and Institute of Food Safety and Toxicology, 2860 Søborg,3 Denmark
Received 2 March 2000/Returned for modification 11 June 2000/Accepted 28 July 2000
Six methods for subtyping of Campylobacter jejuni were compared and evaluated with a collection of 90isolates from poultry, cattle, and sporadic human clinical cases as well as from a waterborne outbreak. Theapplied methods were Penner heat-stable serotyping; automated ribotyping (RiboPrinting); random amplifiedpolymorphic DNA typing (RAPD); pulsed-field gel electrophoresis (PFGE); restriction fragment length poly-morphisms of the flagellin gene, flaA (fla-RFLP); and denaturing gradient gel electrophoresis of flaA (fla-DGGE). The methods were evaluated and compared on the basis of their abilities to identify isolates from oneoutbreak and discriminate between unrelated isolates and the agreement between methods in identifying clonallines. All methods identified the outbreak strain. For a collection of 80 supposedly unrelated isolates, RAPDand PFGE were the most discriminatory methods, followed by fla-RFLP and RiboPrinting. fla-DGGE andserotyping were the least discriminative. All isolates included in this study were found to be typeable by eachof the methods. Thirteen groups of potentially related isolates could be identified using a criterion that at leastfour of the methods agreed on clustering of isolates. None of the subtypes could be related to only one source;rather, these groups represented isolates from different sources. Furthermore, in two cases isolates from cattleand human patients were found to be identical according to all six methods.
Campylobacter spp. are among the most frequently reportedcauses of bacterial enteritis in the developed countries. Thenumber of human cases of campylobacteriosis has increaseddramatically in recent years in many countries. In Denmark,the number of cases has more than tripled during the last 7years from approximately 22 cases/100,000 inhabitants in theyears 1980 to 1992 to 78/100,000 in 1999 (2). Approximately95% of the Danish cases are caused by Campylobacter jejunisubsp. jejuni (hereafter C. jejuni).
A wide range of phenotypic and genotypic typing systemshave been developed and used for epidemiological typing ofCampylobacter spp. The phenotypic methods include serotyp-ing with heat-stable (37) or heat-labile antigens (19), phagetyping (40), and biotyping (4). The phenotypic methods, inparticular the two serotyping systems, are used in laboratoriesworldwide, e.g., for surveillance of a large number of isolates.However, for improved discrimination of isolates one or moreof the genotypic methods are usually selected. Some of themost commonly used genotypic methods for typing of Campy-lobacter are pulsed-field gel electrophoresis (PFGE), ribotyp-ing, flagellin gene typing, random amplified polymorphic DNAtyping (RAPD), and restriction endonuclease analysis (26, 35,49). The genotypic methods have primarily been used for out-breaks and epidemiological studies of poultry flocks, etc.,whereas less has been published regarding typing of sporadichuman cases and surveillance of isolates from animal sources.
Despite the large number of different typing systems forCampylobacter, few studies comparing methods and their effi-cacies exist. Patton et al. (36) evaluated the ability of 10 phe-
notypic and genotypic methods to distinguish C. jejuni strainsfrom animals and humans involved in four milk- and water-borne outbreaks. Three phenotypic methods (the two serotyp-ing systems and multilocus enzyme electrophoresis) and threegenotypic methods (PvuII/PstI ribotyping and two restrictionendonuclease analysis methods) were able to correctly identifyall epidemiologically implicated strains. In contrast, four othermethods (biotyping, phage typing, plasmid profiling, and BglI/XhoI ribotyping) were not sufficiently discriminatory to makethe correct groupings of strains. In another study, PFGE typ-ing, heat-labile serotyping, biotyping, and fatty acid profile typ-ing were compared for typing of C. jejuni and Campylobactercoli isolated from abattoirs (44). PFGE was found to be themost discriminatory method, and biotyping was found to bethe least discriminatory. In two other studies, PFGE was alsothe most discriminatory method, whereas phage typing had lowdiscrimination and typeability, and HaeIII/PstI ribotypingcould differentiate between C. jejuni strains of different se-rotypes but differentiated only to a limited degree betweenstrains of the same serotype (12, 34). In general, PFGE andribotyping with some enzyme combinations showed good dis-criminatory power in the previous studies. In contrast, biotyp-ing, phage typing, and plasmid profiling had low discriminatorypower.
In the present study, we have evaluated one phenotypicmethod and five genotypic methods for subtyping of C. jejuni:heat-stable serotyping (Penner serotyping), PFGE, automatedribotyping (RiboPrinting), RAPD, PCR-restriction fragmentlength polymorphism (RFLP) on the flaA gene, and PCR-de-naturing gradient gel electrophoresis (DGGE) on the flaAgene. Of these typing methods, PFGE, fla-RFLP, and serotyp-ing have been used extensively for typing of C. jejuni by severallaboratories including the laboratories participating in thisstudy (25, 27, 28, 32). RAPD has also been used for typing of
campylobacters previously (10, 14, 21). In the present study,the RAPD protocol included the use of standardized PCRanalysis beads and fluorescence detection of the resulting pro-files on an automated fragment analyzer. RiboPrinting isequivalent to ribotyping, the difference being that most of theprocess is automated. PCR-DGGE has been used for typing ofhuman genes and mixed population fingerprinting (5, 23, 46)but has not previously been used for bacterial subtyping, al-though preliminary results have been presented (6; C.-H. Bro-gren and P. Venema, Abstr. IMBEM IV—4th Int. Meet. Bac-terial Epidemiol. Markers, abstr. S103, 1997).
The six methods were used for subtyping a collection of 90C. jejuni isolates from animal sources, sporadic human cases,and a well-documented waterborne outbreak. The methodswere evaluated and compared on the basis of their abilities toidentify outbreak isolates and discriminate between unrelatedisolates and the agreement between methods in identifyingprobable clones.
MATERIALS AND METHODS
Bacterial isolates. Ninety C. jejuni isolates obtained during an 11-month pe-riod in 1995-1996 were included in the study. Of these, 75 isolates were selectedat random from strain collections at the Danish Veterinary Laboratory and theStatens Serum Institut: 40 human clinical isolates obtained from sporadic casesof campylobacteriosis with no known relation to each other, 20 isolates obtainedfrom broiler chickens, and 15 isolates obtained from cattle. Fifteen isolates wererelated to a waterborne outbreak affecting a small Danish town during Januaryto March 1996 (7). Of these, nine isolates were from patients with a confirmedrelation to the outbreak, two were isolated from the suspected water, and fourwere clinical isolates from the same period and region but apparently epidemi-ologically unrelated to the outbreak (control isolates). Therefore, in total, 80isolates were supposedly unrelated: the 75 randomly selected isolates, the fourcontrol isolates, and one representative of the outbreak isolates (strain 5001). Allhuman strains were isolated from feces at the Statens Serum Institut using astandard procedure (7). All cattle and chicken strains were isolated at the DanishVeterinary Laboratory from fecal samples from healthy animals at slaughteraccording to a previously described procedure (27). Sets of identical cultureswere prepared in 15% glycerol broth and stored at 280°C. One set was thendistributed to each participating laboratory.
