the DIs of six published studies ranged from 0.87 to 0.99, with a pooled DI of 0.96 (Table 5). For PCR-RFLP analysis of ureC digested with Sau3A, the experimental results were discordant from the published results. Results form the three published studies could be analyzed, yielding a pooled DI of 0.86. The experimental results yielded a DI of 0.97. This discrepancy may be explained by differences in the fragments amplified within the ureC gene: we amplified a 1,100-bp fragment of ureC, while a 820-bp fragment was amplified in the three previously described studies whose data were analyzed. A similar discrepancy was observed for PCR-RFLP analysis of ureC digested with HhaI for four published studies, in which 820-bp fragments of ureC were amplified (22, 60–62), and the study of Li et al. (36), in which a fragment of 1.1 kb was amplified (Table 5). The amplification of this longer fragment of the ureC gene allowed a better discrimination of isolates when both Sau3A and HhaI restrictions were used.

The double restriction techniques were less discriminatory: 0.71 for BamHI-HindII restriction digests of the PCR-amplified ureAB genes and 0.89 for HaeIII-HindIII restriction digests of the PCR-amplified ureC gene (Table 2). As observed for RAPD analysis, correlations between PCR-RFLP types were observed. In fact, a triplet of isolates and three pairs of isolates were indistinguishable by three of the four PCR-RFLP techniques used (Table 1). An epidemiological link between these isolates was not likely, since two of these indistinguishable isolates came from two different geographic areas and also because all isolates of the triplet and the three pairs were sampled over a 2-year interval. Moreover, these isolates were clearly distinguished with the other typing systems. A genetic link within the urease operon could explain these correlations between the restriction patterns of fragments amplified into functionally linked genes.

Some studies that used the PCR-RFLP method with other genes or with other enzymes could not be compared with others but are of interest. Moore et al. (42) used three PCR-RFLP typing systems based on restriction digestion of the PCR-amplified ureC (1.1 kb) gene. With 21 unrelated isolates, a DI of 0.71 could be calculated for HindIII restriction, a DI of 0.84 could be calculated for AluI restriction, and a DI of 0.38 could be calculated for PvuII restriction. Li et al. (36) digested a PCR-amplified ureC fragment (1.1 kb) and obtained a DI of 0.91 for MboI and a DI of 0.77 for AluI. Fujimoto et al. (22) applied three PCR-RFLP techniques to 25 unrelated isolates. PCR-RFLP analyses with the ureC-amplified portion (820 bp) digested with HhaI, MboI, or MseI yielded DIs of 0.87, 0.92, and 0.89, respectively. Forbes et al. (19) amplified a portion of flaA from 49 unrelated isolates. By using four restriction enzymes, the DIs obtained were 0.47 for MspI, 0.29 for HindIII, 0.88 for MboI, and 0.80 for AluI. Salaün et al. (58) amplified flaA and restricted it with MboI and HaeIII, yielding DIs of 0.87 and 0.19, respectively. Restriction analysis of urease genes seems to be more discriminatory than restriction analysis of flagellin genes.

Results from only three previously described studies that used REP-PCR could be analyzed, yielding a pooled DI of 0.99. This excellent discriminatory power must be confirmed by more studies with this typing method.

The low discriminatory power of HaeIII ribotyping observed in our study (DI = 0.94) may be explained by the low typeability of this typing system. The 10 untypeable isolates, grouped together into one type, decreased the DI to just under the recommended value of 0.95 (28, 63). The meta-analysis of data from previously published studies confirmed the limited value of the DI of HaeIII ribotyping (Table 4). The pooled DI was 0.94, which was identical to our result and which was less than the recommended value of 0.95. As described by Fraser et al. (21), the meta-analysis confirmed the low discriminatory power of HindIII ribotyping (Table 4). Despite an excellent typeability, the pooled DI was 0.92. In the same way, the low discriminatory power of PFGE calculated in the meta-analysis is explained by the very low typeability of this technique (Table 4). The higher DIs were obtained with the NotI enzyme, but with a pooled DI of 0.88.

This study focused on two performance criteria: typeability and discriminatory power. Performance criteria also include reproducibility and stability (63). Reproducibility refers to the ability of a technique to yield the same result when the same isolate is tested repeatedly (63). Stability refers to the ability of a typing system to recognize the clonal relatedness of strains derived in vitro or in vivo from a common ancestor strain (63). The reproducibilities and stabilities of H. pylori typing systems have been already evaluated in many studies (3, 4, 19–22, 33, 56, 66–69). RAPD analysis, for which reproducibility is critical if technical conditions are not standardized, was especially evaluated for its reproducibility (3, 10, 33, 41, 69).

In this study, evaluation of the reproducibilities and stabilities of the methods were conducted, with the only goal being to confirm acceptable performance. For this purpose, we tested three clonally related familial isolates (isolates 93-28, 92-1, and 95-93) and serially subcultivated strains of isolate 92-1. Our study confirmed the good reproducibilities and stabilities of the three typing techniques (ribotyping, PCR-RFLP analysis, and RAPD analysis). Since reproducibility was tested with only a few isolates and since technical conditions were not exhaustively tested, our evaluation is only qualitative. Our opinion is that the PCR-RFLP analysis is the more easily reproducible technique tested. For ribotyping, the main difficulty in terms of reproducibility was obtaining an identical amount of well-digested total DNA. Different amounts of digested DNA yielded differences in the intensities of the bands constituting ribopatterns. The reproducibility of RAPD analysis has been investigated by several authors (10, 32, 33, 41, 69) and should not represent a problem if the technique is standardized, including the use of standard methods of DNA preparation, the use of consistent volumes and concentrations of reagents, consistent use of the same thermostable DNA polymerase, use of the same thermal cycler, and use of a standard procedure for visualization of the fingerprint. We have observed that when these factors known to contribute to variability were rigidly controlled, the reproducibilities of the RAPD patterns were excellent over the limited period of the study.

Since the three familial isolates, collected over 3 years from three persons who had not lived together for 10 years, yielded the same profiles by all the techniques tested, the stabilities of these markers are excellent. This absence of independent evolution of isolates with a common origin over several years was already detected in studies with familial isolates (5, 7, 68).

Summary.

Our experimental results and the meta-analysis showed that PCR-based typing methods (PCR-RFLP analysis, RAPD analysis, and REP-PCR) have optimal typeabilities. Two PCR-RFLP techniques (PCR-RFLP analyses of ureAB digested with HaeIII and ureC digested with Sau3A) yield excellent discriminatory powers. The low interlaboratory reproducibility of typing by RAPD analysis does not allow an a priori evaluation of its discriminatory power. For this rapid and convenient method, determination of the concordance of the results with those obtained with another typing system is necessary. The excellent discriminatory power of REP-PCR should be confirmed. Chromosome digestion methods (ribotyping and PFGE) are limited by their very low typeabilities, which strongly decrease their discriminatory powers, and therefore, they are not recommended for use in the typing of H. pylori isolates. We recommend the use of the PCR-RFLP method.