ABSTRACT
We examined the sensitivity and specificity of 11 PCR assays described for the species identification of Campylobacter jejuni and Campylobacter coli by using 111 type, reference, and field strains of C. jejuni, C. coli, and Campylobacter lari. For six assays, an additional 21 type strains representing related Campylobacter, Arcobacter, and Helicobacter species were also included. PCR tests were initially established in the laboratory by optimizing conditions with respect to five type and reference strains of C. jejuni, C. coli, and C. lari. One PCR test for C. coli failed to give appropriate results during this initial setup phase and was not evaluated further. The remaining 10 assays were used to examine heated lysate and purified DNA templates as appropriate of well-characterized type, reference, and field strains of C. jejuni (n = 62), C. coli (n = 34), and C. lari (n = 15). The tests varied considerably in their sensitivity and specificity for their respective target species. No assay was found to be 100% sensitive and/or specific for all C. jejuni strains tested, but four assays for C. coli gave appropriate responses for all strains examined. Between one and six strains of C. jejuni gave amplicons in four of seven C. jejuni PCR tests only where purified DNA was used as the template; corresponding results were seen with one strain of C. coli in each of three assays for the latter species. Our findings indicate that a polyphasic strategy for PCR-based identification should be used to identify C. jejuni and C. coli strains. The data may assist laboratories in selecting assays suited for their needs and in designing evaluations of future PCR tests aimed to identify these species.
Campylobacter spp. are gram-negative, microaerophilic and/or anaerobic, mainly spiral-shaped bacteria, most of which are established or suspected human gastrointestinal pathogens (28). Of the 16 species and six subspecies currently known (21), Campylobacter jejuni and Campylobacter coli are those most often isolated from human diarrhea (28). Most developed countries report C. jejuni as predominant, but in other areas, C. coli accounts for up to 50% of human cases (28). The main source of campylobacter infections is considered contaminated foods, since the bacteria are normal flora in animals such as poultry, pigs, and cattle.
The accurate identification of C. jejuni and C. coli provides important data for surveillance and risk assessment studies on which intervention strategies can be based. The principal hosts of C. jejuni and C. coli are widely regarded as poultry and pigs, respectively, but significant proportions of C. coli have been found in poultry (5), and genetically identical clones of C. jejuni have been observed in humans, cattle, and various animals living in the wild, as well as poultry (25, 26). In addition, rapid identification of C. coli may be useful clinically, since up to 68.4% of strains are resistant to erythromycin, the antibiotic of first choice for treatment of severe campylobacter infections (18).
Identification of campylobacters and related bacteria is well known to be problematic, principally because of their complex taxonomy, biochemical inertness, and fastidious growth requirements (20, 21). C. jejuni and C. coli are traditionally differentiated by the hippurate hydrolysis test, for which only C. jejuni gives a positive reaction. However, hippurate-negative strains of this species are well recognized (17, 36), and problems with false positive test results for non-C. jejuni species have also been described (5). Additional phenotypic characters, such as growth on a minimal medium and alpha-hemolytic activity, are useful but do not provide unequivocal discrimination, require stringent standardization, and are seldom used in routine laboratories (20). As a consequence, there has been considerable interest in the development of molecular identification methods for C. jejuni and C. coli.
PCR tests are considered especially attractive due to their relative ease of use, low cost, and potential application in large-scale screening programs by means of automated technologies (12). A wide range of PCR assays for C. jejuni and C. coli have been described, several of which are based on a variety of genes (e.g., 23S rRNA, ceuE, and mapA) (8, 9, 34) and others of which are derived from different randomly generated fragments (4, 32-33, 37) (see Table 1). The sensitivity and specificity of each PCR test have been examined, but the bases of the evaluations differed significantly, particularly with respect to the selection of strains of C. jejuni, C. coli, and Campylobacter lari, a group of species that are closely related by phylogenetic and genetic criteria (21). For example, one study included the reference strains used to establish an international serotyping scheme for C. jejuni and C. coli (13), another tested six C. jejuni strains and one C. coli strain but no C. lari isolates (4), and one was evaluated solely in the context of a specific taxonomic investigation (38). In addition, only one (13) appears to have been applied to hippurate-negative C. jejuni strains. The differences in the numbers and choices of strains used to evaluate the PCR tests make an objective comparison of their efficacy difficult.
PCR tests and assay variables
In this study, we investigated the sensitivity and specificity of 11 PCR assays for identifying C. jejuni and C. coli, mainly by use of 111 well-characterized strains of the most closely related species C. jejuni, C. coli, and C. lari.