Serotyping. For antigen preparation, the bacteria were cultured on blood agarfor 24 to 48 h at 42°C in a microaerobic atmosphere. Heat-stable serotyping (Oserotyping) was performed according to the Penner serotyping scheme (37) withseparate sets of sera for C. jejuni and C. coli (38). The C. jejuni strains were typedusing all 47 C. jejuni antisera in the hemagglutination test. If a strain was non-typeable with these sera, the strain was also tested against the 19 C. coli antisera.The production of antisera has been described previously (27). If a strain reactedin more than one antiserum, it was designated a complex serotype, e.g., O:1,44,O:4,13,16,43,50,64 (5O:4 complex), O:6,7, or O:23,36. All complexes seen in thisstudy were well-known and common C. jejuni serotype complexes (38). Differentcombinations of reactions with the O:4 complex were seen, but these weredisregarded in the data analysis.
PFGE profiling. DNA-agarose samples were prepared from formaldehyde-treated bacterial cells using the protocol of Gibson et al. (11), modified as de-scribed previously (32, 33). DNA was digested with SmaI, and fragments wereseparated in a CHEF-DRIII PFGE system (Bio-Rad Laboratories, Hercules,Calif.), using parameters described previously (32). Any differences betweenPFGE profiles of strains were considered significant, and types were arbitrarilydefined on that basis.
fla-RFLP typing. For fla-RFLP, a 1.7-kb fragment of the flaA gene was am-plified and analyzed after digestion with two restriction enzymes, DdeI and AluI.Bacteria were grown on blood agar (5% cattle blood) overnight in a microaerobicatmosphere. Preparation of template for PCR and the PCRs were carried outlargely as described by Nachamkin and colleagues (24) or, in a few cases, by usingthe commercial DNA isolation kit QIAamp tissue kit (Qiagen, Hilden, Ger-many) by use of the manufacturer’s recommendations for gram-negative bacte-ria, without RNase treatment. The procedure was slightly modified so that each50-ml PCR mixture contained 5.0 ml of Super Taq buffer (HT BiotechnologyLtd., Cambridge, United Kingdom); 5.0 ml of 100 mM Tris-HCl (pH 8.3); 3.0 mlof 25 mM MgCl2, resulting in a total concentration of 3.0 mM Mg21; 0.4 mMdeoxynucleoside triphosphate (Amersham Pharmacia Biotech, Little Chalfont,Buckinghamshire, United Kingdom); 0.25 mM (each) primer; 2.5 U of SuperTaq polymerase (HT Biotechnology Ltd.); and 5 ml of heated bacterial lysate, or2.5 ml of DNA purified by QIAamp.
Computer-assisted analysis using GelCompar (Applied Maths, Kortrijk, Bel-gium) was used for identification of RFLP profiles. Starting with the moredistinct DdeI profiles, profiles were compared with similar patterns derived fromDanish C. jejuni strains contained in existing databases. In cases when a match
could not be found, a new profile type was defined. One band difference or bandshift sufficient to be reproducibly recorded distinguished one profile type fromanother. AluI profiles were identified subsequently and if possible given the samenumber as the DdeI profile of that strain. If more than one AluI profile wereobserved in combination with a DdeI profile, letters were used to indicate therelationship through the DdeI profile (fla-RFLP types 1/1 and 1/1a are distin-guished by the AluI profile). When more than one DdeI profile were observed incombination with one AluI profile, the original AluI profile name was kept(fla-RFLP types 25/25 and 26/25 are distinguished by the DdeI profile).
RiboPrinting. RiboPrinting was performed using the RiboPrinter, as recom-mended by the manufacturer (Qualicon, Wilmington, Del.). In brief, singlecolonies from a 24-h culture on a 5% yeast-enriched blood agar plate weresuspended in a sample buffer and heated at 80°C for 15 min. After addition oflytic enzymes, samples were transferred to the RiboPrinter System. Furtheranalysis, including HaeIII restriction of DNA, was carried out automatically. TheRiboPrint profiles were aligned according to the position of a molecular sizestandard and compared with patterns obtained previously. Profiles were analyzedwith the GelCompar software using the band matching coefficient of Dice andUPGMA (unweighted pair group method with averages) clustering to determineprofile relatedness.
RAPD typing. Template DNA was extracted from a 24-h subculture on ayeast-enriched 5% blood agar plate by picking colony material corresponding toapproximately 1 mg using a 1-ml inoculating loop. The colony material was mixedwith 300 ml of a 20% slurry of Chelex-100 (Bio-Rad) in TE buffer (10 mM Tris[pH 8], 1 mM EDTA) and heated at 95°C for 10 min. The resin was pelleted bycentrifugation at 10,000 rpm (Biofuge 13; Heraeus Sepatech) for 2 min. Twomicroliters of this suspension was used for subsequent amplifications. All PCRamplifications were performed using 25 pmol of primer with Ready-To-GoRAPD analysis beads (Amersham Pharmacia Biotech, Uppsala, Sweden), con-taining premixed, predispensed AmpliTaq DNA polymerase, as well as allnecessary buffer ingredients and nucleotides. The cycling parameters were asfollows: denaturing at 95°C for 30 s, annealing for 1 min at temperatures as statedbelow, and extension at 72°C for 2 min in a total of 31 cycles. Annealing startedat 46°C with a 1°C decrease during 11 cycles until 36°C, followed by 20 cycles at36°C. The ramping was done at 2.5°C/s and 21°C/s. Prior to cycling, sampleswere heated to 95°C for 5 min. Finally, an additional extension step of 72°C for7 min was included. Amplifications were performed using a thermocycler, thePTC-200 Peltier Thermal cycler (MJ Research Inc., Watertown, Mass.), with ahot-start procedure. Fluorescently labeled primers 1281, 1254, and HLWL85 (1,14, 21) were used in three independent amplifications, and the resultant PCRproducts were detected on an ABI PRISM 310 DNA Genetic Analyzer (AppliedBiosystems, Naerum, Denmark) using the manufacturer’s recommendations. Inbrief, 12 ml of deionized formamide, 1 ml of size standard 2500-ROX, and 1 mlof each of the three PCR amplifications were mixed and denatured at 95°C for2 min and subsequently kept on ice until further processing. The ABI-310instrument was prepared with a short capillary (47 cm) and POP4 polymer (4%performance optimized polymer; Applied Biosystems). Running conditions wereas follows: injection time, 10 s; voltage, 15 kV; collection time, 45 min; electro-phoresis voltage, 15 kV; and heat plate temperature, 60°C.
