as new tools for inter- and intra-species differentiation within Bordetella

Transcription

1 Journal of Medical Microbiology (2005), 54, DOI /jmm Amplified fragment length polymorphism (AFLP) versus randomly amplified polymorphic DNA (RAPD) as new tools for inter- and intra-species differentiation within Bordetella Anna Gzyl, 1 Ewa Augustynowicz, 1 Ewa Mosiej, 2 Monika Zawadka, 1 Grzegorz Gniadek, 1 Aneta Nowaczek 1 and Janusz Slusarczyk 1 Correspondence Anna Gzyl agzyl@pzh.gov.pl 1 Department of Sera and Vaccine Evaluation, National Institute of Hygiene, 24 Chocimska Str., Warsaw, Poland 2 Interfaculty Studies of Biotechnology, Warsaw Agricultural University, 159 Nowoursynowska Str., Warsaw, Poland Received 5 April 2004 Accepted 19 November 2004 Automated amplified fragment length polymorphism (AFLP) and randomly amplified polymorphic DNA (RAPD) techniques with fluorescently labelled primers were used to track differences among isolates of the eight known species of the Bordetella genus. Eighty-one representative strains of these species from international and Polish bacterial collections were genotyped according to RAPD protocols using primer 1254 or 1247, and AFLP involving EcoRI/MseI or newly designed SpeI/ApaI restriction/ligation/amplification procedures. By comparing AFLP and RAPD data, it was concluded that the discriminatory power of AFLP is higher in comparison with RAPD for both intraand inter-species differentiation of isolates of the Bordetella genus. The most precise level of interspecies discrimination and the highest level of intra-species discrimination of the Bordetella isolates of the eight species were observed in the AFLP EcoRI/MseI and SpeI/ApaI sets, respectively. Both techniques might provide alternative tools for the identification of Bordetella at the genomic species and strain levels, and thus may be valuable in human and veterinary diagnostics as well as in epidemiology. By applying the AFLP technique presented in this article, more precise data on the emergence of newly acquired and/or on expanded clones and transmission routes of isolates of the Bordetella genus in the human and animal environments might be obtained. INTRODUCTION Currently, three of the eight species of the Bordetella genus that have been described, Bordetella pertussis, Bordetella parapertussis and Bordetella bronchiseptica, present the highest level of genetic relatedness and are associated with infections of the upper respiratory tract of many mammals (Gerlach et al., 2001; Khattak & Matthews, 1993; Musser et al., 1986; van der Zee et al., 1997). B. pertussis is an obligate pathogen of humans and in susceptible individuals induces highly contagious disease so-called whooping cough that still causes over deaths around the world annually (World Health Organization, 2002). B. parapertussis is capable of inducing typical and mild forms of pertussis in Abbreviations: AFLP, amplified fragment length polymorphism; IS, insertion sequence; MLEE, multilocus enzyme electrophoresis; MLST, multilocus sequence typing; RAPD, randomly amplified polymorphic DNA; REA, restriction enzyme analysis; UPGMA, unweighted pair group method using arithmetic averages. humans and has also been recognized as a source of asymptomatic and symptomatic infections of the respiratory tract in ovine animals (Gerlach et al., 2001; Porter et al., 1996). B. bronchiseptica is a common commensal or pathogen in the respiratory tract of several mammalian species. Respiratory or systemic B. bronchiseptica infections in humans have occasionally been described (Dworkin et al., 1999; Woolfrey & Moody, 1991). Most of the B. bronchiseptica human infections described in the literature (septicaemia, pneumonia, pertussis-like illness, tracheobronchitis, sinusitis, peritonitis and meningitis) occurred in immunosuppressed patients (Dworkin et al., 1999) or in individuals with other underlying conditions (Sacco et al., 2000a). Zoonoses from pets might be involved, since B. bronchiseptica is known to produce infection and disease in many mammals (e.g. kennel cough in dogs and atrophic rhinitis in pigs) including those kept in the home (Dworkin et al., 1999; Woolfrey & Moody, 1991). These three species can be differentiated by phenotype. However, the genetic diversity shown with many techniques & 2005 SGM Printed in Great Britain IP:

2 A. Gzyl and others [DNA DNA hybridization, multilocus enzyme electrophoresis (MLEE), and 16S and 23S rrna gene typing] has been found to be limited (Gerlach et al., 2001; Musser et al., 1986; van der Zee et al., 1997). Thus they are commonly considered as members of a single B. bronchiseptica cluster (Gerlach et al., 2001). Bordetella avium colonizes mainly the respiratory tract of birds, inducing bird bordetellosis and turkey coryza (Gerlach et al., 2001). In poultry, B. avium infections result in coryza or rhinotracheitis, highly contagious diseases resulting in economic losses to the turkey/poultry industry (Register et al., 2003). Bordetella hinzii found in birds has recently been also associated with human infections seen in patients with AIDS or cystic fibrosis (Cookson et al., 1994; Dworkin et al., 1999; Funke et al., 1996; Kattar et al., 2000; Vandamme et al., 1995). Bordetella holmesii has been recovered from patients with underlying disorders, including Hodgkin lymphoma, sicklecell anaemia, pulmonary disease and asplenia, septicaemia and endocarditis (Russell et al., 2001; Tang et al., 1998; Yih et al., 1999). Surprisingly, in the last few years it has also been isolated from the nasopharynx of patients with pertussis-like symptoms (Mazengia et al., 2000). Bordetella trematum, commonly isolated in humans from wounds or ears but not from the respiratory tract, is a species of unknown pathogenic potential (Gerlach et al., 2001). Bordetella petrii, the first anaerobic environmental isolate of the Bordetella genus, was described in 2001 and still only a single strain is available (von Wintzingerode et al., 2001). Tracking epidemiological routes is extremely important in the case of B. pertussis since it still affects large proportions of individuals and infection rates are rising unexpectedly in highly vaccinated populations (Mooi et al., 2000; van der Zee et al., 1996a). Similarly, studies of B. parapertussis transmission and co-infection with B. pertussis might elucidate many aspects of the epidemiology of whooping cough. Tracking differences among strains of B. bronchiseptica, B. avium, B. hinzii or B. holmesii as opportunistic infections of humans, together with differences found in strains isolated