• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of jcmPermissionsJournals.ASM.orgJournalJCM ArticleJournal InfoAuthorsReviewers
J Clin Microbiol. Mar 2012; 50(3): 798–809.
PMCID: PMC3295161

Comparison of Molecular Typing Methods Useful for Detecting Clusters of Campylobacter jejuni and C. coli Isolates through Routine Surveillance


Campylobacter spp. may be responsible for unreported outbreaks of food-borne disease. The detection of these outbreaks is made more difficult by the fact that appropriate methods for detecting clusters of Campylobacter have not been well defined. We have compared the characteristics of five molecular typing methods on Campylobacter jejuni and C. coli isolates obtained from human and nonhuman sources during sentinel site surveillance during a 3-year period. Comparative genomic fingerprinting (CGF) appears to be one of the optimal methods for the detection of clusters of cases, and it could be supplemented by the sequencing of the flaA gene short variable region (flaA SVR sequence typing), with or without subsequent multilocus sequence typing (MLST). Different methods may be optimal for uncovering different aspects of source attribution. Finally, the use of several different molecular typing or analysis methods for comparing individuals within a population reveals much more about that population than a single method. Similarly, comparing several different typing methods reveals a great deal about differences in how the methods group individuals within the population.


Despite being the most frequently detected bacterial agent causing gastroenteritis worldwide (4, 9, 55), Campylobacter spp. are infrequently associated with outbreaks of disease (1, 19, 41). Point-source outbreaks have been associated with exposure to contaminated water and milk, as well as contaminated chicken and other foods (13, 19, 42, 55). There is some evidence that outbreaks are more common than is currently thought (20, 21, 36). Eleven European countries reported 154 Campylobacter outbreaks in a 5-year period (54).

The improved detection of clusters of human disease associated with Campylobacter spp. is needed. Previous investigations have identified clusters of Campylobacter isolates in clinical or environmental settings using one or more subtyping methods (8, 15, 22), including pulsed-field gel electrophoresis (PFGE), restriction fragment length polymorphism analysis of PCR amplicons of the flagellin gene (flaA RFLP), and DNA sequencing of the flagellin gene short variable region (flaA SVR sequencing) (16, 56). Systematic typing and subtyping were found to increase the number of common-source outbreaks detected (20), and 25% of culture-positive isolates in two counties in Denmark were part of clusters, suggesting common sources of infection (V. Fussing, unpublished observations [mentioned in reference 38]).

A concern with typing methods that rely on DNA sequence information, such as multilocus sequence typing (MLST), flaA SVR, and the sequencing of the gene encoding the porin/major outer membrane protein (MOMP) (here designated porA sequence typing), is that they appear to be susceptible to the genetic mixing of populations as a consequence of frequent recombination and somewhat less frequent mutation (47, 49). Many investigators therefore have recommended the use of a combination of molecular typing methods for the investigation of Campylobacter populations (10, 37). A combined approach using multilocus sequence typing (MLST) and flaA SVR sequencing was thought to provide an appropriate level of discrimination for outbreak investigations (46). Although microarray-based comparative genomic hybridization (CGH) using microarrays has shown good correspondence with MLST (52), CGH typing using 70-mer oligonucleotides has been found to provide higher resolution than MLST for genotyping C. jejuni (43). These properties suggest that methods based on comparative genomics represent an alternative to MLST.

The problem, then, is to determine the bacterial detection and typing methods optimal for detecting, through routine surveillance, clusters of isolates (cases) that may not be associated with a single point or with a continuous nonpoint source. Such methods would be most useful within a system similar to PulseNet (50), in which molecular typing and subtyping is performed on all isolates at the laboratory responsible for isolation and in which information is exchanged electronically via a central database to allow the rapid detection of clusters that may be outbreaks. Any methods to be implemented also should be useful for outbreak investigations and associated analyses, such as source tracking.

Standardized, curated databases exist for MLST, flaA SVR sequence typing, and porA/MOMP sequence typing. In this work, we have compared the results from the typing of Campylobacter isolates using PFGE; the Oxford MLST method; the Oxford flaA SVR typing method; combined MLST and flaA SVR typing (MLST + flaA SVR); porA sequence typing (6, 28) (http://pubmlst.org/campylobacter/); combined MLST, flaA SVR, and porA sequence typing (MLST + flaA SVR + porA); and comparative genomic fingerprinting (CGF), a newly developed multiplex PCR method designed to detect Campylobacter genes that were identified as having a high degree of intraspecies variability in comparative genomic hybridizations using DNA microarrays (33, 51). Each method or combination of methods appeared to group isolates at different levels within the population and therefore was useful for a different purpose. CGF appeared to be the optimal method for the detection of epidemiologically relevant clusters of isolates (which may represent clusters of cases), while flaA SVR, MLST, and porA sequence typing were useful for the identification of additional clusters that may be linked to the infection of humans from multiple or different sources.


Sentinel site surveillance.

The development of a sentinel site surveillance system for Canada was proposed in 2004 (17) and launched as a pilot in 2005 as the C-EnterNet sentinel site surveillance program. This program was developed to provide enhanced surveillance of human enteric diseases and additional information on isolates from nonhuman sources that may be responsible for human disease. It is currently being implemented in a single sentinel site in the region of Waterloo, Ontario. During the course of the studies reported here, the population of the region was approximately 500,000 individuals. This region contains an urban area surrounded by farmland.

Human enteric samples were collected through the existing passive surveillance system, in which stool samples requested by physicians are submitted to private, hospital, or public health laboratories. C-EnterNet works with these laboratories to ensure that isolates of all reportable pathogens are forwarded to the Ontario Central Public Health Laboratory, Ontario Agency for Health Protection and Promotion, in Toronto for further characterization and submission to the National Microbiology Laboratory in Winnipeg for subtyping. The agricultural and environmental surveillance components of the C-EnterNet program involved the sampling of food animal operations (swine, beef, dairy, and poultry) and surface water in the sentinel site. Meat samples were purchased from randomly selected stores/markets in the sentinel site.

Isolation of Campylobacter spp.

Meat and manure were processed by transferring 1 g of sample into 9 ml of Bolton broth, followed by growth at 42°C for 48 h under microaerophilic conditions. Each enrichment broth was streaked onto modified cefoperazone charcoal deoxycholate agar (mCCDA) agar plates. After growth at 42°C for 48 h under microaerophilic conditions, suspect colonies were identified using dark-field microscopy and the Gram stain. Suspect colonies also were plated to Mueller-Hinton agar containing 5% horse blood (MHB) and a second mCCDA plate. Typical Campylobacter colonies were confirmed using the catalase and oxidase tests. Hippurate hydrolysis, indoxyl acetate hydrolysis tests, and cephalothin susceptibility tests were performed to differentiate suspected Campylobacter samples into C. jejuni and C. coli.

Each 1,000-ml river water sample was passed through a 0.45-μm filter membrane. The filter then was placed in Bolton broth and incubated at 42°C for 48 h in microaerophilic conditions. Cultures then were plated onto mCCDA agar and incubated under microaerophilic conditions at 42°C for 2 to 5 days. Presumptive Campylobacter colonies were plated on MHB agar and incubated under microaerophilic conditions at 37°C for 24 to 48 h and then confirmed by the Gram stain, catalase test, and oxidase test. Campylobacter species were identified using the hippurate hydrolysis, indoxyl acetate hydrolysis, and cephalothin/nalidixic acid susceptibility tests.