The isolates were visually grouped according to combined profiles based oneach of the three primers. A capital letter after the type number indicates thatprofiles were similar with only minor differences in intensity and position ofbanding profile.
fla-DGGE typing. A 702-bp PCR fragment of the Campylobacter flaA gene (29)modified with a GC clamp attached to the 59 end of the reversed primer was usedto optimize the DGGE point mutation analysis (42). The PCR amplicon wasselected after testing 25 PCR systems for Campylobacter based on PCR frag-ments of 16S rRNA, 23S rRNA, flagellin genes flaA and flaB, and the VS1 gene(C.-H. Brogren, unpublished results). All four combinations of clamped systems(42) were tested. Based on band pattern and visual polymorphism, the chosenforward 20-mer primer (p1) was 59-TAC TAC AGG AGT TCA AGC TT-39, andthe 65-mer reversed primer (p4A) was 59-GCG GGC GGG GCG GGG GCACGG GGG GCG CGG CGG GCG GGG CGG GGG GTT GAT GTA ACTTGA TTT TG-39. Template DNA was obtained by boiling a washed suspensionof the isolate harvested from a brain heart infusion plate cultured under mi-croaerobic conditions for 48 h. The DNA template samples were stored in smallaliquots at 280°C. No improvement of the PCRs was observed after QiagenDNA purification, which was therefore omitted. Ready-to-Go PCR beads (Am-ersham Pharmacia Biotech AB) were used for all PCR amplifications withaddition of 5 pmol of each primer and 5 ml of template DNA (25 to 100 pg) ina final volume of 25 ml per reaction. PCR was performed within a Gene AMPSystem 2400 thermocycler (PE Biosystems, Foster City, Calif.) using the follow-ing protocol: 95°C for 5 min; cycling parameters at 95°C for 45 s, 60°C for 45 s,and 72°C for 45 s for 35 cycles; and final elongation at 72°C for 10 min. The finalamplicon (747 bp) was stored at 4°C until analyzed or kept at 220°C for long-term storage. PCR amplicons were size controlled by 1.5% (wt/vol) Sepharide(Gibco-BRL, Glasgow, Scotland) agarose gel electrophorese. Purity and amountof DNA were evaluated visually.
An estimate of melting behavior with or without the various GC clamps wasmade by Melt95 software (Ingeny, Leiden, The Netherlands) on the basis ofDNA sequences of this flaA fragment (GenBank, National Center for Biotech-nology Information, National Institutes of Health, Bethesda, Md.). The theoret-
VOL. 38, 2000 EVALUATION OF SUBTYPING METHODS FOR C. JEJUNI 3801
ical melting curves were compared with the curves experimentally made byperpendicular DGGE, and the denaturing gradient was designed according tothe melting temperature of the main melting domain (18). A 20 to 40% (vol/vol)urea-formamide gradient gel was prepared from a 100% gel stock solution (7 Murea and 400 ml of formamide per liter) with a 6% (wt/vol) polyacrylamide gel(35.5:1 acrylamide/bisacrylamide ratio) prepared in 13 Tris-acetate-EDTAbuffer, pH 8.3. A 1-mm gel (20 by 30 cm) was made by mixing 35 ml of 20%(vol/vol) urea-formamide solution with 35 ml of 40% (vol/vol) urea-formamidedenaturant solution in a linear gradient mixer and adding 382 ml of 10% (wt/vol)ammonium persulfate and 31 ml of N,N,N9,N9-tetramethylethylenediamine(TEMED) into each gel solution. Approximately 10 ml of samples was added toeach of 32 wells. The PCR amplicons and DNA ladder were diluted according toDNA content before mixing 8 ml with 2 ml of 53 gelloader containing bromo-phenol blue and xylene cyanol markers. The DGGE apparatus (2U-Phor; Ing-eny) was run at 60°C for either 17 h at 76 V or 4 to 5 h at 200 V. The gel wasstained with SYBR Green I for 30 min without destaining according to themanufacturer’s protocol (Molecular Probes, Eugene, Oreg.). The PCR-amplifiedhomoduplexes revealed a single band pattern with migration positions corre-sponding to melting behavior (low-melting-point homoduplexes are positionedat a short migration distance, and high-melting-point homoduplexes are posi-tioned at a longer migration distance). The band pattern was photographed andstored as a digital image (Kodak digital camera DC120). Kodak 1-D software wasused to scan the lanes, locate the bands, and compare band positions with markerlanes.
RESULTS
Discriminatory power and typeability. The results of usingsix typing methods on a collection of 80 C. jejuni strains with noknown relationships are presented in Table 1. Isolates groupedtogether by at least four of the six typing methods are markedwith boldface in the table. With the exception of serotyping, allstrains were assigned to types that had been defined arbitrarilyaccording to the individual criteria described above. In Table 2,the methods are evaluated with respect to discriminatory index(D index), number of types obtained, number of unique types,and prevalence of dominant type. Serotyping and fla-DGGE
TABLE 1. Typing of 80 C. jejuni isolates with no known relationa
a Boldface highlighting of source and typing data indicates that at least fourtyping methods agreed on the grouping. Boldface highlighting of the isolatenumber indicates that all six methods agreed on the grouping.
b DdeI/AluI profiles.c Control isolate also included in Table 3.d Outbreak isolate also included in Table 3.
typing were the least discriminatory methods, as they dividedthe 80 strains into 18 and 13 different types, respectively, withD indices of 0.868 and 0.896, respectively. Serotypes 2, 1/44,and 4 complex were the most common serotypes, a result
which is in agreement with the serotype distribution generallyseen for Danish isolates from these sources (27). PFGE andRAPD were the most discriminatory methods (D indices of0.974 and 0.984, respectively), each dividing the 80 strains intomore than 50 different types. RiboPrinting and fla-RFLP eachdivided the 80 strains into 40 types and resulted in D indices of0.945 and 0.960, respectively. In this study, all strains werefound to be typeable with each of the methods used. Repre-sentative pictures of profiles for each of the five genotypicmethods are presented in Fig. 1 to 5.