from natural animal hosts, might be important for human and animal microbiology. Several studies have proved the usefulness of molecular tools in differentiating particular species of the Bordetella genus. Some of them were able to track genetic differences among/ within two to three species simultaneously. For several years, amplified fragment length polymorphism (AFLP) and randomly amplified polymorphic DNA (RAPD) have been found to be very useful in studying the epidemiology of many pathogens, e.g. in order to follow transmission routes (van der Zwet et al., 1999) or to study phylogenetic relationships between organisms (Torriani et al., 2001). In this study we believe we present the first report on the application of AFLP for genotyping isolates within the Bordetella genus. The discriminatory power of AFLP fingerprinting has been compared with RAPD performed with protocols previously applied by Yuk et al. (1998) for B. parapertussis. In addition, we describe a newly designed SpeI/ApaI restriction enzymes/ adapters/primers protocol not used previously for AFLP fingerprinting. Our preliminary study was aimed at investigating the possibility of Bordetella inter- and intra-species genotyping and evaluating the taxonomic potential of AFLP and RAPD in the delineation of species-specific identification. METHODS Bacterial strains and culture media. A total of 81 representative isolates belonging to the Bordetella genus were investigated. The number, origin and host species of all the isolates tested are presented in Table 1. Thirty-five B. pertussis strains collected by the Department of Sera and Vaccine Evaluation of the National Institute of Hygiene, Warsaw, Poland were analysed. Among them, 27 strains were isolated from non-epidemiologically related cases of pertussis that occurred in in Poland. The other isolates of B. pertussis used in the study originated from Hungary (two), the former Soviet Union (two), the former Yugoslavia (one) and Finland (one). Three international strains Tohama I and W28 (both received from Dr N. Guiso, Institut Pasteur, Paris) and (Kendrick) were also studied. Isolates of B. avium (nine), B. hinzii (four) and B. trematum (five) came from the BCCG/LMG Collection, Gent, Belgium. Among the 13 isolates of B. bronchiseptica, nine came from the BCCG/LMG Collection and one from the Institute of Immunology and Experimental Therapy in Wroclaw, Poland. Two B. bronchiseptica strains were obtained from Dr N. Guiso (Institut Pasteur, Paris). Regarding the B. holmesii strains analysed, three isolates came from the BCCG/LMG Collection and two were kindly obtained again from Dr N. Guiso (Institut Pasteur, Paris). Five and four isolates of B. parapertussis originated from the BCGG/ LMG Collection and the parental Bacterial Institute Collection, NIH, Warsaw, Poland. B. petrii was kindly received from Professor R. Gross (Universität Wurzburg, Germany). B. pertussis strains were cultured on Bordet Gengou agar (Difco) supplemented with 15 % sheep blood and incubated at 35 8C for 2 3 days. The B. petrii strain was grown anaerobically on LB agar at 30 8C for 48 h (von Wintzingerode et al., 2001). Isolation of total DNA. DNA from 81 strains belonging to the Bordetella genus was isolated by means of a commercially available Qiagen DNA isolation kit. The concentration of the DNA was measured by electrophoresis of samples on 1 % agarose gel against diluted preparations of º phage DNA of known concentrations. Typing by the RAPD method. Several different short primers designed by Williams et al. (1990) were screened for fingerprinting efficiency of different isolates of the eight Bordetella species. For screening, five different strains of each species were used where possible. The two primers with the best discriminatory potential were chosen to study all of the strains collected: 1247, 59-AAGAGCCCGT-39, and 1254, 59- CCCGTCAGCA-39. Forty-two cycles of denaturation at 94 8C for 1 min, annealing at 36 8C for 1 min and extension at 72 8C for 1 min were applied in the Biometra thermal cycler. Aliquots of amplified PCR products were electrophoresed in 1.5 % agarose gels stained with ethidium bromide. Electrophoresis was carried out in TAE buffer (40 mm Tris/acetate, 1 mm EDTA). Gels were photographed with the Gel Doc 1000 Gel Documentation System (Bio-Rad). Stored patterns were analysed with the GelCompar Software version 3.1 (Applied Maths) after conversion of the data into TIF format. Electrophoresis gels were normalized according to the external standards contained in each fifth lane (1 kb ladder, Gibco). Dendrograms for cluster analysis were based on similarity matrices 334 IP: Journal of Medical Microbiology 54

3 AFLP in genotyping of isolates of Bordetella genus Table 1. Designation, origin and RAPD and AFLP profile types of the Bordetella strains in the study. No. Species Strain/origin Host Host collection Cluster RAPD 1247 RAPD 1254 AFLP EcoRI/ MseI AFLP SpeI/ ApaI 1 B. pertussis 324, Hungary Human NIH, Warsaw, Poland RA AIIa 2 B. pertussis 2896, Hungary Human NIH, Warsaw, Poland RA AIIa 3 B. pertussis 353, former Soviet Union Human NIH, Warsaw, Poland RA AIIa 4 B. pertussis 5375, former Soviet Union Human NIH, Warsaw, Poland RA AIIa 5 B. pertussis Tohama I, Japan Human CIP, The Institut Pasteur, Paris, France RA AIIa 6 B. pertussis 60623/67, former Yugoslavia Human NIH, Warsaw, Poland RA AIIa 7 B. pertussis W 28, England Human CIP, The Institut Pasteur, Paris, France RA AIIa 8 B. pertussis 18530, Finland Human NIH, Warsaw, Poland RA AIIa 9 B. pertussis 3/60, Poland Human NIH, Warsaw, Poland RA AIIa 10 B. pertussis 7/60, Poland Human NIH, Warsaw, Poland RA AIIa 11 B. pertussis 12/60, Poland Human NIH, Warsaw, Poland RA AIIa 12 B. pertussis 2/62, Poland Human NIH, Warsaw, Poland RA AIIa 13 B. pertussis 6/62, Poland Human NIH, Warsaw, Poland RA AIIa 14 B. pertussis A/63, Poland Human NIH, Warsaw, Poland RA AIIa 15 B. pertussis 253/67, Poland Human NIH, Warsaw, Poland RA AIIa 16 B. pertussis 2/68, Poland Human NIH, Warsaw, Poland RA AIIa 17 B. pertussis 9/68, Poland Human NIH, Warsaw, Poland RA AIIa 18 B. pertussis 45/68, Poland Human NIH, Warsaw, Poland RA AIIa 19 B. pertussis 1253/73, Poland Human NIH, Warsaw, Poland RA AIIa 20 B. pertussis 6635/73, Poland Human NIH, Warsaw, Poland RA AIIa 21 B. pertussis 6770/73, Poland Human NIH, Warsaw, Poland RA AIIa 22 B. pertussis 6901/73, Poland Human NIH, Warsaw, Poland RA AIIa 23 B. pertussis 2955/74, Poland Human NIH, Warsaw, Poland RA AIIa 24 B. pertussis 3835/74, Poland Human NIH, Warsaw, Poland RA AIIa 25 B. pertussis 7097/74, Poland