(Preliminary results from this study were presented in the abstracts of the 11th International Workshop for Campylobacter, Helicobacter and Related Organisms, 2 to 4 Sept. 2001, Freiburg, Germany [S. L. W. On, Int. J. Med. Microbiol. 291{Suppl. 31}:106, abstr. L-10].)
MATERIALS AND METHODS
Study design.Six PCR tests for C. jejuni (4, 8, 9, 13, 32, 34), four for C. coli (8, 9, 13, 33), and a multiplex assay designed for concurrent identification and differentiation of both species (38) were examined. For initial PCR setup, all primers and PCR cycling conditions as specified by the respective authors were first examined for their ability to obtain appropriate results with purified DNA template from C. jejuni CCUG 11284T and CCUG 10958, C. coli CCUG 11283T and CCUG 24865, and C. lari CCUG 23947T. The specified reaction conditions were modified by altering key variables (primer annealing temperature, concentration of MgCl2, concentration of Taq polymerase) only where appropriate results were not obtained. Details of the PCR assay variables finally used are given in Table 1. Optimized PCR assays were then applied to heated lysates of 111 strains of C. jejuni, C. coli, and C. lari (see Tables 2 and 3). Strains giving inappropriate results were reexamined by use of the original heated lysate and also purified DNA. For selected C. jejuni (13, 38) and C. coli (9, 13, 33, 38) assays, specificity was reexamined by use of purified DNA extracts from the type strains of related taxa; Campylobacter concisus, Campylobacter curvus, Campylobacter fetus subsp. fetus, Campylobacter fetus subsp. venerealis, Campylobacter gracilis, Campylobacter helveticus, Campylobacter hominis, Campylobacter hyointestinalis subsp. hyointestinalis, Campylobacter hyointestinalis subsp. lawsonii, Campylobacter lanienae, Campylobacter mucosalis, Campylobacter rectus, Campylobacter showae, Campylobacter sputorum, Campylobacter upsaliensis, Arcobacter butzleri, Arcobacter cryaerophilus, Arcobacter nitrofigilis, Arcobacter skirrowii, Helicobacter pametensis, and Helicobacter pullorum.
C. jejuni, C. coli, and C. lari strains giving appropriate (species-specific) results in all 10 PCR tests subjected to full evaluation
Strains for which PCR tests failed to give appropriate results and summary of the sensitivity and specificity of each PCR
Bacterial strains.Tables 2 and 3 list the 57 C. jejuni subsp. jejuni, 5 C. jejuni subsp. doylei, 34 C. coli, and 15 C. lari strains used. Of these, 81 are reference strains from international culture collections. The identities of the remaining strains (including hippurate-negative or variable C. jejuni subsp. jejuni) were verified in previous studies, mainly involving DNA-DNA hybridization and/or amplified fragment length polymorphism (AFLP) technologies (17, 22-23, 31, 36-38).
Preparation of heated lysate and purified DNA templates.Diluted (1:10 in purified distilled water) heated lysates (24) and purified DNA samples (11) from the bacterial strains used here were prepared as described previously. Where purified DNA was employed as the template for the PCR assays described by Gonzalez et al. (9), 100 ng was used per the authors' instructions.
PCR assay conditions.All PCRs were performed in 50-μl volumes containing 5 μl of 10× PCR buffer (N808-0171; Applied Biosystems, Foster City, Calif.), 0.5 μl of each PCR primer (final concentration, 130 μg/ml; DNA Technology A/S, Aarhus, Denmark), 0.5 μl of deoxynucleoside triphosphate mix (final concentration, 1 mM; Amersham Pharmacia Biotech, Hørsholm, Denmark), and 2.0 μl of template. Volumes of DNA polymerase (AmpliTaq; Applied Biosystems), MgCl2 (Applied Biosystems), and DNA-free purified water were used as appropriate to each assay (Table 1). Reaction mixtures were then overlaid with 1 μl of sterile mineral oil and heated for 5 min at 94°C as an initial denaturation step. PCR was then performed using the cycling conditions specified by the original authors of the respective works, with amendments to the annealing temperature as specified in Table 1. All assays were terminated with a 5-min extension period of 72°C and were performed with Trio Biometra 20 thermocyclers (Biometra, Göttingen, Germany), the efficacy of which was regularly tested by in-house quality assurance procedures (10). Amplicons were detected by ethidium bromide staining of electrophoresed samples as described previously (24).