Outbreak isolates. All methods assigned the 11 epidemio-logically implicated isolates (nine clinical isolates and two wa-ter isolates) to the same type (Table 3). However, serotypingalso assigned this type to an isolate from one of the controlpatients originating from the same area. Furthermore, allmethods showed the existence of the epitype among unre-lated human isolates from other areas and isolates from othersources (Table 1).
Sources. For each method, the origin of the most commontypes in the collection of 80 unrelated strains is shown in Fig.6. The majority of these types were found in all three sources,though in different proportions. The dominant type in generalwas also the most common type among human and cattleisolates. Serotype 2 represented 29 and 40% of the human andcattle isolates, respectively, but only 5% (one isolate) of iso-lates from poultry. The same picture was seen for fla-DGGEtypes 3 and 6, RiboGroup 24, fla-RFLP type 1/1, and PFGEtypes 6 and 8. With the exception of the fla-DGGE types, thesegenotypes were mostly represented by isolates of serotype 2.
FIG. 1. PFGE profiles. Lane 1, isolate 5001; lane 2, isolate 4025; lane 3,isolate 5042; lane 4, isolate 5026; lane 5, isolate 733; lane 6, isolate 657; lane 7,isolate 4024. Lanes M, molecular size markers. The PFGE types are 8, 8, 3, 3, 37,35, and 42 for lanes 1 to 7, respectively.
FIG. 2. fla-RFLP profiles. (Left) DdeI restriction; (right) AluI restriction. Lane identities are as for Fig. 1. The fla-RFLP types are 1/1, 1/1, 1/1, 1/1, 7/7, 5/5, and24/24 for lanes 1 to 7, respectively.
VOL. 38, 2000 EVALUATION OF SUBTYPING METHODS FOR C. JEJUNI 3803
Serotype 1,44 was dominant in poultry (30%) but was repre-sented by only one cattle isolate (7%). Dominance in poultrywas also seen for some of the common genotypes: RiboGroup25, fla-RFLP type 7/7, PFGE type 12, and RAPD type 5.
Clonal groups and typing system concordance. By use of thecriterion that the agreement of strain groupings formed by four
or more of the methods indicated a close, probably clonalrelationship, the grouping is indicated in Table 1 by boldface.Thirteen groups were formed in this way, accounting for two toeight isolates each, and in total 38 of the 80 strains were partof such a group. One of the groups consisted of strains from allthree sources (Fig. 7). Isolates from both humans and cattlewere represented in four groups, isolates from humans andpoultry were in two groups, and the remaining six groups wererepresented by one source only (Fig. 7). With the criterion thatall six methods should agree on grouping the isolates, sevengroups of two isolates each were identified (parts of the groupsin Fig. 7; indicated in Table 1 by boldface for isolate number).In four of these groups, both isolates originated from the samesource (poultry or humans), but two pairs consisted of a cattleand a human isolate, and one pair consisted of a poultry and acattle isolate. Interestingly, all methods recognized a cattle iso-late (isolate 4025) as belonging to the epitype from the out-break, and thus this isolate formed an identical pair with theoutbreak strain that is included in Table 1 (human isolate 5001).
As a measure of typing system concordance, the boldfacegroups in Table 1 were also used as an indication of how oftena given method disagreed with the other methods. Of the 38isolates grouped in 13 types with at least four typing methods,
FIG. 3. RiboPrinting. Lane identities are as for Fig. 1 (vertical lanes). TheRiboGroups are 24, 24, 1, 23, 25, 33, and 27 for lanes 1 to 7, respectively.
FIG. 4. RAPD profiles. Lane identities are as for Fig. 1 (vertical lanes). The RAPD types are 1, 1, 2, 10, 5, 7, and 42A for lanes 1 to 7, respectively.
serotyping agreed with the grouping in all cases, whereas fla-DGGE disagreed in grouping 9 isolates, RiboPrinting dis-agreed for 3 isolates, fla-RFLP disagreed for 4 isolates, PFGEdisagreed for 13 isolates, and RAPD disagreed for 8 isolates.
The groups of isolates defined by each of the five genotypicmethods are shown as bars in Fig. 8. In addition, the figureshows the occurrence of different serotypes within these groupsand thereby gives an impression of the serotype variationwithin groups defined by the genotypic methods.
DISCUSSION
Validation of typing methods includes evaluation of theirperformance. Several performance criteria are essential, inparticular, typeability, reproducibility, stability, discriminatorypower, and typing system concordance (45). In the presentstudy, we have evaluated the typeability and discriminatorypower of six methods for typing C. jejuni: Penner serotyping,fla-DGGE, RiboPrinting, fla-RFLP, PFGE, and RAPD. Inaddition, the concordance of these methods was evaluated.The test population consisted of 80 C. jejuni strains that werepresumably unrelated epidemiologically and 11 strains relatedto an outbreak. The discriminatory power differed among thesix marker systems with D indices in the range of 0.868 to0.984. PFGE and RAPD were the most discriminatory meth-ods followed by RiboPrinting and fla-RFLP. Serotyping andfla-DGGE typing were the least discriminatory methods in the
study. All typing methods had a typeability of 100% on thiscollection of isolates. For serotyping, this typeability is higherthan expected but not unusual, as we generally find more than95% of Danish surveillance isolates from human patients, cat-tle, and broiler chickens typeable when using the full set ofunabsorbed antisera (E. M. Nielsen, unpublished data). A high-er proportion of nontypeable isolates has been reported in useof absorbed sera, e.g., 19% of Dutch poultry (16) and 21% ofisolates from clinical cases in the United Kingdom using amodified scheme (9).
Performance of RAPD and PFGE. RAPD and PFGE pro-filing are well recognized as highly discriminatory tools formolecular typing of a wide range of bacteria, including C. jejuni(10, 20, 34, 49). This is reaffirmed in the present study, wherePFGE and RAPD recognized 50 and 56 distinct profiles, re-spectively, among the 80 strains examined. The high discrimi-natory potential of PFGE and RAPD can be attributed to theirability to determine polymorphisms in the entire bacterial ge-nome.
In the RAPD analysis, isolates were visually grouped accord-ing to profiles based on three primers. The G1C content of the10-nucleotide primers, HLWL85, 1281, and 1254, was 50, 60,and 70%, respectively. For the closely related species Helico-bacter pylori (G1C content similar to that of C. jejuni [30 to33%]), it has been shown that 10-nucleotide primers with a 60or 70% G1C content gave better results than those with 50%G1C (3). This was not the case in our study, as HLWL85 and1254 most often produced more informative patterns than did1281 (Fig. 4). A major drawback of RAPD has been reportedto be its reproducibility (22). However, by using Ready-To-GoRAPD analysis beads followed by automated detection of frag-ments on a DNA sequencer, the number of susceptible stepshas largely been reduced (E. M. Nielsen, J. Engberg, and V.Fussing, unpublished data).