Human NIH, Warsaw, Poland RA AIIa 26 B. pertussis 2066/95, Poland Human NIH, Warsaw, Poland RA AIIa 27 B. pertussis 2214/95, Poland Human NIH, Warsaw, Poland RA AIIa 28 B. pertussis 2245/95, Poland Human NIH, Warsaw, Poland RA AIIa 29 B. pertussis 3599/97, Poland Human NIH, Warsaw, Poland RA AIIa a 30 B. pertussis 1038/99, Poland Human NIH, Warsaw, Poland RA AIIa 31 B. pertussis 785/00, Poland Human NIH, Warsaw, Poland RA AIIa 32 B. pertussis 871/00, Poland Human NIH, Warsaw, Poland RA AIIa 33 B. pertussis 1468/00, Poland Human NIH, Warsaw, Poland RA AIIa 34 B. pertussis 2632/00, Poland Human NIH, Warsaw, Poland RA AIIa 35 B. pertussis (Kendrick) Human NIH, Warsaw, Poland RA AIIa 36 B. parapertussis 1816 Human BCCM/LMG, Gent, Belgium RA AIIa 37 B. parapertussis 1820 Human BCCM/LMG, Gent, Belgium RA AIIa 38 B. parapertussis 1821 Human BCCM/LMG, Gent, Belgium RA AIIa 39 B. parapertussis 1823 Human BCCM/LMG, Gent, Belgium RA AIIa 40 B. parapertussis 1829 Human BCCM/LMG, Gent, Belgium RA AIIa 41 B. parapertussis 1251/70, Poland Human NIH, Warsaw, Poland RA AII E5d 42 B. parapertussis 1356/75, Poland Human NIH, Warsaw, Poland RA AII E5d 43 B. parapertussis 762/00, Poland Human NIH, Warsaw, Poland RA AII E5d 44 B. parapertussis 46/00, Poland Human NIH, Warsaw, Poland RA AII E5d 45 B. bronchiseptica 3525, Denmark Rabbit BCCM/LMG, Gent, Belgium RC AI S4a 46 B. bronchiseptica 3530 Monkey BCCM/LMG, Gent, Belgium RC AI S4a 47 B. bronchiseptica 3533, Sweden Horse BCCM/LMG, Gent, Belgium RC AI S4a 48 B. bronchiseptica 3534, Sweden Human BCCM/LMG, Gent, Belgium RC AI 49 B. bronchiseptica 3536 Rat BCCM/LMG, Gent, Belgium RC AIIb S4c 50 B. bronchiseptica 3537 Piglet BCCM/LMG, Gent, Belgium RC AI E5a S4a IP:

4 A. Gzyl and others Table 1. cont. No. Species Strain/origin Host Host collection Cluster RAPD 1247 RAPD 1254 AFLP EcoRI/ MseI AFLP SpeI/ ApaI 51 B. bronchiseptica 3538 Dog BCCM/LMG, Gent, Belgium RC AIIc a 52 B. bronchiseptica 3540 Cat BCCM/LMG, Gent, Belgium RA AI E5a 53 B. bronchiseptica 3543 Guinea pig BCCM/LMG, Gent, Belgium RA AI E5a S4c 54 B. bronchiseptica 1033, Czech Republic Unknown NIH, Warsaw, Poland RA AI S4a 55 B. bronchiseptica Unknown CIP, The Institut Pasteur, Paris, France RA AI S4a 56 B. bronchiseptica Unknown CIP, The Institut Pasteur, Paris, France RA AI S4a 57 B. bronchiseptica 1943 Unknown The Institute of Immunology and Experimental Therapy, Wroclaw, Poland RA AI S4c 58 B. avium 1851, Germany Turkey BCCM/LMG, Gent, Belgium RBa B 59 B. avium 1854, USA Turkey BCCM/LMG, Gent, Belgium RBa B 60 B. avium 1855, Spain Turkey BCCM/LMG, Gent, Belgium RBa B 61 B. avium 1856, Israel Turkey BCCM/LMG, Gent, Belgium RBa B 62 B. avium 3522, Germany Turkey BCCM/LMG, Gent, Belgium RBa B 63 B. avium 3549, UK Turkey BCCM/LMG, Gent, Belgium RBa B 64 B. avium 3556, Germany Goose BCCM/LMG, Gent, Belgium RBa B 65 B. avium 3557, Germany Duck BCCM/LMG, Gent, Belgium RBa B 66 B. avium 10973, Belgium Broiler BCCM/LMG, Gent, Belgium RBa B 67 B. hinzii 10980, Belgium Chicken BCCM/LMG, Gent, Belgium RBb AIIc E4 S2 68 B. hinzii 13494, Australia Chicken BCCM/LMG, Gent, Belgium RBd AIIc E4 S6 69 B. hinzii 14052, USA Human BCCM/LMG, Gent, Belgium RBd AIIa E4 S2 70 B. hinzii 16211, Belgium Human BCCM/LMG, Gent, Belgium RBd AIIc E4 S2 71 B. holmesii 15047, USA Human BCCM/LMG, Gent, Belgium RBc AIIa E3 S5 72 B. holmesii 15945, USA Human BCCM/LMG, Gent, Belgium RBc AIIa E3 S5 73 B. holmesii 15946, USA Human BCCM/LMG, Gent, Belgium RBc AIIa E3 S5 74 B. holmesii Unknown CIP, The Institut Pasteur, Paris, France RBc AIIa E3 S5 75 B. holmesii Unknown CIP, The Institut Pasteur, Paris, France RBc AIIa E3 S5 76 B. trematum 5894 Human BCCM/LMG, Gent, Belgium RBb AI E2 S3 77 B. trematum Human BCCM/LMG, Gent, Belgium RBb AIII E2 S3 78 B. trematum Human BCCM/LMG, Gent, Belgium RBe AIIb E2 S6 79 B. trematum 16652, Switzerland Human BCCM/LMG, Gent, Belgium RBe AI E2 S6 80 B. trematum Human BCCM/LMG, Gent, Belgium RBe AI E2 S6 81 B. petrii 4/03 Isolated from an anaerobic bioreactor Universität Würzburg, Germany RBa AIId E6 S7 calculated from the Pearson product-moment correlation coefficient and the unweighted pair group method using arithmetic averages (UPGMA) algorithm (Everitt, 1993). Typing by the AFLP method. In the preliminary screening DNA from, where possible, five isolates from each of the Bordetella species was used to evaluate the potential to obtain distinctive AFLP patterns. Six different previously described AFLP protocols, using HindIII/TagI, ApaI/TagI, MfeI/BglII, HindIII, PstI/TaqI and EcoRI/MseI, based on different enzyme restriction/specific adapter ligation and primerspecific amplification with/without selective bases complementary to nucleotides flanking the restriction sites, were tested (Grady et al., 1999; Huys et al., 1996; Kokotovic et al., 1999; McLauchlin et al., 2000; van Elderle et al., 1999; Vos & Kuiper, 1998). Additionally, new AFLP sets with SpeI restriction and the frequent-cutter enzymes BglII, HindIII, PstI, ApaI and EcoRI, coupled with ligation by newly designed SpeI adapters and primers, were tested (Table 2). Briefly, in all tested AFLP sets, 200 ng of Bordetella DNA was digested with the following restriction enzymes according to manufacturer s recommendations: HindIII (BioLabs)/TagI (Gibco-BRL); ApaI (Bio- Labs)/TagI; HindIII; MfeI (BioLabs)/BglII (BioLabs); PstI (Eurogentec)/ TagI; EcoRI (Gibco-BRL)/MseI (BioLabs); SpeI (BioLabs)/BglII; SpeI/ HindIII; SpeI/PstI; SpeI/ApaI; and SpeI/EcoRI. Samples were heattreated to inactivate restriction enzymes. Next, adapters, synthesized by Invitrogen, were ligated to the restriction DNA fragments by T4 DNA ligase (BioLabs) according to manufacturer s recommendations. For rare-cutting enzymes, adapters were used at a concentration of 5 pmol ìl 1, and for frequent-cutting enzymes, at 50 pmol ìl 1. After completion of ligation, samples were heat-inactivated, 10-fold diluted in distilled water and refrigerated at 20 8C until the start of PCR. 336 IP: Journal of Medical Microbiology 54