Stool samples from human cases of gastroenteritis were transported to five different primary laboratories and one provincial reference laboratory. Stool was mixed with Cary-Blair transport medium (five laboratories) or Enteric Plus transport medium (1 laboratory) for shipment. Homogenized samples were streaked with sterile swabs onto Campylobacter blood-free agar (four laboratories) or Campylobacter selective medium (two laboratories) and incubated for 18 to 72 h (with the incubation period depending on the laboratory) under microaerophilic conditions. The provincial reference laboratory also cultured stool samples in a liquid enrichment broth, followed by plating to charcoal selective medium. Microaerobic conditions were generated using Oxoid Campygen packs (one laboratory) or gas mixtures of CO2-N2-O2 of either 10:80:10 (one laboratory) or 10:85:5 (three laboratories). Suspect colonies were tested by Gram stain, hippurate hydrolysis, and oxidase tests by all laboratories. Two laboratories also performed the catalase test, nalidixic acid/cephalothin susceptibility testing, and a test for H2S production. The indoxyl acetate hydrolysis test, nitrate reduction test, urea hydrolysis, and tolerance to 1% glycine were done only at the provincial reference laboratory.

Some Campylobacter isolates also were obtained from the New Brunswick Bacterial Enteric Reference Centre as part of an ad hoc investigation to determine whether an increase of cases in their jurisdiction was indicative of an outbreak.

Molecular typing.

MLST was performed according to protocols archived on the Campylobacter jejuni and C. coli MLST website (http://pubmlst.org/campylobacter/) and according to previous work (7, 12). flaA SVR typing was done according to the methods of Meinersmann et al. (35). The sequence typing of the porA (cmp) gene was initially performed in the manner of Huang et al. (28). However, since we had some difficulty obtaining full-length porA sequence using only the F3 and R3 primers for PCR amplification, PCR amplifications and sequencing reactions were done with the primers and methods used previously (6). Subsequent porA sequencing was done using the primers, conditions, and methods set out in the Campylobacter jejuni MOMP database (http://pubmlst.org/campylobacter/), and sequence types were assigned by querying the database with the sequences obtained. Note that the gene name, porA, is used to describe the sequence type throughout.

The DNA used for all amplifications was prepared using a PureGene genomic DNA purification kit (Gentra Systems, Minneapolis, MN). Products were cleaned up using Montage PCR centrifugal filter devices (Fisher Scientific, Edmonton, Canada) according to the manufacturer's instructions. Sequencing was performed in the Genomic Core Facility at the National Microbiology Laboratory, Winnipeg, Canada, with the appropriate primers using BigDye Terminator 3.1 cycle sequencing kits (Applied Biosystems, Streetsville, Canada). Sequencing was done using an ABI 3100 or 3730 DNA analyzer (Applied Biosystems). Sequences were assembled and checked for errors and overall quality in the SeqMan program within the Lasergene software suite (DNASTAR Inc., Madison, WI). Types were identified by using the resulting DNA sequences to query the Campylobacter jejuni and C. coli MLST database (http://pubmlst.org/campylobacter/) and the Campylobacter jejuni flaA SVR database (http://pubmlst.org/campylobacter/).

CGF typing.

Full details of the CGF typing method outlined here will be published elsewhere (53a). Prospective typing markers for inclusion into the CGF typing scheme were identified by microarray-based comparative genomics (5153) and were selected based on a binary distribution (presence or absence) in microarray data from many isolates. Of the genes that exhibited variable carriage in these isolates, 43 were subsequently selected for the development of the CGF typing method. PCR primers (Table 1) were designed for each prospective typing target around regions free of single-nucleotide polymorphisms (SNPs). SNP-free DNA sequences were identified using multiple sequence alignments of homologous targets obtained from genome sequencing projects that either have been completed (strains NCTC 11168, RM1221, and CF93-6) or are ongoing (strains 84-225, HB93-13, and 260.94) (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi). After initial compatibility testing, 40 genes were selected and assembled into 8 multiplex PCRs, each targeting 5 loci, that comprised the CGF assay used here. A more complete description of the development of the CGF assay is the subject of a manuscript in preparation.

Table 1
PCR primers for multiplex PCRs used to determine the CGF types

To generate a CGF fingerprint, the eight multiplex PCRs were performed on each isolate using the primer sets shown in Table 1. Each 5-plex reaction mix contained 1 U MP Taq DNA polymerase (Fisher Scientific, Nepean, Canada), 1× MP buffer, 2.5 mM MgCl2, 0.2 mM each deoxynucleoside triphosphate (dNTP), 0.4 μM each of the 10 primers, and 1 μl DNA template in 25 μl total reaction mix. Amplification cycles were the following: initial denaturation at 94°C for 5 min; 30 cycles of denaturation at 94°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s; final extension at 72°C for 5 min; and a hold at 4°C until samples were removed from the thermocycler. Multiplexes were analyzed using a QIAxcel system (Qiagen, Mississauga, Canada) and visualized using the BioCalculator (v3.0; Qiagen, Mississauga, Canada) software. Separation conditions included using the AM320 method with a 20-s injection time. The 15- to 3,000-bp alignment marker was used as the internal standard marker, and band sizes were determined using the QX 100- to 3,000-bp DNA size marker (Qiagen, Mississauga, Canada). These PCR results were converted to binary values, with 0 representing the absence of the marker and 1 indicating its presence. The clustering of isolates was done based on the binary CGF data using the simple matching distance metric and unweighted-pair group method using average linkages (UPGMA) method of clustering in Bionumerics (version 5.1; Applied Maths, Austin, TX), using 100% fingerprint similarity for cluster definition [designated CGF31 (100%) or CGF40 (100%)]. To obtain an estimate of strain relationships at a slightly lower level of discrimination, data also were analyzed at the 90% fingerprint similarity level [designated CGF31 (90%) or CGF40 (90%)], which organizes all types related at ≥90% into single clusters without recalculating the dendrogram.

The data also were subjected to a retrospective assessment of the relationships between outbreak and nonoutbreak isolates by removing loci that discriminated among outbreak strains. The nine loci removed were Cj0057, Cj0177, Cj0421c, Cj0566, Cj0570, Cj0763, Cj0860, Cj0967, and Cj1585, leaving 31 loci in the analysis that then were used to evaluate the entire database and creating a single CGF31 (100%) type containing all outbreak isolates.

Analysis of data.

Both the means and standard deviations of the number of isolates per type were established for each typing method by using the SigmaStat 3.5 software (Systat Software, Inc., Richmond, CA). The chi-squared test also was performed using this software. Calculations of Simpson's index of diversity (ID) and Wallace coefficients (5) were performed using the online tool at the Comparing Partitions website (http://www.comparingpartitions.info/). The assumption of the epidemiological independence of isolates was not fulfilled within our data set. The term “discriminatory power,” as used in this work, refers to the ability of typing methods to divide the bacterial population into smaller groups, not the ability to distinguish epidemiologically related isolates from epidemiologically unrelated isolates, as in the Simpson's ID method of Hunter and Gaston (29).


Descriptive statistics and discriminatory power of the typing methods.

It is evident from the descriptive statistics associated with the molecular typing methods that the typing methods used describe the Campylobacter population under study in very different ways (Table 2; see also supplementary data file 2 at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls). The least discriminatory methods were flaA peptide type and the clonal complex, both of which created a small number of larger groups. ST and flaA SVR typing resulted in the formation of a few large groups, with many more isolates in smaller groups or constituting the only representative of a type. The remaining methods did not produce any large groups and only a few small groups; most isolates were the sole representative of their type.

