- Top of page
- Materials and methods
Aims: Thirty Campylobacter jejuni strains isolated from fecal samples (n = 94; 32%) from 13 positive farms (n = 17; 76%) from commercial broiler chickens in Puerto Rico were analysed by molecular methods.
Methods and Results: Isolates were identified with multiplex polymerase chain reaction assays, tested for their antimicrobial susceptibility and characterized with pulsed-field gel electrophoresis (PFGE), multilocus sequence typing (MLST), serotyping and bacterial cytotoxicity in mammalian cells. Isolates exhibited high resistance to vancomycin (minimum inhibitory concentration, MIC of >256 μg ml−1) and trimethoprim (MIC of >32 μg ml−1); few were resistant to clindamycin (MIC90 4 μg ml−1), erythromycin (MIC90 8 μg ml−1) and tetracycline (MIC90 8 μg ml−1); but none was resistant to azithromycin (MIC90 4 μg ml−1), ciprofloxacin (MIC90 1 μg ml−1) or gentamycin (MIC90 4 μg ml−1). Most strains restricted with SmaI, but a combination of SmaI–KpnI digestion was more discriminatory. MLST analysis yielded four sequence types (ST), and ST-2624 was the predominant one. Phylogenetic analysis revealed a high degree of recombination for glnA and pgm genes. The predominant serotypes were O:3 and O:5. Most strains had lowest cytotoxicity potential with Caco-2 cells, medium cytotoxicity with INT-407 and Hep-2 cells and high cytotoxicity with CHO cells.
Conclusion: A low degree of antimicrobial resistance, 13 PFGE profiles, 4 ST and a large variability in cytotoxicity assays were found for these strains.
Significance and Impact of the Study: This is the first characterization of C. jejuni strains isolated from broilers in Puerto Rico. The genetic diversity of these strains suggests that several techniques are needed for strain characterization.
- Top of page
- Materials and methods
Little information is available about human campylobacteriosis in Puerto Rico. After Salmonella, Campylobacter jejuni has been reported as the second most common foodborne bacterial pathogen isolated from humans, with an incidence of 3·5%, from a reference laboratory in San Juan (Lopez-Ortiz and Solivan 1999). However, there is no information on the incidence and characteristics of Campylobacter spp. in commercial broilers raised in the island. The poultry industry is the second largest agricultural industry in Puerto Rico, with broiler meat comprising more than 69% of the total meat consumed, with an average of 80 lb of broiler meat consumed per capita for the year 2005. Yet, from the total consumption of broiler meat only 34% is produced locally while the rest is imported as frozen products from the United States (Anon 2005a).
In C. jejuni, epidemic clones responsible for human disease are characterized by unique combinations of virulence genes or their alleles (Maynard Smith et al. 1993; Musser 1996), and some clones have been associated with domestic animals raised for human consumption, primarily broiler chickens. Therefore, it is assumed that contaminated raw poultry products and consuming undercooked poultry are major risk factors for human campylobacteriosis (Butzler and Oosterom 1991; Tauxe 1997; Nadeau et al. 2002). However, drawing conclusions about the pathogenicity potential of C. jejuni strains from different sources has always been a challenge. The weak clonality of C. jejuni brings several limitations when accurate estimations are needed for source tracking efforts aimed at determining the attribution of infections to food sources. Besides the differences in the interpretation of the results from the same techniques applied to different C. jejuni isolates, different molecular approaches may answer specific questions but may not be complementary to each other to help identifying sources and improve the prediction of the appearance of epidemic clones in foods. Few studies have addressed the complementation of methods for epidemiological studies. The use of genomotyping may represent a holistic approach to identifying potentially pathogenic clones. Yet, these methods still do not provide a complete overview on the variability in the genomic profiles among C. jejuni strains from different areas (Kärenlampi et al. 2007). In addition, sequencing methods such as multilocus sequence typing (MLST) are more expensive than molecular typing methods based on restriction profiles, such as pulsed-field gel electrophoresis (PFGE). With this in view, we collected samples from commercial live broilers in Puerto Rico to isolate Campylobacter strains and analyse them with molecular techniques to understand their genomic composition and determine the virulence potential. We collected Campylobacter strains from faecal material taken from commercial broiler chickens. These strains were analysed using different molecular techniques to determine possible traits that would predict their risk for humans. For this purpose, the isolates were: (i) identified to the species level using culture media and multiplex polymerase chain reaction (mPCR) assays; (ii) tested for their antimicrobial susceptibility, and analysed using PFGE, MLST and serotyping. In addition, their cytotoxicity potential in mammalian cell lines was investigated.