RESULTS
Initial PCR setup.Appropriate results in terms of species specificity and amplicon size for 10 of the 11 PCR assays examined for the five type and reference strains used were successfully obtained. Most assays required some alteration to one or more of the specified reaction variables to obtain suitable results (Table 1). The 23S rRNA-based PCR test for C. coli (8) consistently gave amplicons with the strains of C. jejuni and C. lari during the initial setup phase, despite extensive testing (Fig. 1). Table 1 shows the range of different variable parameters used in various combinations in attempts to establish the assay. Consequently, this PCR was not evaluated further. No amplicon was obtained with any of the 21 related Campylobacter, Arcobacter, or Helicobacter taxa in five assays reexamined for species specificity (9, 13, 33, 38).
Selected results from PCR tests. (a) Multiplex PCR for concurrent identification of C. jejuni and C. coli (38); (b to e) tests for C. jejuni (references 13, 8, 4, and 9, respectively); (f and g) tests for C. coli (references 13 and 8, respectively). Lanes: 1, CCUG 11284; 2, CCUG 10966; 3, CCUG 10970; 4, CCUG 10967; 5, CDC D133; 6, CCUG 11283; 7, CCUG 17715; 8, CCUG 10959; 9, LMG 9799; 10, CDC D145; 11, CDC D603; 12, CCUG 10951; 13, CCUG 23947; M, molecular size marker VI (Boehringer-Mannheim, Mannheim, Germany). Lanes 1 to 5 and 11, C. jejuni; lanes 6 to 10 and 12, C. coli; lane 13, C. lari. Results in panels a and e were obtained with heated lysate template; all other results were obtained with purified DNA.
Sensitivity and specificity of PCR tests by evaluation with 111 C. jejuni, C. coli, and C. lari strains.Table 2 lists the strains for which accurate results in all PCR tests evaluated were obtained. Table 3 gives the results for strains for which PCR tests failed in their accuracy and summarizes the sensitivity (percentage of strains of the correct species identified) and specificity (100 − percentage of strains of the nontarget species giving a reaction) of each assay. For the latter, only consistent results obtained from analyses performed on heated lysates and subsequently on purified DNA were considered. All assays tested for identification of C. coli proved both 100% specific and sensitive, although three tests gave amplicons of strains of CCUG 11283 or CDC D145 only when purified DNA was used as the template (Table 2).
The performance of C. jejuni PCR assays varied considerably. Test specificity varied from 84 to 100%, and test sensitivity ranged from 88 to 100%. No test proved 100% accurate; tests that yielded amplicons from all C. jejuni strains proved to be the least specific (Table 3). Conversely, the two assays found to be 100% specific for C. jejuni (13, 38), respectively, detected 91 and 93% of strains of this species. Strains of C. jejuni subsp. doylei contributed significantly to the failure rate of these and two other PCR tests intended to identify C. jejuni; one test derived from an arbitrarily primed PCR product (4) failed to react with a single strain of this taxon (Table 3). In addition, there was an increased proportion of C. jejuni identification assays for which heated lysates did not serve as a suitable reaction template (Tables 2 and 3). Moreover, the use of heated lysate frequently (46% of C. jejuni strains examined) gave rise to additional PCR products of a lower molecular weight than the intended amplicon in one assay (9). These results were obtained with only four strains (7%) when purified DNA of the concentration recommended by these authors was employed as the reaction template. Results obtained with the PCR assay targeting the 23S rRNA gene (8) were noteworthy in that the molecular size of the amplicon in 14 strains of C. jejuni varied between the limits originally described (8); in two strains, two distinct products were observed. Representative results are shown in Fig. 1.
DISCUSSION
We observed considerable variation in the performance of 11 previously described PCR assays for identifying C. jejuni and C. coli. Both physicochemical (i.e., relating to the PCR) and biological (i.e., relating to the diversity of Campylobacter) factors account for this variability.
The quality of the template used for PCR is crucial to the reaction (10). Purified DNA can be regarded as the “gold standard” template for bacterial identification, but for busy routine laboratories the use of simple heated lysates is clearly advantageous. In general, the PCR tests examined here performed well when heated lysates were used, but in several cases amplicons were obtained only with purified DNA. Several factors may explain our results. Some Campylobacter strains do not release PCR-detectable DNA when boiled (16); the fact that PCR tests were not affected equally may reflect differences in the copy number and/or stability of each genetic marker. In addition, some PCR primers are affected more markedly than others by impurities present in crude DNA preparations (7, 27). The fact that nonspecific PCR products were frequently seen in one assay (9) when heated lysate was used as the template but rarely when purified DNA was used may exemplify the influence of high DNA concentrations in heated lysates, which cause some PCR primers to misprime (2). Under such conditions, total inhibition of the PCR can also occur (2, 10).