Table 2
Descriptive statistics for eight typing methods derived from typing 493 Campylobacter isolates

Simpson's ID was calculated for each typing method (Table 3). Since there is a trend toward using more than one type for analysis, all permutations of types were analyzed in combinations of two, three, and four types. CGF40 (100%) alone was more discriminatory than any other single type or combination of methods that did not include this measure. When CGF40 (100%) was excluded, any combination of two types (e.g., ST + flaA SVR, ST + porA, flaA SVR + porA, ST + CGF40 [90%], etc.) resulted in a somewhat higher discriminatory power, with the addition of a third and fourth method resulting in marginal increases in discrimination among types. In contrast, the combination of one or more additional types with CGF40 (100%) did not provide much greater discriminatory power (Table 3). Both clonal complex and Fla peptide typing were notably less discriminatory than other measures when used alone; because these methods grouped ST and flaA SVR into broader categories, their contributions to overall discrimination in combination with other typing methods were not analyzed further.

Table 3
Simpson's index of diversity calculated for the entire set of 493 isolatesa

Because the PFGE patterns produced using SmaI were only available for a subset of the isolates, a second comparison was done using only isolates for which these data were available (Table 4). PFGE was highly discriminatory, with a Simpson's ID only slightly less than that of CGF40 (100%). Despite the lower number of isolates used (152), the Simpson's ID values obtained for other typing methods were quite similar to those of the larger data set. The discriminatory power of different typing methods also ranked similarly in both data sets.

Table 4
Simpson's index of diversity calculated for a subset of 152 isolates for which PFGE SmaI type data were availablea

Assessment of congruence among methods.

Both the strength and directionality of the correspondences or congruence between the various typing methods was assessed using the Wallace coefficient according to the methods of Carriço et al. (5). As those authors stated, given a primary typing method, Wallace's coefficients provide an estimate of how much additional information is provided by a secondary typing method. A value of 1 indicates that the second method does not detect additional groups within the population (i.e., it is completely congruent with the first method), while a value of 0 indicates that all individuals not differentiated by the first method are completely differentiated into different groups by the second method; that is, the methods are completely noncongruent. Compared in this manner, most methods and combinations of methods showed a low to moderate congruence (Fig. 1, green, brown, red-brown, and red cells; also see the green, brown, red-brown, and red cells in File S1 in the supplemental material). As expected, related methods [i.e., Fla peptide and flaA SVR; porA peptide and PorA nucleotide; CC and ST; and CGF40 (90%) and CGF40 (100%)] were completely congruent and had very high Wallace coefficient values when the method of higher resolution was used as a primary typing method (Fig. 1, blue cells; also see File S1).

Fig 1
Comparison of the global congruence (Wallace coefficient [95% confidence interval]) of methods with PFGE included for a subset of 152 isolates. Wi, expected Wallace coefficient value in the case of independence. The first column of the table represents ...

In general, methods with lower discriminatory power as assessed by Simpson's ID (e.g., Fla peptide and clonal complex) had a higher Wallace coefficient and higher congruence with primary typing methods when used as a secondary typing method. The Oxford porA sequence type had intermediate congruence with both ST and flaA SVR and provided additional information when coupled with each of these methods. Oxford porA showed lower Wallace coefficients when used as a secondary method with each of ST and flaA SVR and higher Wallace coefficients when used as the primary method with ST or flaA SVR as the secondary method. ST and flaA SVR had intermediate congruence with each other, with flaA SVR typing providing more additional information when ST was used as the primary typing method than the reverse. It was of interest that all typing methods had greater congruence with CGF40 (100%) when CGF40 (100%) was used as the primary method than with CGF40 (90%) when CGF40 (90%) was used as the primary typing method, despite the fact that CGF40 (90%) was less discriminatory than CGF40 (100%).

Finally, CGF40 (100%) had low congruence with other methods when used as the secondary typing method compared to those methods or combinations of methods that did not include CGF40 (100%) as a component. For instance, when the combination of ST + flaA SVR + porA + CGF40 (90%) was used as the primary typing method and was compared to ST + flaA SVR + porA + CGF40 (100%) as the secondary method, the Wallace coefficient was only 0.335 (Fig. 1). CGF40 (100%) therefore provided most of the discriminatory information obtained from typing. A Wallace coefficient of 0.972 was obtained in the reverse direction. This raised the question of whether CGF40 (100%), in addition to being highly discriminatory, also grouped Campylobacter isolates somewhat differently than the other methods (i.e., was not very congruent with other methods) and provided much more information than the combination of other methods.

Similar, more limited analyses were done to allow comparisons of methods with PFGE; to simplify interpretation, we analyzed only individual typing methods and not combinations of methods. When PFGE was used as the primary typing method, all methods analyzed other than CGF40 (100%) exhibited intermediate or high levels of congruence (Fig. 1). Although CGF40 (100%) showed very low levels of congruence with PFGE compared to those of the other methods, the intermediate coefficient (0.496) obtained at a clustering stringency of 90% [i.e., CGF40 (90%)] suggests that minor variation in CGF40 (100%) profiles among isolates with identical PFGE profiles accounts for the bulk of this effect. In the reverse direction, PFGE added a great deal of information to all methods except CGF40 (100%). These last observations may result partly from the high levels of discriminatory power of both PFGE and CGF40 (100%).

Assessment of the ability of various molecular typing methods to differentiate outbreak from nonoutbreak isolates.

An outbreak of C. jejuni that occurred in the sentinel site during the summer of 2008 provided an opportunity to test the ability of all typing methods to discern outbreak-related isolates. This outbreak was identified as a high proportion of cases of campylobacteriosis among people attending a summer camp within a very short span of time at the beginning of summer vacation. The affected individuals were somewhat sequestered from the general population, suggesting that this was a classic point-source outbreak with a source closely associated with the camp; the source was never found. Cases were identified and isolates obtained within a very short time period, with all molecular typing and subtyping being done after the outbreak was finished. The identification of outbreak-related cases was based on initial microbiological findings of C. jejuni as the pathogen and confirmed by epidemiological investigations.

C. jejuni isolates from all outbreak cases had indistinguishable MLST, flaA SVR, and Oxford porA types (see supplementary data file 2, worksheet 1, at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls), and these types (ST45, flaA SVR 21, and Oxford porA 49) could be found among 18 nonoutbreak-related isolates. For example, while all 22 outbreak isolates were of the ST45 clonal complex, there were an additional 35 human isolates from this clonal complex in the data set. Similarly, there were 22 flaA SVR 21 human isolates and 13 Oxford porA 49 human isolates in the data set in addition to the outbreak isolates. A total of 11 human nonoutbreak isolates had the full ST45, flaA SVR 21, porA 49 genotype. In contrast to the other methods, the CGF40 (100%) types associated with the outbreak isolates were highly specific and found in only two isolates with no apparent epidemiological association with the outbreak [08-4456 and 08-4603, both of CGF40 (100%) type 48]. These isolates showed a close temporal relationship with the outbreak strain and were indistinguishable by MLST, flaA SVR, and Oxford porA. It is therefore possible that the two cases from which they were obtained were infected from the same source. Knowledge of this potential link at the time of the outbreak might have aided the investigation.

The CGF40 (100%) method did not identify all outbreak isolates as a single indistinguishable type, indicating that this method was capable of measuring change within the Campylobacter population that occurred on the time scale of the outbreak, thereby making the method too discriminatory in some ways.