Performance of fla-RFLP and RiboPrinting. fla-RFLP andRiboPrinting were not as discriminatory as PFGE and RAPDbut still identified 40 different types each. Both methodsgrouped the isolates in generally good accordance with theother methods (Table 1). However, several RiboGroups weresubdivided by all other methods, e.g., the 15 isolates of Ribo-Group 23, the most common group, were of three differentserotypes (O:1,44, O:2, and O:4 complex) and eight differentfla-RFLP types. Some of the fla-RFLP types were also subdi-vided by all other methods, but 12 of the 13 isolates of type 1/1,the most common type, were serotype 2. In general, typingbased on the conserved ribosomal genes is considered a stabletyping method. This could be the reason why other typingmethods further divide some RiboGroups, e.g., 23 in this study.
Recently, the validity of fla-RFLP typing has been ques-tioned, due to the potential of the fla genes to undergo recom-bination events, thereby greatly changing the RFLP profile(13). On the other hand, several recent studies conclude thatfla-RFLP types can be linked to evolutionary genetic lineagesof Campylobacter spp. (41–43). In this study, fla typing cor-
FIG. 5. fla-DGGE profiles. Lane identities are as for Fig. 1. The fla-DGGEtypes are, from left to right, 2, 3, 3, 5, 5, 7, and 8, respectively. The position of asingle band is equal to a specific genotype. The DGGE type number is set low forthe fastest-migrating and high for the shortest-migrating band.
TABLE 3. Typing of C. jejuni isolates related to a waterborne outbreak
rectly identified the epitype isolates from the waterborne out-break and succeeded in grouping the strain collection in a waythat seemed reasonable compared to the results from the re-maining five typing tools, indicating that fla typing is overall areliable epidemiological marker for these isolates. However, itwas noted that two isolates (913 and 5025) that were groupedtogether by the other methods examined, and also by SalI-,KpnI-, and BamHI-based PFGE typing (30), were distinct byboth fla-RFLP and fla-DGGE analyses. This suggests thatthese isolates represent a single clone in which the flaA gene
has undergone some spontaneous genetic change. It was evi-dent that the combined use of DdeI and AluI enhanced thediscriminatory power of fla-RFLP typing. In all but one case(AluI profile types 14 and 14a), the AluI profiles that wereassociated with the same DdeI profile were highly similar,distinguished by one or two band differences. As it is impossi-ble to determine if one or two band differences are caused bymajor or minor sequence differences between the flaA genes inquestion, it is not meaningful to interpret similarity between
FIG. 6. For each typing system, the most common types (four or more isolates) are presented as percentages of isolates from each source (human patient, poultry,and cattle). (a) Serotyping; (b) fla-DGGE; (c) RiboPrinting; (d) fla-RFLP; (e) PFGE; (f) RAPD.
profiles as a close interstrain relationship but it is reasonable toregard each fla-RFLP type combination as a separate type.
Performance of serotyping and fla-DGGE. Serotyping wasthe least discriminatory method in this study, although a fairlyhigh D index of 0.868 was still attained. Serotyping was the bestprimary method in the sense that the other methods couldform the best hierarchic structure based on the serotyping, e.g.,only one of the RAPD groups was subdivided by serotyping(Fig. 8). Serotyping never disagreed on the grouping identifiedby at least four of the methods (Table 1). Though serotypingwas the least discriminatory method, this demonstrated thestability of the serotyping system—i.e., serotyping did not sep-arate strains that the genotypic methods grouped together. Inaccordance with other studies (12, 34), strains of serotypesO:1,44 and O:2 were found to be more homologous than werestrains of the O:4 complex, i.e., within serotypes O:1,44 andO:2 several large clonal groups of isolates were identified withthe genotypic methods, whereas none were found in the O:4complex. The use of absorbed antisera may have made it pos-sible to separate isolates of the O:4 complex into more sub-types. The use of absorbed sera for Penner serotyping of poul-try isolates revealed that only 4% of the isolates reacted withany of the antisera comprising the O:4 complex, and half ofthese reacted with O:13,50 (16). The other common complex,O:1,44, could not be separated in that study. A modified sero-typing system based on absorbed antisera and direct aggluti-nation was able to identify isolates with single reactions withthe O:4 complex antisera (the majority of these were serotype50) (9). However, as this modified scheme is not based onpassive hemagglutination, the results are not in complete con-cordance with the traditional scheme (A. N. Oza et al., Abstr.10th Int. Workshop CHRO, abstr. CE15, 1999). The value ofusing absorbed sera for the isolates in the present study istherefore difficult to estimate on the basis of these studies. Inaddition, it must be taken into account that the use of absorbedsera may reduce typeability.
fla-DGGE formed the lowest number of different types, butdue to a more even distribution of types, the D index wasslightly better than that for serotyping. Though fla-DGGE insome cases agreed on the grouping formed by the other meth-ods (Table 1), most of the groups formed by fla-DGGE weresubdivided by all other methods, including the less discrimina-tive serotyping (Fig. 8). DGGE analysis can be sensitive downto single base mutations (8, 18), but the relatively low discrim-inatory power of the present fla-DGGE method may be the
result of many different mutations in the flaA gene fragmentcounteracting each other in melting behavior. A gene with lesspolymorphism might be a better choice. Also, the large size ofthis fragment (747 bp) is not optimal for DGGE typing, whichworks best in the range of 200 to 400 bp. Further developmentof this new DGGE bacterial genotyping method will thereforeinvolve selection of a smaller and less polymorphic DNA frag-ment. Furthermore, the heteroduplex analysis procedure hasbeen shown in recent studies of Salmonella and Legionellatyping to greatly enhance the precision and discriminatorypower in nondenaturing assay systems (17, 39).
Typing system concordance. The more typing systems show-ing the same pattern, the better the predictability of relation-ships between isolates. In this study, the six typing systemspossess different discriminatory powers, which must be consid-ered in the evaluation and comparison of methods. When thegrouping of isolates formed by at least four typing systems wasused for evaluation of concordance of methods, the highlydiscriminatory PFGE most often disagreed with the othermethods, but also fla-DGGE had a high level of disagreementwhen its low discriminatory power was taken into account.Methods with a high level of agreement but different D indicesshow a hierarchic pattern, i.e., the highly discriminatory meth-od split the types formed by the low-discriminatory method,but not vice versa. The most discriminatory typing system,RAPD, showed a hierarchic structure with serotyping as theprimary system, as the majority of RAPD groups consisted ofisolates of only one serotype (Fig. 8). Several of the groupsformed by the other genotypic methods consisted of more thanone serotype (Fig. 8), showing that the markers of these typingsystems often are independent of the serotype. This is not sur-prising when a typing system is based on a single gene, e.g., thefla gene.