Many factors apart from template quality can affect the efficacy of PCR, including source and type of DNA polymerase used, thermal cycler specification, and reaction buffer composition (1, 7, 15). Thus, stringent optimization procedures are recommended for implementation of a given PCR test (7, 10). We optimized 10 of 11 PCR tests under conditions concordant with recommended quality control procedures (10). Nonetheless, given the variables that can influence PCR test efficacy, in the hands of the original developers each PCR assay may perform satisfactorily. However, if such conditions cannot be easily reproduced, portability quickly becomes crucial.
In addition to technical factors contributing to the compromised sensitivity and specificity of certain PCR tests, a biological element is highly likely. Campylobacter is a taxonomically complex genus (21), and both C. jejuni and C. lari are genetically diverse species (6, 19, 21, 23). The finding that four of five PCR tests for identification of C. coli are effective may reflect the implication from AFLP analyses that this species appears to be comparatively homogeneous at the DNA level (6, 23). Our failure to obtain specific results for the C. coli PCR targeting the 23S rRNA gene resembles problems experienced by others (14) and may reflect the high level of conservation in the rRNA operon among closely related species (21).
C. jejuni comprises two genetically distinct but highly related subspecies, C. jejuni subsp. jejuni and C. jejuni subsp. doylei. Since the latter has no known animal reservoir and is infrequently observed in human disease, most attention is focused on C. jejuni subsp. jejuni, and most tests evaluated here were not originally used to examine C. jejuni subsp. doylei strains. The inability of many C. jejuni PCR tests to recognize strains of the latter taxon suggests that the genetic difference between the two subspecies (23, 30) contributes to PCR failure. Furthermore, genetic heterogeneity in C. jejuni subsp. jejuni can arise from various phenomena (19, 22, 35) that can affect the efficacy of a PCR if such change occurs within one or both of the binding sites (20). Mutation in hipO (29) has previously been identified as a source of failure for the PCR assay targeting that gene (13) and for a PCR based on ureC to identify the related organism Helicobacter pylori (3). Some of the negative PCR results we observed probably arose from the natural genetic diversity of C. jejuni. Such population variance is also demonstrated in our results for the C. jejuni subsp. jejuni PCR targeting the 23S rRNA gene (8). Amplicon size variation was also observed among the few strains of C. coli that gave a positive result with this assay (Fig. 1 and Table 3). The wide variation in distribution and sequence of intervening sequences in rRNA genes of C. jejuni and C. coli (37) and related bacteria, including interoperon differences, is now well established, and the effects and implications of intervening sequences for PCR tests have been discussed (21).
Our reexamination of the specificity of the five most accurate PCR tests by use of type strains of 21 related Campylobacter, Arcobacter, and Helicobacter species concurred with results obtained by the original authors (9, 13, 33, 38), emphasizing that more problems are encountered with accurate discrimination of closely related taxa. Our results endorse the use of a strain collection that adequately reflects the diversity and taxonomy of the target species to validate PCR assays. It is noteworthy that two of the PCR assays examined here that were readily established in our laboratory and yielded accurate results were first developed by testing with an extensive reference strain collection (13). Our data also support a polyphasic approach for identification of campylobacters (20), since no single PCR test identified all C. jejuni strains. Suspect C. jejuni subsp. doylei isolates should be confirmed by examining for their inability to reduce nitrate (30) or by alternative genetic methods such as AFLP fingerprinting (23). We now routinely use the multiplex PCR for concurrent identification and discrimination of C. jejuni and C. coli (38) as a first-line identification method and specific assays for these two species (13) as required. We hope the results of this study will assist others in selecting C. jejuni and C. coli PCR assays for their use.
ACKNOWLEDGMENTS
We thank Enevold Falsen (Culture Collection, University of Göteborg, Göteborg, Sweden), Peter Vandamme (University of Ghent, Ghent, Belgium), Collette Fitzgerald (University of Lancaster, Lancaster, United Kingdom), and Bala Swaminathan and Mavis Nicholson (Centers for Disease Control and Prevention, Atlanta, Ga.) for supplying bacterial strains used in this study. We also thank Jeffrey Hoorfar (Danish Veterinary Institute) and Alex van Belkum (Erasmus University Medical Center, Rotterdam, The Netherlands) for useful discussions.
FOOTNOTES
- Received 27 March 2002.
- Returned for modification 15 July 2002.
- Accepted 22 October 2002.
- Copyright © 2003 American Society for Microbiology