In routine surveillance, where clusters are identified initially by typing, the five outbreak isolates with CGF40 (100%) types that were closely related to, but different from, the outbreak type would not initially have been included in the cluster/outbreak because the epidemiological data supporting their inclusion would not yet be available. We therefore were interested in what number and combination of loci used in the CGF40 (100%) typing method would detect the outbreak isolates specifically and how this would change the grouping of isolates within the database as a whole. It should be noted that this was an exercise implemented after the fact as a retrospective analysis, and it was intended only to assess how the interpretation of results may change if the data set were changed slightly and the effects of such changes on the relationships of other nonoutbreak isolates. The nine loci that varied among outbreak isolates were removed from the data, the resulting patterns of gene presence and absence were assigned new CGF31 (100%) types, and the entire database was reanalyzed.

The resulting CGF31 (100%) typing analysis grouped all outbreak isolates into one type. It also proved to be the only analysis that was capable of differentiating outbreak-associated isolates from those with indistinguishable MLST, flaA SVR, and Oxford porA types obtained from other geographic locations, obtained from other sources, or isolated in other years (Table 5). One water isolate and six isolates from chicken meat obtained from retail outlets in the sentinel site at several times during 2006 and 2007 were very closely related to the outbreak strain from the sentinel site in July 2008 (i.e., 30 of 31 matching loci) and also were indistinguishable from three human isolates obtained in the sentinel site and two obtained in New Brunswick (Table 5). These isolates also shared indistinguishable MLST, flaA SVR, and Oxford porA types with the outbreak strain. Taken together, these findings indicated that the analysis of Campylobacter populations by CGF, whether using 40 or 31 alleles, assessed important biological and epidemiological properties of the isolates analyzed. The findings also confirmed that the only property of the CGF40 (100%) analysis that was responsible for the incomplete identification of all outbreak-associated isolates in a single type was a high discriminatory power, not the misclassification of isolates into inappropriate types.

Table 5
Use of CGF31 for discrimination of outbreak-associated isolates from closely related nonoutbreak isolates

The analysis of the data set using CGF31 (100%) was not always helpful in detecting clusters of isolates related by clonal descent. Most members of a group of 9 isolates with CGF31 (100%) type 14_4 had different STs, flaA SVR types, and porA types (see supplementary data file 2, worksheet 1, at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls). These particular isolates all were C. coli, and most were from animal sources. However, all three CGF analyses detected a large cluster of 17 isolates [CGF40 (100%) 288; CGF40 (90%) 117; and CGF31 (100%) 28_4] that were of clonal complex ST21 (18 isolates) and highly related sequence types ST982 (16 isolates), ST21 (1 isolate), and ST141 (1 isolate). In contrast, this group of isolates displayed a variety of flaA SVR types (49, n = 9; 46, n = 5; 36, n = 1; 362, n = 1; 371, n = 1; and 838, n = 1) and porA types (738, n = 12; 737, n = 2; 736, n = 1; 740, n = 1; 753, n = 1; and 765, n = 1). Assessing multiple types created an extremely complex picture, even with this small number of isolates. These data clearly support a scenario in which the exchange of DNA and recombination are occurring independently for each gene assessed (see supplementary data file 2 at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls).

Characteristics of clusters detected using one or more typing schemes.

The analysis of the entire data set provided further evidence that changes in types due to mutation or recombination occurred within all typing methods (see supplementary data file 2, worksheet 2 to 14, at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls). This could be observed by examining the allele data for flaA and the seven MLST loci among the 23 multiple-isolate CGF40 (100%) clusters in the data set and in the allele data for porA, as well as the 7 MLST loci among the 35 multiple-isolate porA clusters. Changes in variable gene content as measured by CGF40 (100%) also were seen in ST clusters, including clusters formed by ST + flaA SVR types and in porA clusters. Types appeared to change somewhat independently compared to findings with other typing methods. This provides a visual explanation of, and an elaboration on, the congruence analysis presented earlier. In most cases recombination events resulted in very closely related isolates within a primary type. However, each method, including the CGF analyses and PFGE, on occasion showed changes in a single type that could have resulted only from the pattern of loci present or combination of alleles characteristic of the primary type having become established in a genetically distinct background.

Other interesting properties of the typing methods became apparent after further analyses (see supplementary data file 2 at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls). ST, Oxford porA type, and CGF40 (100%) discriminated between C. jejuni and C. coli, whereas flaA SVR and CGF40 (90%) did not. In addition, there were some types in which isolates appeared to have been exchanging alleles quite freely among themselves. CGF40 (100%) type clusters did not contain any STs that differed by more than 3 alleles, confirming the close concordance, in this direction, of the two typing methods.

Some clusters of isolates were formed when all typing methods were combined, very strongly suggesting a clonal descent for each cluster (Table 6). The largest group found by using this analytic strategy was a group of 17/22 of the outbreak C. jejuni isolates from 2008. This was expected given the extremely short duration and very precise location of this point-source outbreak, which would have been detected by any one of the typing methods. These findings validate the use of these typing methods for detecting epidemiologically related isolates. The next largest group was a set of 10 isolates closely related to the outbreak isolates, differing only in their CGF40 (100%) type. CGF40 (100%) was clearly differentiating important epidemiological differences in at least some of these isolates; while 2 isolates from humans were obtained in 2008, the year of the C. jejuni outbreak in the sentinel site, the other 8 were obtained in 2006 or 2007. Furthermore, while 4 isolates were from humans, the other 6 were obtained from chicken breasts obtained from retail outlets, suggesting retail chicken breasts as a possible source of the human cases. Most of the other clusters in Table 6 contained isolates from both human and animal sources, further suggesting that the molecular subtyping of isolates is an effective way of performing analyses for source attribution. Furthermore, the fact that isolates with identical combined types were sometimes found in 3 years of the 4-year period indicates that these types can be quite stable.

Table 6
Properties of clusters of isolates with completely indistinguishable types when all typing methods were combined

Use of typing for source attribution.

These observations led us to ask whether Campylobacter genotypes found in isolates from human cases also were found in nonhuman sources. These data have been summarized in Table 7. Different methods differed in the extent to which they were capable of assigning isolates to only a single source. CGF40 (100%) assigned 93% of types to a single source, whereas ST, Oxford porA, and flaA SVR assigned 85, 75, and 70% of types, respectively, to a single source. This property of the typing methods did not correlate perfectly with the discriminatory power of each method, as ST was the least discriminatory of the four methods (Table 3). Furthermore, the difference in the frequencies with which each method detected types associated with humans only, with humans and retail chicken meat together, and with humans, retail chicken meat, and bovine manure was significant at a high P value (Table 7). It was evident that each typing method associated isolates with sources in somewhat different ways. For example, flaA SVR associated isolates from a single type with human, chicken, bovine, and water sources, while none of the other typing methods did. Other examples can be found in Table 7.

Table 7
Number of unique types found for each source or combination of sources


Optimal typing methods for C. jejuni and C. coli need to have appropriate discriminatory power, high typeability, a minimum of mistaken types, and a high degree of correspondence with the characteristics of the consensus bacterial type. This could include genomic content, as in the case of typing to estimate phylogenetic relationships, or epidemiological relatedness, as in the case of typing to analyze outbreaks or to ascertain whether there are clusters of cases that suggest an outbreak is occurring. For many applications, typing methods must also be assessed for turnaround time, throughput, cost, and technical difficulty.

In this work, we have assessed and compared several characteristics of a number of typing methods for use with Campylobacter in the context of the detection of outbreaks by subtyping isolates obtained through routine surveillance. The optimal methods would necessarily have fast turnaround, high throughput, and low cost, and they would need to produce groups of isolates that correspond as closely as possible to groups obtained by using epidemiological criteria.