- Top of page
- Materials and methods
We pursued this study to determine the prevalent Campylobacter species that colonizes commercial broiler flocks in Puerto Rico and to characterize these isolates using PFGE, MLST, antimicrobial susceptibility, heat-stable serotyping and the invasiveness in mammalian cell cultures. The year average temperature in Puerto Rico is between 70° and 90°F, and the temperature recorded outside the farms was between 75° and 80°F. Only few samples were positive by direct plating, but 32% of the samples were positive after enrichment in Preston broth. This number of positive flocks is similar to the 57% reported for commercial broilers in the southeast United States (Stern et al. 2001; Potturi-Venkata et al. 2007a). Therefore, the shipping of the samples under refrigeration for several days may have accounted for a reduction in the number of campylobacters in the sample, but did not result in a detectable reduction in the prevalence of positive flocks (56%) or farms (76%). The flocks were sampled two weeks before processing, when Campylobacter is usually present in high numbers (Potturi-Venkata et al. 2007a; b). The methodology used was optimized with combinations of plates and enrichment broth for the isolation of Campylobacter spp. from faecal material collected from live broilers (Potturi-Venkata et al. 2007b).
The general incidence of resistance among the antimicrobials tested with the disc diffusion method was low. Multiresistance was not found among these isolates and may be related to the low usage of antimicrobials for disease treatment in animal production in Puerto Rico. The disc diffusion method appears to be reliable for the determination of MIC for erythromycin, ciprofloxacin and tetracycline for C. jejuni and C. coli (Gaudreau and Gilbert 1997). Although gentamicin and erythromycin resistance is usually low in C. jejuni, erythromycin resistance in C. jejuni from chicken appears to be increasing in Canada (Lévesque et al. 2007). Erythromycin is still an important antimicrobial for treatment of campylobacteriosis in humans. The increase in the inoculum used to seed the plates for the disc diffusion method resulted in more consistent and reliable results recorded at 24 h of incubation, and similar suggestions on the preparation of the inocula and MIC interpretation have been reported to improve the per cent agreement between the disc diffusion method and the agar dilution method for erythromycin (Luangtongkum et al. 2007).
None of the isolates exhibited resistance to ciprofloxacin. Our results differed from the antimicrobial resistance reported for C. jejuni in Trinidad and Tobago, where 87% of the C. jejuni broiler strains were found to be resistant to sulfa/trimethoprim and ciprofloxacin, three antimicrobials that are apparently used routinely by the poultry industry in that country (Rodrigo et al. 2007). In France, the resistance to ciprofloxacin by C. jejuni from broilers decreased from 31% in 2002 to 9% in 2004, a decrease that has been suggested to be the benefits of policies limiting the use of antimicrobials in food animals (Gallay et al. 2007).
The incidence of tetracycline resistance was lower than the incidence recorded for organic or conventionally grown poultry in the United States (Luangtongkum et al. 2006), or in Canada, where more than 60% of the C. jejuni isolates from chickens are resistant to tetracycline and 8–40% are resistant to ciprofloxacin (Anon 2005b; Kos et al. 2006). The sequencing of c. 90% of the chromosomally encoded gene that confers tetracycline resistance confirmed that this gene is highly conserved in C. jejuni strains isolated from humans and domestic animals (Gibreel et al. 2004).
Studies that compared the restriction patterns of SmaI and KpnI of human isolates of C. jejuni have concluded that KpnI is the enzyme of choice for molecular epidemiology studies of C. jejuni (Michaud et al. 2001; Kärenlampi et al. 2003). However, we found that the use of SmaI appears to be better than KpnI for PFGE analysis of C. jejuni from live broilers. SmaI always restricts more C. jejuni strains than KpnI (Michaud et al. 2001), and it is the primary restriction enzyme for the protocols developed by the CDC, CampyNet (http://campynet.vetinst.dk/PFGE.html) and studies in Denmark (Nielsen et al. 2006). SalI normally generates partial digest despite the experimental conditions (Ho and Monaco 1995), a feature that we noticed by the presence of unresolved fragments in the compression zone of the gels for all the strains. Therefore, this enzyme should be limited in its use for restriction profiling for fingerprinting purposes. The combination of restriction with SmaI followed by restriction with KpnI resulted in a larger number of bands and resulted in more PFGE profiles (Table 5). In addition, the combination of molecular methods demonstrated that more than one PFGE profile may be present in one farm, in different houses (F10-H1-S3 and F10-H2-S4) or in the same house or flock (F11-H2-S4 and F11-H2-S5) (Table 5). It is important to restrict the plugs first with SmaI and then with KpnI to guarantee that a larger number of strains will be restricted. The increase in the number of fragments with the SmaI/KpnI combination was more obvious in the area below 250 kbp (Fig. 2); an element that made the PFGE profiles more challenging to analyse using the visual scoring system suggested by Tenover et al. (1995). Therefore, the use of fingerprinting analysis software, such as BioNumerics, appears to be indispensable for a reliable analysis of the results obtained with this combination.