The most discriminatory methods, PFGE and RAPD, showedsome level of agreement in terms of strain differentiation andgrouping, but for about 40% of the isolates, the two methodsdisagreed. Both methods subdivided groups formed by theother method. Although both methods detect whole-genomepolymorphisms, the principles underlying each method arequite different and different genetic variations may be de-tected. It is well established that PFGE profiles of relatedstrains can be altered by a variety of genetic phenomena, in-cluding point mutations in restriction sites and genomic rear-rangements (31, 48). Such phenomena may account for thedifferentiation of RAPD groups by PFGE profiling, especiallywhere other markers are concordant with RAPD groupings(e.g., RAPD types 5, 7, 31, and 49). Furthermore, since thediscriminatory potential of PFGE is dependent upon the re-striction enzyme used, it is conceivable that the use of a more-frequent-cutting enzyme (e.g., KpnI) would further distinguishthe SmaI-based PFGE types that were subdivided by RAPD,thereby yielding equivalent results.
fla-DGGE is based on polymorphism on a smaller part ofthe flaA gene than the one used for fla-RFLP in this study.Though fla-DGGE and fla-RFLP are based on parts of thesame gene, they measure different parameters (melting pointof the whole amplicon versus position of restriction sites), andthis is likely to be the reason for the lack of correlation be-tween the two methods.
Identification of outbreak isolates and sources of sporadichuman infection. The 11 isolates related to a waterborne out-break were clearly identified by all six typing methods. Thetyping methods included in this study are thus sufficiently sta-ble to correctly group isolates of clonal origin, even though theisolates were sampled over a period of 2 1/2 months fromhuman diarrheal cases and from the contaminated water.
FIG. 7. Groups of isolates formed by at least four typing methods (boldfacegroups in Table 1). Origins of isolates are indicated. The groups are numberedaccording to positions in Table 1, starting from the top, i.e., group 1 includesisolates 913, 943, 5025, and 5130.
VOL. 38, 2000 EVALUATION OF SUBTYPING METHODS FOR C. JEJUNI 3807
The sporadic nature of human campylobacteriosis and theubiquitous distribution of the bacteria have traditionally hin-dered the unequivocal identification of sources of infection. Inour study, we found six groups each consisting of two suppos-edly unrelated strains where groupings from each of the typing
methods used were concordant. Three of the aforementionedgroups contained isolates from more than one source: twogroups comprised cattle and human isolates, and one groupcontained one isolate each from cattle and poultry (Table 1).The application of stringent criteria for strain identity haspreviously shown that isolates obtained from cattle and poultryare genetically similar to isolates from cases of human diarrhea(32). The agreement of six different phenotypic and genotypicmarkers, as described here, can similarly be said to be a strin-gent criterion for strain identity and thus provide further doc-umentation of the presence of strains from cattle in humanenteric disease. Although contaminated, undercooked poultrymeat is believed to be a significant vector of sporadically de-tected human disease (47), these data show that the impor-tance of other animal reservoirs such as cattle requires furtherstudy.
Applicability of typing systems. Depending on the nature ofthe bacterial species under investigation, more or less discrim-inatory methods are suitable for studying the epidemiology ofthe bacteria. Highly discriminatory methods or combinations
FIG. 8. Groups of isolates defined by each of the five genotypic methods.Numbers of isolates within each group are indicated on the y axis. Stacked barsshow the serotype distribution within each bar group. For the four most commonserotypes, the serotype is indicated by hatching of the bar, and the remainingserotypes are shown by serotype number in the bar.
of such are necessary for typing of clonal population, whereasstable and perhaps less discriminatory methods are necessaryfor typing panmictic populations in order to determine thecorrect relations between isolates. All typing systems evaluatedin this study were able to identify the outbreak isolates, andnone of the systems failed with respect to typeability and dis-criminatory power. However, the methods clearly showed dif-ferent discriminatory powers and different levels of agreementin identifying clonal lines. The methods can be recommendedfor different uses on the basis of the results of this comparativestudy, combined with considerations of the costs and laborassociated with the methods, as covered by recent reviews (30,49). As a definitive typing system, Penner serotyping proved tobe useful for typing of large numbers of isolates to obtain acoarse grouping of isolates and comparing the serotype distri-bution to other sources, other time periods, other countriesand regions, etc. Serotyping can be supplemented with morediscriminatory methods; e.g., serotyping can be used for aninitial screening of isolates and then isolates for further typingcan be selected. fla-DGGE showed a discriminatory power atthe same level as that of serotyping, but the method was notuseful as a primary selection method, as isolates in the groupsformed by the more discriminatory methods were generallyspread among several DGGE groups. The method needs to befurther developed and evaluated. fla-RFLP and RiboPrintingare both fairly discriminative and can be used for screeninghigh numbers of isolates. Furthermore, due to standardizationand automation, RiboPrinting can be regarded as a definitivetyping system. PFGE and RAPD are highly discriminatorymethods, based on the whole genome, and these methods aretherefore useful for ensuring genotypic similarity in cases ofoutbreaks.
Typing of bacterial isolates from different sources is a pre-requisite for intervention and infection control and to contrib-ute to risk assessment studies of sources of human campylobac-teriosis. A comparison of Campylobacter types from foodanimals and foods of animal origin with isolates from humansmakes it possible to produce estimates for the number ofhuman cases attributable to certain animal sources. The morelaboratories and countries involved in such surveillance, thebetter the knowledge of the global epidemiology of campy-lobacters that can be obtained. Standardization and harmoni-zation of typing methods between laboratories involved in suchstudies are of utmost importance. Harmonization of the geno-typic methods used in the present study (except fla-DGGE)has been initiated in a program of cooperation among sev-eral European laboratories (CAMPYNET; http://www.svs.dk/campynet). The applied method or methods must be definitiveto render results comparable over time, area, etc. Serotyping isthe only method that has been generally used for this purposein typing of Campylobacter. However, by the construction ofdatabases in a program suitable for the assimilation and anal-ysis of molecular fingerprints, our results indicate that themolecular methods used in the present setup may be applica-ble as definitive typing tools.