Assessment of methods based on discriminatory power.

Based on Simpson's ID, neither the flaA peptide determined by sequencing the flaA SVR region nor the clonal complex determined by assessing the larger scale population structure resulting from MLST analysis was discriminatory enough to be useful for typing Campylobacter in the context we have defined. Furthermore, each measure produced a single predominant type; for Fla peptide typing, a single type comprised 149/493 (30%) of the entire population analyzed. This would make it difficult to detect clusters of cases, because the baseline level of incidence of that particular type would be so high. While other typing measures provided a much higher discriminatory power, when used alone, ST, flaA SVR typing, porA typing, or CGF40 (90%) typing still did not produce optimal discrimination. The use of a single molecular typing method generally is thought to be suboptimal for typing or subtyping bacterial populations (9a). Typing results therefore were combined, and a combination of any two of these methods appeared to provide a satisfactory level of discrimination. However, the use of more than two measures did not result in much additional discriminatory power; the time, cost, and effort of adding the results from a third typing method did not appear to be worthwhile, at least for the detection of clusters of isolates during routine surveillance.

CGF40 (100%) alone was more discriminatory than all other measures singly or in combination, and the addition of data from other methods did not appear to greatly increase the discriminatory power obtained. On the basis of discriminatory power, CGF40 (100%) appeared to be the preferred method to use for the detection of clusters of C. jejuni and C. coli cases in humans through routine surveillance. This method also was relatively inexpensive and had very high throughput and a short turnaround time when used in conjunction with the QIAxcel capillary electrophoresis system. If this system were not used, although the throughput would be reduced and turnaround time increased, the method still would retain a throughput that is comparable to that of other molecular typing methods assessed here. The multiplex methods used for CGF40 (100%) are not technically difficult and would be accessible to most primary laboratories and reference laboratories. Furthermore, all CGF types are in binary format, and systems for numbering types that do not require querying a centralized data repository can be devised. These considerations make the method extremely portable within laboratory networks. Only a minimal amount of standardization would be necessary.

Assessment of the congruence of the methods.

Discriminatory power alone is not necessarily the only, or the most appropriate, measure for choosing a typing method (40). It also is worth knowing whether the methods used provide clustering that orders populations similarly to other typing methods. In other words, if typing by three or four independent methods suggests that a group of isolates constitutes a clone, any new method considered also should detect similar clones at some level of analysis. Tools to test the congruence of typing methods have been developed recently and were used to assess relationships among typing methods.

Most of the typing methods showed intermediate congruence with each other, suggesting that they group isolates in the population under study in similar ways but with different levels of discrimination. This interpretation is supported by the existence of groups of isolates with indistinguishable ST + flaA SVR + porA types (see supplementary data file 2 at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls). The striking exception was CGF40 (100%). When analysis was performed with CGF40 (100%) as the primary typing method, the highest congruence was found with CGF40 (90%), which was expected since the latter measure was derived from the former. The next highest congruence (0.827) was found with ST, indicating that ST did not provide much information in addition to what was found with CGF40 (100%). Since ST is widely regarded as a highly informative phylogenetic tool for the analysis of Campylobacter populations, this indicates that CGF40 (100%) is also a reliable tool for assessing the phylogenetic structure of C. jejuni, though at much higher levels of discrimination.

Molecular typing methods and source attribution.

One of the surprises that arose in the analysis of these data was the observation that the specific methods used for source tracking affected the results obtained, significantly in some cases. This may be the result of the small size of the data set, in relative terms, resulting in a bias in results that would disappear if thousands or tens of thousands of isolates were analyzed. However, it also may be due to the biological properties of the markers assessed in each typing method. Whereas the MLST alleles are housekeeping genes not subject to diversifying selection (12), the flaA locus is; it is also associated with the virulence of the organism and could theoretically be associated with differences in host specificity, although there are few data to support this hypothesis. It is possible that differences in the protein product of the porA gene, the MOMP, result in differences in function that affect host range or environmental survival, and this could change the distribution of alleles in different hosts. The MOMP is also highly antigenic and immunogenic. Finally, while MLST assesses variability in the core genome present in all C. jejuni or C. coli isolates, CGF40 (100%) and related methods analyze the variable gene content of these species, a fundamental biological and phylogenetic difference. Despite this, there is a very high congruence of the two methods at the phylogenetic level. However, the differences in how these two methods subdivide Campylobacter populations during source attribution studies might stem from this underlying difference. Uncovering why different typing methods lead to differences in source attribution could be a fruitful avenue for future research.

If the goal is to unambiguously and correctly assign types to specific hosts, the CGF40 (100%) method is clearly superior. However, if the goal is to show which sources could be responsible for human infections by including both source and human isolates in a common type, one of the less discriminatory measures may be optimal. The conclusion to be drawn from the results summarized here is that it may not be possible for both goals to be accomplished using only one method. This is an area for further investigation.

We found that a significant proportion of clusters of multiple isolates indistinguishable by multiple types was comprised of isolates from both human and nonhuman sources. Although the strongest association found was with retail chicken, other potential matches also were observed. The detection of C. jejuni in a ground beef sample was unusual. This isolate was indistinguishable by multiple typing methods from one human isolate and closely related to three additional isolates, strongly suggesting that ground beef is a source of human infection with C. jejuni (8, 38), especially in the absence of detection of the same type in chicken or any other source. The data presented in the current study support a growing opinion that cattle contribute significantly in some way to the burden of human disease caused by Campylobacter (26, 39).

Assessment of different typing methods.

PFGE has been successful for the detection of clusters that could represent outbreaks in a few instances (24, 27) and has been of use in outbreak investigations (14, 46).

However, the diversity of PFGE patterns is thought to limit the usefulness of PFGE for outbreak detection (24). In addition to being highly discriminatory, perhaps too much so, we found novel data in this work that PFGE, on occasion, inappropriately groups Campylobacter isolates that appear to be genetically widely divergent by producing an indistinguishable banding pattern. This may provide an explanation for the finding that indistinguishable PFGE patterns have occasionally been found in different outbreaks (46) and probably renders the method unsuitable for longitudinal epidemiologic studies (16).

MLST provides a reliable predictor of clonality (12), correlates closely with the results from multilocus enzyme electrophoresis (45), and has been considered relatively stable (10, 11). There has been an increase in the frequency of the use of the Oxford MLST typing system for the characterization of Campylobacter populations (31). The measure most frequently used to describe these populations has been the clonal complex, which describes groups of related STs (10, 12) and which therefore has relatively lower discriminatory power but can describe changes taking place over longer times and/or larger geographical areas. We found additionally that clonal complexes can be very large, potentially limiting their utility for case cluster detection (see supplementary data file 2 at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls). The ST provided a second, more discriminatory output of the MLST. There were some larger clusters, such as the ST45 group, that could be further subdivided by additional methods. It therefore was not clear whether ST alone would be adequate for providing data appropriate for the epidemiological analysis of clusters that would allow outbreak detection. These outbreak investigation methods, including hypothesis-generating interviews, case-control studies, and cohort studies, can be sensitive to the case definitions, which include the specific identification of the genus, species, type(s), and subtype(s) of the etiologic agent involved.

porA sequence typing could provide some additional discrimination (44) in some cases and valuable support for clusters obtained using other methods. The results presented here also confirmed our earlier observations (6) that, in general, a different set of porins are present in C. jejuni than in C. coli. This typing method may best be used by reference or research laboratories interested in studies of Campylobacter populations. Because the MOMP, or porin, produced by this gene is essential for the survival of the bacterium, and because changes in the amino acid sequence of the MOMP may affect the permeability of the bacterial outer membrane and hence adaptation to different niches (6), it is expected that further investigations into porA sequence variability in Campylobacter populations will lead to an improved understanding of the biology of this organism. The recent development of standard methods for porA typing and the implementation of the Oxford MOMP/porA online database will greatly facilitate research efforts in this area.