5 AFLP in genotyping of isolates of Bordetella genus Table 2. Characteristics of the AFLP sets AFLP set Strains studied* Reference or adaptor/primer sequence HindIII/TagI 5 van Elderle et al. (1999) ApaI/TagI 5 Grady et al. (1999) ApaI/TagI 5 Huys et al. (1996) HindIII 5 McLauchlin et al. (2000) MfeI/BgllI 5 Kokotovic et al. (1999) PstI/TagI 5 Vos & Kuiper (1998) EcoRI/MseI 81 Vos et al. (1998) Adaptor sequence Cy5-labelled primer Unlabelled primer (59 39) SpeI/ApaI 81 AD1 SpeI 59-CTCGTAGACTGCGTACC-39 AD2 SpeI 59-CTAGGGTACGCAGTCTAC-39 AD1 ApaI 59-CATCTGACGCATGT-39 AD2 ApaI 59-TCGTAGACTGCGTACAGGCC-39 SpeI/BglII 5 AD1 SpeI 59-CTCGTAGACTGCGTACC-39 AD2 SpeI 59-CTAGGGTACGCAGTCTAC-39 AD1 BglII 59-CGGACTAGAGTACACTGTC-39 AD2 BglII 59-GATCGACAGTGTACTCTAGTC-39 SpeI/HindIII 5 AD1 SpeI 59-CTCGTAGACTGCGTACC-39 AD2 SpeI 59-CTAGGGTACGCAGTCTAC-39 AD1 HindIII 59-CTCGTAGACTGCGTACC-39 AD2 HindIII 59-AGCTGGTACGCAGTC-39 SpeI/PstI 5 AD1 SpeI 59-CTCGTAGACTGCGTACC-39 AD2 SpeI 59-CTAGGGTACGCAGTCTAC-39 AD1 PstI 59-CTCGTAGACTGCGTACATGCA-39 AD2 PstI 59-TGTACGCAGTCTAC-39 SpeI/EcoRI 5 AD1 SpeI 59-CTCGTAGACTGCGTACC-39 AD2 SpeI 59-CTAGGGTACGCAGTCTAC-39 AD1 EcoRI 59-CTCGTAGACTGCGTACC-39 AD2 EcoRI 59-AATTGGTACGCAGTCTAC-39 ApaI+0 59-GACTGCGTACAGGCCC-39 BglII+0 59-GAGTACACTGTCGATCT-39 HindIII+1 59-GACTGCGTACCAGCTTA-39 PstI+0 59-GACTGCGTACATGCAG-39 EcoRI+0 59-GACTGCGTACCAATT C-39 SpeI+0 59-GACTGCGTACCCTAGT-39 *5 indicates tested on five strains of each species of Bordetella (except a single isolate of B. petrii); 81 indicates tested on all 81 Bordetella strains in the study. IP:

6 A. Gzyl and others Restriction fragments tagged with specific adapters were selectively amplified with relevant primers. PCR was performed in a 20 ìl volume in a Biometra thermal cycler. Each amplification reaction mixture contained 5 ìl of a 10-fold diluted ligation reaction sample, 12 ng Cy5- labelled primer, 60 ng non-labelled primer, 120 ìm each of datp, dgtp, dctp and dttp, 1. 0 U AccuTaq-LA DNA polymerase (Sigma), 3. 0 mm MgCl 2, 50 mm Tris/HCl, 15 mm ammonium sulphate and 0.1 % Tween 20. Amplification for previously described sets was performed according to recommended conditions (Table 2). In the AFLP amplification step with SpeI primers tested with BglII, HindIII, PstI, ApaI or EcoRI ones, the following conditions were used: cycle 1, 94 8C for 30 s, 65 8C for 30 s, 72 8C for 60 s; cycles 2 13, the same PCR profile as seen in the first cycle except for a stepwise 0.7 8C decrease in the annealing temperature in each subsequent cycle of the 12 cycles; cycles 14 36, 94 8C for 30 s, 56 8C for 30 s, 72 8C for 60 s. After completion of the PCR, 5 ìl of each reaction tube mixture was mixed with 3 ìl loading dye (Amersham Pharmacia), denatured at 90 8C for 3 min and applied to the gel. Selectively amplified fragments were separated through a ReproGel High Resolution (Amersham Pharmacia) in 0.53 TBE buffer on an ALFexpress DNA sequencer. Separation was done at 1500 V, 60 ma, 25 W for 420 min at 55 8C. To evaluate intra- and inter-gel differences and identity levels, a fluorescein-labelled molecular size marker (ALFexpress Sizer ) and an external reference strain were used as external size markers. Stored fluorograms were analysed with the GelCompar Software version 3.1 (Applied Maths) after conversion into TIF format. The track resolution was reduced, excluding the primer front and fragments higher than 500 bp. Dendrograms for cluster analysis were based on similarity matrices calculated from the Pearson product-moment correlation coefficient and the UPGMA algorithm (Everitt, 1993). RESULTS RAPD analysis of isolates at the Bordetella genus level For RAPD fingerprinting, six short (10 bp) primers previously described by Williams et al. (1990) were used. Of all the primers tested, two (1254 and 1247) were chosen to screen all of the collected isolates, since they enabled the most informative DNA profiles to be observed. The Bordetella strain-specific DNA arrays consisted of approximately 2 12 bands with a fragment-length distribution in the kb range. The level of similarity for reference strain patterns was at least 94 % within the same gel and at least 90 % between gels. The number of RAPD patterns was evaluated after consideration of the identity level of 90 %. The overall genetic similarities of all the RAPD patterns obtained using primers 1254 and 1247 within the dendrograms, as defined by the Pearson product-moment correlation coefficient, were % and %, respectively. RAPD arrays classified Bordetella isolates into several clusters. Yet, exact species-specific differentiation of patterns was not found, however B. avium isolates were found exclusively in single clusters in both of the dendrograms. Two clusters were found for the RAPD protocol using the 1254 primer: A, containing 72 non-bordetella avium isolates, and B, containing nine B. avium isolates. Cluster A, at the 63 % overall similarity level, consisted of two subclusters, AI and AII, and a single pattern that was specific for B. trematum no , defined as AIII. Eleven B. bronchiseptica and four B. trematum isolates clustered at a linkage level of 77 % within the AI subcluster. Subcluster AII patterns clustered at a level of 76 % and contained four clades AIIa d. The AIIa clade contained groups that were related to each other, including patterns specific for B. pertussis and B. holmesii strains, and a group with five B. parapertussis and a single B. hinzii isolate (no ). Polish isolates of B. parapertussis, B. trematum no , and B. bronchiseptica no formed the AIIb clade. Clade AIIc contained strains of B. hinzii (nos 10980, and 13494) and B. bronchiseptica no The AIId pattern was characteristic of B. petrii. In the 1247 RAPD set, at the 72 % similarity level, three main clusters were identified. Cluster RA contained patterns of all of the B. pertussis and B. parapertussis strains, and six isolates of B. bronchiseptica (nos 3540, 3543, 1033, , 1943 and ). Cluster RB, contained groups of related patterns for nine B. avium isolates and B. petrii (RBa), five B. holmesii isolates (RBc), B. hinzii nos 16211, and (RBd), B. trematum nos 5894 and 14993, and B. hinzii no (RBb), and B. trematum nos 16877, and (RBe). Cluster RC contained seven B. bronchiseptica isolates presenting almost 100 % as the similarity value. The B. petriispecific pattern has been classified as directly adjacent to the B. avium cluster (RBa). Within the B. bronchiseptica cluster, four non-species-specific pattern types were observed. Similarly, for B. hinzii and B. trematum three non-species-specific pattern types were found. For B. holmesii and for B. avium isolates single clusters have been generated (however, the RBa B. avium-like cluster contained B. petrii). No band variation was observed in RAPD band patterns obtained for DNA isolated from seven passages of a single strain or with DNA content in the range ng. Data obtained show that typing with the RAPD 1254 primer can easily identify B. avium from other Bordetella species, as species-specific patterns classified them into a separate branch of the dendrogram. Similarly, typing with 1247 primer may be of value for confirming B. avium and B. holmesii species. Analysis of RAPD profiling within species We found that primers used in RAPD typing hardly differentiated B. pertussis, B. parapertussis, B. avium and B. holmesii isolates (Table 3). Relevant isolates of these four species tested with RAPD using primer 1254 or 1247 presented similarity levels near identity value: 87. 9/86. 7%, 85. 6/91. 5%,95. 5/94. 8 % and 95. 0/97. 4 %, respectively. High homogeneity of patterns was also seen in the case of B. hinzii in RAPD fingerprinting with primer 1247 (91. 6 %). Typing allowed differentiation of B. bronchiseptica and B. trematum isolates at the level of 64. 3/49 % and 82. 2/66. 8 % in RAPD profiling using the of 1247/1254 primers, respectively. Some potential for differentiation using RAPD with primer 1254 was seen in the case of B. hinzii (80. 8 %). The greatest value for intra-species RAPD typing using these primers was for B. 338 IP: Journal of Medical Microbiology 54