ACKNOWLEDGMENTS
We gratefully thank all the persons at the participating laboratorieswho contributed with technical assistance in this study.
REFERENCES
1. Akopianz, N., N. O. Bukanov, T. U. Westblom, S. Kresovich, and D. E. Berg.1992. DNA diversity among clinical isolates of Helicobacter pylori detected byPCR-based RAPD fingerprinting. Nucleic Acids Res. 20:5137–5142.
2. Anonymous. 2000. Annual report on zoonoses in Denmark 1999, p. 1–28.Danish Zoonosis Centre, Danish Veterinary Laboratory, Copenhagen, Den-mark.
3. Berg, D., N. S. Akopianz, and D. Kersulyte. 1994. Fingerprinting microbialgenomes using the RAPD or AP-PCR method. Methods Mol. Cell. Biol. 5:13–24.
4. Bolton, F. J., A. V. Holt, and D. N. Hutchinson. 1984. Campylobacter bio-typing scheme of epidemiological value. J. Clin. Pathol. 37:677–681.
5. Børresen-Dale, A. L., S. Lystad, and A. Langerød. 1998. Temporal temper-ature gradient gel electrophoresis (TTGE) compared with denaturing gra-dient gel electrophoresis (DGGE) and constant denaturant gel electro-phoresis (CDGE) in mutation screening. BioRadiation 99:12–13.
6. Brogren, C.-H., and L. Bergstrom. 1999. New bacterial genotyping methodsusing mutation analysis in denaturing slap-gel and capillary electrophoresis.Jpn. J. Electrophor. 43:83.
7. Engberg, J., P. Gerner-Smidt, F. Scheutz, E. M. Nielsen, S. L. W. On, and K.Mølbak. 1998. Water-borne Campylobacter jejuni infection in a Danishtown—a 6-week continuous source outbreak. Clin. Microbiol. Infect. 4:648–656.
8. Fisher, S. G., and L. S. Lerman. 1983. DNA fragments differing by singlebase-pair substitutions are separated in denaturing gradient gels: correspon-dence with theory. Proc. Natl. Acad. Sci. USA 80:1579–1583.
9. Frost, J. A., A. N. Oza, R. T. Thwaites, and B. Rowe. 1998. Serotyping schemefor Campylobacter jejuni and Campylobacter coli based on direct agglutina-tion of heat-stable antigens. J. Clin. Microbiol. 36:335–339.
10. Fujimoto, S., B. M. Allos, N. Misawa, C. M. Patton, and M. J. Blaser. 1997.Restriction fragment length polymorphism analysis and random amplifiedpolymorphic DNA analysis of Campylobacter jejuni strains isolated frompatients with Guillain-Barre syndrome. J. Infect. Dis. 176:1105–1108.
11. Gibson, J., K. Sutherland, and R. J. Owen. 1994. Inhibition of DNAseactivity in PFGE analysis of DNA from Campylobacter jejuni. Lett. Appl.Microbiol. 19:357–358.
12. Gibson, J. R., C. Fitzegerald, and R. J. Owen. 1995. Comparison of PFGE,ribotyping and phage-typing in the epidemiological analysis of Campy-lobacter jejuni serotype HS2 infections. Epidemiol. Infect. 115:215–225.
13. Harrington, C. S., F. M. Thomson-Carter, and P. E. Carter. 1997. Evidencefor recombination in the flagellin locus of Campylobacter jejuni: implicationsfor the flagellin gene typing scheme. J. Clin. Microbiol. 35:2386–2392.
14. Hilton, A. C., D. Mortiboy, J. F. Banks, and C. W. Penn. 1997. RAPDanalysis of environmental, food and clinical isolates of Campylobacter spp.FEMS Immunol. Med. Microbiol. 18:199–124.
15. Hunter, P. R., and M. A. Gaston. 1988. Numerical index of the discrimina-tory ability of typing systems: an application of Simpson’s index of diversity.J. Clin. Microbiol. 26:2465–2466.
16. Jacobs-Reitsma, W. F., H. M. E. Maas, and W. H. Jansen. 1995. Pennerserotyping of Campylobacter isolates from poultry, with absorbed pooledantisera. J. Appl. Bacteriol. 79:286–291.
17. Jensen, M. A., and R. J. Hubner. 1996. Use of homoduplex ribosomal DNAspacer amplification products and heteroduplex cross-hybridization productsin the identification of Salmonella serovars. Appl. Environ. Microbiol. 62:2741–2746.
18. Lerman, L. S., and C. Beldjord. 1999. Comprehensive mutation detectionwith denaturing gradient gel electrophoresis, p. 35–61. In R. G. H. Cotton,E. Edkins, and S. Forrest (ed.), Mutation detection. Oxford University PressInc., New York, N.Y.
19. Lior, H., D. L. Woodward, J. A. Edgar, L. J. LaRoche, and P. Gill. 1982.Serotyping of Campylobacter jejuni by slide agglutination based on heat-labile antigenic factors. J. Clin. Microbiol. 15:761–768.
20. Madden, R. H., L. Moran, and P. Scates. 1998. Frequency of occurrence ofCampylobacter spp. in red meats and poultry in Northern Ireland and theirsubsequent subtyping using polymerase chain reaction-restriction fragmentlength polymorphism and the random amplified polymorphic DNA method.J. Appl. Microbiol. 84:703–708.
21. Mazurier, S., A. van de Giessen, K. Heuvelman, and K. Wernars. 1992.RAPD analysis of Campylobacter isolates: DNA fingerprinting without theneed to purify DNA. Lett. Appl. Microbiol. 14:260–262.
22. Meunier, J., and P. A. Grimont. 1993. Factors affecting reproducibility ofrandom amplified polymorphic DNA fingerprinting. Res. Microbiol. 144:373–379.
23. Muyer, G., E. C. de Waal, and A. G. Uitterlinden. 1993. Profiling of complexmicrobial populations by denaturing gradient gel electrophoresis analysis ofpolymerase chain reaction-amplified genes coding for 16S rRNA. Appl.Environ. Microbiol. 59:695–700.
24. Nachamkin, I., K. Bohachick, and C. M. Patton. 1993. Flagellin gene typingof Campylobacter jejuni by restriction fragment length polymorphism analy-sis. J. Clin. Microbiol. 31:1531–1536.
25. Nachamkin, I., H. Ung, and C. M. Patton. 1996. Analysis of HL and Oserotypes of Campylobacter strains by flagellin gene typing system. J. Clin.Microbiol. 34:277–281.