The sequencing of the flaA SVR region provided information at two phylogenetic levels within the Campylobacter population without additional analysis or interpretation, with the Fla peptide sequence of quite low discriminatory power and the flaA SVR DNA sequence type of higher discriminatory power. We did not find Fla peptide types to be of much use for our purposes and did not include them in subsequent analyses. While flaA sequence typing also has been used to infer phylogenetic relationships among Campylobacter isolates (25, 35), it is quite well accepted that the frequency of genetic exchange renders results from this method unsuitable for long-term population studies (23). Furthermore, although evidence here and elsewhere (23) clearly indicates that recombination at the flaA locus is relatively frequent, studies of isolates from chickens also clearly indicate that subtypes of Campylobacter spp. can persist throughout the year and in different locations (25). In an earlier study, the flaA SVR was only moderately predictive of ST (13), and our results supported these findings. Recombination was found to be frequent in the alleles targeted by both the Oxford MLST system and by the alleles analyzed by Suerbaum and colleagues (47, 49). Our data support this conclusion and further suggest that alleles detected by the Oxford MLST are not much more stable than those detected by flaA SVR and porA sequencing (see supplementary data file 2 at www.corefacility.ca/supplementary_data/Supplementary_Data_2_Nov14_11_Final.xls).

The creation of the CGF methods has included a detailed theoretical treatment of the optimal number and choice of loci, as well as extensive experimental development and testing, both of which will be dealt with elsewhere. The basis for these methods is the demonstration, using multiplex PCRs, of the presence or absence of genes found to be variable in earlier comparative genomic hybridization/DNA microarray experiments involving multiple C. jejuni isolates (5153). The CGF type of approach has recently been shown to generate epidemiologically relevant clusters in the analysis of verocytotoxin-producing Escherichia coli (VTEC) isolates (33). It now can be used prospectively for sentinel site surveillance of Campylobacter to determine how robust it is as a real-time cluster detection method.

The CGF40 suite of methods or analyses, i.e., CGF40 (100%), CGF40 (90%), and CGF31 (100%), provided a flexible tool for acquiring data that could be analyzed in different ways to achieve different goals. We found the CGF40 (100%) measure to be useful for predicting epidemiological relationships and to be phylogenetically informative. The very high discriminatory power of the CGF40 (100%) analysis is, in some ways, an advantage in the context of providing information on clusters of cases that might be outbreaks. For confirmation, such clusters would need to be subjected to epidemiological investigation with associated statistical analyses. Assuming that case clusters were large enough for detection and analysis, the failure to detect a few associated cases with closely related types would not have a significant effect on the statistical power of methods designed for use in case-control studies, for instance. In contrast, typing methods such as MLST, flaA SVR sequencing, and Oxford porA sequence typing, which were demonstrated in this work to include nonoutbreak cases with outbreak cases, could reduce the statistical power of such methods to a point where an outbreak is occurring but is not identified. Using these methods could result in large outbreaks being missed, while the use of CGF40 (100%) would likely cause public health surveillance systems to miss only smaller outbreaks. The implementation of a second typing system in conjunction with CGF40 (100%) would reduce the likelihood of missing outbreaks. In this case, the simplest, least expensive, highest throughput, fastest turnaround time, and easiest method technically would be flaA SVR sequence typing.

It was of interest to see what would happen with the discriminatory loci of the CGF40 assay removed from the analysis, which can be done only when the outbreak identification and investigation have been completed and control measures have been implemented. When this was done and the resulting CGF31 types applied to all isolates in the database, CGF31 typing clearly discriminated outbreak isolates from similar nonoutbreak isolates by the absence of a single locus. A small group of C. jejuni isolates that were very similar to those associated with the outbreak were seen in New Brunswick in September 2008, although the presence of Cj1224 differentiated these isolates. The information suggested that the outbreak isolates from Ontario and the related isolates from New Brunswick came from the same source. This information might have been valuable if it were available at the time of the outbreak investigation. Because the epidemiological relationship of the outbreak-associated isolates was the only basis for removing the CGF40 loci, the resulting CGF31 data would not be expected to have the same phylogenetic validity as the CGF40 data. There is also an element of chance associated with which particular subset of bacteria manage to contaminate sources of infection responsible for outbreaks. It therefore seems unlikely that the CGF31 test utilized here would be useful in any other context. A new retrospective analysis based on CGF40 loci common to outbreak-associated isolates would need to be developed for each distinct outbreak, if required, after the outbreak strain(s) has been characterized and the results used to aid in outbreak investigations, traceback investigations, or in source tracking. The amount of work is not onerous, however, and the potential rewards of doing so may justify the investment in time.

It should be emphasized that the CGF40 test was not developed randomly, as are many molecular typing methods, but was developed to replicate the phylogenetic clustering of isolates based on much larger gene data sets. We have shown here that changing the loci used, while useful, can change the relationships among isolates. Each set of loci used for assessing the epidemiological relevance of relationships between isolates derived from molecular subtyping data must be validated separately against the collection of isolates acquired through surveillance.

Aside from the well-characterized outbreak of 2008, there were no clusters of isolates large enough to apply epidemiological methods for outbreak detection, although there were several small, discrete clusters of isolates that may have been associated with outbreaks. This was analogous to previous results in Denmark (37). Evidence for temporally limited clusters of cases has also been found in England (48). The relatively small population size of the sentinel site, approximately 500,000 people during the period of investigation, was likely the limiting factor in the size of isolate clusters detected. It is therefore vital that this work be continued with a much larger human population, for instance, through the expansion of the number of sentinel sites associated with the C-EnterNet surveillance system. The great utility of data from even a very limited number of isolates from New Brunswick strongly supports this position.

All typing methods ultimately were useful. The challenge is to find out what methods are useful for specific purposes. In this work, we have evaluated the utility of a number of methods for the detection of clusters of isolates and have further determined whether any methods, or combinations of methods, are useful for tracing the sources of human Campylobacter infections. Furthermore, as has been seen in work by other researchers (37), the comparison of a number of methods has allowed the evaluation of the relative strengths and weaknesses of each method in grouping isolates within the Campylobacter population. We think that the data acquired are sufficient to recommend the implementation of the CGF method for cluster detection as part of routine surveillance, with flaA SVR as a secondary method in support of CGF, as flaA SVR sequencing is a relatively rapid and inexpensive typing method that does not impose the burden of analytic effort characteristic of MLST. For routine sentinel site surveillance to detect clusters, MLST could be applied in addition to CGF and flaA SVR when warranted by sufficient numbers of indistinguishable types within a limited time frame. Targeting each method in this way will maximize the scientific returns on the investment and provide a firm basis for understanding how Campylobacter isolates interact with their various ecological and animal niches.