Table 5. Pulsed-field gel electrophoresis (PFGE) macrorestriction profiles, sequence type (ST) allelic profiles and serotyping results of Campylobacter jejuni strains isolated from five different farms
|Isolate ID||Isolation (day/place)||PFGE profiles||ST allelic profiles||Serotyping|
The lack of improvement by adding formaldehyde in the PFGE protocol strongly suggests that the lack of restriction patterns found for some isolates and specific enzymes is not related to the presence of DNAses in those strains (Gibson et al. 1994), but may be related to the lack of restriction sites or restriction site methylation. In large eukaryotic genomes (>108 base pairs), this lack of restriction, especially the CpG nucleotide deficiency, has been usually associated with DNA methylation, which is almost universal in these organisms (Bestor 1990). DNA methylation involves the addition of a methyl group to DNA bases, primarily to carbon #5 of the cytosine pyrimidine ring. This theory has traditionally been applied to explain the lack of restriction of bacterial genome. However, recent findings question this assertion and suggest that cytosine methylation is not the primary reason for the CpG dinucleotide deficiency in bacterial genomes (Wang et al. 2004).
PFGE restriction with SmaI has been reported to be more discriminating than MLST for outbreak investigations, although a combination of MLST plus sequencing of the flaA short variable region may provide a level of discrimination equivalent to PFGE for outbreak investigations (Sails et al. 2003). Our comparison of the PFGE, MLST and serotyping profiles of C. jejuni strains isolated in the same geographical area suggests that the PFGE profiles may change quickly within C. jejuni strains and may account for the variety of genotypes frequently found when analysing C. jejuni with PFGE (Table 5).
It is apparent that a ST profile of C. jejuni, or clone, appears to dominate in a geographic area for a variable period. The prevalent ST profile in our report is a new combination of previously described alleles. The other three MLST profiles found have been identified in C. jejuni isolated from humans, animals and the environment. ST-460 has been reported by the Centers for Disease Control and Prevention, and ST-48 and ST-353 were described in the United Kingdom (Dingle et al. 2001, 2002). ST-48 was the most predominant ST in a study done in New South Wales, Australia (Mickan et al. 2007), and has been associated to serotype O:4 of the Penner serotyping system (Penner and Hennessey 1980). ST-353 has been associated with O:3, O:11 and O:37 of the heat-resistant serotyping system (Dingle et al. 2002). In our study ST-48 was represented by O:3 and O:5 serotypes, while ST-353 was represented only by serotype O:3. ST-2624 was represented by all the serotypes found in this study, including the minor O:21.
Within the individual genes, glnA and pgm showed the highest diversity (four alleles) and the analysis of their sequence diversity showed a variability that has been described for other genes in C. jejuni (Suerbaum et al. 2001). Both the diversity within genes and the analysis of the allelic profile data using SplitsTree and the UPGMA cluster analysis resulted in graphs with alleles connected to each other by multiple pathways, with a network-like structure suggesting recombination (Suerbaum et al. 2001). pgm has been used to analyse the recombination of C. jejuni from MLST data, and along with unc (atpA, ATP synthase alpha subunit) they appear to come to C. jejuni from other Campylobacter spp. (Fearnhead et al. 2005). In our study, the diversity of the allelic profiles for unc was minimal, with two alleles and only one base difference (C vs T) in base 864 of the atpA gene (NCBI accession number AL111168). This is different from the variety of allelic profiles, with multiple polymorphic sites, reported by French et al. (2005) for unc alleles analysed from C. jejuni associated to cattle in the United Kingdom. It appears that the allelic variability for each of the seven housekeeping genes for MLST profiles varies according to the origin of the strains and different geographical areas.
We calculated both the index of association (IA), as described for the analysis of the degree of linkage in multilocus enzyme electrophoresis (Maynard Smith et al. 1993; Maynard Smith 1999), and the standardized IA (sIA) that corrects for the number of loci analysed (Haubold and Hudson 2000). The sIA value of 0·73 and a calculated variance that was significantly higher than the expected variance (P ≤ 0·0127) suggested a high degree of recombination among the MLST genes within this population of C. jejuni isolates (Table 3).