26. Newell, D. G., and S. L. W. On. 1998. Speciation and subtyping of campy-lobacters, helicobacters and related organisms. A survey of current applica-tions, p. 213–215. In A. Lastovica, D. G. Newell, and E. E. Lastovica (ed.),Proceedings of the 9th International Workshop on Campylobacter, Helico-bacter and Related Organisms. Institute of Child Health, University of CapeTown, Cape Town, South Africa.
VOL. 38, 2000 EVALUATION OF SUBTYPING METHODS FOR C. JEJUNI 3809
27. Nielsen, E. M., J. Engberg, and M. Madsen. 1997. Distribution of serotypesof Campylobacter jejuni and C. coli from Danish patients, poultry, cattle andswine. FEMS Immunol. Med. Microbiol. 19:47–56.
28. Nielsen, E. M., and N. L. Nielsen. 1999. Serotypes and typability of Campy-lobacter jejuni and Campylobacter coli isolated from poultry products. Int. J.Food Microbiol. 46:199–205.
29. Nishimura, M., M. Nukima, J. M. Yuan, B. O. Shen, J. J. Ma, M. Shta, T.Saida, and T. Uchiyama. 1996. PCR-based restriction fragment length poly-morphism (RFLP) analysis and serotyping of Campylobacter jejuni isolatesfrom diarrheic patients in China and Japan. FEMS Microbiol. Lett. 142:133–138.
30. Olive, D. M., and P. Bean. 1999. Principles and applications of methods forDNA-based typing of microbial organisms. J. Clin. Microbiol. 37:1661–1669.
31. On, S. L. W. 1998. In vitro genotypic variation of Campylobacter coli docu-mented by pulsed-field gel electrophoretic DNA profiling: implications forepidemiological studies. FEMS Microbiol. Lett. 165:341–346.
32. On, S. L. W., E. M. Nielsen, J. Engberg, and M. Madsen. 1998. Validity ofSmaI-defined genotypes of Campylobacter jejuni examined by SalI, KpnI, andBamHI polymorphisms: evidence of identical clones infecting humans, poul-try, and cattle. Epidemiol. Infect. 120:231–237.
33. On, S. L. W., and P. Vandamme. 1997. Identification and epidemiologicaltyping of Campylobacter hyointestinalis subspecies by phenotypic and geno-typic methods and description of novel subgroups. Syst. Appl. Microbiol. 20:238–247.
34. Owen, R. J., K. Sutherland, C. Fitzgerald, J. Gibson, P. Borman, and J.Stanley. 1995. Molecular subtyping scheme for serotypes HS1 and HS4 ofCampylobacter jejuni. J. Clin. Microbiol. 33:872–877.
35. Patton, C. M., and I. K. Wachsmuth. 1992. Typing schemes: are currentmethods useful?, p. 110–128. In I. Nachamkin, M. J. Blaser, and L. S.Tompkins (ed.), Campylobacter jejuni: current status and future trends.American Society for Microbiology, Washington, D.C.
36. Patton, C. M., I. K. Wachsmuth, G. M. Evins, J. A. Kiehlbauch, B. D.Plikaytis, N. Troup, L. Tompkins, and H. Lior. 1991. Evaluation of 10methods to distinguish epidemic-associated Campylobacter strains. J. Clin.Microbiol. 29:680–688.
37. Penner, J. L., and J. N. Hennessey. 1980. Passive hemagglutination tech-nique for serotyping Campylobacter fetus subsp. jejuni on the basis of solubleheat-stable antigens. J. Clin. Microbiol. 12:732–737.
38. Penner, J. L., J. N. Hennessey, and R. V. Congi. 1983. Serotyping of Campy-
lobacter jejuni and Campylobacter coli on the basis of thermostable antigens.Eur. J. Clin. Microbiol. 2:378–383.
39. Pinar, A., S. Ajkee, R. D. Miller, J. A. Ramirez, and J. T. Summersgill. 1997.Use of heteroduplex analysis to classify legionellae on the basis of 5S rRNAgene sequences. J. Clin. Microbiol. 35:1607–1611.
40. Salama, S. M., F. J. Bolton, and D. N. Hutchinson. 1990. Application of anew phagetyping scheme to campylobacters isolated during outbreaks. Epi-demiol. Infect. 104:405–411.
41. Santesteban, E., J. Gibson, and R. J. Owen. 1996. Flagellin gene profiling ofCampylobacter jejuni heat-stable serotype 1 and 4 complex. Res. Microbiol.147:641–649.
42. Sheffield, V. C., D. R. Cox, L. S. Lerman, and R. M. Myers. 1993. Attachmentof a 40-base-pair G1C-rich sequence (GC-clamp) to genomic DNA frag-ments by the polymerase chain reaction results in improved detection ofsingle-base changes. Proc. Natl. Acad. Sci. USA 86:232–236.
43. Slater, E., and R. J. Owen. 1998. Subtyping of Campylobacter jejuni Pennerheat-stable (HS) serotype 11 isolates from human infections. J. Med. Micro-biol. 47:353–357.
44. Steele, M., B. McNab, L. Fruhner, S. DeGrandis, D. Woodward, and J. A.Odumeru. 1998. Epidemiological typing of Campylobacter isolates from meatprocessing plants by pulsed-field gel electrophoresis, fatty acid profile typing,serotyping, and biotyping. Appl. Environ. Microbiol. 64:2346–2349.
45. Struelens, M. J., and the Members of the European Study Group on Epi-demiological Markers of the European Society for Clinical Microbiology andInfectious Diseases. 1996. Consensus guidelines for appropriate use andevaluation of microbial epidemiologic typing systems. Clin. Microbiol. Infect.2:2–11.
46. Tamouza, R., P. Marzais, R. Krishnamoorthy, E. Gordien, C. Besmond, C.Raffoux, and D. Charron. 1996. HLA B44 subtyping by DGGE: a sensitiveand rapid new approach to HLA class I typing. Hum. Immunol. 47:43.
47. Tauxe, R. V. 1992. Epidemiology of Campylobacter jejuni infections in theUnited States and other industrialized nations, p. 9–19. In I. Nachamkin,M. J. Blaser, and L. S. Tompkins (ed.), Campylobacter jejuni: current statusand future trends. American Society for Microbiology, Washington, D.C.
48. Wassenaar, T. M., B. Geilhausen, and D. G. Newell. 1998. Evidence ofgenomic instability in Campylobacter jejuni isolated from poultry. Appl. En-viron. Microbiol. 64:1816–1821.
49. Wassenaar, T. M., and D. G. Newell. 2000. Genotyping of Campylobacterspp. Appl. Environ. Microbiol. 66:1–9.