Supplementary Material

Supplemental material:


Analyses of congruence of typing methods were done using the Comparing Partitions tool, developed and maintained by João Carriço and Ana Severiano and found at http://darwin.phyloviz.net/ComparingPartitions/index.php?link=Home. This publication made use of the Campylobacter jejuni MLST website (http://pubmlst.org/campylobacter/), the Campylobacter jejuni flaA SVR database (http://pubmlst.org/campylobacter/), and the Campylobacter jejuni MOMP database (http://pubmlst.org/campylobacter/), developed by Keith Jolley and Man-Suen Chan and sited at the University of Oxford (30). The development of this site has been funded by the Wellcome Trust. We also acknowledge the contribution of A. Cody in the assignment of a rather large number of new porA (MOMP) alleles. We acknowledge the provision of strains by the New Brunswick Bacterial Enteric Reference Centre and permission from this center to use data associated with these strains in this publication. We thank Celine Nadon for providing all PFGE data, for providing sage advice during the completion of this work, and for critically reading the manuscript. We also acknowledge the contribution of the NML Genomics Core for DNA sequencing and PCR primer synthesis, specifically Brynn Kaplen, Kimberly Melnychuk, Erika Landry, Shari Tyson, and Travis Murphy. Nancy Sittler, Waterloo, Ontario, was a key person who was responsible for the collection of data at the sentinel site, and we thank Andrea Nesbitt, Public Health Agency of Canada, for database management. We thank Cynthia Misfeldt and Lorelee Tschetter for performing PFGE analysis and database curation and management. Finally, we thank Cai Guan for help with assigning CGF40 and CGF31 pattern designations and for carefully checking the data set.


Published ahead of print 7 December 2011

Supplemental material for this article may be found at http://jcm.asm.org/.