7 AFLP in genotyping of isolates of Bordetella genus Table 3. Overall genetic similarity values as defined by the Pearson product-moment correlation coefficient and the UPGMA algorithm for Bordetella patterns obtained with RAPD or AFLP at the genus and species levels Species No. of isolates Overall genetic similarity (%)/number of patterns differentiated after consideration of the relevant identity levels obtained under study conditions RAPD 1254 RAPD 1247 AFLP EcoRI/MseI AFLP SpeI/ApaI Bordetella species / / / /51 B. pertussis / / / /16 B. parapertussis / / / /6 B. bronchiseptica /7 64.3/3 70.9/ /10 B. avium /1 94.8/1 56.6/9 47.6/8 B. hinzii /3 91.6/1 78.3/4 42.1/4 B. holmesii /1 97.4/1 72.8/4 76.1/4 B. trematum /3 82.2/2 71.4/5 10.7/4 B. petrii 1 bronchiseptica and B. trematum, since for 13 and five strains tested, seven and three RAPD patterns, respectively, could be differentiated. Screening of AFLP sets Preliminary screening of 12 different AFLP schemes revealed that all of them except the HindIII AFLP set were able to fingerprint Bordetella isolates efficiently (Table 2). Profiles obtained for screened isolates visually appeared to differentiate particular species. For genotyping of all the strains in the study, the EcoRI/MseI and SpeI/ApaI AFLP sets were chosen. The profiles of these sets presented two options of band distribution: higher (EcoRI/MseI, 16 39) and lower (SpeI/ApaI, 6 25) numbers of bands within the profiles. Overall intra-/inter-gel similarity levels for the external reference strain were 97/92 %, respectively, and thus the number of patterns obtained relates to the identity level evaluated at 92 %. Under the conditions tested, the method was able to discriminate the analysed strains at both the species and the inter-species level. Overall genetic similarities defined by Pearson product-moment correlation coefficient for banding patterns of Bordetella isolates, reached values of % and 4. 3 % for the EcoRI/MseI and SpeI/ApaI AFLP sets, respectively. Cluster analysis of the AFLP EcoRI/MseI patterns identified five pattern groups, at the 50 % overall similarity level, each consisting of strains belonging to a single species with the exception of B. pertussis, B. parapertussis and B. bronchiseptica, which all clustered together (Fig. 1). The clusters identified were:, nine isolates of B. avium; E2, five isolates of B. trematum; E3, five isolates of B. holmesii; E4, four isolates of B. hinzii; E5, 57 isolates of B. bronchiseptica, B. pertussis and B. parapertussis; and a singly classified B. petriispecific pattern (E6). The closest relationship and lowest diversity were found for patterns clustering at the 67 % level of similarity within E5. Four subclusters were found within E5: E5a, three isolates of B. bronchiseptica;, 10 isolates of B. bronchiseptica and five B. parapertussis;, containing exclusively B. pertussis strains; and E5d, four Polish isolates of B. parapertussis. The subcluster specific for B. pertussis showed a 79 % similarity value. The nine isolates of B. avium showed nine AFLP patterns, which clustered at a linkage level of 74 % into the cluster. The E2 cluster contained five different AFLP patterns for the five B. trematum strains, which associated together at a linkage level of 72 %. The five isolates of B. holmesii, presenting four different patterns, clustered into E3 at a linkage level of 73 %. The four analysed B. hinzii isolates revealed four AFLP patterns which grouped at a linkage level of 78 %. The B. pertussis, B. parapertussis and B. bronchiseptica isolates showed six, five and 12 different AFLP patterns, respectively, and clustered at a linkage level of 69 %. Dendrograms constructed separately for isolates of eight particular species from the EcoRI/MseI AFLP patterns presented similarity values in the range of % except for B. avium (56. 6 %) (Table 3). The highest number of different patterns was observed within the SpeI/ApaI AFLP set evaluated as the most discriminatory (Table 2, Fig. 2). At the 30 % similarity level, eight clusters ( ) were differentiated. Within the cluster, nine B. avium isolates clustered at a linkage level of 46 %. Two clusters S2 (three isolates of B. hinzii, nos 10980, and 14052) and S3 (B. trematum and 5894) clustered together at a linkage level of 20 %. S6 (three isolates of B. trematum nos 15543, and and B. hinzii no ) and S5 (all B. holmesii isolates in the study) clustered at linkage levels of 69 % and 78 %, respectively. contained B. pertussis strains that were highly related to each other and clustered at a linkage level of 59 %. Within also, isolates of IP:

8 A. Gzyl and others Similarity (%) Species Strain No Cluster E2 E E E E E E E E E3 E E E E E5a 3543 E5a 3537 E5a / / / /60 7/ / / /74 W /97 785/ / / /73 253/ /68 A/63 45/68 2/62 Tohama I 9/68 12/ / / / / /74 871/ / /00 E5d 46/00 E5d 1251/ /75 E5d E5d B.petrii 4/03 E6 Fig. 1. Genotypic relationships among AFLP patterns obtained with EcoRI/MseI set for strains of Bordetella genus. The dendrogram was generated with similarity matrices calculated from the Pearson product-moment correlation coefficient and the UPGMA algorithm. 340 IP: Journal of Medical Microbiology 54

9 AFLP in genotyping of isolates of Bordetella genus Similarity (%) Species Strain No. Cluster 3538 a 3599/ / /73 a 6901/73 3/60 2/ / / / /00 9/68 2/68 W /74 12/60 7/ / / / /73 A/63 785/ B.petrii 2245/ / /99 6/62 Tohama I / /67 4/ S7 S6 S6 S6 S6 S S5 S5 S / / /00 S5 S4a S4a S4a S4a S4a S4a S4a 46/ S4c S4c S4c S4c S4c S2 S2 S Fig. 2. Genotypic relationships among AFLP patterns obtained with SpeI/ApaI set for strains of Bordetella genus. The dendrogram was generated with similarity matrices calculated from the Pearson product-moment correlation coefficient and the UPGMA algorithm IP:

10 A. Gzyl and others B. bronchiseptica (no. 3538) and B. pertussis (no. 3599/97) were classified as clade a and clustered with the B. pertussis clade at a linkage level of of 32 %. B. petrii was classified as a single type pattern (S7) related to the cluster. Cluster S4, with the overall similarity value of 35 %, formed three subclusters: S4a, seven isolates of B. bronchiseptica;, nine B. parapertussis isolates and two B. bronchiseptica, nos 3540 and 3534; and S4c, B. bronchiseptica nos 1943, 3543 and Within the SpeI/ApaI AFLP set, 16, 10 and six different patterns for the 35, 13 and nine isolates of B. pertussis, B. bronchiseptica and B. parapertussis were found, respectively. For B. trematum, B. hinzii, B. holmesii and B. avium, five, four, four and eight patterns were identified, respectively. DNA isolates obtained from seven subcultures of two different strains repeated over time were used to compare specific pattern reproducibility. The band patterns did not change after seven passages in any of the strains tested (data not shown). Similarly, differing quantities of DNA in the range of ng did not induce band variation in the AFLP profile of a single strain. DISCUSSION In this study, we describe AFLP as a new tool for generating the detailed genotyping data of isolates belonging to the Bordetella genus. The purpose of the study was to evaluate the discrimination level of intra- and inter-species fingerprinting by AFLP/RAPD methods for use in epidemiology and in searching for real associations between Bordetella species in terms of previously described genetic relationships. For RAPD fingerprinting we have chosen two primers previously found by Yuk et al. (1998) to be useful in the study of genotypes of B. parapertussis. From six previously described (Grady et al., 1999; Huys et al., 1996; Kokotovic et al., 1999; McLauchlin et al., 2000; van Elderle et al., 1999; Vos & Kuiper, 1998), and five newly designed and applied AFLP protocols, two were chosen to test discrimination ability of representative isolates of the Bordetella genus. In order to overcome the liability of the RAPD results according to the procedures and reagents used (Heersma et al., 2001; Olive & Bean, 1999) and to achieve higher levels of AFLP reproducibility, a standardized methodology, including commercially available kits for DNA isolation and gels, was applied. General approaches for reliable and accurate intra- and inter-gel comparison of banding patterns were included (Desai et al., 1998; Janssen, 2001; Savelkoul et al., 1999). The reproducibility was tested by generating multiple subcultures on which DNA extraction, RAPD and AFLP procedures were performed. Intra-/inter-gel reproducibility found identity levels for the RAPD and AFLP protocols as 94/90 % and 97/91 %, respectively. The overall genetic similarities, defined by the Pearson product-moment correlation coefficient, were % and 64 % for RAPD using primers 1254 and 1247, respectively, and % and 4. 4 % for AFLP with EcoRI/MseI and SpeI/ ApaI protocols, respectively. Although RAPD can provide analysis data very rapidly, the discriminatory power in comparison to AFLP, as predicted, was much lower. Although the limitations of the RAPD method are wellknown, it seems that it has the potential to be a quick tool for intra-species differentiation of some species of Bordetella. For B. bronchiseptica, RAPD was found to be highly accurate in tracking differences among isolates coming from different mammalian species; thus in this case it might well provide an additional tool to study genetic diversity. In addition, RAPD can also be useful for confirming B. avium, B. parapertussis and B. holmesii identification. Previously, AFLP has been found to be useful for both identifying and typing many micro-organisms after defining windows of similarity (Savelkoul et al., 1999). The definition of the similarity value ranges in AFLP genotyping was found to be useful in the taxonomic studies of for example Klebsiella, Mycoplasma, Acinetobacter, Aeromonas and Xanthomonas species (Huys et al., 1996, Kokotovic et al., 1999). Moreover, in some cases, the discriminatory power has been evaluated as equal to or higher than that obtained with the gold standard method of PFGE (van Elderle et al., 1999), and congruent with similar grouping through multilocus sequence typing (MLST; Schouls et al., 2003). AFLP has been widely accepted as an excellent genotyping tool because of its typability and discriminatory power, and good reproducibility; likewise, it is easy to perform and to draw interpretations from (van Belkum et al., 2001). AFLP fingerprints of the isolates of the Bordetella genus, as revealed by cluster analysis, have retained attributes of species specificity, leading to the conclusion that the method may be used as an additional tool for species-specific identification. In our study, four B. parapertussis isolates from Poland were first incorrectly identified as B. pertussis in the diagnostic laboratory. The AFLP technique clustered them not within the highly related group of B. pertussis but within the B. bronchiseptica cluster. By applying PCR (van der Zee et al., 1996a), their identification as B. parapertussis was confirmed. Additionally, because of high reproducibility and easy performance AFLP has value in epidemiological studies of the Bordetella. Slightly different classifications of Bordetella isolates in the two AFLP sets screened might result from the different levels of polymorphism detected. Greater levels of discrimination of fingerprinting in the case of AFLP with the SpeI/ApaI protocol might allow for a more in-depth examination of polymorphism compared to AFLP with the EcoRI/MseI set. The AFLP set with the EcoRI/MseI protocol seems to be a potential alternative for studying genetic relationships among Bordetella species. The DNA DNA hybridization of nucleotide sequences of the pertussis toxin operon and MLEE found B. pertussis, B. parapertussis and B. bronchiseptica to be highly related and with limited diversity(arico et al., 1987; Khattak& Matthews, 1993; Muller & Hildebrandt, 1993; Musser et al., 1986; van der Zee et al., 1997). Genetic differences within the B. bronchiseptica cluster were suggested to be difficult to detect because the mechanism of regulation of expression 342 IP: Journal of Medical Microbiology 54