MLST results have revealed that populations of different bacteria frequently exchange genetic material, and produce interclonal variance in virulence traits. These findings based on molecular population genetic studies have resulted in the tenet that the unit of bacterial pathogenicity is the clone, or cell line (Musser 1996). However, the analyses of linkage disequilibrium on comparison of gene trees have revealed that few bacteria are indeed clonal (Maynard Smith et al. 1993; Spratt and Maiden 1999; Feil et al. 2001; Supply et al. 2003). In C. jejuni, a high frequency of inter- and intraspecies recombination may account for the high degree of variation (addition and or deletions) of restriction sites and PFGE profiles found in C. jejuni isolated from live broilers chickens (Dingle et al. 2001; Schouls et al. 2003; Potturi-Venkata et al. 2007a,b) and retail broiler products (Dickins et al. 2002; Oyarzabal et al. 2007a). Therefore, MLST profiling, although useful in determining temporal phylogenetic comparison, may be less discriminating for spatial, short-term epidemiological studies of C. jejuni strains from live broilers. The fact that we found only four ST allelic profiles, with one being completely new, vs 10 different PFGE profiles suggest that microrestriction profiling is still a powerful system for the tracking of C. jejuni strains.
Most of the studies of cytotoxicity on Campylobacter spp. make use of tissue culture cell models, with a subjective, qualitative characterization of the degree of morphological cell rounding. We measured instead the supernatant LDH that was released upon cell lysis. Our results measuring LDH from the cytotoxicity assays in cell lines showed a wide range of variation that has been described using different mammalian cell lines on C. jejuni isolated from different sources, including broiler chickens (Gilbert and Slavik 2004; Coote et al. 2007). The variability of these assays is dependent on the way the assays are performed, may appear to even change with strains over time (Coote et al. 2007), and demonstrate that cytotoxin production by C. jejuni is quite complex when compared with that of other enteric pathogens (Misawa et al. 1995). Our results showed that some strains had a high agreement among the results from the same cell line, while other strains showed no agreement among the results from the same cell line. The number of isolates that grouped in the low cytotoxicity category was much smaller than previously reported (Gilbert and Slavik 2004). Although there is no direct correlation between the pathogenicity of C. jejuni and the presence of the toxins (Lam 1993), cytotoxicity assays have been used to infer the potential pathogenicity of C. jejuni isolates. For instance, there is an important correlation between invasiveness and colitis in patients from which the isolates were obtained, although some strains attach to cell lines but do not invade the cells (Everest et al. 1992). The findings of low toxicity reported with CHO (Nadeau et al. 2003; Coote et al. 2007) do not correlate with the present results.
Globally, there are differences in Campylobacter antigens and on the typability of C. jejuni strains with Penner antisera and it is common to encounter strains that are untypable with this method (Eyles et al. 2006; Nielsen et al. 2006). Some of the strains from this study were not recognized by any antisera, and a weak serotyping found with O:21 makes this serotype less significant, as the titre was too low to be compared with the titres obtained with O:3 and O:5. Serotype O:5 has been found in different geographical areas, and has recently been described in C. jejuni isolates from humans and water sources in New Zealand (Eyles et al. 2006). Campylobacter jejuni serotypes O:3 and O:5 have been found in the blood cultures and diarrhoetic stools of paediatric patients (Lastovica 1996). Serotype O:21 has been also found in paediatric diarrhoeic patients, in chickens without disease and in terminally ill ostriches with enteritis in South Africa (A.J. Lastovica, unpublished results).
The molecular characterization of bacterial pathogens is essential to understand the short- and long-term epidemiological implications of a given foodborne pathogen (Spratt and Maiden 1999). The long-term goal of an active surveillance programme is to predict, with high degree of certainty, the rise and spread of an epidemic clone with high pathogenicity potential for humans (Maynard Smith et al. 1993). For weakly clonal organisms molecular methods have yielded different data, and there appears to be a lack of agreement among methods applied to the same group of strains. For instance, different databases have been collected and maintained for C. jejuni, such as fingerprinting profiles based on PFGE or sequencing data based on MLST. MLST results in 100% typability of the isolates, but the cost of this method (Dingle et al. 2002) and the apparent low number of alleles found in a relatively small population of isolates may limit its application for epidemiological studies. Yet, the prediction of the appearance of epidemic clones in a given geographic area may still be limited even if an active surveillance is in place. The difficulty in interpretation of a wide range of results is a severe limitation in studies on pathogens with high capacity for horizontal gene transfer (Oyarzabal et al. 2007b), where even typability results may vary according to the different methods. The small population of C. jejuni that we sampled may be a limiting factor of this study, but at the same time exemplifies some of the limitation of each of the methods employed for strain characterization. The diversity found within C. jejuni and the lack of agreement among the different methods suggests that studies of larger populations of strains will result in more complex information to analyse. In conclusion, these results suggest that the use of more than two typing methods may be necessary to fully understand the variability of C. jejuni strains colonizing commercial broilers in Puerto Rico.