1. Allos BM. 2001. Campylobacter jejuni infections: update on emerging issues and trends. Clin. Infect. Dis. 32:1201–1206 [PubMed]
2. Baker M, Wilson N, Edwards R. 2007. Campylobacter infection and chicken: an update on New Zealand's largest “common source outbreak.” N. Z. Med. J. 120:U2717. [PubMed]
3. Black AP, Kirk MD, Millard G. 2006. Campylobacter outbreak due to chicken consumption at an Australian Capital Territory restaurant. Commun. Dis. Intell. 30:373–377 [PubMed]
4. Blaser MJ. 1997. Epidemiologic and clinical features of Campylobacter jejuni infections. J. Infect. Dis. 176(Suppl. 2):S103–S105 [PubMed]
5. Carriço JA, et al. 2006. Illustration of a common framework for relating multiple typing methods by application to macrolide-resistant Streptococcus pyogenes. J. Clin. Microbiol. 44:2524–2532 [PMC free article] [PubMed]
6. Clark CG, et al. 2007. Phylogenetic relationships of Campylobacter jejuni based on porA sequences. Can. J. Microbiol. 53:27–38 [PubMed]
7. Clark CG, et al. 2005. Use of the Oxford multilocus sequence typing protocol and sequencing of the flagellin short variable region to characterize isolates from a large outbreak of waterborne Campylobacter sp. strains in Walkerton, Ontario, Canada. J. Clin. Microbiol. 43:2080–2091 [PMC free article] [PubMed]
8. Clark CG, et al. 2003. Characterization of waterborne outbreak-associated Campylobacter jejuni, Walkerton, Ontario. Emerg. Infect. Dis. 9:1232–1241 [PMC free article] [PubMed]
9. Coker AO, Isokpehi RD, Thomas BN, Amisu KO, Obi CL. 2002. Human campylobacteriosis in developing countries. Emerg. Infect. Dis. 8:237–244 [PMC free article] [PubMed]
9a. de Boer P, et al. 2000. Computer-assisted analysis and epidemiological value of genotyping methods for Campylobacter jejuni and Campylobacter coli. J. Clin. Microbiol. 38:1940–1946 [PMC free article] [PubMed]
10. Dingle KE, Colles FM, Falush D, Maiden MC. 2005. Sequence typing and comparison of population biology of Campylobacter coli and Campylobacter jejuni. J. Clin. Microbiol. 43:340–347 [PMC free article] [PubMed]
11. Dingle KE, et al. 2002. Molecular characterization of Campylobacter jejuni clones: a basis for epidemiologic investigation. Emerg. Infect. Dis. 8:949–955 [PMC free article] [PubMed]
12. Dingle KE, et al. 2001. Multilocus sequence typing system for Campylobacter jejuni. J. Clin. Microbiol. 39:14–23 [PMC free article] [PubMed]
13. Djordjevic SP, Unicomb LE, Adamson PJ, Mickan L, Rios R. 2007. Clonal complexes of Campylobacter jejuni identified by multilocus sequence typing are reliably predicted by restriction fragment length polymorphism analyses of the flaA gene. J. Clin. Microbiol. 45:102–108 [PMC free article] [PubMed]
14. Fitzgerald C, et al. 2001. Evaluation of methods for subtyping Campylobacter jejuni during an outbreak involving a food handler. J. Clin. Microbiol. 39:2386–2390 [PMC free article] [PubMed]
15. Fitzgerald C, Stanley K, Andrew S, Jones K. 2001. Use of pulsed-field gel electrophoresis and flagellin gene typing in identifying clonal groups of Campylobacter jejuni and Campylobacter coli in farm and clinical environments. Appl. Environ. Microbiol. 67:1429–1436 [PMC free article] [PubMed]
16. Fitzgerald C, Sails AD, Fields PI. 2005. Campylobacter jejuni strain variation, p. 59–77 In Ketley JM, Konkel ME, editors. (ed.), Campylobacter jejuni: new perspectives in molecular and cellular biology. Horizon Bioscience, Wymondham, United Kingdom
17. Flint JA, Dore K, Majowicz SE, Edge VL, Sockett P. 2004. From stool to statistics: reporting of acute gastrointestinal illnesses in Canada. Can. J. Public Health 95:309–313 [PubMed]
18. Reference deleted.
19. Frost JA, Gillespie IA, O'Brien SJ. 2002. Public health implications of campylobacter outbreaks in England and Wales, 1995–9: epidemiological and microbiological investigations. Epidemiol. Infect. 128:111–118 [PMC free article] [PubMed]
20. Fussing V, Moller NE, Neimann J, Engberg J. 2007. Systematic serotyping and riboprinting of Campylobacter spp. improves surveillance: experiences from two Danish counties. Clin. Microbiol. Infect. 13:635–642 [PubMed]
21. Gillespie IA, et al. 2003. Point source outbreaks of Campylobacter jejuni infection–are they more common than we think and what might cause them? Epidemiol. Infect. 130:367–375 [PMC free article] [PubMed]
22. Hänninen ML, Pajarre S, Klossner ML, Rautelin H. 1998. Typing of human Campylobacter jejuni isolates in Finland by pulsed-field gel electrophoresis. J. Clin. Microbiol. 36:1787–1789 [PMC free article] [PubMed]
23. Harrington CS, Thomson-Carter FM, Carter PE. 1997. Evidence for recombination in the flagellin locus of Campylobacter jejuni: implications for the flagellin gene typing scheme. J. Clin. Microbiol. 35:2386–2392 [PMC free article] [PubMed]
24. Hedberg CW, et al. 2001. Limitations of pulsed-field gel electrophoresis for the routine surveillance of Campylobacter infections. J. Infect. Dis. 184:242–244 [PubMed]
25. Hiett KL, Stern NJ, Fedorka-Cray P, Cox NA, Seal BS. 2007. Molecular phylogeny of the flaA short variable region among Campylobacter jejuni isolates collected during an annual evaluation of poultry flocks in the Southeastern United States. Foodborne Pathog. Dis. 4:339–347 [PubMed]
26. Hoffmann S, Fischbeck P, Krupnick A, McWilliams M. 2007. Using expert elicitation to link foodborne illnesses in the United States to foods. J. Food Prot. 70:1220–1229 [PubMed]
27. Hook H, Ekegren MB, Ericsson H, Vagsholm I, Danielsson-Tham ML. 2004. Genetic and epidemiological relationships among Campylobacter isolates from humans. Scand. J. Infect. Dis. 36:435–442 [PubMed]
28. Huang S, Luangtongkum T, Morishita TY, Zhang Q. 2005. Molecular typing of Campylobacter strains using the cmp gene encoding the major outer membrane protein. Foodborne Pathog. Dis. 2:12–23 [PubMed]
29. Hunter PR, Gaston MA. 1988. Numerical index of the discriminatory ability of typing systems: an application of Simpson's index of diversity. J. Clin. Microbiol. 26:2465–2466 [PMC free article] [PubMed]
30. Jolley KA, Chan MS, Maiden MC. 2004. mlstdbNet–distributed multi-locus sequence typing (MLST) databases. BMC Bioinformatics 5:86. [PMC free article] [PubMed]
31. Kärenlampi R, Rautelin H, Hänninen ML. 2007. Evaluation of genetic markers and molecular typing methods for prediction of sources of Campylobacter jejuni and C. coli infections. Appl. Environ. Microbiol. 73:1683–1685 [PMC free article] [PubMed]
32. Reference deleted.
33. Laing C, et al. 2008. Rapid determination of Escherichia coli O157:H7 lineage types and molecular subtypes by using comparative genomic fingerprinting (CGF). Appl. Environ. Microbiol. 74:6606–6615 [PMC free article] [PubMed]
34. Reference deleted.
35. Meinersmann RJ, Helsel LO, Fields PI, Hiett KL. 1997. Discrimination of Campylobacter jejuni isolates by fla gene sequencing. J. Clin. Microbiol. 35:2810–2814 [PMC free article] [PubMed]
36. Miller G, Dunn GM, Smith-Palmer A, Ogden ID, Strachan NJ. 2004. Human campylobacteriosis in Scotland: seasonality, regional trends and bursts of infection. Epidemiol. Infect. 132:585–593 [PMC free article] [PubMed]
37. Nielsen EM, et al. 2000. Evaluation of phenotypic and genotypic methods for subtyping Campylobacter jejuni isolates from humans, poultry, and cattle. J. Clin. Microbiol. 38:3800–3810 [PMC free article] [PubMed]
38. Nielsen EM, Fussing V, Engberg J, Nielsen NL, Neimann J. 2006. Most Campylobacter subtypes from sporAdic infections can be found in retail poultry products and food animals. Epidemiol. Infect. 134:758–767 [PMC free article] [PubMed]
39. On SL, Nielsen EM, Engberg J, Madsen M. 1998. Validity of SmaI-defined genotypes of Campylobacter jejuni examined by SalI, KpnI, and BamHI polymorphisms: evidence of identical clones infecting humans, poultry, and cattle. Epidemiol. Infect. 120:231–237 [PMC free article] [PubMed]
40. Pagotto F, et al. 2005. Molecular typing and differentiation of foodborne pathogens, p. 51–76 In Fratamico PM, Bhunia AK, Smith JL, editors. (ed.), Foodborne pathogens: microbiology and molecular biology. Caister Academic Press, Norfolk, United Kingdom
41. Pebody RG, Ryan MJ, Wall PG. 1997. Outbreaks of Campylobacter infection: rare events for a common pathogen. CDR Rev. 7:R33–R37 [PubMed]
42. Rautelin H, Hänninen M-L. 2000. Campylobacters: the most common bacterial enteropathogens in the Nordic countries. Ann. Med. 32:440–445 [PubMed]
43. Rodin S, et al. 2008. Performance of a 70-mer oligonucleotide microarray for genotyping of Campylobacter jejuni. BMC Microbiol. 8:73. [PMC free article] [PubMed]
44. Sahin O, et al. 2008. Emergence of a tetracycline-resistant Campylobacter jejuni clone associated with outbreaks of ovine abortion in the United States. J. Clin. Microbiol. 46:1663–1671 [PMC free article] [PubMed]
45. Sails AD, Swaminathan B, Fields PI. 2003. Clonal complexes of Campylobacter jejuni identified by multilocus sequence typing correlate with strain associations identified by multilocus enzyme electrophoresis. J. Clin. Microbiol. 41:4058–4067 [PMC free article] [PubMed]
46. Sails AD, Swaminathan B, Fields PI. 2003. Utility of multilocus sequence typing as an epidemiological tool for investigation of outbreaks of gastroenteritis caused by Campylobacter jejuni. J. Clin. Microbiol. 41:4733–4739 [PMC free article] [PubMed]
47. Schouls LM, et al. 2003. Comparative genotyping of Campylobacter jejuni by amplified fragment length polymorphism, multilocus sequence typing, and short repeat sequencing: strain diversity, host range, and recombination. J. Clin. Microbiol. 41:15–26 [PMC free article] [PubMed]
48. Sopwith W, et al. 2006. Campylobacter jejuni multilocus sequence types in humans, northwest England, 2003–2004. Emerg. Infect. Dis. 12:1500–1507 [PMC free article] [PubMed]
49. Suerbaum S, Lohrengel M, Sonnevend A, Ruberg F, Kist M. 2001. Allelic diversity and recombination in Campylobacter jejuni. J. Bacteriol. 183:2553–2559 [PMC free article] [PubMed]
50. Swaminathan B, Barrett TJ, Hunter SB, Tauxe RV., and the PulseNet Task Force CDC 2001. PulseNet: the molecular subtyping network for foodborne bacterial disease surveillance, United States. Emerg. Infect. Dis. 7:382–389 [PMC free article] [PubMed]
51. Taboada EN, et al. 2004. Large-scale comparative genomics meta-analysis of Campylobacter jejuni isolates reveals low level of genome plasticity. J. Clin. Microbiol. 42:4566–4576 [PMC free article] [PubMed]
52. Taboada EN, et al. 2008. Comparative genomic assessment of multi-locus sequence typing: rapid accumulation of genomic heterogeneity among clonal isolates of Campylobacter jejuni. BMC Evol. Biol. 8:229. [PMC free article] [PubMed]
53. Taboada EN, et al. 2007. Comparative genomic analysis of Campylobacter jejuni associated with Guillain-Barré and Miller Fisher syndromes: neuropathogenic and enteritis-associated isolates can share high levels of genomic similarity. BMC Genomics 8:359. [PMC free article] [PubMed]
53a. Tabaoda EN, et al. 2012. Development and validation of a comparative genomic fingerprinting method for high-resolution genotyping of Campylobacter jejuni. J. Clin. Microbiol. 50:788–797 [PMC free article] [PubMed]
54. Takkinen J, Ammon A, Robstad O, Breuer T., and the Campylobacter Working Group 2003. European survey on Campylobacter surveillance and diagnosis 2001. Euro Surveill. 8:207–213 [PubMed]
55. Tauxe RV. 1992. Epidemiology of Campylobacter jejuni infections in the United States and other industrialized countries, p. 9–19 In Blaser MJ, Tompkins L, editors. (ed.), Campylobacter jejuni: current status and future trends. American Society for Microbiology, Washington, DC
56. Wassenaar TM, Newell DG. 2000. Genotyping of Campylobacter spp. Appl. Environ. Microbiol. 66:1–9 [PMC free article] [PubMed]

Articles from Journal of Clinical Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...