11 AFLP in genotyping of isolates of Bordetella genus seems to be species-specific. Different strategies in regulating gene expression have been suggested to be more important for the adaptation of particular lineages of the B. bronchiseptica cluster to the particular host than horizontal gene transfer (Gerlach et al., 2001). On the other hand, adaptation to different hosts has been suggested to be a force capable of inducing differences, which could be observed by some tools (Gerlach et al., 2001). Insertion sequence (IS)-dependent rearrangements responsible for different genome organization and size were assumed to be the most important changes possible to observe (Cummings et al., 2004). As for the B. bronchiseptica cluster and other Bordetella isolates, the genetic differences generated could be observed through AFLP and the method can be considered as an additional tool to study their relationships. The different extents of intra-species polymorphism detected seem to confirm the varying degrees of genetic diversity among analysed species (Kokotovic et al., 1999). For B. pertussis, substantial homogeneity was found both in RAPD and AFLP genotyping compared with B. bronchiseptica. As such, in MLEE in the case of B. bronchiseptica as many as 38 polymorphic types have been identified in comparison to the four to five found for B. pertussis and B. parapertussis (van der Zee et al., 1997). Higher genetic variation seen in B. bronchiseptica versus B. pertussis and B. parapertussis has also been found in other studies (Boursaux-Eude & Guiso, 2000; van der Zee et al., 1996b). The AFLP fingerprinting data are in agreement with previously published data on the close relationship of the B. bronchiseptica cluster, since B. pertussis, B. parapertussis and B. bronchiseptica were strictly classified as a single branch during the clusterization process. Moreover, fingerprints of B. parapertussis and B. bronchiseptica isolates were mixed together in the clade, unlike the highly related group of B. pertussis isolates. Comparative analyses of the genome sequences revealed that B. parapertussis and B. bronchiseptica are more similar in overall genome organization than B. bronchiseptica and B. pertussis (Cummings et al., 2004; Parkhill et al., 2003). Phenotype analysis by in vitro and in vivo assays also has shown that B. parapertussis is more similar to B. bronchiseptica than to B. pertussis, thus the phenotype is related to the genetic species-specific differences (Heininger et al., 2002). B. pertussis isolates were found to be hardly differentiated by RAPD and AFLP using the EcoRI/MseI protocol compared with AFLP using the SpeI/ApaI protocol. MLEE, PFGE and IS-RFLP studies have shown the population of B. pertussis to be homogeneous or clonal, resulting most probably from high adaptation to humans and limited opportunity for horizontal genetic exchange (Musser et al., 1986; van Loo et al., 1999; van der Zee et al., 1996a). IS sequences are considered to be reasons for genome plasticity of B. pertussis coupled with high-genetic-relatedness (Gerlach et al., 2001; Stibitz & Yang, 1997). PFGE as a gold standard has been successfully used to study B. pertussis challenges with time, identify outbreakassociated isolates, monitor transmission and evaluate genetic relatedness (Hardwick et al., 2002a; Khattak et al., 1992; Mooi et al., 2000; Peppler et al., 2003; Weber et al., 2001). For strains belonging to the B. bronchiseptica cluster, using PFGE it was possible to differentiate isolates between and within species (Khattak & Matthews, 1993). Moreover, it has been successfully applied for differentiation of isolates of B. parapertussis, B. bronchiseptica and B. holmesii (Hardwick et al., 2002a, 2002b; Khattak etal., 1992; Makinen et al., 2003; Mastrantonio et al., 1998; Mazengia et al., 2000; Peppler et al., 2003; Weber et al., 2001). Although PFGE is the most reproducible method, it is time-consuming and labour-intensive. MLST studies (van Loo et al., 2002), although highly useful for tracing the movements of B. pertussis strains within and between individuals and communities, need sophistically equipped laboratories and highly experienced personnel. We postulate that AFLP using the SpeI/ApaI protocol may be a new, rapid and reproducible option for the genotyping of B. pertussis. The epidemiology of B. bronchiseptica is still not well enough elucidated, especially in terms of the possibility of host interspecies transmission. Some studies have documented cospecies transmission between dogs and cats (Binns et al., 1998; Dawson et al., 2000; Foley et al., 2002), from rabbits to humans (Gueirard et al., 1995) and from rats/cats to pigs (Switzer et al., 1966). Available genotyping tools involve restriction enzyme analysis (REA) of chromosomal DNA, ribotyping, PFGE and RAPD (described for canine isolates) (Keil & Fenwick, 1999; Sacco et al., 2000b). Genetic diversity of B. bronchiseptica isolates from 12 different host animals measured by REA was considerable in the study by Sacco et al. (2000b), with similarity ranges of % for HinfI and % for AluI. Thus AFLP using SpeI/ApeI might present a better alternative because of its higher discrimination level (30. 1 %). In our study, we found similar heterogeneity of B. bronchiseptica strains isolated from different host animals using the AFLP SpeI/ApaI protocol and even using as simple an approach as RAPD. Further studies are expected to show a degree of variation detected by AFLP in a higher number of B. bronchiseptica strains exclusively isolated from particular hosts, e.g. swine and dogs. In addition, AFLP SpeI/ApaI fingerprinting might be a new option to the previously described alternatives to follow the genotype trends for vaccine and clinical strains in the population of B. bronchiseptica in dogs, similarly as it is possible for B. pertussis in humans (Keil & Fenwick, 1999; Mooi et al., 1998, 1999; van der Zee et al., 1996a; Fry et al., 2001; Gzyl et al., 2001). Among nine B. parapertussis isolates, four and six different patterns in the AFLP EcoRI/MseI and SpeI/ApaI sets were observed, respectively, proving its high capability for differentiation. A higher number of B. parapertussis isolates is expected to be studied and compared with AFLP results with previously described data on limited genetic variability (Heininger et al., 2002; Khattak & Matthews, 1993; Makinen et al., 2003; Mastrantonio et al., 1998; Porter et al., 1996). For 32 B. holmesii isolates recovered from nasopharyngeal specimens up to five pulsofield types were identified, which IP:

12 A. Gzyl and others supports restricted diversity of this species (Mazengia et al., 2000). In our study, although only five isolates of B. holmesii were available for testing, four different patterns in both AFLP protocols were observed. REA and ribotyping have been described as useful tools for discriminating between B. avium and B. hinzii species and for epidemiological purposes (Register et al., 2003; Sacco et al., 2000a). Similarly, in our study we found AFLP and RAPD to be useful tools for identification of B. avium identification among other species of the Bordetella genus; for the nine strains in the study, seven and six different AFLP fingerprints obtained with EcoRI/MseI and SpeI/ApaI protocols were observed, respectively. In the case of B. hinzii, for both AFLP sets tested, the four isolates presented four different patterns. It has been proposed that the evolution of Bordetella is associated with the adaptation to particular hosts of environmental bacteria (Gerlach et al., 2001). Isolation of B. petrii as the first environmental Bordetella of anaerobic metabolism (von Wintzingerode et al., 2001) has further supported the evolutionary relationship of Bordetella with its environmental ancestor. In the AFLP EcoRI/MseI protocol, with the highest potential for taxonomic discrimination of the two studied, the B. petrii isolate came last in the relationship with other Bordetellae. Our study has described the potential of AFLP for discrimination of isolates of the Bordetella genus. However, a higher number of strains is expected to be analysed for a precise confirmation of these preliminary data. The newly designed AFLP protocol with SpeI/ApaI seems to possess the highest discrimination potential, and especially as regards B. pertussis might become a first-choice method after exact standardization with gold-standard methods. Furthermore, it seems that AFLP has some potential for the genetic phylogenetic studies of Bordetella; however, studies with higher numbers of isolates and standardization using tools such as 16S rdna, PFGE, MLEE, MLST and IS-RFLP should first be performed for exact confirmation. Finally, AFLP data obtained with two different sets of restriction enzymes should be compared to well-established methods such as MLEE, MLST and PFGE (Mooi et al., 2000; van Loo et al., 2002; van der Zee et al., 1997; Advani et al., 2004) in order to establish the mostprobable valid evolutionary pathway. Additionally, some other different AFLP procedures, tested in the preliminary screening on only a limited number of strains and not on the whole strain collection used here, remain available for assessment of their value for genotyping of isolates of different species of Bordetella. ACKNOWLEDGEMENTS This work was partly supported by a grant from the State Committee for Scientific Research (2PO5D04726). We would like to express our gratitude to Miss Marta Balińska for help with English version of the manuscript and to Ms Barbara Husejnow for her excellent technical assistance. REFERENCES Advani, A., Donnelly, D., Hallander, H. (2004). Reference system for characterization of Boredetella pertussis pulsed-field gel electrophoresis profiles. J Clin Microbiol 42, Arico, B., Gross, R., Smida, J. & Rappuoli, R. (1987). Evolutionary relationships in the genus Bordetella. Mol Microbiol 1, Binns, S. H., Speakman, A. J., Dawson, S., Bennett, M., Gaskell, R. M. & Hart, C. A. (1998). The use of pulsed-field gel electrophoresis to examine the epidemiology of Bordetella bronchiseptica isolated from cats and other species. Epidemiol Infect 120, Boursaux-Eude, C. & Guiso, N. (2000). Polymorphism of repeated regions of pertactin in Bordetella pertussis, Bordetella parapertussis, and Bordetella bronchiseptica. Infect Immun 68, Cookson, B. T., Vandamme, P., Carlson, L. C., Larson, A. M., Sheffield, J. V., Kersters, K. & Spach, D. H. (1994). Bacteriemia caused by a novel Bordetella species B. hinzii. J Clin Microbiol 32, Cummings, C. A., Brinig, M. M., Lepp, P. W., van de Pas, S. & Relman, D. A. (2004). Bordetella species are distinguished by patterns of substantial gene loss and host adaptation. J Bacteriol 186, Dawson, S., Jones, D., McCracken, C. M., Gaskell, R. M., Hart, C. A. & Gaskell, C. J. (2000). Bordetella bronchiseptica infection in cats following contact with infected dogs. Vet Rec 146, Desai, M., Tanna, A., Wall, R., Efstratiou, A., Georgie, R. & Stanley, J. (1998). Fluorescent amplified-fragment length polymorphism analysis of an outbreak of group A Streptococcal invasive disease. J Clin Microbiol 36, Dworkin, M. S., Sullivan, P. S., Buskin, S. E., Harrington, R. D., Olliffe, J., MacArthur, R. D. & Lopez, C. E. (1999). Bordetella bronchiseptica infection in human immunodeficiency virus-infected patients. Clin Infect Dis 28, Everitt, B. (1993). Cluster Analysis, 3rd edn. London: Arnold. Foley, J. E., Rand, C., Bannasch, M. J., Norris, C. R. & Milan, J. (2002). Molecular epidemiology of feline bordetellosis in two animal shelters in California, USA. Prev Vet Med 54, Fry, N. K., Neal, S., Harrison, T. G., Miller, E., Matthews, R. & George, R. C. (2001). Genotypic variation in the Bordetella pertussis virulence factors pertactin and pertussis toxin in historical and recent clinical isolates in the United Kingdom. Infect Immun 69, Funke, G., Hess, T., von Graevenitz, A. & Vandamme, P. (1996). Characteristics of Bordetella hinzii strains isolated from a cystic fibrosis patient over a 3-year period. J Clin Microbiol 34, Gerlach, G., von Wintzingerode, F., Middendorf, B. & Gross, R. (2001). Evolutionary trends in the genus Bordetella. Microbes Infect 3, Grady, R., Desai, M., O Neill, G., Cookson, B. & Stanley, J. (1999). Genotyping of epidemic methicillin-resistant Staphylococcus aureus phage type 15 isolates by fluorescent amplified-fragment length polymorphism analysis. J Clin Microbiol 37, Gueirard, P., Weber, C., Le Coustumier, A. & Guiso, N. (1995). Human Bordetella bronchiseptica infection related to contact with infected animals: persistence of bacteria in host. J Clin Microbiol 33, Gzyl, A., Augustynowicz, E., van Loo, I. & Slusarczyk, J. (2001). Temporal nucleotide changes in pertactin and pertussis toxin genes in Bordetella pertussis strains isolated from clinical cases in Poland. Vaccine 20, Hardwick, T. H., Cassiday, P. K., Weyant, R. S., Bisgard, K. M. & Sanden, G. N. (2002a). Changes in predominance and diversity of genomic subtypes of Bordetella pertussis isolated in the United States, 1935 to Emerg Infect Dis 8, Hardwick, T. H., Plikaytis, B., Cassiday, P. K., Cage, G., Peppler, M. S., Shea, D., Boxrud, D. & Sanden, G. N. (2002b). Reproducibility of 344 IP: Journal of Medical Microbiology 54